From: Sascha Steinbiss Date: Sun, 15 Jan 2017 08:08:53 +0000 (+0000) Subject: Import libzstd_1.1.2.orig.tar.gz X-Git-Tag: archive/raspbian/1.3.3+dfsg-2+rpi1~2^2^2~10 X-Git-Url: https://dgit.raspbian.org/?a=commitdiff_plain;h=abfb703840a88cf03b9f3c1fa054c7513b9429eb;p=libzstd.git Import libzstd_1.1.2.orig.tar.gz [dgit import orig libzstd_1.1.2.orig.tar.gz] --- abfb703840a88cf03b9f3c1fa054c7513b9429eb diff --git a/.gitattributes b/.gitattributes new file mode 100644 index 0000000..6212bd4 --- /dev/null +++ b/.gitattributes @@ -0,0 +1,21 @@ +# Set the default behavior +* text eol=lf + +# Explicitly declare source files +*.c text eol=lf +*.h text eol=lf + +# Denote files that should not be modified. +*.odt binary +*.png binary + +# Visual Studio +*.sln text eol=crlf +*.vcxproj* text eol=crlf +*.vcproj* text eol=crlf +*.suo binary +*.rc text eol=crlf + +# Windows +*.bat text eol=crlf +*.cmd text eol=crlf diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000..796a696 --- /dev/null +++ b/.gitignore @@ -0,0 +1,38 @@ +# Object files +*.o +*.ko +*.dSYM + +# Libraries +*.lib +*.a + +# Shared objects (inc. Windows DLLs) +*.dll +*.so +*.so.* +*.dylib + +# Executables +zstd +*.exe +*.out +*.app + +# Test artefacts +tmp* +dictionary* + +# Other files +.directory +_codelite/ +_zstdbench/ +.clang_complete +*.idea +*.swp +.DS_Store +googletest/ +*.d + +# Directories +bin/ diff --git a/.travis.yml b/.travis.yml new file mode 100644 index 0000000..148a98f --- /dev/null +++ b/.travis.yml @@ -0,0 +1,148 @@ +language: c +matrix: + fast_finish: true + include: + # OS X Mavericks + - env: Ubu=OS_X_Mavericks Cmd="make gnu90test && make clean && make test && make clean && make travis-install" + os: osx + + + # Container-based Ubuntu 12.04 LTS Server Edition 64 bit (doesn't support 32-bit includes) + - env: Ubu=12.04cont Cmd="make test && make clean && make travis-install" + os: linux + sudo: false + + - env: Ubu=12.04cont Cmd="make zlibwrapper && make clean && make -C tests test-zstd-nolegacy && make clean && make cmaketest && make clean && make -C contrib/pzstd googletest pzstd tests check && make -C contrib/pzstd clean" + os: linux + sudo: false + language: cpp + install: + - export CXX="g++-4.8" CC="gcc-4.8" + - export TESTFLAGS='--gtest_filter=-*ExtremelyLarge*' + addons: + apt: + sources: + - ubuntu-toolchain-r-test + packages: + - gcc-4.8 + - g++-4.8 + + - env: Ubu=12.04cont Cmd="make usan" + os: linux + sudo: false + + - env: Ubu=12.04cont Cmd="make asan" + os: linux + sudo: false + + + # Standard Ubuntu 12.04 LTS Server Edition 64 bit + - env: Ubu=12.04 Cmd="make -C programs zstd-small zstd-decompress zstd-compress && make -C tests test-gzstd && make -C programs clean && make -C tests versionsTest" + os: linux + sudo: required + + - env: Ubu=12.04 Cmd="make asan32" + os: linux + sudo: required + addons: + apt: + sources: + - ubuntu-toolchain-r-test + packages: + - libc6-dev-i386 + - gcc-multilib + + - env: Ubu=12.04 Cmd='cd contrib/pzstd && make googletest && make tsan && make check && make clean && make asan && make check && make clean && cd ../..' + os: linux + sudo: required + install: + - export CXX="g++-6" CC="gcc-6" + - export LDFLAGS="-fuse-ld=gold" + - export TESTFLAGS='--gtest_filter=-*ExtremelyLarge*' + addons: + apt: + sources: + - ubuntu-toolchain-r-test + packages: + - gcc-6 + - g++-6 + + + # Ubuntu 14.04 LTS Server Edition 64 bit + - env: Ubu=14.04 Cmd="make armtest && make clean && make aarch64test" + dist: trusty + sudo: required + addons: + apt: + packages: + - qemu-system-arm + - qemu-user-static + - gcc-arm-linux-gnueabi + - libc6-dev-armel-cross + - gcc-aarch64-linux-gnu + - libc6-dev-arm64-cross + + - env: Ubu=14.04 Cmd='make ppctest && make clean && make ppc64test' + dist: trusty + sudo: required + addons: + apt: + packages: + - qemu-system-ppc + - qemu-user-static + - gcc-powerpc-linux-gnu + + - env: Ubu=14.04 Cmd='make -C lib all && CFLAGS="-O1 -g" make -C zlibWrapper valgrindTest && make -C tests valgrindTest' + os: linux + dist: trusty + sudo: required + addons: + apt: + packages: + - valgrind + + - env: Ubu=14.04 Cmd="make gpptest && make clean && make gnu90test && make clean && make c99test && make clean && make gnu99test && make clean && make clangtest && make clean && make -C contrib/pzstd googletest32 && make -C contrib/pzstd all32 && make -C contrib/pzstd check && make -C contrib/pzstd clean" + os: linux + dist: trusty + sudo: required + install: + - export CXX="g++-4.8" CC="gcc-4.8" + addons: + apt: + packages: + - libc6-dev-i386 + - g++-multilib + - gcc-4.8 + - gcc-4.8-multilib + - g++-4.8 + - g++-4.8-multilib + + - env: Ubu=14.04 Cmd="make -C tests test32" + os: linux + dist: trusty + sudo: required + addons: + apt: + packages: + - libc6-dev-i386 + - gcc-multilib + + - env: Ubu=14.04 Cmd="make gcc5test && make clean && make gcc6test" + os: linux + dist: trusty + sudo: required + addons: + apt: + sources: + - ubuntu-toolchain-r-test + packages: + - gcc-multilib + - gcc-5 + - gcc-5-multilib + - gcc-6 + - gcc-6-multilib + +script: + - JOB_NUMBER=$(echo $TRAVIS_JOB_NUMBER | sed -e 's:[0-9][0-9]*\.\(.*\):\1:') + # - if [ $JOB_NUMBER -eq 9 ] || [ $JOB_NUMBER -eq 10 ]; then sh -c "$Cmd"; fi + - sh -c "$Cmd" diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md new file mode 100644 index 0000000..edf5b7f --- /dev/null +++ b/CONTRIBUTING.md @@ -0,0 +1,42 @@ +# Contributing to Zstandard +We want to make contributing to this project as easy and transparent as +possible. + +## Our Development Process +New versions are being developed in the "dev" branch, +or in their own feature branch. +When they are deemed ready for a release, they are merged into "master". + +As a consequences, all contributions must stage first through "dev" +or their own feature branch. + +## Pull Requests +We actively welcome your pull requests. + +1. Fork the repo and create your branch from `dev`. +2. If you've added code that should be tested, add tests. +3. If you've changed APIs, update the documentation. +4. Ensure the test suite passes. +5. Make sure your code lints. +6. If you haven't already, complete the Contributor License Agreement ("CLA"). + +## Contributor License Agreement ("CLA") +In order to accept your pull request, we need you to submit a CLA. You only need +to do this once to work on any of Facebook's open source projects. + +Complete your CLA here: + +## Issues +We use GitHub issues to track public bugs. Please ensure your description is +clear and has sufficient instructions to be able to reproduce the issue. + +Facebook has a [bounty program](https://www.facebook.com/whitehat/) for the safe +disclosure of security bugs. In those cases, please go through the process +outlined on that page and do not file a public issue. + +## Coding Style +* 4 spaces for indentation rather than tabs + +## License +By contributing to Zstandard, you agree that your contributions will be licensed +under the [LICENSE](LICENSE) file in the root directory of this source tree. diff --git a/LICENSE b/LICENSE new file mode 100644 index 0000000..a793a80 --- /dev/null +++ b/LICENSE @@ -0,0 +1,30 @@ +BSD License + +For Zstandard software + +Copyright (c) 2016-present, Facebook, Inc. All rights reserved. + +Redistribution and use in source and binary forms, with or without modification, +are permitted provided that the following conditions are met: + + * Redistributions of source code must retain the above copyright notice, this + list of conditions and the following disclaimer. + + * Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the documentation + and/or other materials provided with the distribution. + + * Neither the name Facebook nor the names of its contributors may be used to + endorse or promote products derived from this software without specific + prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. diff --git a/LICENSE-examples b/LICENSE-examples new file mode 100644 index 0000000..1de7813 --- /dev/null +++ b/LICENSE-examples @@ -0,0 +1,11 @@ +Copyright (c) 2016-present, Facebook, Inc. All rights reserved. + +The examples provided by Facebook are for non-commercial testing and evaluation +purposes only. Facebook reserves all rights not expressly granted. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL +FACEBOOK BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. diff --git a/Makefile b/Makefile new file mode 100644 index 0000000..bb3a4e4 --- /dev/null +++ b/Makefile @@ -0,0 +1,170 @@ +# ################################################################ +# Copyright (c) 2016-present, Yann Collet, Facebook, Inc. +# All rights reserved. +# +# This source code is licensed under the BSD-style license found in the +# LICENSE file in the root directory of this source tree. An additional grant +# of patent rights can be found in the PATENTS file in the same directory. +# ################################################################ + +PRGDIR = programs +ZSTDDIR = lib +BUILDIR = build +ZWRAPDIR = zlibWrapper +TESTDIR = tests + +# Define nul output +VOID = /dev/null + +ifneq (,$(filter Windows%,$(OS))) +EXT =.exe +else +EXT = +endif + +.PHONY: default +default: lib zstd + +.PHONY: all +all: + $(MAKE) -C $(ZSTDDIR) $@ + $(MAKE) -C $(PRGDIR) $@ zstd32 + $(MAKE) -C $(TESTDIR) $@ all32 + +.PHONY: lib +lib: + @$(MAKE) -C $(ZSTDDIR) + +.PHONY: zstd +zstd: + @$(MAKE) -C $(PRGDIR) $@ + cp $(PRGDIR)/zstd$(EXT) . + +.PHONY: zlibwrapper +zlibwrapper: + $(MAKE) -C $(ZWRAPDIR) test + +.PHONY: test +test: + $(MAKE) -C $(TESTDIR) $@ + +.PHONY: clean +clean: + @$(MAKE) -C $(ZSTDDIR) $@ > $(VOID) + @$(MAKE) -C $(PRGDIR) $@ > $(VOID) + @$(MAKE) -C $(TESTDIR) $@ > $(VOID) + @$(MAKE) -C $(ZWRAPDIR) $@ > $(VOID) + @$(RM) zstd$(EXT) tmp* + @echo Cleaning completed + + +#------------------------------------------------------------------------------ +# make install is validated only for Linux, OSX, Hurd and some BSD targets +#------------------------------------------------------------------------------ +ifneq (,$(filter $(shell uname),Linux Darwin GNU/kFreeBSD GNU FreeBSD DragonFly NetBSD)) +HOST_OS = POSIX +.PHONY: install uninstall travis-install clangtest gpptest armtest usan asan uasan + +install: + @$(MAKE) -C $(ZSTDDIR) $@ + @$(MAKE) -C $(PRGDIR) $@ + +uninstall: + @$(MAKE) -C $(ZSTDDIR) $@ + @$(MAKE) -C $(PRGDIR) $@ + +travis-install: + $(MAKE) install PREFIX=~/install_test_dir + +gpptest: clean + $(MAKE) -C programs all CC=g++ CFLAGS="-O3 -Wall -Wextra -Wundef -Wshadow -Wcast-align -Werror" + +gcc5test: clean + gcc-5 -v + $(MAKE) all CC=gcc-5 MOREFLAGS="-Werror" + +gcc6test: clean + gcc-6 -v + $(MAKE) all CC=gcc-6 MOREFLAGS="-Werror" + +clangtest: clean + clang -v + $(MAKE) all CC=clang MOREFLAGS="-Werror -Wconversion -Wno-sign-conversion -Wdocumentation" + +armtest: clean + $(MAKE) -C $(TESTDIR) datagen # use native, faster + $(MAKE) -C $(TESTDIR) test CC=arm-linux-gnueabi-gcc QEMU_SYS=qemu-arm-static ZSTDRTTEST= MOREFLAGS="-Werror -static" + +aarch64test: + $(MAKE) -C $(TESTDIR) datagen # use native, faster + $(MAKE) -C $(TESTDIR) test CC=aarch64-linux-gnu-gcc QEMU_SYS=qemu-aarch64-static ZSTDRTTEST= MOREFLAGS="-Werror -static" + +ppctest: clean + $(MAKE) -C $(TESTDIR) datagen # use native, faster + $(MAKE) -C $(TESTDIR) test CC=powerpc-linux-gnu-gcc QEMU_SYS=qemu-ppc-static ZSTDRTTEST= MOREFLAGS="-Werror -Wno-attributes -static" + +ppc64test: clean + $(MAKE) -C $(TESTDIR) datagen # use native, faster + $(MAKE) -C $(TESTDIR) test CC=powerpc-linux-gnu-gcc QEMU_SYS=qemu-ppc64-static ZSTDRTTEST= MOREFLAGS="-m64 -static" + +usan: clean + $(MAKE) test CC=clang MOREFLAGS="-g -fsanitize=undefined" + +asan: clean + $(MAKE) test CC=clang MOREFLAGS="-g -fsanitize=address" + +msan: clean + $(MAKE) test CC=clang MOREFLAGS="-g -fsanitize=memory -fno-omit-frame-pointer" # datagen.c fails this test for no obvious reason + +asan32: clean + $(MAKE) -C $(TESTDIR) test32 CC=clang MOREFLAGS="-g -fsanitize=address" + +uasan: clean + $(MAKE) test CC=clang MOREFLAGS="-g -fsanitize=address -fsanitize=undefined" + +endif + + +ifneq (,$(filter MSYS%,$(shell uname))) +HOST_OS = MSYS +CMAKE_PARAMS = -G"MSYS Makefiles" +endif + + +#------------------------------------------------------------------------ +#make tests validated only for MSYS, Linux, OSX, kFreeBSD and Hurd targets +#------------------------------------------------------------------------ +ifneq (,$(filter $(HOST_OS),MSYS POSIX)) +cmaketest: + cmake --version + $(RM) -r $(BUILDIR)/cmake/build + mkdir $(BUILDIR)/cmake/build + cd $(BUILDIR)/cmake/build ; cmake -DPREFIX:STRING=~/install_test_dir $(CMAKE_PARAMS) .. ; $(MAKE) install ; $(MAKE) uninstall + +c90test: clean + CFLAGS="-std=c90" $(MAKE) all # will fail, due to // and long long + +gnu90test: clean + CFLAGS="-std=gnu90" $(MAKE) all + +c99test: clean + CFLAGS="-std=c99" $(MAKE) all + +gnu99test: clean + CFLAGS="-std=gnu99" $(MAKE) all + +c11test: clean + CFLAGS="-std=c11" $(MAKE) all + +bmix64test: clean + CFLAGS="-O3 -mbmi -Werror" $(MAKE) -C $(TESTDIR) test + +bmix32test: clean + CFLAGS="-O3 -mbmi -mx32 -Werror" $(MAKE) -C $(TESTDIR) test + +bmi32test: clean + CFLAGS="-O3 -mbmi -m32 -Werror" $(MAKE) -C $(TESTDIR) test + +staticAnalyze: clean + CPPFLAGS=-g scan-build --status-bugs -v $(MAKE) all +endif diff --git a/NEWS b/NEWS new file mode 100644 index 0000000..9b00016 --- /dev/null +++ b/NEWS @@ -0,0 +1,229 @@ +v1.1.2 +API : streaming : decompression : changed : automatic implicit reset when chain-decoding new frames without init +API : experimental : added : dictID retrieval functions, and ZSTD_initCStream_srcSize() +API : zbuff : changed : prototypes now generate deprecation warnings +lib : improved : faster decompression speed at ultra compression settings and 32-bits mode +lib : changed : only public ZSTD_ symbols are now exposed +lib : changed : reduced usage of stack memory +lib : fixed : several corner case bugs, by Nick Terrell +cli : new : gzstd, experimental version able to decode .gz files, by Przemyslaw Skibinski +cli : new : preserve file attributes +cli : new : added zstdless and zstdgrep tools +cli : fixed : status displays total amount decoded, even for file consisting of multiple frames (like pzstd) +cli : fixed : zstdcat +zlib_wrapper : added support for gz* functions, by Przemyslaw Skibinski +install : better compatibility with FreeBSD, by Dimitry Andric +source tree : changed : zbuff source files moved to lib/deprecated + +v1.1.1 +New : command -M#, --memory=, --memlimit=, --memlimit-decompress= to limit allowed memory consumption +New : doc/zstd_manual.html, by Przemyslaw Skibinski +Improved : slightly better compression ratio at --ultra levels (>= 20) +Improved : better memory usage when using streaming compression API, thanks to @Rogier-5 report +Added : API : ZSTD_initCStream_usingCDict(), ZSTD_initDStream_usingDDict() (experimental section) +Added : example/multiple_streaming_compression.c +Changed : zstd_errors.h is now installed within /include (and replaces errors_public.h) +Updated man page +Fixed : zstd-small, zstd-compress and zstd-decompress compilation targets + +v1.1.0 +New : contrib/pzstd, parallel version of zstd, by Nick Terrell +added : NetBSD install target (#338) +Improved : speed for batches of small files +Improved : speed of zlib wrapper, by Przemyslaw Skibinski +Changed : libzstd on Windows supports legacy formats, by Christophe Chevalier +Fixed : CLI -d output to stdout by default when input is stdin (#322) +Fixed : CLI correctly detects console on Mac OS-X +Fixed : CLI supports recursive mode `-r` on Mac OS-X +Fixed : Legacy decoders use unified error codes, reported by benrg (#341), fixed by Przemyslaw Skibinski +Fixed : compatibility with OpenBSD, reported by Juan Francisco Cantero Hurtado (#319) +Fixed : compatibility with Hurd, by Przemyslaw Skibinski (#365) +Fixed : zstd-pgo, reported by octoploid (#329) + +v1.0.0 +Change Licensing, all project is now BSD, Copyright Facebook +Small decompression speed improvement +API : Streaming API supports legacy format +API : ZDICT_getDictID(), ZSTD_sizeof_{CCtx, DCtx, CStream, DStream}(), ZSTD_setDStreamParamter() +CLI supports legacy formats v0.4+ +Fixed : compression fails on certain huge files, reported by Jesse McGrew +Enhanced documentation, by Przemyslaw Skibinski + +v0.8.1 +New streaming API +Changed : --ultra now enables levels beyond 19 +Changed : -i# now selects benchmark time in second +Fixed : ZSTD_compress* can now compress > 4 GB in a single pass, reported by Nick Terrell +Fixed : speed regression on specific patterns (#272) +Fixed : support for Z_SYNC_FLUSH, by Dmitry Krot (#291) +Fixed : ICC compilation, by Przemyslaw Skibinski + +v0.8.0 +Improved : better speed on clang and gcc -O2, thanks to Eric Biggers +New : Build on FreeBSD and DragonFly, thanks to JrMarino +Changed : modified API : ZSTD_compressEnd() +Fixed : legacy mode with ZSTD_HEAPMODE=0, by Christopher Bergqvist +Fixed : premature end of frame when zero-sized raw block, reported by Eric Biggers +Fixed : large dictionaries (> 384 KB), reported by Ilona Papava +Fixed : checksum correctly checked in single-pass mode +Fixed : combined --test amd --rm, reported by Andreas M. Nilsson +Modified : minor compression level adaptations +Updated : compression format specification to v0.2.0 +changed : zstd.h moved to /lib directory + +v0.7.5 +Transition version, supporting decoding of v0.8.x + +v0.7.4 +Added : homebrew for Mac, by Daniel Cade +Added : more examples +Fixed : segfault when using small dictionaries, reported by Felix Handte +Modified : default compression level for CLI is now 3 +Updated : specification, to v0.1.1 + +v0.7.3 +New : compression format specification +New : `--` separator, stating that all following arguments are file names. Suggested by Chip Turner. +New : `ZSTD_getDecompressedSize()` +New : OpenBSD target, by Juan Francisco Cantero Hurtado +New : `examples` directory +fixed : dictBuilder using HC levels, reported by Bartosz Taudul +fixed : legacy support from ZSTD_decompress_usingDDict(), reported by Felix Handte +fixed : multi-blocks decoding with intermediate uncompressed blocks, reported by Greg Slazinski +modified : removed "mem.h" and "error_public.h" dependencies from "zstd.h" (experimental section) +modified : legacy functions no longer need magic number + +v0.7.2 +fixed : ZSTD_decompressBlock() using multiple consecutive blocks. Reported by Greg Slazinski. +fixed : potential segfault on very large files (many gigabytes). Reported by Chip Turner. +fixed : CLI displays system error message when destination file cannot be created (#231). Reported by Chip Turner. + +v0.7.1 +fixed : ZBUFF_compressEnd() called multiple times with too small `dst` buffer, reported by Christophe Chevalier +fixed : dictBuilder fails if first sample is too small, reported by Руслан Ковалёв +fixed : corruption issue, reported by cj +modified : checksum enabled by default in command line mode + +v0.7.0 +New : Support for directory compression, using `-r`, thanks to Przemyslaw Skibinski +New : Command `--rm`, to remove source file after successful de/compression +New : Visual build scripts, by Christophe Chevalier +New : Support for Sparse File-systems (do not use space for zero-filled sectors) +New : Frame checksum support +New : Support pass-through mode (when using `-df`) +API : more efficient Dictionary API : `ZSTD_compress_usingCDict()`, `ZSTD_decompress_usingDDict()` +API : create dictionary files from custom content, by Giuseppe Ottaviano +API : support for custom malloc/free functions +New : controllable Dictionary ID +New : Support for skippable frames + +v0.6.1 +New : zlib wrapper API, thanks to Przemyslaw Skibinski +New : Ability to compile compressor / decompressor separately +Changed : new lib directory structure +Fixed : Legacy codec v0.5 compatible with dictionary decompression +Fixed : Decoder corruption error (#173) +Fixed : null-string roundtrip (#176) +New : benchmark mode can select directory as input +Experimental : midipix support, VMS support + +v0.6.0 +Stronger high compression modes, thanks to Przemyslaw Skibinski +API : ZSTD_getFrameParams() provides size of decompressed content +New : highest compression modes require `--ultra` command to fully unleash their capacity +Fixed : zstd cli return error code > 0 and removes dst file artifact when decompression fails, thanks to Chip Turner + +v0.5.1 +New : Optimal parsing => Very high compression modes, thanks to Przemyslaw Skibinski +Changed : Dictionary builder integrated into libzstd and zstd cli +Changed (!) : zstd cli now uses "multiple input files" as default mode. See `zstd -h`. +Fix : high compression modes for big-endian platforms +New : zstd cli : `-t` | `--test` command + +v0.5.0 +New : dictionary builder utility +Changed : streaming & dictionary API +Improved : better compression of small data + +v0.4.7 +Improved : small compression speed improvement in HC mode +Changed : `zstd_decompress.c` has ZSTD_LEGACY_SUPPORT to 0 by default +fix : bt search bug + +v0.4.6 +fix : fast compression mode on Windows +New : cmake configuration file, thanks to Artyom Dymchenko +Improved : high compression mode on repetitive data +New : block-level API +New : ZSTD_duplicateCCtx() + +v0.4.5 +new : -m/--multiple : compress/decompress multiple files + +v0.4.4 +Fixed : high compression modes for Windows 32 bits +new : external dictionary API extended to buffered mode and accessible through command line +new : windows DLL project, thanks to Christophe Chevalier + +v0.4.3 : +new : external dictionary API +new : zstd-frugal + +v0.4.2 : +Generic minor improvements for small blocks +Fixed : big-endian compatibility, by Peter Harris (#85) + +v0.4.1 +Fixed : ZSTD_LEGACY_SUPPORT=0 build mode (reported by Luben) +removed `zstd.c` + +v0.4.0 +Command line utility compatible with high compression levels +Removed zstdhc => merged into zstd +Added : ZBUFF API (see zstd_buffered.h) +Rolling buffer support + +v0.3.6 +small blocks params + +v0.3.5 +minor generic compression improvements + +v0.3.4 +Faster fast cLevels + +v0.3.3 +Small compression ratio improvement + +v0.3.2 +Fixed Visual Studio + +v0.3.1 : +Small compression ratio improvement + +v0.3 +HC mode : compression levels 2-26 + +v0.2.2 +Fix : Visual Studio 2013 & 2015 release compilation, by Christophe Chevalier + +v0.2.1 +Fix : Read errors, advanced fuzzer tests, by Hanno Böck + +v0.2.0 +**Breaking format change** +Faster decompression speed +Can still decode v0.1 format + +v0.1.3 +fix uninitialization warning, reported by Evan Nemerson + +v0.1.2 +frame concatenation support + +v0.1.1 +fix compression bug +detects write-flush errors + +v0.1.0 +first release diff --git a/PATENTS b/PATENTS new file mode 100644 index 0000000..15b4a2e --- /dev/null +++ b/PATENTS @@ -0,0 +1,33 @@ +Additional Grant of Patent Rights Version 2 + +"Software" means the Zstandard software distributed by Facebook, Inc. + +Facebook, Inc. ("Facebook") hereby grants to each recipient of the Software +("you") a perpetual, worldwide, royalty-free, non-exclusive, irrevocable +(subject to the termination provision below) license under any Necessary +Claims, to make, have made, use, sell, offer to sell, import, and otherwise +transfer the Software. For avoidance of doubt, no license is granted under +Facebook’s rights in any patent claims that are infringed by (i) modifications +to the Software made by you or any third party or (ii) the Software in +combination with any software or other technology. + +The license granted hereunder will terminate, automatically and without notice, +if you (or any of your subsidiaries, corporate affiliates or agents) initiate +directly or indirectly, or take a direct financial interest in, any Patent +Assertion: (i) against Facebook or any of its subsidiaries or corporate +affiliates, (ii) against any party if such Patent Assertion arises in whole or +in part from any software, technology, product or service of Facebook or any of +its subsidiaries or corporate affiliates, or (iii) against any party relating +to the Software. Notwithstanding the foregoing, if Facebook or any of its +subsidiaries or corporate affiliates files a lawsuit alleging patent +infringement against you in the first instance, and you respond by filing a +patent infringement counterclaim in that lawsuit against that party that is +unrelated to the Software, the license granted hereunder will not terminate +under section (i) of this paragraph due to such counterclaim. + +A "Necessary Claim" is a claim of a patent owned by Facebook that is +necessarily infringed by the Software standing alone. + +A "Patent Assertion" is any lawsuit or other action alleging direct, indirect, +or contributory infringement or inducement to infringe any patent, including a +cross-claim or counterclaim. diff --git a/README.md b/README.md new file mode 100644 index 0000000..694bfd5 --- /dev/null +++ b/README.md @@ -0,0 +1,124 @@ + __Zstandard__, or `zstd` as short version, is a fast lossless compression algorithm, + targeting real-time compression scenarios at zlib-level and better compression ratios. + +It is provided as an open-source BSD-licensed **C** library, +and a command line utility producing and decoding `.zst` compressed files. +For other programming languages, +you can consult a list of known ports on [Zstandard homepage](http://www.zstd.net/#other-languages). + +|Branch |Status | +|------------|---------| +|master | [![Build Status](https://travis-ci.org/facebook/zstd.svg?branch=master)](https://travis-ci.org/facebook/zstd) | +|dev | [![Build Status](https://travis-ci.org/facebook/zstd.svg?branch=dev)](https://travis-ci.org/facebook/zstd) | + +As a reference, several fast compression algorithms were tested and compared on a Core i7-3930K CPU @ 4.5GHz, using [lzbench], an open-source in-memory benchmark by @inikep compiled with GCC 5.4.0, with the [Silesia compression corpus]. + +[lzbench]: https://github.com/inikep/lzbench +[Silesia compression corpus]: http://sun.aei.polsl.pl/~sdeor/index.php?page=silesia + + +| Name | Ratio | C.speed | D.speed | +|-------------------------|-------|--------:|--------:| +| | | MB/s | MB/s | +| **zstd 0.8.2 -1** |**2.877**| **330** | **940** | +| [zlib] 1.2.8 deflate -1 | 2.730 | 95 | 360 | +| brotli 0.4 -0 | 2.708 | 320 | 375 | +| QuickLZ 1.5 | 2.237 | 510 | 605 | +| LZO 2.09 | 2.106 | 610 | 870 | +| [LZ4] r131 | 2.101 | 620 | 3100 | +| Snappy 1.1.3 | 2.091 | 480 | 1600 | +| LZF 3.6 | 2.077 | 375 | 790 | + +[zlib]:http://www.zlib.net/ +[LZ4]: http://www.lz4.org/ + +Zstd can also offer stronger compression ratios at the cost of compression speed. +Speed vs Compression trade-off is configurable by small increments. Decompression speed is preserved and remains roughly the same at all settings, a property shared by most LZ compression algorithms, such as [zlib] or lzma. + +The following tests were run on a Core i7-3930K CPU @ 4.5GHz, using [lzbench], an open-source in-memory benchmark by @inikep compiled with GCC 5.2.1, on the [Silesia compression corpus]. + +Compression Speed vs Ratio | Decompression Speed +---------------------------|-------------------- +![Compression Speed vs Ratio](doc/images/Cspeed4.png "Compression Speed vs Ratio") | ![Decompression Speed](doc/images/Dspeed4.png "Decompression Speed") + +Several algorithms can produce higher compression ratios, but at slower speeds, falling outside of the graph. +For a larger picture including very slow modes, [click on this link](doc/images/DCspeed5.png) . + + +### The case for Small Data compression + +Previous charts provide results applicable to typical file and stream scenarios (several MB). Small data comes with different perspectives. The smaller the amount of data to compress, the more difficult it is to achieve any significant compression. + +This problem is common to many compression algorithms. The reason is, compression algorithms learn from past data how to compress future data. But at the beginning of a new file, there is no "past" to build upon. + +To solve this situation, Zstd offers a __training mode__, which can be used to tune the algorithm for a selected type of data, by providing it with a few samples. The result of the training is stored in a file called "dictionary", which can be loaded before compression and decompression. Using this dictionary, the compression ratio achievable on small data improves dramatically: + +![Compressing Small Data](doc/images/smallData.png "Compressing Small Data") + +These compression gains are achieved while simultaneously providing faster compression and decompression speeds. + +Dictionary works if there is some correlation in a family of small data (there is no _universal dictionary_). +Hence, deploying one dictionary per type of data will provide the greatest benefits. Dictionary gains are mostly effective in the first few KB. Then, the compression algorithm will rely more and more on previously decoded content to compress the rest of the file. + +#### Dictionary compression How To : + +1) Create the dictionary + +`zstd --train FullPathToTrainingSet/* -o dictionaryName` + +2) Compress with dictionary + +`zstd FILE -D dictionaryName` + +3) Decompress with dictionary + +`zstd --decompress FILE.zst -D dictionaryName` + +### Build + +Once you have the repository cloned, there are multiple ways provided to build Zstandard. + +#### Makefile + +If your system is compatible with a standard `make` (or `gmake`) binary generator, +you can simply run it at the root directory. +It will generate `zstd` within root directory. + +Other available options include : +- `make install` : create and install zstd binary, library and man page +- `make test` : create and run `zstd` and test tools on local platform + +#### cmake + +A `cmake` project generator is provided within `build/cmake`. +It can generate Makefiles or other build scripts +to create `zstd` binary, and `libzstd` dynamic and static libraries. + +#### Visual (Windows) + +Going into `build` directory, you will find additional possibilities : +- Projects for Visual Studio 2005, 2008 and 2010 + + VS2010 project is compatible with VS2012, VS2013 and VS2015 +- Automated build scripts for Visual compiler by @KrzysFR , in `build/VS_scripts`, + which will build `zstd` cli and `libzstd` library without any need to open Visual Studio solution. + + +### Status + +Zstandard is currently deployed within Facebook. It is used daily to compress and decompress very large amounts of data in multiple formats and use cases. +Zstandard is considered safe for production environments. + +### License + +Zstandard is [BSD-licensed](LICENSE). We also provide an [additional patent grant](PATENTS). + +### Contributing + +The "dev" branch is the one where all contributions will be merged before reaching "master". +If you plan to propose a patch, please commit into the "dev" branch or its own feature branch. +Direct commit to "master" are not permitted. +For more information, please read [CONTRIBUTING](CONTRIBUTING.md). + +### Miscellaneous + +Zstd entropy stage is provided by [Huff0 and FSE, from Finite State Entropy library](https://github.com/Cyan4973/FiniteStateEntropy). diff --git a/appveyor.yml b/appveyor.yml new file mode 100644 index 0000000..27c83c0 --- /dev/null +++ b/appveyor.yml @@ -0,0 +1,171 @@ +version: 1.0.{build} +environment: + matrix: + - COMPILER: "gcc" + MAKE_PARAMS: "test" + PLATFORM: "mingw64" + - COMPILER: "gcc" + MAKE_PARAMS: "test" + PLATFORM: "mingw32" + - COMPILER: "visual" + CONFIGURATION: "Debug" + PLATFORM: "x64" + - COMPILER: "visual" + CONFIGURATION: "Debug" + PLATFORM: "Win32" + - COMPILER: "visual" + CONFIGURATION: "Release" + PLATFORM: "x64" + - COMPILER: "visual" + CONFIGURATION: "Release" + PLATFORM: "Win32" + +install: + - ECHO Installing %COMPILER% %PLATFORM% %CONFIGURATION% + - MKDIR bin + - if [%COMPILER%]==[gcc] SET PATH_ORIGINAL=%PATH% + - if [%COMPILER%]==[gcc] ( + SET "CLANG_PARAMS=-C tests zstd fullbench fuzzer paramgrill datagen CC=clang MOREFLAGS="--target=x86_64-w64-mingw32 -Werror -Wconversion -Wno-sign-conversion"" && + SET "PATH_MINGW32=c:\MinGW\bin;c:\MinGW\usr\bin" && + SET "PATH_MINGW64=c:\msys64\mingw64\bin;c:\msys64\usr\bin" && + COPY C:\msys64\usr\bin\make.exe C:\MinGW\bin\make.exe && + COPY C:\MinGW\bin\gcc.exe C:\MinGW\bin\cc.exe + ) else ( + IF [%PLATFORM%]==[x64] (SET ADDITIONALPARAM=/p:LibraryPath="C:\Program Files\Microsoft SDKs\Windows\v7.1\lib\x64;c:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\lib\amd64;C:\Program Files (x86)\Microsoft Visual Studio 10.0\;C:\Program Files (x86)\Microsoft Visual Studio 10.0\lib\amd64;") + ) + +build_script: + - ECHO Building %COMPILER% %PLATFORM% %CONFIGURATION% + - if [%PLATFORM%]==[mingw32] SET PATH=%PATH_MINGW32%;%PATH_ORIGINAL% + - if [%PLATFORM%]==[mingw64] SET PATH=%PATH_MINGW64%;%PATH_ORIGINAL% + - if [%PLATFORM%]==[mingw64] ( + make clean && + ECHO *** && + ECHO *** Building clang && + ECHO *** && + ECHO make %CLANG_PARAMS% && + make %CLANG_PARAMS% && + COPY tests\fuzzer.exe tests\fuzzer_clang.exe && + make clean && + ECHO *** && + ECHO *** Building pzstd for %PLATFORM% && + ECHO *** && + make -C contrib\pzstd googletest-mingw64 && + make -C contrib\pzstd pzstd.exe && + make -C contrib\pzstd tests && + make -C contrib\pzstd check && + make -C contrib\pzstd clean + ) + - if [%COMPILER%]==[gcc] ( + ECHO *** && + ECHO *** Building %PLATFORM% && + ECHO *** && + make -v && + cc -v && + ECHO make %MAKE_PARAMS% && + make %MAKE_PARAMS% + ) + - if [%COMPILER%]==[gcc] if [%PLATFORM%]==[mingw64] ( + COPY programs\zstd.exe bin\zstd.exe && + appveyor PushArtifact bin\zstd.exe + ) + - if [%COMPILER%]==[gcc] if [%PLATFORM%]==[mingw32] ( + COPY programs\zstd.exe bin\zstd32.exe && + appveyor PushArtifact bin\zstd32.exe + ) + - if [%COMPILER%]==[gcc] make clean + - if [%COMPILER%]==[visual] ( + ECHO *** && + ECHO *** Building Visual Studio 2008 %PLATFORM%\%CONFIGURATION% in %APPVEYOR_BUILD_FOLDER% && + ECHO *** && + msbuild "build\VS2008\zstd.sln" /m /verbosity:minimal /property:PlatformToolset=v90 /t:Clean,Build /p:Platform=%PLATFORM% /p:Configuration=%CONFIGURATION% /logger:"C:\Program Files\AppVeyor\BuildAgent\Appveyor.MSBuildLogger.dll" && + DIR build\VS2008\bin\%PLATFORM%\%CONFIGURATION%\*.exe && + MD5sum build/VS2008/bin/%PLATFORM%/%CONFIGURATION%/*.exe && + COPY build\VS2008\bin\%PLATFORM%\%CONFIGURATION%\fuzzer.exe tests\fuzzer_VS2008_%PLATFORM%_%CONFIGURATION%.exe && + ECHO *** && + ECHO *** Building Visual Studio 2010 %PLATFORM%\%CONFIGURATION% && + ECHO *** && + msbuild "build\VS2010\zstd.sln" %ADDITIONALPARAM% /m /verbosity:minimal /property:PlatformToolset=v100 /p:ForceImportBeforeCppTargets=%APPVEYOR_BUILD_FOLDER%\build\VS2010\CompileAsCpp.props /t:Clean,Build /p:Platform=%PLATFORM% /p:Configuration=%CONFIGURATION% /logger:"C:\Program Files\AppVeyor\BuildAgent\Appveyor.MSBuildLogger.dll" && + DIR build\VS2010\bin\%PLATFORM%_%CONFIGURATION%\*.exe && + MD5sum build/VS2010/bin/%PLATFORM%_%CONFIGURATION%/*.exe && + msbuild "build\VS2010\zstd.sln" %ADDITIONALPARAM% /m /verbosity:minimal /property:PlatformToolset=v100 /t:Clean,Build /p:Platform=%PLATFORM% /p:Configuration=%CONFIGURATION% /logger:"C:\Program Files\AppVeyor\BuildAgent\Appveyor.MSBuildLogger.dll" && + DIR build\VS2010\bin\%PLATFORM%_%CONFIGURATION%\*.exe && + MD5sum build/VS2010/bin/%PLATFORM%_%CONFIGURATION%/*.exe && + COPY build\VS2010\bin\%PLATFORM%_%CONFIGURATION%\fuzzer.exe tests\fuzzer_VS2010_%PLATFORM%_%CONFIGURATION%.exe && + ECHO *** && + ECHO *** Building Visual Studio 2012 %PLATFORM%\%CONFIGURATION% && + ECHO *** && + msbuild "build\VS2010\zstd.sln" /m /verbosity:minimal /property:PlatformToolset=v110 /p:ForceImportBeforeCppTargets=%APPVEYOR_BUILD_FOLDER%\build\VS2010\CompileAsCpp.props /t:Clean,Build /p:Platform=%PLATFORM% /p:Configuration=%CONFIGURATION% /logger:"C:\Program Files\AppVeyor\BuildAgent\Appveyor.MSBuildLogger.dll" && + DIR build\VS2010\bin\%PLATFORM%_%CONFIGURATION%\*.exe && + MD5sum build/VS2010/bin/%PLATFORM%_%CONFIGURATION%/*.exe && + msbuild "build\VS2010\zstd.sln" /m /verbosity:minimal /property:PlatformToolset=v110 /t:Clean,Build /p:Platform=%PLATFORM% /p:Configuration=%CONFIGURATION% /logger:"C:\Program Files\AppVeyor\BuildAgent\Appveyor.MSBuildLogger.dll" && + DIR build\VS2010\bin\%PLATFORM%_%CONFIGURATION%\*.exe && + MD5sum build/VS2010/bin/%PLATFORM%_%CONFIGURATION%/*.exe && + COPY build\VS2010\bin\%PLATFORM%_%CONFIGURATION%\fuzzer.exe tests\fuzzer_VS2012_%PLATFORM%_%CONFIGURATION%.exe && + ECHO *** && + ECHO *** Building Visual Studio 2013 %PLATFORM%\%CONFIGURATION% && + ECHO *** && + msbuild "build\VS2010\zstd.sln" /m /verbosity:minimal /property:PlatformToolset=v120 /p:ForceImportBeforeCppTargets=%APPVEYOR_BUILD_FOLDER%\build\VS2010\CompileAsCpp.props /t:Clean,Build /p:Platform=%PLATFORM% /p:Configuration=%CONFIGURATION% /logger:"C:\Program Files\AppVeyor\BuildAgent\Appveyor.MSBuildLogger.dll" && + DIR build\VS2010\bin\%PLATFORM%_%CONFIGURATION%\*.exe && + MD5sum build/VS2010/bin/%PLATFORM%_%CONFIGURATION%/*.exe && + msbuild "build\VS2010\zstd.sln" /m /verbosity:minimal /property:PlatformToolset=v120 /t:Clean,Build /p:Platform=%PLATFORM% /p:Configuration=%CONFIGURATION% /logger:"C:\Program Files\AppVeyor\BuildAgent\Appveyor.MSBuildLogger.dll" && + DIR build\VS2010\bin\%PLATFORM%_%CONFIGURATION%\*.exe && + MD5sum build/VS2010/bin/%PLATFORM%_%CONFIGURATION%/*.exe && + COPY build\VS2010\bin\%PLATFORM%_%CONFIGURATION%\fuzzer.exe tests\fuzzer_VS2013_%PLATFORM%_%CONFIGURATION%.exe && + ECHO *** && + ECHO *** Building Visual Studio 2015 %PLATFORM%\%CONFIGURATION% && + ECHO *** && + msbuild "build\VS2010\zstd.sln" /m /verbosity:minimal /property:PlatformToolset=v140 /p:ForceImportBeforeCppTargets=%APPVEYOR_BUILD_FOLDER%\build\VS2010\CompileAsCpp.props /t:Clean,Build /p:Platform=%PLATFORM% /p:Configuration=%CONFIGURATION% /logger:"C:\Program Files\AppVeyor\BuildAgent\Appveyor.MSBuildLogger.dll" && + DIR build\VS2010\bin\%PLATFORM%_%CONFIGURATION%\*.exe && + MD5sum build/VS2010/bin/%PLATFORM%_%CONFIGURATION%/*.exe && + msbuild "build\VS2010\zstd.sln" /m /verbosity:minimal /property:PlatformToolset=v140 /t:Clean,Build /p:Platform=%PLATFORM% /p:Configuration=%CONFIGURATION% /logger:"C:\Program Files\AppVeyor\BuildAgent\Appveyor.MSBuildLogger.dll" && + DIR build\VS2010\bin\%PLATFORM%_%CONFIGURATION%\*.exe && + MD5sum build/VS2010/bin/%PLATFORM%_%CONFIGURATION%/*.exe && + COPY build\VS2010\bin\%PLATFORM%_%CONFIGURATION%\fuzzer.exe tests\fuzzer_VS2015_%PLATFORM%_%CONFIGURATION%.exe && + COPY build\VS2010\bin\%PLATFORM%_%CONFIGURATION%\*.exe tests\ + ) + +test_script: + - ECHO Testing %COMPILER% %PLATFORM% %CONFIGURATION% + - SET FUZZERTEST=-T1mn + - if [%COMPILER%]==[gcc] ( + if [%PLATFORM%]==[mingw64] tests\fuzzer_clang.exe %FUZZERTEST% + ) + - if [%COMPILER%]==[visual] if [%CONFIGURATION%]==[Release] ( + CD tests && + SET ZSTD=./zstd.exe && + sh -e playTests.sh --test-large-data && + fullbench.exe -i1 && + fullbench.exe -i1 -P0 && + fuzzer_VS2008_%PLATFORM%_Release.exe %FUZZERTEST% && + fuzzer_VS2010_%PLATFORM%_Release.exe %FUZZERTEST% && + fuzzer_VS2012_%PLATFORM%_Release.exe %FUZZERTEST% && + fuzzer_VS2013_%PLATFORM%_Release.exe %FUZZERTEST% && + fuzzer_VS2015_%PLATFORM%_Release.exe %FUZZERTEST% + ) + +artifacts: + - path: bin\zstd.exe + - path: bin\zstd32.exe + +deploy: +- provider: GitHub + auth_token: + secure: LgJo8emYc3sFnlNWkGl4/VYK3nk/8+RagcsqDlAi3xeqNGNutnKjcftjg84uJoT4 + artifact: bin\zstd.exe + force_update: true + on: + branch: autobuild + COMPILER: gcc + PLATFORM: "mingw64" + appveyor_repo_tag: true +- provider: GitHub + auth_token: + secure: LgJo8emYc3sFnlNWkGl4/VYK3nk/8+RagcsqDlAi3xeqNGNutnKjcftjg84uJoT4 + artifact: bin\zstd32.exe + force_update: true + on: + branch: autobuild + COMPILER: gcc + PLATFORM: "mingw32" + appveyor_repo_tag: true diff --git a/build/.gitignore b/build/.gitignore new file mode 100644 index 0000000..f03aac8 --- /dev/null +++ b/build/.gitignore @@ -0,0 +1,20 @@ +# Visual C++ +.vs/ +*Copy +*.db +*.opensdf +*.sdf +*.suo +*.user +*.opendb + +VS2005/ +VS2008/ +VS2010/bin/ +VS2010/zwrapbench/ +VS2012/bin/ +VS2013/bin/ +VS2015/bin/ + +# CMake +cmake/ diff --git a/build/README.md b/build/README.md new file mode 100644 index 0000000..c4abe9e --- /dev/null +++ b/build/README.md @@ -0,0 +1,56 @@ +Projects for various integrated development environments (IDE) +============================================================== + +#### Included projects + +The following projects are included with the zstd distribution: +- `cmake` - CMake project contributed by Artyom Dymchenko +- `VS2005` - Visual Studio 2005 project +- `VS2008` - Visual Studio 2008 project +- `VS2010` - Visual Studio 2010 project (which also works well with Visual Studio 2012, 2013, 2015) +- `VS_scripts` - command line scripts prepared for Visual Studio compilation without IDE + + +#### How to compile zstd with Visual Studio + +1. Install Visual Studio e.g. VS 2015 Community Edition (it's free). +2. Download the latest version of zstd from https://github.com/facebook/zstd/releases +3. Decompress ZIP archive. +4. Go to decompressed directory then to `projects` then `VS2010` and open `zstd.sln` +5. Visual Studio will ask about converting VS2010 project to VS2015 and you should agree. +6. Change `Debug` to `Release` and if you have 64-bit Windows change also `Win32` to `x64`. +7. Press F7 on keyboard or select `BUILD` from the menu bar and choose `Build Solution`. +8. If compilation will be fine a compiled executable will be in `projects\VS2010\bin\x64\Release\zstd.exe` + + +#### Projects available within zstd.sln + +The Visual Studio solution file `visual\VS2010\zstd.sln` contains many projects that will be compiled to the +`visual\VS2010\bin\$(Platform)_$(Configuration)` directory. For example `zstd` set to `x64` and +`Release` will be compiled to `visual\VS2010\bin\x64_Release\zstd.exe`. The solution file contains the +following projects: + +- `zstd` : Command Line Utility, supporting gzip-like arguments +- `datagen` : Synthetic and parametrable data generator, for tests +- `fullbench` : Precisely measure speed for each zstd inner functions +- `fuzzer` : Test tool, to check zstd integrity on target platform +- `libzstd` : A static ZSTD library compiled to `libzstd_static.lib` +- `libzstd-dll` : A dynamic ZSTD library (DLL) compiled to `libzstd.dll` with the import library `libzstd.lib` +- `fullbench-dll` : The fullbench program compiled with the import library; the executable requires ZSTD DLL + + +#### Using ZSTD DLL with Microsoft Visual C++ project + +The header file `lib\zstd.h` and the import library +`visual\VS2010\bin\$(Platform)_$(Configuration)\libzstd.lib` are required to compile +a project using Visual C++. + +1. The path to header files should be added to `Additional Include Directories` that can + be found in Project Properties of Visual Studio IDE in the `C/C++` Property Pages on the `General` page. +2. The import library has to be added to `Additional Dependencies` that can + be found in Project Properties in the `Linker` Property Pages on the `Input` page. + If one will provide only the name `libzstd.lib` without a full path to the library + then the directory has to be added to `Linker\General\Additional Library Directories`. + +The compiled executable will require ZSTD DLL which is available at +`visual\VS2010\bin\$(Platform)_$(Configuration)\libzstd.dll`. diff --git a/build/VS2005/fullbench/fullbench.vcproj b/build/VS2005/fullbench/fullbench.vcproj new file mode 100644 index 0000000..ca235d0 --- /dev/null +++ b/build/VS2005/fullbench/fullbench.vcproj @@ -0,0 +1,440 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/build/VS2005/fuzzer/fuzzer.vcproj b/build/VS2005/fuzzer/fuzzer.vcproj new file mode 100644 index 0000000..1534587 --- /dev/null +++ b/build/VS2005/fuzzer/fuzzer.vcproj @@ -0,0 +1,460 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/build/VS2005/zstd.sln b/build/VS2005/zstd.sln new file mode 100644 index 0000000..dfc73b4 --- /dev/null +++ b/build/VS2005/zstd.sln @@ -0,0 +1,55 @@ +Microsoft Visual Studio Solution File, Format Version 9.00 +# Visual C++ Express 2005 +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "zstd", "zstd\zstd.vcproj", "{1A2AB08E-5CE7-4C5B-BE55-458157C14051}" +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "fuzzer", "fuzzer\fuzzer.vcproj", "{A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}" +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "fullbench", "fullbench\fullbench.vcproj", "{CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}" +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "zstdlib", "zstdlib\zstdlib.vcproj", "{99DE2A79-7298-4004-A0ED-030D7A3796CA}" +EndProject +Global + GlobalSection(SolutionConfigurationPlatforms) = preSolution + Debug|Win32 = Debug|Win32 + Debug|x64 = Debug|x64 + Release|Win32 = Release|Win32 + Release|x64 = Release|x64 + EndGlobalSection + GlobalSection(ProjectConfigurationPlatforms) = postSolution + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Debug|Win32.ActiveCfg = Debug|Win32 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Debug|Win32.Build.0 = Debug|Win32 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Debug|x64.ActiveCfg = Debug|x64 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Debug|x64.Build.0 = Debug|x64 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Release|Win32.ActiveCfg = Release|Win32 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Release|Win32.Build.0 = Release|Win32 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Release|x64.ActiveCfg = Release|x64 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Release|x64.Build.0 = Release|x64 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Debug|Win32.ActiveCfg = Debug|Win32 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Debug|Win32.Build.0 = Debug|Win32 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Debug|x64.ActiveCfg = Debug|x64 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Debug|x64.Build.0 = Debug|x64 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Release|Win32.ActiveCfg = Release|Win32 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Release|Win32.Build.0 = Release|Win32 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Release|x64.ActiveCfg = Release|x64 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Release|x64.Build.0 = Release|x64 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Debug|Win32.ActiveCfg = Debug|Win32 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Debug|Win32.Build.0 = Debug|Win32 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Debug|x64.ActiveCfg = Debug|x64 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Debug|x64.Build.0 = Debug|x64 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Release|Win32.ActiveCfg = Release|Win32 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Release|Win32.Build.0 = Release|Win32 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Release|x64.ActiveCfg = Release|x64 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Release|x64.Build.0 = Release|x64 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Debug|Win32.ActiveCfg = Debug|Win32 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Debug|Win32.Build.0 = Debug|Win32 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Debug|x64.ActiveCfg = Debug|x64 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Debug|x64.Build.0 = Debug|x64 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Release|Win32.ActiveCfg = Release|Win32 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Release|Win32.Build.0 = Release|Win32 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Release|x64.ActiveCfg = Release|x64 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Release|x64.Build.0 = Release|x64 + EndGlobalSection + GlobalSection(SolutionProperties) = preSolution + HideSolutionNode = FALSE + EndGlobalSection +EndGlobal diff --git a/build/VS2005/zstd/zstd.vcproj b/build/VS2005/zstd/zstd.vcproj new file mode 100644 index 0000000..7f1ab96 --- /dev/null +++ b/build/VS2005/zstd/zstd.vcproj @@ -0,0 +1,536 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/build/VS2005/zstdlib/zstdlib.vcproj b/build/VS2005/zstdlib/zstdlib.vcproj new file mode 100644 index 0000000..da94157 --- /dev/null +++ b/build/VS2005/zstdlib/zstdlib.vcproj @@ -0,0 +1,530 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/build/VS2008/fullbench/fullbench.vcproj b/build/VS2008/fullbench/fullbench.vcproj new file mode 100644 index 0000000..8679a12 --- /dev/null +++ b/build/VS2008/fullbench/fullbench.vcproj @@ -0,0 +1,441 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/build/VS2008/fuzzer/fuzzer.vcproj b/build/VS2008/fuzzer/fuzzer.vcproj new file mode 100644 index 0000000..9fb5fa3 --- /dev/null +++ b/build/VS2008/fuzzer/fuzzer.vcproj @@ -0,0 +1,461 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/build/VS2008/zstd.sln b/build/VS2008/zstd.sln new file mode 100644 index 0000000..89b6993 --- /dev/null +++ b/build/VS2008/zstd.sln @@ -0,0 +1,56 @@ + +Microsoft Visual Studio Solution File, Format Version 10.00 +# Visual Studio 2008 +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "zstd", "zstd\zstd.vcproj", "{1A2AB08E-5CE7-4C5B-BE55-458157C14051}" +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "fuzzer", "fuzzer\fuzzer.vcproj", "{A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}" +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "fullbench", "fullbench\fullbench.vcproj", "{CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}" +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "zstdlib", "zstdlib\zstdlib.vcproj", "{99DE2A79-7298-4004-A0ED-030D7A3796CA}" +EndProject +Global + GlobalSection(SolutionConfigurationPlatforms) = preSolution + Debug|Win32 = Debug|Win32 + Debug|x64 = Debug|x64 + Release|Win32 = Release|Win32 + Release|x64 = Release|x64 + EndGlobalSection + GlobalSection(ProjectConfigurationPlatforms) = postSolution + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Debug|Win32.ActiveCfg = Debug|Win32 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Debug|Win32.Build.0 = Debug|Win32 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Debug|x64.ActiveCfg = Debug|x64 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Debug|x64.Build.0 = Debug|x64 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Release|Win32.ActiveCfg = Release|Win32 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Release|Win32.Build.0 = Release|Win32 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Release|x64.ActiveCfg = Release|x64 + {1A2AB08E-5CE7-4C5B-BE55-458157C14051}.Release|x64.Build.0 = Release|x64 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Debug|Win32.ActiveCfg = Debug|Win32 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Debug|Win32.Build.0 = Debug|Win32 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Debug|x64.ActiveCfg = Debug|x64 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Debug|x64.Build.0 = Debug|x64 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Release|Win32.ActiveCfg = Release|Win32 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Release|Win32.Build.0 = Release|Win32 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Release|x64.ActiveCfg = Release|x64 + {A62E89D2-9DDE-42BA-8F9B-9DA74889A6B0}.Release|x64.Build.0 = Release|x64 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Debug|Win32.ActiveCfg = Debug|Win32 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Debug|Win32.Build.0 = Debug|Win32 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Debug|x64.ActiveCfg = Debug|x64 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Debug|x64.Build.0 = Debug|x64 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Release|Win32.ActiveCfg = Release|Win32 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Release|Win32.Build.0 = Release|Win32 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Release|x64.ActiveCfg = Release|x64 + {CC8F1D1B-BA2F-43E3-A71F-FA415D81AAD3}.Release|x64.Build.0 = Release|x64 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Debug|Win32.ActiveCfg = Debug|Win32 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Debug|Win32.Build.0 = Debug|Win32 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Debug|x64.ActiveCfg = Debug|x64 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Debug|x64.Build.0 = Debug|x64 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Release|Win32.ActiveCfg = Release|Win32 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Release|Win32.Build.0 = Release|Win32 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Release|x64.ActiveCfg = Release|x64 + {99DE2A79-7298-4004-A0ED-030D7A3796CA}.Release|x64.Build.0 = Release|x64 + EndGlobalSection + GlobalSection(SolutionProperties) = preSolution + HideSolutionNode = FALSE + EndGlobalSection +EndGlobal diff --git a/build/VS2008/zstd/zstd.vcproj b/build/VS2008/zstd/zstd.vcproj new file mode 100644 index 0000000..006c3dc --- /dev/null +++ b/build/VS2008/zstd/zstd.vcproj @@ -0,0 +1,537 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/build/VS2008/zstdlib/zstdlib.vcproj b/build/VS2008/zstdlib/zstdlib.vcproj new file mode 100644 index 0000000..e6e7f79 --- /dev/null +++ b/build/VS2008/zstdlib/zstdlib.vcproj @@ -0,0 +1,519 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/build/VS2010/CompileAsCpp.props b/build/VS2010/CompileAsCpp.props new file mode 100644 index 0000000..372a94b --- /dev/null +++ b/build/VS2010/CompileAsCpp.props @@ -0,0 +1,8 @@ + + + + + CompileAsCpp + + + diff --git a/build/VS2010/datagen/datagen.vcxproj b/build/VS2010/datagen/datagen.vcxproj new file mode 100644 index 0000000..280a56e --- /dev/null +++ b/build/VS2010/datagen/datagen.vcxproj @@ -0,0 +1,167 @@ + + + + + Debug + Win32 + + + Debug + x64 + + + Release + Win32 + + + Release + x64 + + + + {037E781E-81A6-494B-B1B3-438AB1200523} + Win32Proj + datagen + $(SolutionDir)bin\$(Platform)_$(Configuration)\ + $(SolutionDir)bin\obj\$(RootNamespace)_$(Platform)_$(Configuration)\ + + + + Application + true + MultiByte + + + Application + true + MultiByte + + + Application + false + true + MultiByte + + + Application + false + true + MultiByte + + + + + + + + + + + + + + + + + + + true + false + $(IncludePath);$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath); + + + true + false + $(IncludePath);$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath); + + + false + false + $(IncludePath);$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath); + + + false + false + $(IncludePath);$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath); + + + + + + Level4 + Disabled + WIN32;_DEBUG;_CONSOLE;%(PreprocessorDefinitions) + true + false + + + Console + true + + + + + + + Level4 + Disabled + WIN32;_DEBUG;_CONSOLE;%(PreprocessorDefinitions) + true + false + + + Console + true + + + + + Level4 + + + MaxSpeed + true + true + WIN32;NDEBUG;_CONSOLE;%(PreprocessorDefinitions) + false + false + MultiThreaded + + + Console + true + true + true + + + + + Level4 + + + MaxSpeed + true + true + WIN32;NDEBUG;_CONSOLE;%(PreprocessorDefinitions) + false + false + MultiThreaded + + + Console + true + true + true + + + + + + + + + + + + + \ No newline at end of file diff --git a/build/VS2010/fullbench-dll/fullbench-dll.vcxproj b/build/VS2010/fullbench-dll/fullbench-dll.vcxproj new file mode 100644 index 0000000..70b9ac2 --- /dev/null +++ b/build/VS2010/fullbench-dll/fullbench-dll.vcxproj @@ -0,0 +1,183 @@ + + + + + Debug + Win32 + + + Debug + x64 + + + Release + Win32 + + + Release + x64 + + + + {00000000-1CC8-4FD7-9281-6B8DBB9D3DF8} + Win32Proj + fullbench-dll + $(SolutionDir)bin\$(Platform)_$(Configuration)\ + $(SolutionDir)bin\obj\$(RootNamespace)_$(Platform)_$(Configuration)\ + + + + Application + true + MultiByte + + + Application + true + MultiByte + + + Application + false + true + MultiByte + + + Application + false + true + MultiByte + + + + + + + + + + + + + + + + + + + true + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath); + false + + + true + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath); + false + + + false + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath); + false + + + false + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath); + false + + + + + + Level4 + Disabled + WIN32;_DEBUG;_CONSOLE;ZSTD_DLL_IMPORT=1;%(PreprocessorDefinitions) + true + false + + + Console + true + $(SolutionDir)bin\$(Platform)_$(Configuration);%(AdditionalLibraryDirectories) + libzstd.lib;%(AdditionalDependencies) + + + + + + + Level4 + Disabled + WIN32;_DEBUG;_CONSOLE;ZSTD_DLL_IMPORT=1;%(PreprocessorDefinitions) + true + false + + + Console + true + $(SolutionDir)bin\$(Platform)_$(Configuration);%(AdditionalLibraryDirectories) + libzstd.lib;%(AdditionalDependencies) + + + + + Level4 + + + MaxSpeed + true + true + WIN32;_DEBUG;_CONSOLE;ZSTD_DLL_IMPORT=1;%(PreprocessorDefinitions) + false + false + MultiThreaded + + + Console + true + true + true + $(SolutionDir)bin\$(Platform)_$(Configuration);%(AdditionalLibraryDirectories) + libzstd.lib;%(AdditionalDependencies) + + + + + Level4 + + + MaxSpeed + true + true + WIN32;_DEBUG;_CONSOLE;ZSTD_DLL_IMPORT=1;%(PreprocessorDefinitions) + false + false + MultiThreaded + + + Console + true + true + true + $(SolutionDir)bin\$(Platform)_$(Configuration);%(AdditionalLibraryDirectories) + libzstd.lib;%(AdditionalDependencies) + + + + + + + + + + + + + + + {00000000-94d5-4bf9-8a50-7bd9929a0850} + + + + + + \ No newline at end of file diff --git a/build/VS2010/fullbench/fullbench.vcxproj b/build/VS2010/fullbench/fullbench.vcxproj new file mode 100644 index 0000000..7766c94 --- /dev/null +++ b/build/VS2010/fullbench/fullbench.vcxproj @@ -0,0 +1,186 @@ + + + + + Debug + Win32 + + + Debug + x64 + + + Release + Win32 + + + Release + x64 + + + + {61ABD629-1CC8-4FD7-9281-6B8DBB9D3DF8} + Win32Proj + fullbench + $(SolutionDir)bin\$(Platform)_$(Configuration)\ + $(SolutionDir)bin\obj\$(RootNamespace)_$(Platform)_$(Configuration)\ + + + + Application + true + MultiByte + + + Application + true + MultiByte + + + Application + false + true + MultiByte + + + Application + false + true + MultiByte + + + + + + + + + + + + + + + + + + + true + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath); + false + + + true + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath); + false + + + false + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath); + false + + + false + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath); + false + + + + + + Level4 + Disabled + WIN32;_DEBUG;_CONSOLE;%(PreprocessorDefinitions) + true + false + + + Console + true + + + + + + + Level4 + Disabled + WIN32;_DEBUG;_CONSOLE;%(PreprocessorDefinitions) + true + false + + + Console + true + + + + + Level4 + + + MaxSpeed + true + true + WIN32;NDEBUG;_CONSOLE;%(PreprocessorDefinitions) + false + false + MultiThreaded + + + Console + true + true + true + + + + + Level4 + + + MaxSpeed + true + true + WIN32;NDEBUG;_CONSOLE;%(PreprocessorDefinitions) + false + false + MultiThreaded + + + Console + true + true + true + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/build/VS2010/fuzzer/fuzzer.vcxproj b/build/VS2010/fuzzer/fuzzer.vcxproj new file mode 100644 index 0000000..82c905c --- /dev/null +++ b/build/VS2010/fuzzer/fuzzer.vcxproj @@ -0,0 +1,190 @@ + + + + + Debug + Win32 + + + Debug + x64 + + + Release + Win32 + + + Release + x64 + + + + {6FD4352B-346C-4703-96EA-D4A8B9A6976E} + Win32Proj + fuzzer + $(SolutionDir)bin\$(Platform)_$(Configuration)\ + $(SolutionDir)bin\obj\$(RootNamespace)_$(Platform)_$(Configuration)\ + + + + Application + true + MultiByte + + + Application + true + MultiByte + + + Application + false + true + MultiByte + + + Application + false + true + MultiByte + + + + + + + + + + + + + + + + + + + true + false + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + + + true + false + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + + + false + false + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + + + false + false + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + + + + + + Level4 + Disabled + WIN32;_DEBUG;_CONSOLE;%(PreprocessorDefinitions) + true + false + + + Console + true + + + + + + + Level4 + Disabled + WIN32;_DEBUG;_CONSOLE;%(PreprocessorDefinitions) + true + false + + + Console + true + + + + + Level4 + + + MaxSpeed + true + true + WIN32;NDEBUG;_CONSOLE;%(PreprocessorDefinitions) + false + false + MultiThreaded + + + Console + true + true + true + + + + + Level4 + + + MaxSpeed + true + true + WIN32;NDEBUG;_CONSOLE;%(PreprocessorDefinitions) + false + false + MultiThreaded + + + Console + true + true + true + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/build/VS2010/libzstd-dll/libzstd-dll.rc b/build/VS2010/libzstd-dll/libzstd-dll.rc new file mode 100644 index 0000000..e9a3bb8 --- /dev/null +++ b/build/VS2010/libzstd-dll/libzstd-dll.rc @@ -0,0 +1,51 @@ +// Microsoft Visual C++ generated resource script. +// + +#include "zstd.h" /* ZSTD_VERSION_STRING */ +#define APSTUDIO_READONLY_SYMBOLS +#include "verrsrc.h" +#undef APSTUDIO_READONLY_SYMBOLS + + +#if !defined(AFX_RESOURCE_DLL) || defined(AFX_TARG_ENU) +LANGUAGE 9, 1 + +///////////////////////////////////////////////////////////////////////////// +// +// Version +// + +VS_VERSION_INFO VERSIONINFO + FILEVERSION ZSTD_VERSION_MAJOR,ZSTD_VERSION_MINOR,ZSTD_VERSION_RELEASE,0 + PRODUCTVERSION ZSTD_VERSION_MAJOR,ZSTD_VERSION_MINOR,ZSTD_VERSION_RELEASE,0 + FILEFLAGSMASK VS_FFI_FILEFLAGSMASK +#ifdef _DEBUG + FILEFLAGS VS_FF_DEBUG +#else + FILEFLAGS 0x0L +#endif + FILEOS VOS_NT_WINDOWS32 + FILETYPE VFT_DLL + FILESUBTYPE VFT2_UNKNOWN +BEGIN + BLOCK "StringFileInfo" + BEGIN + BLOCK "040904B0" + BEGIN + VALUE "CompanyName", "Yann Collet" + VALUE "FileDescription", "Fast and efficient compression algorithm" + VALUE "FileVersion", ZSTD_VERSION_STRING + VALUE "InternalName", "libzstd.dll" + VALUE "LegalCopyright", "Copyright (C) 2013-2016, Yann Collet" + VALUE "OriginalFilename", "libzstd.dll" + VALUE "ProductName", "Zstandard" + VALUE "ProductVersion", ZSTD_VERSION_STRING + END + END + BLOCK "VarFileInfo" + BEGIN + VALUE "Translation", 0x0409, 1200 + END +END + +#endif diff --git a/build/VS2010/libzstd-dll/libzstd-dll.vcxproj b/build/VS2010/libzstd-dll/libzstd-dll.vcxproj new file mode 100644 index 0000000..5e8d870 --- /dev/null +++ b/build/VS2010/libzstd-dll/libzstd-dll.vcxproj @@ -0,0 +1,225 @@ + + + + + Debug + Win32 + + + Debug + x64 + + + Release + Win32 + + + Release + x64 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + {00000000-94D5-4BF9-8A50-7BD9929A0850} + Win32Proj + libzstd-dll + $(SolutionDir)bin\$(Platform)_$(Configuration)\ + $(SolutionDir)bin\obj\$(RootNamespace)_$(Platform)_$(Configuration)\ + + + + DynamicLibrary + true + MultiByte + + + DynamicLibrary + true + MultiByte + + + DynamicLibrary + false + true + MultiByte + + + DynamicLibrary + false + true + MultiByte + + + + + + + + + + + + + + + + + + + true + libzstd + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs\legacy;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + false + + + true + libzstd + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs\legacy;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + false + + + false + libzstd + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs\legacy;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + false + + + false + libzstd + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs\legacy;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + false + + + + + + Level4 + Disabled + ZSTD_DLL_EXPORT=1;ZSTD_LEGACY_SUPPORT=1;WIN32;_DEBUG;_CONSOLE;_CRT_SECURE_NO_WARNINGS;%(PreprocessorDefinitions) + true + EnableFastChecks + MultiThreadedDebugDLL + EditAndContinue + true + false + + + Console + true + MachineX86 + + + + + + + Level4 + Disabled + ZSTD_DLL_EXPORT=1;ZSTD_LEGACY_SUPPORT=1;WIN32;_DEBUG;_CONSOLE;_CRT_SECURE_NO_WARNINGS;%(PreprocessorDefinitions) + true + EnableFastChecks + MultiThreadedDebugDLL + ProgramDatabase + false + + + Console + true + + + + + Level4 + + + MaxSpeed + true + true + ZSTD_DLL_EXPORT=1;ZSTD_LEGACY_SUPPORT=1;WIN32;NDEBUG;_CONSOLE;_CRT_SECURE_NO_WARNINGS;%(PreprocessorDefinitions) + false + MultiThreaded + ProgramDatabase + All + + + Console + true + true + true + MachineX86 + + + + + Level4 + + + MaxSpeed + true + true + ZSTD_DLL_EXPORT=1;ZSTD_LEGACY_SUPPORT=1;WIN32;NDEBUG;_CONSOLE;_CRT_SECURE_NO_WARNINGS;%(PreprocessorDefinitions) + false + false + MultiThreaded + ProgramDatabase + true + true + All + + + Console + true + true + true + + + + + + diff --git a/build/VS2010/libzstd/libzstd.vcxproj b/build/VS2010/libzstd/libzstd.vcxproj new file mode 100644 index 0000000..4f3d4f9 --- /dev/null +++ b/build/VS2010/libzstd/libzstd.vcxproj @@ -0,0 +1,222 @@ + + + + + Debug + Win32 + + + Debug + x64 + + + Release + Win32 + + + Release + x64 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + {8BFD8150-94D5-4BF9-8A50-7BD9929A0850} + Win32Proj + libzstd + $(SolutionDir)bin\$(Platform)_$(Configuration)\ + $(SolutionDir)bin\obj\$(RootNamespace)_$(Platform)_$(Configuration)\ + + + + StaticLibrary + true + MultiByte + + + StaticLibrary + true + MultiByte + + + StaticLibrary + false + true + MultiByte + + + StaticLibrary + false + true + MultiByte + + + + + + + + + + + + + + + + + + + true + libzstd_static + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs\legacy;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + false + + + true + libzstd_static + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs\legacy;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + false + + + false + libzstd_static + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs\legacy;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + false + + + false + libzstd_static + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs\legacy;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + false + + + + + + Level4 + Disabled + ZSTD_DLL_EXPORT=1;ZSTD_LEGACY_SUPPORT=1;WIN32;_DEBUG;_CONSOLE;_CRT_SECURE_NO_WARNINGS;%(PreprocessorDefinitions) + true + EnableFastChecks + MultiThreadedDebugDLL + EditAndContinue + true + false + + + Console + true + MachineX86 + + + + + + + Level4 + Disabled + ZSTD_DLL_EXPORT=1;ZSTD_LEGACY_SUPPORT=1;WIN32;_DEBUG;_CONSOLE;_CRT_SECURE_NO_WARNINGS;%(PreprocessorDefinitions) + true + EnableFastChecks + MultiThreadedDebugDLL + ProgramDatabase + false + + + Console + true + + + + + Level4 + + + MaxSpeed + true + true + ZSTD_DLL_EXPORT=1;ZSTD_LEGACY_SUPPORT=1;WIN32;NDEBUG;_CONSOLE;_CRT_SECURE_NO_WARNINGS;%(PreprocessorDefinitions) + false + MultiThreaded + ProgramDatabase + All + + + Console + true + true + true + MachineX86 + + + + + Level4 + + + MaxSpeed + true + true + ZSTD_DLL_EXPORT=1;ZSTD_LEGACY_SUPPORT=1;WIN32;NDEBUG;_CONSOLE;_CRT_SECURE_NO_WARNINGS;%(PreprocessorDefinitions) + false + false + MultiThreaded + ProgramDatabase + true + true + All + + + Console + true + true + true + + + + + + diff --git a/build/VS2010/zstd.sln b/build/VS2010/zstd.sln new file mode 100644 index 0000000..ddcf840 --- /dev/null +++ b/build/VS2010/zstd.sln @@ -0,0 +1,89 @@ + +Microsoft Visual Studio Solution File, Format Version 12.00 +# Visual Studio Express 2012 for Windows Desktop +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "zstd", "zstd\zstd.vcxproj", "{4E52A41A-F33B-4C7A-8C36-A1A6B4F4277C}" +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "fuzzer", "fuzzer\fuzzer.vcxproj", "{6FD4352B-346C-4703-96EA-D4A8B9A6976E}" +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "fullbench", "fullbench\fullbench.vcxproj", "{61ABD629-1CC8-4FD7-9281-6B8DBB9D3DF8}" +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "fullbench-dll", "fullbench-dll\fullbench-dll.vcxproj", "{00000000-1CC8-4FD7-9281-6B8DBB9D3DF8}" + ProjectSection(ProjectDependencies) = postProject + {00000000-94D5-4BF9-8A50-7BD9929A0850} = {00000000-94D5-4BF9-8A50-7BD9929A0850} + EndProjectSection +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "datagen", "datagen\datagen.vcxproj", "{037E781E-81A6-494B-B1B3-438AB1200523}" +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "libzstd", "libzstd\libzstd.vcxproj", "{8BFD8150-94D5-4BF9-8A50-7BD9929A0850}" +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "libzstd-dll", "libzstd-dll\libzstd-dll.vcxproj", "{00000000-94D5-4BF9-8A50-7BD9929A0850}" +EndProject +Global + GlobalSection(SolutionConfigurationPlatforms) = preSolution + Debug|Win32 = Debug|Win32 + Debug|x64 = Debug|x64 + Release|Win32 = Release|Win32 + Release|x64 = Release|x64 + EndGlobalSection + GlobalSection(ProjectConfigurationPlatforms) = postSolution + {4E52A41A-F33B-4C7A-8C36-A1A6B4F4277C}.Debug|Win32.ActiveCfg = Debug|Win32 + {4E52A41A-F33B-4C7A-8C36-A1A6B4F4277C}.Debug|Win32.Build.0 = Debug|Win32 + {4E52A41A-F33B-4C7A-8C36-A1A6B4F4277C}.Debug|x64.ActiveCfg = Debug|x64 + {4E52A41A-F33B-4C7A-8C36-A1A6B4F4277C}.Debug|x64.Build.0 = Debug|x64 + {4E52A41A-F33B-4C7A-8C36-A1A6B4F4277C}.Release|Win32.ActiveCfg = Release|Win32 + {4E52A41A-F33B-4C7A-8C36-A1A6B4F4277C}.Release|Win32.Build.0 = Release|Win32 + {4E52A41A-F33B-4C7A-8C36-A1A6B4F4277C}.Release|x64.ActiveCfg = Release|x64 + {4E52A41A-F33B-4C7A-8C36-A1A6B4F4277C}.Release|x64.Build.0 = Release|x64 + {6FD4352B-346C-4703-96EA-D4A8B9A6976E}.Debug|Win32.ActiveCfg = Debug|Win32 + {6FD4352B-346C-4703-96EA-D4A8B9A6976E}.Debug|Win32.Build.0 = Debug|Win32 + {6FD4352B-346C-4703-96EA-D4A8B9A6976E}.Debug|x64.ActiveCfg = Debug|x64 + {6FD4352B-346C-4703-96EA-D4A8B9A6976E}.Debug|x64.Build.0 = Debug|x64 + {6FD4352B-346C-4703-96EA-D4A8B9A6976E}.Release|Win32.ActiveCfg = Release|Win32 + {6FD4352B-346C-4703-96EA-D4A8B9A6976E}.Release|Win32.Build.0 = Release|Win32 + {6FD4352B-346C-4703-96EA-D4A8B9A6976E}.Release|x64.ActiveCfg = Release|x64 + {6FD4352B-346C-4703-96EA-D4A8B9A6976E}.Release|x64.Build.0 = Release|x64 + {61ABD629-1CC8-4FD7-9281-6B8DBB9D3DF8}.Debug|Win32.ActiveCfg = Debug|Win32 + {61ABD629-1CC8-4FD7-9281-6B8DBB9D3DF8}.Debug|Win32.Build.0 = Debug|Win32 + {61ABD629-1CC8-4FD7-9281-6B8DBB9D3DF8}.Debug|x64.ActiveCfg = Debug|x64 + {61ABD629-1CC8-4FD7-9281-6B8DBB9D3DF8}.Debug|x64.Build.0 = Debug|x64 + {61ABD629-1CC8-4FD7-9281-6B8DBB9D3DF8}.Release|Win32.ActiveCfg = Release|Win32 + {61ABD629-1CC8-4FD7-9281-6B8DBB9D3DF8}.Release|Win32.Build.0 = Release|Win32 + {61ABD629-1CC8-4FD7-9281-6B8DBB9D3DF8}.Release|x64.ActiveCfg = Release|x64 + {61ABD629-1CC8-4FD7-9281-6B8DBB9D3DF8}.Release|x64.Build.0 = Release|x64 + {00000000-1CC8-4FD7-9281-6B8DBB9D3DF8}.Debug|Win32.ActiveCfg = Debug|Win32 + {00000000-1CC8-4FD7-9281-6B8DBB9D3DF8}.Debug|Win32.Build.0 = Debug|Win32 + {00000000-1CC8-4FD7-9281-6B8DBB9D3DF8}.Debug|x64.ActiveCfg = Debug|x64 + {00000000-1CC8-4FD7-9281-6B8DBB9D3DF8}.Debug|x64.Build.0 = Debug|x64 + {00000000-1CC8-4FD7-9281-6B8DBB9D3DF8}.Release|Win32.ActiveCfg = Release|Win32 + {00000000-1CC8-4FD7-9281-6B8DBB9D3DF8}.Release|Win32.Build.0 = Release|Win32 + {00000000-1CC8-4FD7-9281-6B8DBB9D3DF8}.Release|x64.ActiveCfg = Release|x64 + {00000000-1CC8-4FD7-9281-6B8DBB9D3DF8}.Release|x64.Build.0 = Release|x64 + {037E781E-81A6-494B-B1B3-438AB1200523}.Debug|Win32.ActiveCfg = Debug|Win32 + {037E781E-81A6-494B-B1B3-438AB1200523}.Debug|Win32.Build.0 = Debug|Win32 + {037E781E-81A6-494B-B1B3-438AB1200523}.Debug|x64.ActiveCfg = Debug|x64 + {037E781E-81A6-494B-B1B3-438AB1200523}.Debug|x64.Build.0 = Debug|x64 + {037E781E-81A6-494B-B1B3-438AB1200523}.Release|Win32.ActiveCfg = Release|Win32 + {037E781E-81A6-494B-B1B3-438AB1200523}.Release|Win32.Build.0 = Release|Win32 + {037E781E-81A6-494B-B1B3-438AB1200523}.Release|x64.ActiveCfg = Release|x64 + {037E781E-81A6-494B-B1B3-438AB1200523}.Release|x64.Build.0 = Release|x64 + {8BFD8150-94D5-4BF9-8A50-7BD9929A0850}.Debug|Win32.ActiveCfg = Debug|Win32 + {8BFD8150-94D5-4BF9-8A50-7BD9929A0850}.Debug|Win32.Build.0 = Debug|Win32 + {8BFD8150-94D5-4BF9-8A50-7BD9929A0850}.Debug|x64.ActiveCfg = Debug|x64 + {8BFD8150-94D5-4BF9-8A50-7BD9929A0850}.Debug|x64.Build.0 = Debug|x64 + {8BFD8150-94D5-4BF9-8A50-7BD9929A0850}.Release|Win32.ActiveCfg = Release|Win32 + {8BFD8150-94D5-4BF9-8A50-7BD9929A0850}.Release|Win32.Build.0 = Release|Win32 + {8BFD8150-94D5-4BF9-8A50-7BD9929A0850}.Release|x64.ActiveCfg = Release|x64 + {8BFD8150-94D5-4BF9-8A50-7BD9929A0850}.Release|x64.Build.0 = Release|x64 + {00000000-94D5-4BF9-8A50-7BD9929A0850}.Debug|Win32.ActiveCfg = Debug|Win32 + {00000000-94D5-4BF9-8A50-7BD9929A0850}.Debug|Win32.Build.0 = Debug|Win32 + {00000000-94D5-4BF9-8A50-7BD9929A0850}.Debug|x64.ActiveCfg = Debug|x64 + {00000000-94D5-4BF9-8A50-7BD9929A0850}.Debug|x64.Build.0 = Debug|x64 + {00000000-94D5-4BF9-8A50-7BD9929A0850}.Release|Win32.ActiveCfg = Release|Win32 + {00000000-94D5-4BF9-8A50-7BD9929A0850}.Release|Win32.Build.0 = Release|Win32 + {00000000-94D5-4BF9-8A50-7BD9929A0850}.Release|x64.ActiveCfg = Release|x64 + {00000000-94D5-4BF9-8A50-7BD9929A0850}.Release|x64.Build.0 = Release|x64 + EndGlobalSection + GlobalSection(SolutionProperties) = preSolution + HideSolutionNode = FALSE + EndGlobalSection +EndGlobal diff --git a/build/VS2010/zstd/zstd.rc b/build/VS2010/zstd/zstd.rc new file mode 100644 index 0000000..4a8aef3 --- /dev/null +++ b/build/VS2010/zstd/zstd.rc @@ -0,0 +1,51 @@ +// Microsoft Visual C++ generated resource script. +// + +#include "zstd.h" /* ZSTD_VERSION_STRING */ +#define APSTUDIO_READONLY_SYMBOLS +#include "verrsrc.h" +#undef APSTUDIO_READONLY_SYMBOLS + + +#if !defined(AFX_RESOURCE_DLL) || defined(AFX_TARG_ENU) +LANGUAGE 9, 1 + +///////////////////////////////////////////////////////////////////////////// +// +// Version +// + +VS_VERSION_INFO VERSIONINFO + FILEVERSION ZSTD_VERSION_MAJOR,ZSTD_VERSION_MINOR,ZSTD_VERSION_RELEASE,0 + PRODUCTVERSION ZSTD_VERSION_MAJOR,ZSTD_VERSION_MINOR,ZSTD_VERSION_RELEASE,0 + FILEFLAGSMASK VS_FFI_FILEFLAGSMASK +#ifdef _DEBUG + FILEFLAGS VS_FF_DEBUG +#else + FILEFLAGS 0x0L +#endif + FILEOS VOS_NT_WINDOWS32 + FILETYPE VFT_DLL + FILESUBTYPE VFT2_UNKNOWN +BEGIN + BLOCK "StringFileInfo" + BEGIN + BLOCK "040904B0" + BEGIN + VALUE "CompanyName", "Yann Collet, Facebook, Inc." + VALUE "FileDescription", "Zstandard - Fast and efficient compression algorithm" + VALUE "FileVersion", ZSTD_VERSION_STRING + VALUE "InternalName", "zstd.exe" + VALUE "LegalCopyright", "Copyright (c) 2013-present, Yann Collet, Facebook, Inc." + VALUE "OriginalFilename", "zstd.exe" + VALUE "ProductName", "Zstandard" + VALUE "ProductVersion", ZSTD_VERSION_STRING + END + END + BLOCK "VarFileInfo" + BEGIN + VALUE "Translation", 0x0409, 1200 + END +END + +#endif diff --git a/build/VS2010/zstd/zstd.vcxproj b/build/VS2010/zstd/zstd.vcxproj new file mode 100644 index 0000000..e76e090 --- /dev/null +++ b/build/VS2010/zstd/zstd.vcxproj @@ -0,0 +1,221 @@ + + + + + Debug + Win32 + + + Debug + x64 + + + Release + Win32 + + + Release + x64 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + {4E52A41A-F33B-4C7A-8C36-A1A6B4F4277C} + Win32Proj + zstd + $(SolutionDir)bin\$(Platform)_$(Configuration)\ + $(SolutionDir)bin\obj\$(RootNamespace)_$(Platform)_$(Configuration)\ + + + + Application + true + MultiByte + + + Application + true + MultiByte + + + Application + false + true + MultiByte + + + Application + false + true + MultiByte + + + + + + + + + + + + + + + + + + + true + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + false + $(LibraryPath) + + + true + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + false + $(LibraryPath); + + + false + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + false + $(LibraryPath) + + + false + $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(SolutionDir)..\..\lib\dictBuilder;$(UniversalCRT_IncludePath); + false + $(LibraryPath); + + + + + + Level4 + Disabled + ZSTD_LEGACY_SUPPORT=1;WIN32;_DEBUG;_CONSOLE;%(PreprocessorDefinitions) + true + false + + + Console + true + setargv.obj;%(AdditionalDependencies) + + + + + + + Level4 + Disabled + ZSTD_LEGACY_SUPPORT=1;WIN32;_DEBUG;_CONSOLE;%(PreprocessorDefinitions) + true + false + + + Console + true + setargv.obj;%(AdditionalDependencies) + + + + + Level4 + + + MaxSpeed + true + true + ZSTD_LEGACY_SUPPORT=1;WIN32;NDEBUG;_CONSOLE;%(PreprocessorDefinitions) + false + false + MultiThreaded + + + Console + true + true + true + setargv.obj;%(AdditionalDependencies) + + + + + Level4 + + + MaxSpeed + true + true + ZSTD_LEGACY_SUPPORT=1;WIN32;NDEBUG;_CONSOLE;%(PreprocessorDefinitions) + false + false + MultiThreaded + + + Console + true + true + true + setargv.obj;%(AdditionalDependencies) + + + + + + diff --git a/build/VS_scripts/README.md b/build/VS_scripts/README.md new file mode 100644 index 0000000..3942e38 --- /dev/null +++ b/build/VS_scripts/README.md @@ -0,0 +1,54 @@ +Command line scripts for Visual Studio compilation without IDE +============================================================== + +Here are a few command lines for reference : + +### Build with Visual Studio 2013 for msvcr120.dll + +Running the following command will build both the `Release Win32` and `Release x64` versions: +```batch +build\build.VS2013.cmd +``` +The result of each build will be in the corresponding `build\bin\Release\{ARCH}\` folder. + +If you want to only need one architecture: +- Win32: `build\build.generic.cmd VS2013 Win32 Release v120` +- x64: `build\build.generic.cmd VS2013 x64 Release v120` + +If you want a Debug build: +- Win32: `build\build.generic.cmd VS2013 Win32 Debug v120` +- x64: `build\build.generic.cmd VS2013 x64 Debug v120` + +### Build with Visual Studio 2015 for msvcr140.dll + +Running the following command will build both the `Release Win32` and `Release x64` versions: +```batch +build\build.VS2015.cmd +``` +The result of each build will be in the corresponding `build\bin\Release\{ARCH}\` folder. + +If you want to only need one architecture: +- Win32: `build\build.generic.cmd VS2015 Win32 Release v140` +- x64: `build\build.generic.cmd VS2015 x64 Release v140` + +If you want a Debug build: +- Win32: `build\build.generic.cmd VS2015 Win32 Debug v140` +- x64: `build\build.generic.cmd VS2015 x64 Debug v140` + +### Build with Visual Studio 2015 for msvcr120.dll + +You need to invoke `build\build.generic.cmd` with the proper arguments: + +**For Win32** +```batch +build\build.generic.cmd VS2015 Win32 Release v120 +``` +The result of the build will be in the `build\bin\Release\Win32\` folder. + +**For x64** +```batch +build\build.generic.cmd VS2015 x64 Release v120 +``` +The result of the build will be in the `build\bin\Release\x64\` folder. + +If you want Debug builds, replace `Release` with `Debug`. diff --git a/build/VS_scripts/build.VS2010.cmd b/build/VS_scripts/build.VS2010.cmd new file mode 100644 index 0000000..60fbbf6 --- /dev/null +++ b/build/VS_scripts/build.VS2010.cmd @@ -0,0 +1,7 @@ +@echo off + +rem build 32-bit +call "%~p0%build.generic.cmd" VS2010 Win32 Release v100 + +rem build 64-bit +call "%~p0%build.generic.cmd" VS2010 x64 Release v100 \ No newline at end of file diff --git a/build/VS_scripts/build.VS2012.cmd b/build/VS_scripts/build.VS2012.cmd new file mode 100644 index 0000000..4328bae --- /dev/null +++ b/build/VS_scripts/build.VS2012.cmd @@ -0,0 +1,6 @@ +@echo off + +rem build 32-bit +call "%~p0%build.generic.cmd" VS2012 Win32 Release v110 +rem build 64-bit +call "%~p0%build.generic.cmd" VS2012 x64 Release v110 \ No newline at end of file diff --git a/build/VS_scripts/build.VS2013.cmd b/build/VS_scripts/build.VS2013.cmd new file mode 100644 index 0000000..f906d90 --- /dev/null +++ b/build/VS_scripts/build.VS2013.cmd @@ -0,0 +1,7 @@ +@echo off + +rem build 32-bit +call "%~p0%build.generic.cmd" VS2013 Win32 Release v120 + +rem build 64-bit +call "%~p0%build.generic.cmd" VS2013 x64 Release v120 \ No newline at end of file diff --git a/build/VS_scripts/build.VS2015.cmd b/build/VS_scripts/build.VS2015.cmd new file mode 100644 index 0000000..798e9dc --- /dev/null +++ b/build/VS_scripts/build.VS2015.cmd @@ -0,0 +1,7 @@ +@echo off + +rem build 32-bit +call "%~p0%build.generic.cmd" VS2015 Win32 Release v140 + +rem build 64-bit +call "%~p0%build.generic.cmd" VS2015 x64 Release v140 \ No newline at end of file diff --git a/build/VS_scripts/build.generic.cmd b/build/VS_scripts/build.generic.cmd new file mode 100644 index 0000000..7c31d55 --- /dev/null +++ b/build/VS_scripts/build.generic.cmd @@ -0,0 +1,52 @@ +@echo off + +IF "%1%" == "" GOTO display_help + +SETLOCAL + +SET msbuild_version=%1 + +SET msbuild_platform=%2 +IF "%msbuild_platform%" == "" SET msbuild_platform=x64 + +SET msbuild_configuration=%3 +IF "%msbuild_configuration%" == "" SET msbuild_configuration=Release + +SET msbuild_toolset=%4 + +GOTO build + +:display_help + +echo Syntax: build.generic.cmd msbuild_version msbuild_platform msbuild_configuration msbuild_toolset +echo msbuild_version: VS installed version (VS2012, VS2013, VS2015, ...) +echo msbuild_platform: Platform (x64 or Win32) +echo msbuild_configuration: VS configuration (Release or Debug) +echo msbuild_toolset: Platform Toolset (v100, v110, v120, v140) + +EXIT /B 1 + +:build + +SET msbuild="%windir%\Microsoft.NET\Framework\v4.0.30319\MSBuild.exe" +IF %msbuild_version% == VS2013 SET msbuild="%programfiles(x86)%\MSBuild\12.0\Bin\MSBuild.exe" +IF %msbuild_version% == VS2015 SET msbuild="%programfiles(x86)%\MSBuild\14.0\Bin\MSBuild.exe" +rem TODO: Visual Studio "15" (vNext) will use MSBuild 15.0 ? + +SET project="%~p0\..\VS2010\zstd.sln" + +SET msbuild_params=/verbosity:minimal /nologo /t:Clean,Build /p:Platform=%msbuild_platform% /p:Configuration=%msbuild_configuration% +IF NOT "%msbuild_toolset%" == "" SET msbuild_params=%msbuild_params% /p:PlatformToolset=%msbuild_toolset% + +SET output=%~p0%bin +SET output="%output%/%msbuild_configuration%/%msbuild_platform%/" +SET msbuild_params=%msbuild_params% /p:OutDir=%output% + +echo ### Building %msbuild_version% project for %msbuild_configuration% %msbuild_platform% (%msbuild_toolset%)... +echo ### Build Params: %msbuild_params% + +%msbuild% %project% %msbuild_params% +IF ERRORLEVEL 1 EXIT /B 1 +echo # Success +echo # OutDir: %output% +echo # diff --git a/build/cmake/.gitignore b/build/cmake/.gitignore new file mode 100644 index 0000000..98f29c7 --- /dev/null +++ b/build/cmake/.gitignore @@ -0,0 +1,6 @@ +# cmake producted +CMakeCache.txt +CMakeFiles +Makefile +cmake_install.cmake +cmake_uninstall.cmake diff --git a/build/cmake/CMakeLists.txt b/build/cmake/CMakeLists.txt new file mode 100644 index 0000000..b8f5d18 --- /dev/null +++ b/build/cmake/CMakeLists.txt @@ -0,0 +1,36 @@ +# ################################################################ +# Copyright (c) 2016-present, Yann Collet, Facebook, Inc. +# All rights reserved. +# +# This source code is licensed under the BSD-style license found in the +# LICENSE file in the root directory of this source tree. An additional grant +# of patent rights can be found in the PATENTS file in the same directory. +# ################################################################ + +PROJECT(zstd) +CMAKE_MINIMUM_REQUIRED(VERSION 2.8.7) + +OPTION(ZSTD_LEGACY_SUPPORT "LEGACY SUPPORT" OFF) + +IF (ZSTD_LEGACY_SUPPORT) + MESSAGE(STATUS "ZSTD_LEGACY_SUPPORT defined!") + ADD_DEFINITIONS(-DZSTD_LEGACY_SUPPORT=1) +ELSE (ZSTD_LEGACY_SUPPORT) + MESSAGE(STATUS "ZSTD_LEGACY_SUPPORT not defined!") + ADD_DEFINITIONS(-DZSTD_LEGACY_SUPPORT=0) +ENDIF (ZSTD_LEGACY_SUPPORT) + +ADD_SUBDIRECTORY(lib) +ADD_SUBDIRECTORY(programs) +ADD_SUBDIRECTORY(tests) + +#----------------------------------------------------------------------------- +# Add extra compilation flags +#----------------------------------------------------------------------------- +INCLUDE(CMakeModules/AddExtraCompilationFlags.cmake) +ADD_EXTRA_COMPILATION_FLAGS() + +ADD_CUSTOM_TARGET(clean-all + COMMAND ${CMAKE_BUILD_TOOL} clean + COMMAND rm -rf ${CMAKE_BINARY_DIR}/ +) diff --git a/build/cmake/CMakeModules/AddExtraCompilationFlags.cmake b/build/cmake/CMakeModules/AddExtraCompilationFlags.cmake new file mode 100644 index 0000000..2d59fab --- /dev/null +++ b/build/cmake/CMakeModules/AddExtraCompilationFlags.cmake @@ -0,0 +1,331 @@ +MACRO(ADD_EXTRA_COMPILATION_FLAGS) + include(CheckCXXCompilerFlag) + include(CheckCCompilerFlag) + if (CMAKE_COMPILER_IS_GNUCXX OR MINGW) #Not only UNIX but also WIN32 for MinGW + + set(POSITION_INDEPENDENT_CODE_FLAG "-fPIC") + CHECK_C_COMPILER_FLAG(${POSITION_INDEPENDENT_CODE_FLAG} POSITION_INDEPENDENT_CODE_FLAG_ALLOWED) + if (POSITION_INDEPENDENT_CODE_FLAG_ALLOWED) + MESSAGE("Compiler flag ${POSITION_INDEPENDENT_CODE_FLAG} allowed") + set(ACTIVATE_POSITION_INDEPENDENT_CODE_FLAG "ON" CACHE BOOL "activate -fPIC flag") + else () + MESSAGE("Compiler flag ${POSITION_INDEPENDENT_CODE_FLAG} not allowed") + endif (POSITION_INDEPENDENT_CODE_FLAG_ALLOWED) + + set(WARNING_UNDEF "-Wundef") + CHECK_C_COMPILER_FLAG(${WARNING_UNDEF} WARNING_UNDEF_ALLOWED) + if (WARNING_UNDEF_ALLOWED) + MESSAGE("Compiler flag ${WARNING_UNDEF} allowed") + set(ACTIVATE_WARNING_UNDEF "ON" CACHE BOOL "activate -Wundef flag") + else () + MESSAGE("Compiler flag ${WARNING_UNDEF} not allowed") + endif (WARNING_UNDEF_ALLOWED) + + set(WARNING_SHADOW "-Wshadow") + CHECK_C_COMPILER_FLAG(${WARNING_SHADOW} WARNING_SHADOW_ALLOWED) + if (WARNING_SHADOW_ALLOWED) + MESSAGE("Compiler flag ${WARNING_SHADOW} allowed") + set(ACTIVATE_WARNING_SHADOW "ON" CACHE BOOL "activate -Wshadow flag") + else () + MESSAGE("Compiler flag ${WARNING_SHADOW} not allowed") + endif (WARNING_SHADOW_ALLOWED) + + set(WARNING_CAST_ALIGN "-Wcast-align") + CHECK_C_COMPILER_FLAG(${WARNING_CAST_ALIGN} WARNING_CAST_ALIGN_ALLOWED) + if (WARNING_CAST_ALIGN_ALLOWED) + MESSAGE("Compiler flag ${WARNING_CAST_ALIGN} allowed") + set(ACTIVATE_WARNING_CAST_ALIGN "ON" CACHE BOOL "activate -Wcast-align flag") + else () + MESSAGE("Compiler flag ${WARNING_CAST_ALIGN} not allowed") + endif (WARNING_CAST_ALIGN_ALLOWED) + + set(WARNING_CAST_QUAL "-Wcast-qual") + CHECK_C_COMPILER_FLAG(${WARNING_CAST_QUAL} WARNING_CAST_QUAL_ALLOWED) + if (WARNING_CAST_QUAL_ALLOWED) + MESSAGE("Compiler flag ${WARNING_CAST_QUAL} allowed") + set(ACTIVATE_WARNING_CAST_QUAL "ON" CACHE BOOL "activate -Wcast-qual flag") + else () + MESSAGE("Compiler flag ${WARNING_CAST_QUAL} not allowed") + endif (WARNING_CAST_QUAL_ALLOWED) + + set(WARNING_STRICT_PROTOTYPES "-Wstrict-prototypes") + CHECK_C_COMPILER_FLAG(${WARNING_STRICT_PROTOTYPES} WARNING_STRICT_PROTOTYPES_ALLOWED) + if (WARNING_STRICT_PROTOTYPES_ALLOWED) + MESSAGE("Compiler flag ${WARNING_STRICT_PROTOTYPES} allowed") + set(ACTIVATE_WARNING_STRICT_PROTOTYPES "ON" CACHE BOOL "activate -Wstrict-prototypes flag") + else () + MESSAGE("Compiler flag ${WARNING_STRICT_PROTOTYPES} not allowed") + endif (WARNING_STRICT_PROTOTYPES_ALLOWED) + + set(WARNING_ALL "-Wall") + CHECK_C_COMPILER_FLAG(${WARNING_ALL} WARNING_ALL_ALLOWED) + if (WARNING_ALL_ALLOWED) + MESSAGE("Compiler flag ${WARNING_ALL} allowed") + set(ACTIVATE_WARNING_ALL "ON" CACHE BOOL "activate -Wall flag") + else () + MESSAGE("Compiler flag ${WARNING_ALL} not allowed") + endif (WARNING_ALL_ALLOWED) + + set(WARNING_EXTRA "-Wextra") + CHECK_C_COMPILER_FLAG(${WARNING_EXTRA} WARNING_EXTRA_ALLOWED) + if (WARNING_EXTRA_ALLOWED) + MESSAGE("Compiler flag ${WARNING_EXTRA} allowed") + set(ACTIVATE_WARNING_EXTRA "ON" CACHE BOOL "activate -Wextra flag") + else () + MESSAGE("Compiler flag ${WARNING_EXTRA} not allowed") + endif (WARNING_EXTRA_ALLOWED) + + set(WARNING_FLOAT_EQUAL "-Wfloat-equal") + CHECK_C_COMPILER_FLAG(${WARNING_FLOAT_EQUAL} WARNING_FLOAT_EQUAL_ALLOWED) + if (WARNING_FLOAT_EQUAL_ALLOWED) + MESSAGE("Compiler flag ${WARNING_FLOAT_EQUAL} allowed") + set(ACTIVATE_WARNING_FLOAT_EQUAL "OFF" CACHE BOOL "activate -Wfloat-equal flag") + else () + MESSAGE("Compiler flag ${WARNING_FLOAT_EQUAL} not allowed") + endif (WARNING_FLOAT_EQUAL_ALLOWED) + + set(WARNING_SIGN_CONVERSION "-Wsign-conversion") + CHECK_C_COMPILER_FLAG(${WARNING_SIGN_CONVERSION} WARNING_SIGN_CONVERSION_ALLOWED) + if (WARNING_SIGN_CONVERSION_ALLOWED) + MESSAGE("Compiler flag ${WARNING_SIGN_CONVERSION} allowed") + set(ACTIVATE_WARNING_SIGN_CONVERSION "OFF" CACHE BOOL "activate -Wsign-conversion flag") + else () + MESSAGE("Compiler flag ${WARNING_SIGN_CONVERSION} not allowed") + endif (WARNING_SIGN_CONVERSION_ALLOWED) + + if (ACTIVATE_POSITION_INDEPENDENT_CODE_FLAG) + list(APPEND CMAKE_C_FLAGS ${POSITION_INDEPENDENT_CODE_FLAG}) + else () + string(REPLACE ${POSITION_INDEPENDENT_CODE_FLAG} "" CMAKE_C_FLAGS "${POSITION_INDEPENDENT_CODE_FLAG}") + endif (ACTIVATE_POSITION_INDEPENDENT_CODE_FLAG) + + if (ACTIVATE_WARNING_UNDEF) + list(APPEND CMAKE_CXX_FLAGS ${WARNING_UNDEF}) + list(APPEND CMAKE_C_FLAGS ${WARNING_UNDEF}) + else () + string(REPLACE ${WARNING_UNDEF} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${WARNING_UNDEF} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_WARNING_UNDEF) + + if (ACTIVATE_WARNING_SHADOW) + list(APPEND CMAKE_CXX_FLAGS ${WARNING_SHADOW}) + list(APPEND CMAKE_C_FLAGS ${WARNING_SHADOW}) + else () + string(REPLACE ${WARNING_SHADOW} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${WARNING_SHADOW} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_WARNING_SHADOW) + + if (ACTIVATE_WARNING_CAST_QUAL) + list(APPEND CMAKE_CXX_FLAGS ${WARNING_CAST_QUAL}) + list(APPEND CMAKE_C_FLAGS ${WARNING_CAST_QUAL}) + else () + string(REPLACE ${WARNING_CAST_QUAL} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${WARNING_CAST_QUAL} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_WARNING_CAST_QUAL) + + if (ACTIVATE_WARNING_CAST_ALIGN) + list(APPEND CMAKE_CXX_FLAGS ${WARNING_CAST_ALIGN}) + list(APPEND CMAKE_C_FLAGS ${WARNING_CAST_ALIGN}) + else () + string(REPLACE ${WARNING_CAST_ALIGN} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${WARNING_CAST_ALIGN} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_WARNING_CAST_ALIGN) + + if (ACTIVATE_WARNING_STRICT_PROTOTYPES) + list(APPEND CMAKE_CXX_FLAGS ${WARNING_STRICT_PROTOTYPES}) + list(APPEND CMAKE_C_FLAGS ${WARNING_STRICT_PROTOTYPES}) + else () + string(REPLACE ${WARNING_STRICT_PROTOTYPES} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${WARNING_STRICT_PROTOTYPES} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_WARNING_STRICT_PROTOTYPES) + + if (ACTIVATE_WARNING_ALL) + list(APPEND CMAKE_CXX_FLAGS ${WARNING_ALL}) + list(APPEND CMAKE_C_FLAGS ${WARNING_ALL}) + else () + string(REPLACE ${WARNING_ALL} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${WARNING_ALL} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_WARNING_ALL) + + if (ACTIVATE_WARNING_EXTRA) + list(APPEND CMAKE_CXX_FLAGS ${WARNING_EXTRA}) + list(APPEND CMAKE_C_FLAGS ${WARNING_EXTRA}) + else () + string(REPLACE ${WARNING_EXTRA} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${WARNING_EXTRA} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_WARNING_EXTRA) + + if (ACTIVATE_WARNING_FLOAT_EQUAL) + list(APPEND CMAKE_CXX_FLAGS ${WARNING_FLOAT_EQUAL}) + list(APPEND CMAKE_C_FLAGS ${WARNING_FLOAT_EQUAL}) + else () + string(REPLACE ${WARNING_FLOAT_EQUAL} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${WARNING_FLOAT_EQUAL} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_WARNING_FLOAT_EQUAL) + + if (ACTIVATE_WARNING_SIGN_CONVERSION) + list(APPEND CMAKE_CXX_FLAGS ${WARNING_SIGN_CONVERSION}) + list(APPEND CMAKE_C_FLAGS ${WARNING_SIGN_CONVERSION}) + else () + string(REPLACE ${WARNING_SIGN_CONVERSION} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${WARNING_SIGN_CONVERSION} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_WARNING_SIGN_CONVERSION) + + #Set c++11 by default + list(APPEND CMAKE_CXX_FLAGS "-std=c++11") + + #Set c99 by default + list(APPEND CMAKE_C_FLAGS "-std=c99") + + elseif (MSVC) + # Add specific compilation flags for Windows Visual + + set(WARNING_ALL "/Wall") + CHECK_C_COMPILER_FLAG(${WARNING_ALL} WARNING_ALL_ALLOWED) + if (WARNING_ALL_ALLOWED) + MESSAGE("Compiler flag ${WARNING_ALL} allowed") + set(ACTIVATE_WARNING_ALL "OFF" CACHE BOOL "activate /Wall flag") + else () + MESSAGE("Compiler flag ${WARNING_ALL} not allowed") + endif (WARNING_ALL_ALLOWED) + + set(RTC_FLAG "/RTC1") + CHECK_C_COMPILER_FLAG(${RTC_FLAG} RTC_FLAG_ALLOWED) + if (RTC_FLAG_ALLOWED) + MESSAGE("Compiler flag ${RTC_FLAG} allowed") + set(ACTIVATE_RTC_FLAG "ON" CACHE BOOL "activate /RTC1 flag") + else () + MESSAGE("Compiler flag ${RTC_FLAG} not allowed") + endif (RTC_FLAG_ALLOWED) + + set(ZC_FLAG "/Zc:forScope") + CHECK_C_COMPILER_FLAG(${ZC_FLAG} ZC_FLAG_ALLOWED) + if (ZC_FLAG_ALLOWED) + MESSAGE("Compiler flag ${ZC_FLAG} allowed") + set(ACTIVATE_ZC_FLAG "ON" CACHE BOOL "activate /Zc:forScope flag") + else () + MESSAGE("Compiler flag ${ZC_FLAG} not allowed") + endif (ZC_FLAG_ALLOWED) + + set(GD_FLAG "/Gd") + CHECK_C_COMPILER_FLAG(${GD_FLAG} GD_FLAG_ALLOWED) + if (GD_FLAG_ALLOWED) + MESSAGE("Compiler flag ${GD_FLAG} allowed") + set(ACTIVATE_GD_FLAG "ON" CACHE BOOL "activate /Gd flag") + else () + MESSAGE("Compiler flag ${GD_FLAG} not allowed") + endif (GD_FLAG_ALLOWED) + + set(ANALYZE_FLAG "/analyze:stacksize25000") + CHECK_C_COMPILER_FLAG(${ANALYZE_FLAG} ANALYZE_FLAG_ALLOWED) + if (ANALYZE_FLAG_ALLOWED) + MESSAGE("Compiler flag ${ANALYZE_FLAG} allowed") + set(ACTIVATE_ANALYZE_FLAG "ON" CACHE BOOL "activate /ANALYZE flag") + else () + MESSAGE("Compiler flag ${ANALYZE_FLAG} not allowed") + endif (ANALYZE_FLAG_ALLOWED) + + if (ACTIVATE_WARNING_ALL) + list(APPEND CMAKE_CXX_FLAGS ${WARNING_ALL}) + list(APPEND CMAKE_C_FLAGS ${WARNING_ALL}) + else () + string(REPLACE ${WARNING_ALL} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${WARNING_ALL} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_WARNING_ALL) + + # Only for DEBUG version + if (ACTIVATE_RTC_FLAG) + list(APPEND CMAKE_CXX_FLAGS_DEBUG ${RTC_FLAG}) + list(APPEND CMAKE_C_FLAGS_DEBUG ${RTC_FLAG}) + else () + string(REPLACE ${RTC_FLAG} "" CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG}") + string(REPLACE ${RTC_FLAG} "" CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG}") + endif (ACTIVATE_RTC_FLAG) + + if (ACTIVATE_ZC_FLAG) + list(APPEND CMAKE_CXX_FLAGS ${ZC_FLAG}) + list(APPEND CMAKE_C_FLAGS ${ZC_FLAG}) + else () + string(REPLACE ${ZC_FLAG} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${ZC_FLAG} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_ZC_FLAG) + + if (ACTIVATE_GD_FLAG) + list(APPEND CMAKE_CXX_FLAGS ${GD_FLAG}) + list(APPEND CMAKE_C_FLAGS ${GD_FLAG}) + else () + string(REPLACE ${GD_FLAG} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${GD_FLAG} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_GD_FLAG) + + if (ACTIVATE_ANALYZE_FLAG) + list(APPEND CMAKE_CXX_FLAGS ${ANALYZE_FLAG}) + list(APPEND CMAKE_C_FLAGS ${ANALYZE_FLAG}) + else () + string(REPLACE ${ANALYZE_FLAG} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${ANALYZE_FLAG} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_ANALYZE_FLAG) + + if (MSVC80 OR MSVC90 OR MSVC10 OR MSVC11) + # To avoid compiler warning (level 4) C4571, compile with /EHa if you still want + # your catch(...) blocks to catch structured exceptions. + list(APPEND CMAKE_CXX_FLAGS "/EHa") + endif (MSVC80 OR MSVC90 OR MSVC10 OR MSVC11) + + set(MULTITHREADED_COMPILATION "/MP") + MESSAGE("Compiler flag ${MULTITHREADED_COMPILATION} allowed") + set(ACTIVATE_MULTITHREADED_COMPILATION "ON" CACHE BOOL "activate /MP flag") + + if (ACTIVATE_MULTITHREADED_COMPILATION) + list(APPEND CMAKE_CXX_FLAGS ${MULTITHREADED_COMPILATION}) + list(APPEND CMAKE_C_FLAGS ${MULTITHREADED_COMPILATION}) + else () + string(REPLACE ${MULTITHREADED_COMPILATION} "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") + string(REPLACE ${MULTITHREADED_COMPILATION} "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}") + endif (ACTIVATE_MULTITHREADED_COMPILATION) + + #For exceptions + list(APPEND CMAKE_CXX_FLAGS "/EHsc") + list(APPEND CMAKE_C_FLAGS "/EHsc") + + # UNICODE SUPPORT + list(APPEND CMAKE_CXX_FLAGS "/D_UNICODE /DUNICODE") + list(APPEND CMAKE_C_FLAGS "/D_UNICODE /DUNICODE") + endif () + + # Remove duplicates compilation flags + FOREACH (flag_var CMAKE_C_FLAGS CMAKE_C_FLAGS_DEBUG CMAKE_C_FLAGS_RELEASE + CMAKE_C_FLAGS_MINSIZEREL CMAKE_C_FLAGS_RELWITHDEBINFO + CMAKE_CXX_FLAGS CMAKE_CXX_FLAGS_DEBUG CMAKE_CXX_FLAGS_RELEASE + CMAKE_CXX_FLAGS_MINSIZEREL CMAKE_CXX_FLAGS_RELWITHDEBINFO) + separate_arguments(${flag_var}) + list(REMOVE_DUPLICATES ${flag_var}) + string(REPLACE ";" " " ${flag_var} "${${flag_var}}") + set(${flag_var} "${${flag_var}}" CACHE STRING "common build flags" FORCE) + ENDFOREACH (flag_var) + + if (MSVC) + # Replace /MT to /MD flag + # Replace /O2 to /O3 flag + FOREACH (flag_var CMAKE_C_FLAGS CMAKE_C_FLAGS_DEBUG CMAKE_C_FLAGS_RELEASE + CMAKE_C_FLAGS_MINSIZEREL CMAKE_C_FLAGS_RELWITHDEBINFO + CMAKE_CXX_FLAGS CMAKE_CXX_FLAGS_DEBUG CMAKE_CXX_FLAGS_RELEASE + CMAKE_CXX_FLAGS_MINSIZEREL CMAKE_CXX_FLAGS_RELWITHDEBINFO) + STRING(REGEX REPLACE "/MT" "/MD" ${flag_var} "${${flag_var}}") + STRING(REGEX REPLACE "/O2" "/Ox" ${flag_var} "${${flag_var}}") + ENDFOREACH (flag_var) + endif () + + set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}" CACHE STRING "Updated flags" FORCE) + set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG}" CACHE STRING "Updated flags" FORCE) + set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE}" CACHE STRING "Updated flags" FORCE) + set(CMAKE_CXX_FLAGS_MINSIZEREL "${CMAKE_CXX_FLAGS_MINSIZEREL}" CACHE STRING "Updated flags" FORCE) + set(CMAKE_CXX_FLAGS_RELWITHDEBINFO "${CMAKE_CXX_FLAGS_RELWITHDEBINFO}" CACHE STRING "Updated flags" FORCE) + + set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS}" CACHE STRING "Updated flags" FORCE) + set(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG}" CACHE STRING "Updated flags" FORCE) + set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE}" CACHE STRING "Updated flags" FORCE) + set(CMAKE_C_FLAGS_MINSIZEREL "${CMAKE_C_FLAGS_MINSIZEREL}" CACHE STRING "Updated flags" FORCE) + set(CMAKE_C_FLAGS_RELWITHDEBINFO "${CMAKE_C_FLAGS_RELWITHDEBINFO}" CACHE STRING "Updated flags" FORCE) + +ENDMACRO(ADD_EXTRA_COMPILATION_FLAGS) diff --git a/build/cmake/cmake_uninstall.cmake.in b/build/cmake/cmake_uninstall.cmake.in new file mode 100644 index 0000000..e3774dc --- /dev/null +++ b/build/cmake/cmake_uninstall.cmake.in @@ -0,0 +1,22 @@ + +if(NOT EXISTS "@CMAKE_BINARY_DIR@/install_manifest.txt") + message(FATAL_ERROR "Cannot find install manifest: @CMAKE_BINARY_DIR@/install_manifest.txt") +endif(NOT EXISTS "@CMAKE_BINARY_DIR@/install_manifest.txt") + +file(READ "@CMAKE_BINARY_DIR@/install_manifest.txt" files) +string(REGEX REPLACE "\n" ";" files "${files}") +foreach(file ${files}) + message(STATUS "Uninstalling $ENV{DESTDIR}${file}") + if(IS_SYMLINK "$ENV{DESTDIR}${file}" OR EXISTS "$ENV{DESTDIR}${file}") + exec_program( + "@CMAKE_COMMAND@" ARGS "-E remove \"$ENV{DESTDIR}${file}\"" + OUTPUT_VARIABLE rm_out + RETURN_VALUE rm_retval + ) + if(NOT "${rm_retval}" STREQUAL 0) + message(FATAL_ERROR "Problem when removing $ENV{DESTDIR}${file}") + endif(NOT "${rm_retval}" STREQUAL 0) + else(IS_SYMLINK "$ENV{DESTDIR}${file}" OR EXISTS "$ENV{DESTDIR}${file}") + message(STATUS "File $ENV{DESTDIR}${file} does not exist.") + endif(IS_SYMLINK "$ENV{DESTDIR}${file}" OR EXISTS "$ENV{DESTDIR}${file}") +endforeach(file) diff --git a/build/cmake/lib/CMakeLists.txt b/build/cmake/lib/CMakeLists.txt new file mode 100644 index 0000000..abcf4ff --- /dev/null +++ b/build/cmake/lib/CMakeLists.txt @@ -0,0 +1,210 @@ +# ################################################################ +# zstd - Makefile +# Copyright (C) Yann Collet 2014-2016 +# All rights reserved. +# +# BSD license +# +# Redistribution and use in source and binary forms, with or without modification, +# are permitted provided that the following conditions are met: +# +# * Redistributions of source code must retain the above copyright notice, this +# list of conditions and the following disclaimer. +# +# * Redistributions in binary form must reproduce the above copyright notice, this +# list of conditions and the following disclaimer in the documentation and/or +# other materials provided with the distribution. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +# ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +# ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +# (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +# ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +# +# You can contact the author at : +# - zstd homepage : http://www.zstd.net/ +# ################################################################ + +# Get library version based on information from input content (use regular exp) +function(GetLibraryVersion _content _outputVar1 _outputVar2 _outputVar3) + string(REGEX MATCHALL ".*define ZSTD_VERSION_MAJOR+.* ([0-9]+).*define ZSTD_VERSION_MINOR+.* ([0-9]+).*define ZSTD_VERSION_RELEASE+.* ([0-9]+)" VERSION_REGEX "${_content}") + SET(${_outputVar1} ${CMAKE_MATCH_1} PARENT_SCOPE) + SET(${_outputVar2} ${CMAKE_MATCH_2} PARENT_SCOPE) + SET(${_outputVar3} ${CMAKE_MATCH_3} PARENT_SCOPE) +endfunction() + +PROJECT(libzstd) + +SET(CMAKE_INCLUDE_CURRENT_DIR TRUE) + +# Define project root directory +SET(ROOT_DIR ../../..) + +# Define library directory, where sources and header files are located +SET(LIBRARY_DIR ${ROOT_DIR}/lib) +INCLUDE_DIRECTORIES(${LIBRARY_DIR} ${LIBRARY_DIR}/common) + +# Read file content +FILE(READ ${LIBRARY_DIR}/zstd.h HEADER_CONTENT) + +# Parse version +GetLibraryVersion("${HEADER_CONTENT}" LIBVER_MAJOR LIBVER_MINOR LIBVER_RELEASE) +MESSAGE("ZSTD VERSION ${LIBVER_MAJOR}.${LIBVER_MINOR}.${LIBVER_RELEASE}") + +SET(Sources + ${LIBRARY_DIR}/common/entropy_common.c + ${LIBRARY_DIR}/common/zstd_common.c + ${LIBRARY_DIR}/common/error_private.c + ${LIBRARY_DIR}/common/xxhash.c + ${LIBRARY_DIR}/common/fse_decompress.c + ${LIBRARY_DIR}/compress/fse_compress.c + ${LIBRARY_DIR}/compress/huf_compress.c + ${LIBRARY_DIR}/compress/zstd_compress.c + ${LIBRARY_DIR}/decompress/huf_decompress.c + ${LIBRARY_DIR}/decompress/zstd_decompress.c + ${LIBRARY_DIR}/dictBuilder/divsufsort.c + ${LIBRARY_DIR}/dictBuilder/zdict.c + ${LIBRARY_DIR}/deprecated/zbuff_compress.c + ${LIBRARY_DIR}/deprecated/zbuff_decompress.c) + +SET(Headers + ${LIBRARY_DIR}/zstd.h + ${LIBRARY_DIR}/common/bitstream.h + ${LIBRARY_DIR}/common/error_private.h + ${LIBRARY_DIR}/common/zstd_errors.h + ${LIBRARY_DIR}/common/fse.h + ${LIBRARY_DIR}/common/huf.h + ${LIBRARY_DIR}/common/mem.h + ${LIBRARY_DIR}/common/zstd_internal.h + ${LIBRARY_DIR}/dictBuilder/zdict.h + ${LIBRARY_DIR}/deprecated/zbuff.h) + +IF (ZSTD_LEGACY_SUPPORT) + SET(LIBRARY_LEGACY_DIR ${LIBRARY_DIR}/legacy) + INCLUDE_DIRECTORIES(${LIBRARY_LEGACY_DIR}) + + SET(Sources ${Sources} + ${LIBRARY_LEGACY_DIR}/zstd_v01.c + ${LIBRARY_LEGACY_DIR}/zstd_v02.c + ${LIBRARY_LEGACY_DIR}/zstd_v03.c + ${LIBRARY_LEGACY_DIR}/zstd_v04.c + ${LIBRARY_LEGACY_DIR}/zstd_v05.c + ${LIBRARY_LEGACY_DIR}/zstd_v06.c + ${LIBRARY_LEGACY_DIR}/zstd_v07.c) + + SET(Headers ${Headers} + ${LIBRARY_LEGACY_DIR}/zstd_legacy.h + ${LIBRARY_LEGACY_DIR}/zstd_v01.h + ${LIBRARY_LEGACY_DIR}/zstd_v02.h + ${LIBRARY_LEGACY_DIR}/zstd_v03.h + ${LIBRARY_LEGACY_DIR}/zstd_v04.h + ${LIBRARY_LEGACY_DIR}/zstd_v05.h + ${LIBRARY_LEGACY_DIR}/zstd_v06.h + ${LIBRARY_LEGACY_DIR}/zstd_v07.h) +ENDIF (ZSTD_LEGACY_SUPPORT) + +IF (MSVC) + SET(MSVC_RESOURCE_DIR ${ROOT_DIR}/build/VS2010/zstdlib) + SET(PlatformDependResources ${MSVC_RESOURCE_DIR}/zstdlib.rc) +ENDIF (MSVC) + +# Split project to static and shared libraries build +ADD_LIBRARY(libzstd_static STATIC ${Sources} ${Headers} ${PlatformDependResources}) +ADD_LIBRARY(libzstd_shared SHARED ${Sources} ${Headers} ${PlatformDependResources}) + +# Add specific compile definitions for MSVC project +IF (MSVC) + SET_TARGET_PROPERTIES(libzstd_static PROPERTIES COMPILE_DEFINITIONS "ZSTD_HEAPMODE=0;_CRT_SECURE_NO_WARNINGS") + SET_TARGET_PROPERTIES(libzstd_shared PROPERTIES COMPILE_DEFINITIONS "ZSTD_DLL_EXPORT=1;ZSTD_HEAPMODE=0;_CONSOLE;_CRT_SECURE_NO_WARNINGS") +ENDIF (MSVC) + +# Define library base name +IF (MSVC) + SET(LIBRARY_BASE_NAME zstdlib) +ELSE () + SET(LIBRARY_BASE_NAME libzstd) +ENDIF (MSVC) + +IF (MSVC) + IF (CMAKE_SIZEOF_VOID_P MATCHES "8") + SET(LIBRARY_ARCH_SUFFIX "_x64") + ELSE () + SET(LIBRARY_ARCH_SUFFIX "_x86") + ENDIF (CMAKE_SIZEOF_VOID_P MATCHES "8") +ELSE () + SET(LIBRARY_ARCH_SUFFIX "") +ENDIF (MSVC) + +# Define static and shared library names +SET(STATIC_LIBRARY_OUTPUT_NAME ${LIBRARY_BASE_NAME}${LIBRARY_ARCH_SUFFIX} CACHE STRING "Static library output name") +SET(SHARED_LIBRARY_OUTPUT_NAME ${LIBRARY_BASE_NAME}.${LIBVER_MAJOR}.${LIBVER_MINOR}.${LIBVER_RELEASE}${LIBRARY_ARCH_SUFFIX} CACHE STRING "Shared library output name") + +SET_TARGET_PROPERTIES( + libzstd_static + PROPERTIES + PREFIX "" + OUTPUT_NAME ${STATIC_LIBRARY_OUTPUT_NAME}) + +SET_TARGET_PROPERTIES( + libzstd_shared + PROPERTIES + PREFIX "" + OUTPUT_NAME ${SHARED_LIBRARY_OUTPUT_NAME}) + +IF (UNIX) + IF ("${PREFIX}" STREQUAL "") + SET(PREFIX /usr/local) + ENDIF() + MESSAGE("the variable PREFIX=${PREFIX}") + SET(INSTALL_LIBRARY_DIR ${PREFIX}/lib) + SET(INSTALL_INCLUDE_DIR ${PREFIX}/include) + + # install target + INSTALL(FILES ${LIBRARY_DIR}/zstd.h ${LIBRARY_DIR}/deprecated/zbuff.h ${LIBRARY_DIR}/dictBuilder/zdict.h DESTINATION ${INSTALL_INCLUDE_DIR}) + INSTALL(TARGETS libzstd_static DESTINATION ${INSTALL_LIBRARY_DIR}) + INSTALL(TARGETS libzstd_shared LIBRARY DESTINATION ${INSTALL_LIBRARY_DIR}) + + # Create symlinks and setup this files + SET(SHARED_LIBRARY_LINK ${SHARED_LIBRARY_OUTPUT_NAME}${CMAKE_SHARED_LIBRARY_SUFFIX}) + SET(SHARED_LIBRARY_SYMLINK1 ${LIBRARY_BASE_NAME}${CMAKE_SHARED_LIBRARY_SUFFIX}) + SET(SHARED_LIBRARY_SYMLINK2 ${LIBRARY_BASE_NAME}${CMAKE_SHARED_LIBRARY_SUFFIX}.${LIBVER_MAJOR}) + + SET(SHARED_LIBRARY_LINK_PATH ${CMAKE_CURRENT_BINARY_DIR}/${SHARED_LIBRARY_LINK}) + SET(SHARED_LIBRARY_SYMLINK1_PATH ${CMAKE_CURRENT_BINARY_DIR}/${SHARED_LIBRARY_SYMLINK1}) + SET(SHARED_LIBRARY_SYMLINK2_PATH ${CMAKE_CURRENT_BINARY_DIR}/${SHARED_LIBRARY_SYMLINK2}) + + if (EXISTS ${SHARED_LIBRARY_SYMLINK1_PATH}) + FILE(REMOVE ${SHARED_LIBRARY_SYMLINK1_PATH}) + endif (EXISTS ${SHARED_LIBRARY_SYMLINK1_PATH}) + + if (EXISTS ${SHARED_LIBRARY_SYMLINK2_PATH}) + FILE(REMOVE ${SHARED_LIBRARY_SYMLINK2_PATH}) + endif (EXISTS ${SHARED_LIBRARY_SYMLINK2_PATH}) + + ADD_CUSTOM_COMMAND(TARGET libzstd_shared POST_BUILD + COMMAND ln -s ${SHARED_LIBRARY_LINK} ${SHARED_LIBRARY_SYMLINK1} + DEPENDS ${SHARED_LIBRARY_LINK_PATH} + COMMENT "Generating symbolic link") + + ADD_CUSTOM_COMMAND(TARGET libzstd_shared POST_BUILD + COMMAND ln -s ${SHARED_LIBRARY_LINK} ${SHARED_LIBRARY_SYMLINK2} + DEPENDS ${SHARED_LIBRARY_LINK_PATH} + COMMENT "Generating symbolic link") + + INSTALL(FILES ${SHARED_LIBRARY_SYMLINK1_PATH} DESTINATION ${INSTALL_LIBRARY_DIR}) + INSTALL(FILES ${SHARED_LIBRARY_SYMLINK2_PATH} DESTINATION ${INSTALL_LIBRARY_DIR}) + + # uninstall target + CONFIGURE_FILE( + "${CMAKE_SOURCE_DIR}/cmake_uninstall.cmake.in" + "${CMAKE_BINARY_DIR}/cmake_uninstall.cmake" + IMMEDIATE @ONLY) + + ADD_CUSTOM_TARGET(uninstall + COMMAND ${CMAKE_COMMAND} -P ${CMAKE_BINARY_DIR}/cmake_uninstall.cmake) +ENDIF (UNIX) diff --git a/build/cmake/programs/.gitignore b/build/cmake/programs/.gitignore new file mode 100644 index 0000000..f04c5b4 --- /dev/null +++ b/build/cmake/programs/.gitignore @@ -0,0 +1,3 @@ +# produced by make +zstd +zstd-frugal diff --git a/build/cmake/programs/CMakeLists.txt b/build/cmake/programs/CMakeLists.txt new file mode 100644 index 0000000..26cd6c0 --- /dev/null +++ b/build/cmake/programs/CMakeLists.txt @@ -0,0 +1,57 @@ +# ################################################################ +# zstd - Makefile +# Copyright (C) Yann Collet 2014-2016 +# All rights reserved. +# +# BSD license +# +# Redistribution and use in source and binary forms, with or without modification, +# are permitted provided that the following conditions are met: +# +# * Redistributions of source code must retain the above copyright notice, this +# list of conditions and the following disclaimer. +# +# * Redistributions in binary form must reproduce the above copyright notice, this +# list of conditions and the following disclaimer in the documentation and/or +# other materials provided with the distribution. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +# ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +# ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +# (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +# ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +# +# You can contact the author at : +# - zstd homepage : http://www.zstd.net/ +# ################################################################ + +PROJECT(programs) + +SET(CMAKE_INCLUDE_CURRENT_DIR TRUE) + +# Define project root directory +SET(ROOT_DIR ../../..) + +# Define programs directory, where sources and header files are located +SET(LIBRARY_DIR ${ROOT_DIR}/lib) +SET(PROGRAMS_DIR ${ROOT_DIR}/programs) +INCLUDE_DIRECTORIES(${PROGRAMS_DIR} ${LIBRARY_DIR} ${LIBRARY_DIR}/common ${LIBRARY_DIR}/dictBuilder) + +IF (ZSTD_LEGACY_SUPPORT) + SET(PROGRAMS_LEGACY_DIR ${PROGRAMS_DIR}/legacy) + INCLUDE_DIRECTORIES(${PROGRAMS_LEGACY_DIR} ${LIBRARY_DIR}/legacy) +ENDIF (ZSTD_LEGACY_SUPPORT) + +ADD_EXECUTABLE(zstd ${PROGRAMS_DIR}/zstdcli.c ${PROGRAMS_DIR}/fileio.c ${PROGRAMS_DIR}/bench.c ${PROGRAMS_DIR}/datagen.c ${PROGRAMS_DIR}/dibio.c) +TARGET_LINK_LIBRARIES(zstd libzstd_static) + +IF (UNIX) + ADD_EXECUTABLE(zstd-frugal ${PROGRAMS_DIR}/zstdcli.c ${PROGRAMS_DIR}/fileio.c) + TARGET_LINK_LIBRARIES(zstd-frugal libzstd_static) + SET_TARGET_PROPERTIES(zstd-frugal PROPERTIES COMPILE_DEFINITIONS "ZSTD_NOBENCH;ZSTD_NODICT") +ENDIF (UNIX) diff --git a/build/cmake/tests/.gitignore b/build/cmake/tests/.gitignore new file mode 100644 index 0000000..2ab62a3 --- /dev/null +++ b/build/cmake/tests/.gitignore @@ -0,0 +1,7 @@ +# produced by make +datagen +fullbench +fuzzer +paramgrill +zbufftest + diff --git a/build/cmake/tests/CMakeLists.txt b/build/cmake/tests/CMakeLists.txt new file mode 100644 index 0000000..7f9c38e --- /dev/null +++ b/build/cmake/tests/CMakeLists.txt @@ -0,0 +1,58 @@ +# ################################################################ +# zstd - Makefile +# Copyright (C) Yann Collet 2014-2016 +# All rights reserved. +# +# BSD license +# +# Redistribution and use in source and binary forms, with or without modification, +# are permitted provided that the following conditions are met: +# +# * Redistributions of source code must retain the above copyright notice, this +# list of conditions and the following disclaimer. +# +# * Redistributions in binary form must reproduce the above copyright notice, this +# list of conditions and the following disclaimer in the documentation and/or +# other materials provided with the distribution. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +# ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +# ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +# (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +# ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +# +# You can contact the author at : +# - zstd homepage : http://www.zstd.net/ +# ################################################################ + +PROJECT(tests) + +SET(CMAKE_INCLUDE_CURRENT_DIR TRUE) + +# Define project root directory +SET(ROOT_DIR ../../..) + +# Define programs directory, where sources and header files are located +SET(LIBRARY_DIR ${ROOT_DIR}/lib) +SET(PROGRAMS_DIR ${ROOT_DIR}/programs) +SET(TESTS_DIR ${ROOT_DIR}/tests) +INCLUDE_DIRECTORIES(${TESTS_DIR} ${PROGRAMS_DIR} ${LIBRARY_DIR} ${LIBRARY_DIR}/common ${LIBRARY_DIR}/dictBuilder) + +ADD_EXECUTABLE(fullbench ${PROGRAMS_DIR}/datagen.c ${TESTS_DIR}/fullbench.c) +TARGET_LINK_LIBRARIES(fullbench libzstd_static) + +ADD_EXECUTABLE(fuzzer ${PROGRAMS_DIR}/datagen.c ${TESTS_DIR}/fuzzer.c) +TARGET_LINK_LIBRARIES(fuzzer libzstd_static) + +IF (UNIX) + ADD_EXECUTABLE(paramgrill ${PROGRAMS_DIR}/datagen.c ${TESTS_DIR}/paramgrill.c) + TARGET_LINK_LIBRARIES(paramgrill libzstd_static m) #m is math library + + ADD_EXECUTABLE(datagen ${PROGRAMS_DIR}/datagen.c ${TESTS_DIR}/datagencli.c) + TARGET_LINK_LIBRARIES(datagen libzstd_static) +ENDIF (UNIX) diff --git a/contrib/gen_html/.gitignore b/contrib/gen_html/.gitignore new file mode 100644 index 0000000..3446114 --- /dev/null +++ b/contrib/gen_html/.gitignore @@ -0,0 +1,3 @@ +# make artefact +gen_html +zstd_manual.html diff --git a/contrib/gen_html/Makefile b/contrib/gen_html/Makefile new file mode 100644 index 0000000..c68e560 --- /dev/null +++ b/contrib/gen_html/Makefile @@ -0,0 +1,36 @@ +# ########################################################################## +# Copyright (c) 2016-present, Facebook, Inc. +# All rights reserved. +# +# This source code is licensed under the BSD-style license found in the +# LICENSE file in the root directory of this source tree. An additional grant +# of patent rights can be found in the PATENTS file in the same directory. +# ########################################################################## + + +CFLAGS ?= -O3 +CFLAGS += -Wall -Wextra -Wcast-qual -Wcast-align -Wshadow -Wstrict-aliasing=1 -Wswitch-enum -Wno-comment +CFLAGS += $(MOREFLAGS) +FLAGS = $(CPPFLAGS) $(CFLAGS) $(LDFLAGS) + + + +# Define *.exe as extension for Windows systems +ifneq (,$(filter Windows%,$(OS))) +EXT =.exe +else +EXT = +endif + + +.PHONY: default gen_html + +default: gen_html + +gen_html: gen_html.cpp + $(CXX) $(FLAGS) $^ -o $@$(EXT) + + +clean: + @$(RM) gen_html$(EXT) + @echo Cleaning completed diff --git a/contrib/gen_html/README.md b/contrib/gen_html/README.md new file mode 100644 index 0000000..63a4caa --- /dev/null +++ b/contrib/gen_html/README.md @@ -0,0 +1,31 @@ +gen_html - a program for automatic generation of zstd manual +============================================================ + +#### Introduction + +This simple C++ program generates a single-page HTML manual from `zstd.h`. + +The format of recognized comment blocks is following: +- comments of type `/*!` mean: this is a function declaration; switch comments with declarations +- comments of type `/**` and `/*-` mean: this is a comment; use a `

` header for the first line +- comments of type `/*=` and `/**=` mean: use a `

` header and show also all functions until first empty line +- comments of type `/*X` where `X` is different from above-mentioned are ignored + +Moreover: +- `ZSTDLIB_API` is removed to improve readability +- `typedef` are detected and included even if uncommented +- comments of type `/**<` and `/*!<` are detected and only function declaration is highlighted (bold) + + +#### Usage + +The program requires 3 parameters: +``` +gen_html [zstd_version] [input_file] [output_html] +``` + +To compile program and generate zstd manual we have used: +``` +make +./gen_html.exe 1.1.1 ../../lib/zstd.h zstd_manual.html +``` diff --git a/contrib/gen_html/gen-zstd-manual.sh b/contrib/gen_html/gen-zstd-manual.sh new file mode 100755 index 0000000..57a8b6e --- /dev/null +++ b/contrib/gen_html/gen-zstd-manual.sh @@ -0,0 +1,9 @@ +#!/bin/sh + +LIBVER_MAJOR_SCRIPT=`sed -n '/define ZSTD_VERSION_MAJOR/s/.*[[:blank:]]\([0-9][0-9]*\).*/\1/p' < ../../lib/zstd.h` +LIBVER_MINOR_SCRIPT=`sed -n '/define ZSTD_VERSION_MINOR/s/.*[[:blank:]]\([0-9][0-9]*\).*/\1/p' < ../../lib/zstd.h` +LIBVER_PATCH_SCRIPT=`sed -n '/define ZSTD_VERSION_RELEASE/s/.*[[:blank:]]\([0-9][0-9]*\).*/\1/p' < ../../lib/zstd.h` +LIBVER_SCRIPT=$LIBVER_MAJOR_SCRIPT.$LIBVER_MINOR_SCRIPT.$LIBVER_PATCH_SCRIPT + +echo ZSTD_VERSION=$LIBVER_SCRIPT +./gen_html $LIBVER_SCRIPT ../../lib/zstd.h ./zstd_manual.html diff --git a/contrib/gen_html/gen_html.cpp b/contrib/gen_html/gen_html.cpp new file mode 100644 index 0000000..22ff65b --- /dev/null +++ b/contrib/gen_html/gen_html.cpp @@ -0,0 +1,220 @@ +/* + * Copyright (c) 2016-present, Przemyslaw Skibinski, Facebook, Inc. + * All rights reserved. + * + * This source code is licensed under the BSD-style license found in the + * LICENSE file in the root directory of this source tree. An additional grant + * of patent rights can be found in the PATENTS file in the same directory. + */ + +#include +#include +#include +#include +using namespace std; + + +/* trim string at the beginning and at the end */ +void trim(string& s, string characters) +{ + size_t p = s.find_first_not_of(characters); + s.erase(0, p); + + p = s.find_last_not_of(characters); + if (string::npos != p) + s.erase(p+1); +} + + +/* trim C++ style comments */ +void trim_comments(string &s) +{ + size_t spos, epos; + + spos = s.find("/*"); + epos = s.find("*/"); + s = s.substr(spos+3, epos-(spos+3)); +} + + +/* get lines until a given terminator */ +vector get_lines(vector& input, int& linenum, string terminator) +{ + vector out; + string line; + size_t epos; + + while ((size_t)linenum < input.size()) { + line = input[linenum]; + + if (terminator.empty() && line.empty()) { linenum--; break; } + + epos = line.find(terminator); + if (!terminator.empty() && epos!=string::npos) { + out.push_back(line); + break; + } + out.push_back(line); + linenum++; + } + return out; +} + + +/* print line with ZSTDLIB_API removed and C++ comments not bold */ +void print_line(stringstream &sout, string line) +{ + size_t spos; + + if (line.substr(0,12) == "ZSTDLIB_API ") line = line.substr(12); + spos = line.find("/*"); + if (spos!=string::npos) { + sout << line.substr(0, spos); + sout << "" << line.substr(spos) << "" << endl; + } else { + // fprintf(stderr, "lines=%s\n", line.c_str()); + sout << line << endl; + } +} + + +int main(int argc, char *argv[]) { + char exclam; + int linenum, chapter = 1; + vector input, lines, comments, chapters; + string line, version; + size_t spos, l; + stringstream sout; + ifstream istream; + ofstream ostream; + + if (argc < 4) { + cout << "usage: " << argv[0] << " [zstd_version] [input_file] [output_html]" << endl; + return 1; + } + + version = "zstd " + string(argv[1]) + " Manual"; + + istream.open(argv[2], ifstream::in); + if (!istream.is_open()) { + cout << "Error opening file " << argv[2] << endl; + return 1; + } + + ostream.open(argv[3], ifstream::out); + if (!ostream.is_open()) { + cout << "Error opening file " << argv[3] << endl; + return 1; + } + + while (getline(istream, line)) { + input.push_back(line); + } + + for (linenum=0; (size_t)linenum < input.size(); linenum++) { + line = input[linenum]; + + /* typedefs are detected and included even if uncommented */ + if (line.substr(0,7) == "typedef" && line.find("{")!=string::npos) { + lines = get_lines(input, linenum, "}"); + sout << "
";
+            for (l=0; l

" << endl; + continue; + } + + /* comments of type /**< and /*!< are detected and only function declaration is highlighted (bold) */ + if ((line.find("/**<")!=string::npos || line.find("/*!<")!=string::npos) && line.find("*/")!=string::npos) { + sout << "
";
+            print_line(sout, line);
+            sout << "

" << endl; + continue; + } + + spos = line.find("/**="); + if (spos==string::npos) { + spos = line.find("/*!"); + if (spos==string::npos) + spos = line.find("/**"); + if (spos==string::npos) + spos = line.find("/*-"); + if (spos==string::npos) + spos = line.find("/*="); + if (spos==string::npos) + continue; + exclam = line[spos+2]; + } + else exclam = '='; + + comments = get_lines(input, linenum, "*/"); + if (!comments.empty()) comments[0] = line.substr(spos+3); + if (!comments.empty()) comments[comments.size()-1] = comments[comments.size()-1].substr(0, comments[comments.size()-1].find("*/")); + for (l=0; l"; + for (l=0; l

"; + for (l=0; l
" << endl << endl; + } else if (exclam == '=') { /* comments of type /*= and /**= mean: use a

header and show also all functions until first empty line */ + trim(comments[0], " "); + sout << "

" << comments[0] << "

";
+            for (l=1; l
";
+            lines = get_lines(input, ++linenum, "");
+            for (l=0; l
" << endl;
+        } else { /* comments of type /** and /*- mean: this is a comment; use a 
 header for the first line */
+            if (comments.empty()) continue;
+
+            trim(comments[0], " ");
+            sout << "
" << comments[0] << "
";
+            chapters.push_back(comments[0]);
+            chapter++;
+
+            for (l=1; l 1)
+                sout << "
" << endl << endl;
+            else
+                sout << "
" << endl << endl;
+        }
+    }
+
+    ostream << "\n\n\n" << version << "\n\n" << endl;
+    ostream << "
" << version << "
\n";
+
+    ostream << "
\n
Contents
\n
    \n";
+    for (size_t i=0; i" << chapters[i].c_str() << "\n";
+    ostream << "
\n
\n";
+
+    ostream << sout.str();
+    ostream << "" << endl << "" << endl;
+
+    return 0;
+}
\ No newline at end of file
diff --git a/contrib/pzstd/.gitignore b/contrib/pzstd/.gitignore
new file mode 100644
index 0000000..84e68fb
--- /dev/null
+++ b/contrib/pzstd/.gitignore
@@ -0,0 +1,2 @@
+# compilation result
+pzstd
diff --git a/contrib/pzstd/ErrorHolder.h b/contrib/pzstd/ErrorHolder.h
new file mode 100644
index 0000000..188badc
--- /dev/null
+++ b/contrib/pzstd/ErrorHolder.h
@@ -0,0 +1,54 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#pragma once
+
+#include 
+#include 
+#include 
+#include 
+
+namespace pzstd {
+
+// Coordinates graceful shutdown of the pzstd pipeline
+class ErrorHolder {
+  std::atomic error_;
+  std::string message_;
+
+ public:
+  ErrorHolder() : error_(false) {}
+
+  bool hasError() noexcept {
+    return error_.load();
+  }
+
+  void setError(std::string message) noexcept {
+    // Given multiple possibly concurrent calls, exactly one will ever succeed.
+    bool expected = false;
+    if (error_.compare_exchange_strong(expected, true)) {
+      message_ = std::move(message);
+    }
+  }
+
+  bool check(bool predicate, std::string message) noexcept {
+    if (!predicate) {
+      setError(std::move(message));
+    }
+    return !hasError();
+  }
+
+  std::string getError() noexcept {
+    error_.store(false);
+    return std::move(message_);
+  }
+
+  ~ErrorHolder() {
+    assert(!hasError());
+  }
+};
+}
diff --git a/contrib/pzstd/Logging.h b/contrib/pzstd/Logging.h
new file mode 100644
index 0000000..76c982a
--- /dev/null
+++ b/contrib/pzstd/Logging.h
@@ -0,0 +1,72 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#pragma once
+
+#include 
+#include 
+
+namespace pzstd {
+
+constexpr int ERROR = 1;
+constexpr int INFO = 2;
+constexpr int DEBUG = 3;
+constexpr int VERBOSE = 4;
+
+class Logger {
+  std::mutex mutex_;
+  FILE* out_;
+  const int level_;
+
+  using Clock = std::chrono::system_clock;
+  Clock::time_point lastUpdate_;
+  std::chrono::milliseconds refreshRate_;
+
+ public:
+  explicit Logger(int level, FILE* out = stderr)
+      : out_(out), level_(level), lastUpdate_(Clock::now()),
+        refreshRate_(150) {}
+
+
+  bool logsAt(int level) {
+    return level <= level_;
+  }
+
+  template 
+  void operator()(int level, const char *fmt, Args... args) {
+    if (level > level_) {
+      return;
+    }
+    std::lock_guard lock(mutex_);
+    std::fprintf(out_, fmt, args...);
+  }
+
+  template 
+  void update(int level, const char *fmt, Args... args) {
+    if (level > level_) {
+      return;
+    }
+    std::lock_guard lock(mutex_);
+    auto now = Clock::now();
+    if (now - lastUpdate_ > refreshRate_) {
+      lastUpdate_ = now;
+      std::fprintf(out_, "\r");
+      std::fprintf(out_, fmt, args...);
+    }
+  }
+
+  void clear(int level) {
+    if (level > level_) {
+      return;
+    }
+    std::lock_guard lock(mutex_);
+    std::fprintf(out_, "\r%79s\r", "");
+  }
+};
+
+}
diff --git a/contrib/pzstd/Makefile b/contrib/pzstd/Makefile
new file mode 100644
index 0000000..99d955e
--- /dev/null
+++ b/contrib/pzstd/Makefile
@@ -0,0 +1,252 @@
+# ##########################################################################
+# Copyright (c) 2016-present, Facebook, Inc.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree. An additional grant
+# of patent rights can be found in the PATENTS file in the same directory.
+# ##########################################################################
+
+# Standard variables for installation
+DESTDIR ?=
+PREFIX  ?= /usr/local
+BINDIR  := $(DESTDIR)$(PREFIX)/bin
+
+ZSTDDIR = ../../lib
+PROGDIR = ../../programs
+
+# External program to use to run tests, e.g. qemu or valgrind
+TESTPROG  ?=
+# Flags to pass to the tests
+TESTFLAGS ?=
+
+# We use gcc/clang to generate the header dependencies of files
+DEPFLAGS = -MMD -MP -MF $*.Td
+POSTCOMPILE = mv -f $*.Td $*.d
+
+# CFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS are for the users to override
+CFLAGS   ?= -O3 -Wall -Wextra
+CXXFLAGS ?= -O3 -Wall -Wextra -pedantic -std=c++11
+CPPFLAGS ?=
+LDFLAGS  ?=
+
+# Include flags
+PZSTD_INC  = -I$(ZSTDDIR) -I$(ZSTDDIR)/common -I$(PROGDIR) -I.
+GTEST_INC  = -isystem googletest/googletest/include
+
+PZSTD_CPPFLAGS  = $(PZSTD_INC)
+PZSTD_CCXXFLAGS =
+PZSTD_CFLAGS    = $(PZSTD_CCXXFLAGS)
+PZSTD_CXXFLAGS  = $(PZSTD_CCXXFLAGS)
+PZSTD_LDFLAGS   =
+EXTRA_FLAGS     =
+ALL_CFLAGS      = $(EXTRA_FLAGS) $(CPPFLAGS) $(PZSTD_CPPFLAGS) $(CFLAGS)   $(PZSTD_CFLAGS)
+ALL_CXXFLAGS    = $(EXTRA_FLAGS) $(CPPFLAGS) $(PZSTD_CPPFLAGS) $(CXXFLAGS) $(PZSTD_CXXFLAGS)
+ALL_LDFLAGS     = $(EXTRA_FLAGS) $(LDFLAGS) $(PZSTD_LDFLAGS)
+
+
+# gtest libraries need to go before "-lpthread" because they depend on it.
+GTEST_LIB  = -L googletest/build/googlemock/gtest
+LIBS       =
+
+# Compilation commands
+LD_COMMAND  = $(CXX) $^          $(ALL_LDFLAGS) $(LIBS) -lpthread -o $@
+CC_COMMAND  = $(CC)  $(DEPFLAGS) $(ALL_CFLAGS)   -c $<  -o $@
+CXX_COMMAND = $(CXX) $(DEPFLAGS) $(ALL_CXXFLAGS) -c $<  -o $@
+
+# Get a list of all zstd files so we rebuild the static library when we need to
+ZSTDCOMMON_FILES := $(wildcard $(ZSTDDIR)/common/*.c) \
+                    $(wildcard $(ZSTDDIR)/common/*.h)
+ZSTDCOMP_FILES   := $(wildcard $(ZSTDDIR)/compress/*.c) \
+                    $(wildcard $(ZSTDDIR)/compress/*.h)
+ZSTDDECOMP_FILES := $(wildcard $(ZSTDDIR)/decompress/*.c) \
+                    $(wildcard $(ZSTDDIR)/decompress/*.h)
+ZSTDPROG_FILES   := $(wildcard $(PROGDIR)/*.c) \
+                    $(wildcard $(PROGDIR)/*.h)
+ZSTD_FILES       := $(wildcard $(ZSTDDIR)/*.h) \
+                    $(ZSTDDECOMP_FILES) $(ZSTDCOMMON_FILES) $(ZSTDCOMP_FILES) \
+                    $(ZSTDPROG_FILES)
+
+# List all the pzstd source files so we can determine their dependencies
+PZSTD_SRCS  := $(wildcard *.cpp)
+PZSTD_TESTS := $(wildcard test/*.cpp)
+UTILS_TESTS := $(wildcard utils/test/*.cpp)
+ALL_SRCS    := $(PZSTD_SRCS) $(PZSTD_TESTS) $(UTILS_TESTS)
+
+
+# Define *.exe as extension for Windows systems
+ifneq (,$(filter Windows%,$(OS)))
+EXT =.exe
+else
+EXT =
+endif
+
+# Standard targets
+.PHONY: default
+default: all
+
+.PHONY: check
+check:
+	$(TESTPROG) ./utils/test/BufferTest$(EXT) $(TESTFLAGS)
+	$(TESTPROG) ./utils/test/RangeTest$(EXT) $(TESTFLAGS)
+	$(TESTPROG) ./utils/test/ResourcePoolTest$(EXT) $(TESTFLAGS)
+	$(TESTPROG) ./utils/test/ScopeGuardTest$(EXT) $(TESTFLAGS)
+	$(TESTPROG) ./utils/test/ThreadPoolTest$(EXT) $(TESTFLAGS)
+	$(TESTPROG) ./utils/test/WorkQueueTest$(EXT) $(TESTFLAGS)
+	$(TESTPROG) ./test/OptionsTest$(EXT) $(TESTFLAGS)
+	$(TESTPROG) ./test/PzstdTest$(EXT) $(TESTFLAGS)
+
+.PHONY: install
+install: PZSTD_CPPFLAGS += -DNDEBUG
+install: pzstd$(EXT)
+	install -d -m 755 $(BINDIR)/
+	install -m 755 pzstd$(EXT) $(BINDIR)/pzstd$(EXT)
+
+.PHONY: uninstall
+uninstall:
+	$(RM) $(BINDIR)/pzstd$(EXT)
+
+# Targets for many different builds
+.PHONY: all
+all: PZSTD_CPPFLAGS += -DNDEBUG
+all: pzstd$(EXT)
+
+.PHONY: debug
+debug: EXTRA_FLAGS += -g
+debug: pzstd$(EXT) tests roundtrip
+
+.PHONY: tsan
+tsan: PZSTD_CCXXFLAGS += -fsanitize=thread -fPIC
+tsan: PZSTD_LDFLAGS   += -fsanitize=thread -pie
+tsan: debug
+
+.PHONY: asan
+asan: EXTRA_FLAGS += -fsanitize=address
+asan: debug
+
+.PHONY: ubsan
+ubsan: EXTRA_FLAGS += -fsanitize=undefined
+ubsan: debug
+
+.PHONY: all32
+all32: EXTRA_FLAGS += -m32
+all32: all tests roundtrip
+
+.PHONY: debug32
+debug32: EXTRA_FLAGS += -m32
+debug32: debug
+
+.PHONY: asan32
+asan32: EXTRA_FLAGS += -m32
+asan32: asan
+
+.PHONY: tsan32
+tsan32: EXTRA_FLAGS += -m32
+tsan32: tsan
+
+.PHONY: ubsan32
+ubsan32: EXTRA_FLAGS += -m32
+ubsan32: ubsan
+
+# Run long round trip tests
+.PHONY: roundtripcheck
+roundtripcheck: roundtrip check
+	$(TESTPROG) ./test/RoundTripTest$(EXT) $(TESTFLAGS)
+
+# Build the main binary
+pzstd$(EXT): main.o Options.o Pzstd.o SkippableFrame.o $(ZSTDDIR)/libzstd.a
+	$(LD_COMMAND)
+
+# Target that depends on all the tests
+.PHONY: tests
+tests: EXTRA_FLAGS += -Wno-deprecated-declarations
+tests: $(patsubst %,%$(EXT),$(basename $(PZSTD_TESTS) $(UTILS_TESTS)))
+
+# Build the round trip tests
+.PHONY: roundtrip
+roundtrip: EXTRA_FLAGS += -Wno-deprecated-declarations
+roundtrip: test/RoundTripTest$(EXT)
+
+# Use the static library that zstd builds for simplicity and
+# so we get the compiler options correct
+$(ZSTDDIR)/libzstd.a: $(ZSTD_FILES)
+	CFLAGS="$(ALL_CFLAGS)" LDFLAGS="$(ALL_LDFLAGS)" $(MAKE) -C $(ZSTDDIR) libzstd.a
+
+# Rules to build the tests
+test/RoundTripTest$(EXT): test/RoundTripTest.o $(PROGDIR)/datagen.o Options.o \
+                          Pzstd.o SkippableFrame.o $(ZSTDDIR)/libzstd.a
+	$(LD_COMMAND)
+
+test/%Test$(EXT): PZSTD_LDFLAGS += $(GTEST_LIB)
+test/%Test$(EXT): LIBS += -lgtest -lgtest_main
+test/%Test$(EXT): test/%Test.o $(PROGDIR)/datagen.o Options.o Pzstd.o  \
+                  SkippableFrame.o $(ZSTDDIR)/libzstd.a
+	$(LD_COMMAND)
+
+utils/test/%Test$(EXT): PZSTD_LDFLAGS += $(GTEST_LIB)
+utils/test/%Test$(EXT): LIBS += -lgtest -lgtest_main
+utils/test/%Test$(EXT): utils/test/%Test.o
+	$(LD_COMMAND)
+
+
+GTEST_CMAKEFLAGS =
+
+# Install googletest
+.PHONY: googletest
+googletest: PZSTD_CCXXFLAGS += -fPIC
+googletest:
+	@$(RM) -rf googletest
+	@git clone https://github.com/google/googletest
+	@mkdir -p googletest/build
+	@cd googletest/build && cmake $(GTEST_CMAKEFLAGS) -DCMAKE_CXX_FLAGS="$(ALL_CXXFLAGS)" .. && $(MAKE)
+
+.PHONY: googletest32
+googletest32: PZSTD_CCXXFLAGS  += -m32
+googletest32: googletest
+
+.PHONY: googletest-mingw64
+googletest-mingw64: GTEST_CMAKEFLAGS += -G "MSYS Makefiles"
+googletest-mingw64: googletest
+
+.PHONY: clean
+clean:
+	$(RM) -f *.o pzstd$(EXT) *.Td *.d
+	$(RM) -f test/*.o test/*Test$(EXT) test/*.Td test/*.d
+	$(RM) -f utils/test/*.o utils/test/*Test$(EXT) utils/test/*.Td utils/test/*.d
+	$(RM) -f $(PROGDIR)/*.o $(PROGDIR)/*.Td $(PROGDIR)/*.d
+	$(MAKE) -C $(ZSTDDIR) clean
+	@echo Cleaning completed
+
+
+# Cancel implicit rules
+%.o: %.c
+%.o: %.cpp
+
+# Object file rules
+%.o: %.c
+	$(CC_COMMAND)
+	$(POSTCOMPILE)
+
+$(PROGDIR)/%.o: $(PROGDIR)/%.c
+	$(CC_COMMAND)
+	$(POSTCOMPILE)
+
+%.o: %.cpp
+	$(CXX_COMMAND)
+	$(POSTCOMPILE)
+
+test/%.o: PZSTD_CPPFLAGS += $(GTEST_INC)
+test/%.o: test/%.cpp
+	$(CXX_COMMAND)
+	$(POSTCOMPILE)
+
+utils/test/%.o: PZSTD_CPPFLAGS += $(GTEST_INC)
+utils/test/%.o: utils/test/%.cpp
+	$(CXX_COMMAND)
+	$(POSTCOMPILE)
+
+# Dependency file stuff
+.PRECIOUS: %.d test/%.d utils/test/%.d
+
+# Include rules that specify header file dependencies
+-include $(patsubst %,%.d,$(basename $(ALL_SRCS)))
diff --git a/contrib/pzstd/Options.cpp b/contrib/pzstd/Options.cpp
new file mode 100644
index 0000000..0b14033
--- /dev/null
+++ b/contrib/pzstd/Options.cpp
@@ -0,0 +1,421 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#include "Options.h"
+#include "utils/ScopeGuard.h"
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#if defined(MSDOS) || defined(OS2) || defined(WIN32) || defined(_WIN32) ||     \
+    defined(__CYGWIN__)
+#include  /* _isatty */
+#define IS_CONSOLE(stdStream) _isatty(_fileno(stdStream))
+#elif defined(_POSIX_C_SOURCE) || defined(_XOPEN_SOURCE) || defined(_POSIX_SOURCE) || (defined(__APPLE__) && defined(__MACH__)) || \
+      defined(__DragonFly__) || defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__)  /* https://sourceforge.net/p/predef/wiki/OperatingSystems/ */
+#include  /* isatty */
+#define IS_CONSOLE(stdStream) isatty(fileno(stdStream))
+#else
+#define IS_CONSOLE(stdStream) 0
+#endif
+
+namespace pzstd {
+
+namespace {
+unsigned defaultNumThreads() {
+#ifdef PZSTD_NUM_THREADS
+  return PZSTD_NUM_THREADS;
+#else
+  return std::thread::hardware_concurrency();
+#endif
+}
+
+unsigned parseUnsigned(const char **arg) {
+  unsigned result = 0;
+  while (**arg >= '0' && **arg <= '9') {
+    result *= 10;
+    result += **arg - '0';
+    ++(*arg);
+  }
+  return result;
+}
+
+const char *getArgument(const char *options, const char **argv, int &i,
+                        int argc) {
+  if (options[1] != 0) {
+    return options + 1;
+  }
+  ++i;
+  if (i == argc) {
+    std::fprintf(stderr, "Option -%c requires an argument, but none provided\n",
+                 *options);
+    return nullptr;
+  }
+  return argv[i];
+}
+
+const std::string kZstdExtension = ".zst";
+constexpr char kStdIn[] = "-";
+constexpr char kStdOut[] = "-";
+constexpr unsigned kDefaultCompressionLevel = 3;
+constexpr unsigned kMaxNonUltraCompressionLevel = 19;
+
+#ifdef _WIN32
+const char nullOutput[] = "nul";
+#else
+const char nullOutput[] = "/dev/null";
+#endif
+
+void notSupported(const char *option) {
+  std::fprintf(stderr, "Operation not supported: %s\n", option);
+}
+
+void usage() {
+  std::fprintf(stderr, "Usage:\n");
+  std::fprintf(stderr, "  pzstd [args] [FILE(s)]\n");
+  std::fprintf(stderr, "Parallel ZSTD options:\n");
+  std::fprintf(stderr, "  -p, --processes   #    : number of threads to use for (de)compression (default:%d)\n", defaultNumThreads());
+
+  std::fprintf(stderr, "ZSTD options:\n");
+  std::fprintf(stderr, "  -#                     : # compression level (1-%d, default:%d)\n", kMaxNonUltraCompressionLevel, kDefaultCompressionLevel);
+  std::fprintf(stderr, "  -d, --decompress       : decompression\n");
+  std::fprintf(stderr, "  -o                file : result stored into `file` (only if 1 input file)\n");
+  std::fprintf(stderr, "  -f, --force            : overwrite output without prompting\n");
+  std::fprintf(stderr, "      --rm               : remove source file(s) after successful (de)compression\n");
+  std::fprintf(stderr, "  -k, --keep             : preserve source file(s) (default)\n");
+  std::fprintf(stderr, "  -h, --help             : display help and exit\n");
+  std::fprintf(stderr, "  -V, --version          : display version number and exit\n");
+  std::fprintf(stderr, "  -v, --verbose          : verbose mode; specify multiple times to increase log level (default:2)\n");
+  std::fprintf(stderr, "  -q, --quiet            : suppress warnings; specify twice to suppress errors too\n");
+  std::fprintf(stderr, "  -c, --stdout           : force write to standard output, even if it is the console\n");
+#ifdef UTIL_HAS_CREATEFILELIST
+  std::fprintf(stderr, "  -r                     : operate recursively on directories\n");
+#endif
+  std::fprintf(stderr, "      --ultra            : enable levels beyond %i, up to %i (requires more memory)\n", kMaxNonUltraCompressionLevel, ZSTD_maxCLevel());
+  std::fprintf(stderr, "  -C, --check            : integrity check (default)\n");
+  std::fprintf(stderr, "      --no-check         : no integrity check\n");
+  std::fprintf(stderr, "  -t, --test             : test compressed file integrity\n");
+  std::fprintf(stderr, "  --                     : all arguments after \"--\" are treated as files\n");
+}
+} // anonymous namespace
+
+Options::Options()
+    : numThreads(defaultNumThreads()), maxWindowLog(23),
+      compressionLevel(kDefaultCompressionLevel), decompress(false),
+      overwrite(false), keepSource(true), writeMode(WriteMode::Auto),
+      checksum(true), verbosity(2) {}
+
+Options::Status Options::parse(int argc, const char **argv) {
+  bool test = false;
+  bool recursive = false;
+  bool ultra = false;
+  bool forceStdout = false;
+  // Local copy of input files, which are pointers into argv.
+  std::vector localInputFiles;
+  for (int i = 1; i < argc; ++i) {
+    const char *arg = argv[i];
+    // Protect against empty arguments
+    if (arg[0] == 0) {
+      continue;
+    }
+    // Everything after "--" is an input file
+    if (!std::strcmp(arg, "--")) {
+      ++i;
+      std::copy(argv + i, argv + argc, std::back_inserter(localInputFiles));
+      break;
+    }
+    // Long arguments that don't have a short option
+    {
+      bool isLongOption = true;
+      if (!std::strcmp(arg, "--rm")) {
+        keepSource = false;
+      } else if (!std::strcmp(arg, "--ultra")) {
+        ultra = true;
+        maxWindowLog = 0;
+      } else if (!std::strcmp(arg, "--no-check")) {
+        checksum = false;
+      } else if (!std::strcmp(arg, "--sparse")) {
+        writeMode = WriteMode::Sparse;
+        notSupported("Sparse mode");
+        return Status::Failure;
+      } else if (!std::strcmp(arg, "--no-sparse")) {
+        writeMode = WriteMode::Regular;
+        notSupported("Sparse mode");
+        return Status::Failure;
+      } else if (!std::strcmp(arg, "--dictID")) {
+        notSupported(arg);
+        return Status::Failure;
+      } else if (!std::strcmp(arg, "--no-dictID")) {
+        notSupported(arg);
+        return Status::Failure;
+      } else {
+        isLongOption = false;
+      }
+      if (isLongOption) {
+        continue;
+      }
+    }
+    // Arguments with a short option simply set their short option.
+    const char *options = nullptr;
+    if (!std::strcmp(arg, "--processes")) {
+      options = "p";
+    } else if (!std::strcmp(arg, "--version")) {
+      options = "V";
+    } else if (!std::strcmp(arg, "--help")) {
+      options = "h";
+    } else if (!std::strcmp(arg, "--decompress")) {
+      options = "d";
+    } else if (!std::strcmp(arg, "--force")) {
+      options = "f";
+    } else if (!std::strcmp(arg, "--stdout")) {
+      options = "c";
+    } else if (!std::strcmp(arg, "--keep")) {
+      options = "k";
+    } else if (!std::strcmp(arg, "--verbose")) {
+      options = "v";
+    } else if (!std::strcmp(arg, "--quiet")) {
+      options = "q";
+    } else if (!std::strcmp(arg, "--check")) {
+      options = "C";
+    } else if (!std::strcmp(arg, "--test")) {
+      options = "t";
+    } else if (arg[0] == '-' && arg[1] != 0) {
+      options = arg + 1;
+    } else {
+      localInputFiles.emplace_back(arg);
+      continue;
+    }
+    assert(options != nullptr);
+
+    bool finished = false;
+    while (!finished && *options != 0) {
+      // Parse the compression level
+      if (*options >= '0' && *options <= '9') {
+        compressionLevel = parseUnsigned(&options);
+        continue;
+      }
+
+      switch (*options) {
+      case 'h':
+      case 'H':
+        usage();
+        return Status::Message;
+      case 'V':
+        std::fprintf(stderr, "PZSTD version: %s.\n", ZSTD_VERSION_STRING);
+        return Status::Message;
+      case 'p': {
+        finished = true;
+        const char *optionArgument = getArgument(options, argv, i, argc);
+        if (optionArgument == nullptr) {
+          return Status::Failure;
+        }
+        if (*optionArgument < '0' || *optionArgument > '9') {
+          std::fprintf(stderr, "Option -p expects a number, but %s provided\n",
+                       optionArgument);
+          return Status::Failure;
+        }
+        numThreads = parseUnsigned(&optionArgument);
+        if (*optionArgument != 0) {
+          std::fprintf(stderr,
+                       "Option -p expects a number, but %u%s provided\n",
+                       numThreads, optionArgument);
+          return Status::Failure;
+        }
+        break;
+      }
+      case 'o': {
+        finished = true;
+        const char *optionArgument = getArgument(options, argv, i, argc);
+        if (optionArgument == nullptr) {
+          return Status::Failure;
+        }
+        outputFile = optionArgument;
+        break;
+      }
+      case 'C':
+        checksum = true;
+        break;
+      case 'k':
+        keepSource = true;
+        break;
+      case 'd':
+        decompress = true;
+        break;
+      case 'f':
+        overwrite = true;
+        forceStdout = true;
+        break;
+      case 't':
+        test = true;
+        decompress = true;
+        break;
+#ifdef UTIL_HAS_CREATEFILELIST
+      case 'r':
+        recursive = true;
+        break;
+#endif
+      case 'c':
+        outputFile = kStdOut;
+        forceStdout = true;
+        break;
+      case 'v':
+        ++verbosity;
+        break;
+      case 'q':
+        --verbosity;
+        // Ignore them for now
+        break;
+      // Unsupported options from Zstd
+      case 'D':
+      case 's':
+        notSupported("Zstd dictionaries.");
+        return Status::Failure;
+      case 'b':
+      case 'e':
+      case 'i':
+      case 'B':
+        notSupported("Zstd benchmarking options.");
+        return Status::Failure;
+      default:
+        std::fprintf(stderr, "Invalid argument: %s\n", arg);
+        return Status::Failure;
+      }
+      if (!finished) {
+        ++options;
+      }
+    } // while (*options != 0);
+  }   // for (int i = 1; i < argc; ++i);
+
+  // Set options for test mode
+  if (test) {
+    outputFile = nullOutput;
+    keepSource = true;
+  }
+
+  // Input file defaults to standard input if not provided.
+  if (localInputFiles.empty()) {
+    localInputFiles.emplace_back(kStdIn);
+  }
+
+  // Check validity of input files
+  if (localInputFiles.size() > 1) {
+    const auto it = std::find(localInputFiles.begin(), localInputFiles.end(),
+                              std::string{kStdIn});
+    if (it != localInputFiles.end()) {
+      std::fprintf(
+          stderr,
+          "Cannot specify standard input when handling multiple files\n");
+      return Status::Failure;
+    }
+  }
+  if (localInputFiles.size() > 1 || recursive) {
+    if (!outputFile.empty() && outputFile != nullOutput) {
+      std::fprintf(
+          stderr,
+          "Cannot specify an output file when handling multiple inputs\n");
+      return Status::Failure;
+    }
+  }
+
+  // Translate input files/directories into files to (de)compress
+  if (recursive) {
+    char *scratchBuffer = nullptr;
+    unsigned numFiles = 0;
+    const char **files =
+        UTIL_createFileList(localInputFiles.data(), localInputFiles.size(),
+                            &scratchBuffer, &numFiles);
+    if (files == nullptr) {
+      std::fprintf(stderr, "Error traversing directories\n");
+      return Status::Failure;
+    }
+    auto guard =
+        makeScopeGuard([&] { UTIL_freeFileList(files, scratchBuffer); });
+    if (numFiles == 0) {
+      std::fprintf(stderr, "No files found\n");
+      return Status::Failure;
+    }
+    inputFiles.resize(numFiles);
+    std::copy(files, files + numFiles, inputFiles.begin());
+  } else {
+    inputFiles.resize(localInputFiles.size());
+    std::copy(localInputFiles.begin(), localInputFiles.end(),
+              inputFiles.begin());
+  }
+  localInputFiles.clear();
+  assert(!inputFiles.empty());
+
+  // If reading from standard input, default to standard output
+  if (inputFiles[0] == kStdIn && outputFile.empty()) {
+    assert(inputFiles.size() == 1);
+    outputFile = "-";
+  }
+
+  if (inputFiles[0] == kStdIn && IS_CONSOLE(stdin)) {
+    assert(inputFiles.size() == 1);
+    std::fprintf(stderr, "Cannot read input from interactive console\n");
+    return Status::Failure;
+  }
+  if (outputFile == "-" && IS_CONSOLE(stdout) && !(forceStdout && decompress)) {
+    std::fprintf(stderr, "Will not write to console stdout unless -c or -f is "
+                         "specified and decompressing\n");
+    return Status::Failure;
+  }
+
+  // Check compression level
+  {
+    unsigned maxCLevel =
+        ultra ? ZSTD_maxCLevel() : kMaxNonUltraCompressionLevel;
+    if (compressionLevel > maxCLevel || compressionLevel == 0) {
+      std::fprintf(stderr, "Invalid compression level %u.\n", compressionLevel);
+      return Status::Failure;
+    }
+  }
+
+  // Check that numThreads is set
+  if (numThreads == 0) {
+    std::fprintf(stderr, "Invalid arguments: # of threads not specified "
+                         "and unable to determine hardware concurrency.\n");
+    return Status::Failure;
+  }
+
+  // Modify verbosity
+  // If we are piping input and output, turn off interaction
+  if (inputFiles[0] == kStdIn && outputFile == kStdOut && verbosity == 2) {
+    verbosity = 1;
+  }
+  // If we are in multi-file mode, turn off interaction
+  if (inputFiles.size() > 1 && verbosity == 2) {
+    verbosity = 1;
+  }
+
+  return Status::Success;
+}
+
+std::string Options::getOutputFile(const std::string &inputFile) const {
+  if (!outputFile.empty()) {
+    return outputFile;
+  }
+  // Attempt to add/remove zstd extension from the input file
+  if (decompress) {
+    int stemSize = inputFile.size() - kZstdExtension.size();
+    if (stemSize > 0 && inputFile.substr(stemSize) == kZstdExtension) {
+      return inputFile.substr(0, stemSize);
+    } else {
+      return "";
+    }
+  } else {
+    return inputFile + kZstdExtension;
+  }
+}
+}
diff --git a/contrib/pzstd/Options.h b/contrib/pzstd/Options.h
new file mode 100644
index 0000000..97c3885
--- /dev/null
+++ b/contrib/pzstd/Options.h
@@ -0,0 +1,68 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#pragma once
+
+#define ZSTD_STATIC_LINKING_ONLY
+#include "zstd.h"
+#undef ZSTD_STATIC_LINKING_ONLY
+
+#include 
+#include 
+#include 
+
+namespace pzstd {
+
+struct Options {
+  enum class WriteMode { Regular, Auto, Sparse };
+
+  unsigned numThreads;
+  unsigned maxWindowLog;
+  unsigned compressionLevel;
+  bool decompress;
+  std::vector inputFiles;
+  std::string outputFile;
+  bool overwrite;
+  bool keepSource;
+  WriteMode writeMode;
+  bool checksum;
+  int verbosity;
+
+  enum class Status {
+    Success, // Successfully parsed options
+    Failure, // Failure to parse options
+    Message  // Options specified to print a message (e.g. "-h")
+  };
+
+  Options();
+  Options(unsigned numThreads, unsigned maxWindowLog, unsigned compressionLevel,
+          bool decompress, std::vector inputFiles,
+          std::string outputFile, bool overwrite, bool keepSource,
+          WriteMode writeMode, bool checksum, int verbosity)
+      : numThreads(numThreads), maxWindowLog(maxWindowLog),
+        compressionLevel(compressionLevel), decompress(decompress),
+        inputFiles(std::move(inputFiles)), outputFile(std::move(outputFile)),
+        overwrite(overwrite), keepSource(keepSource), writeMode(writeMode),
+        checksum(checksum), verbosity(verbosity) {}
+
+  Status parse(int argc, const char **argv);
+
+  ZSTD_parameters determineParameters() const {
+    ZSTD_parameters params = ZSTD_getParams(compressionLevel, 0, 0);
+    params.fParams.contentSizeFlag = 1;
+    params.fParams.checksumFlag = checksum;
+    if (maxWindowLog != 0 && params.cParams.windowLog > maxWindowLog) {
+      params.cParams.windowLog = maxWindowLog;
+      params.cParams = ZSTD_adjustCParams(params.cParams, 0, 0);
+    }
+    return params;
+  }
+
+  std::string getOutputFile(const std::string &inputFile) const;
+};
+}
diff --git a/contrib/pzstd/Pzstd.cpp b/contrib/pzstd/Pzstd.cpp
new file mode 100644
index 0000000..f4cb19d
--- /dev/null
+++ b/contrib/pzstd/Pzstd.cpp
@@ -0,0 +1,615 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#include "Pzstd.h"
+#include "SkippableFrame.h"
+#include "utils/FileSystem.h"
+#include "utils/Range.h"
+#include "utils/ScopeGuard.h"
+#include "utils/ThreadPool.h"
+#include "utils/WorkQueue.h"
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#if defined(MSDOS) || defined(OS2) || defined(WIN32) || defined(_WIN32) || defined(__CYGWIN__)
+#  include     /* _O_BINARY */
+#  include        /* _setmode, _isatty */
+#  define SET_BINARY_MODE(file) { if (_setmode(_fileno(file), _O_BINARY) == -1) perror("Cannot set _O_BINARY"); }
+#else
+#  include    /* isatty */
+#  define SET_BINARY_MODE(file)
+#endif
+
+namespace pzstd {
+
+namespace {
+#ifdef _WIN32
+const std::string nullOutput = "nul";
+#else
+const std::string nullOutput = "/dev/null";
+#endif
+}
+
+using std::size_t;
+
+static std::uintmax_t fileSizeOrZero(const std::string &file) {
+  if (file == "-") {
+    return 0;
+  }
+  std::error_code ec;
+  auto size = file_size(file, ec);
+  if (ec) {
+    size = 0;
+  }
+  return size;
+}
+
+static std::uint64_t handleOneInput(const Options &options,
+                             const std::string &inputFile,
+                             FILE* inputFd,
+                             const std::string &outputFile,
+                             FILE* outputFd,
+                             SharedState& state) {
+  auto inputSize = fileSizeOrZero(inputFile);
+  // WorkQueue outlives ThreadPool so in the case of error we are certain
+  // we don't accidently try to call push() on it after it is destroyed
+  WorkQueue> outs{options.numThreads + 1};
+  std::uint64_t bytesRead;
+  std::uint64_t bytesWritten;
+  {
+    // Initialize the (de)compression thread pool with numThreads
+    ThreadPool executor(options.numThreads);
+    // Run the reader thread on an extra thread
+    ThreadPool readExecutor(1);
+    if (!options.decompress) {
+      // Add a job that reads the input and starts all the compression jobs
+      readExecutor.add(
+          [&state, &outs, &executor, inputFd, inputSize, &options, &bytesRead] {
+            bytesRead = asyncCompressChunks(
+                state,
+                outs,
+                executor,
+                inputFd,
+                inputSize,
+                options.numThreads,
+                options.determineParameters());
+          });
+      // Start writing
+      bytesWritten = writeFile(state, outs, outputFd, options.decompress);
+    } else {
+      // Add a job that reads the input and starts all the decompression jobs
+      readExecutor.add([&state, &outs, &executor, inputFd, &bytesRead] {
+        bytesRead = asyncDecompressFrames(state, outs, executor, inputFd);
+      });
+      // Start writing
+      bytesWritten = writeFile(state, outs, outputFd, options.decompress);
+    }
+  }
+  if (!state.errorHolder.hasError()) {
+    std::string inputFileName = inputFile == "-" ? "stdin" : inputFile;
+    std::string outputFileName = outputFile == "-" ? "stdout" : outputFile;
+    if (!options.decompress) {
+      double ratio = static_cast(bytesWritten) /
+                     static_cast(bytesRead + !bytesRead);
+      state.log(INFO, "%-20s :%6.2f%%   (%6" PRIu64 " => %6" PRIu64
+                   " bytes, %s)\n",
+                   inputFileName.c_str(), ratio * 100, bytesRead, bytesWritten,
+                   outputFileName.c_str());
+    } else {
+      state.log(INFO, "%-20s: %" PRIu64 " bytes \n",
+                   inputFileName.c_str(),bytesWritten);
+    }
+  }
+  return bytesWritten;
+}
+
+static FILE *openInputFile(const std::string &inputFile,
+                           ErrorHolder &errorHolder) {
+  if (inputFile == "-") {
+    SET_BINARY_MODE(stdin);
+    return stdin;
+  }
+  // Check if input file is a directory
+  {
+    std::error_code ec;
+    if (is_directory(inputFile, ec)) {
+      errorHolder.setError("Output file is a directory -- ignored");
+      return nullptr;
+    }
+  }
+  auto inputFd = std::fopen(inputFile.c_str(), "rb");
+  if (!errorHolder.check(inputFd != nullptr, "Failed to open input file")) {
+    return nullptr;
+  }
+  return inputFd;
+}
+
+static FILE *openOutputFile(const Options &options,
+                            const std::string &outputFile,
+                            SharedState& state) {
+  if (outputFile == "-") {
+    SET_BINARY_MODE(stdout);
+    return stdout;
+  }
+  // Check if the output file exists and then open it
+  if (!options.overwrite && outputFile != nullOutput) {
+    auto outputFd = std::fopen(outputFile.c_str(), "rb");
+    if (outputFd != nullptr) {
+      std::fclose(outputFd);
+      if (!state.log.logsAt(INFO)) {
+        state.errorHolder.setError("Output file exists");
+        return nullptr;
+      }
+      state.log(
+          INFO,
+          "pzstd: %s already exists; do you wish to overwrite (y/n) ? ",
+          outputFile.c_str());
+      int c = getchar();
+      if (c != 'y' && c != 'Y') {
+        state.errorHolder.setError("Not overwritten");
+        return nullptr;
+      }
+    }
+  }
+  auto outputFd = std::fopen(outputFile.c_str(), "wb");
+  if (!state.errorHolder.check(
+          outputFd != nullptr, "Failed to open output file")) {
+    return nullptr;
+  }
+  return outputFd;
+}
+
+int pzstdMain(const Options &options) {
+  int returnCode = 0;
+  SharedState state(options);
+  for (const auto& input : options.inputFiles) {
+    // Setup the shared state
+    auto printErrorGuard = makeScopeGuard([&] {
+      if (state.errorHolder.hasError()) {
+        returnCode = 1;
+        state.log(ERROR, "pzstd: %s: %s.\n", input.c_str(),
+                  state.errorHolder.getError().c_str());
+      }
+    });
+    // Open the input file
+    auto inputFd = openInputFile(input, state.errorHolder);
+    if (inputFd == nullptr) {
+      continue;
+    }
+    auto closeInputGuard = makeScopeGuard([&] { std::fclose(inputFd); });
+    // Open the output file
+    auto outputFile = options.getOutputFile(input);
+    if (!state.errorHolder.check(outputFile != "",
+                           "Input file does not have extension .zst")) {
+      continue;
+    }
+    auto outputFd = openOutputFile(options, outputFile, state);
+    if (outputFd == nullptr) {
+      continue;
+    }
+    auto closeOutputGuard = makeScopeGuard([&] { std::fclose(outputFd); });
+    // (de)compress the file
+    handleOneInput(options, input, inputFd, outputFile, outputFd, state);
+    if (state.errorHolder.hasError()) {
+      continue;
+    }
+    // Delete the input file if necessary
+    if (!options.keepSource) {
+      // Be sure that we are done and have written everything before we delete
+      if (!state.errorHolder.check(std::fclose(inputFd) == 0,
+                             "Failed to close input file")) {
+        continue;
+      }
+      closeInputGuard.dismiss();
+      if (!state.errorHolder.check(std::fclose(outputFd) == 0,
+                             "Failed to close output file")) {
+        continue;
+      }
+      closeOutputGuard.dismiss();
+      if (std::remove(input.c_str()) != 0) {
+        state.errorHolder.setError("Failed to remove input file");
+        continue;
+      }
+    }
+  }
+  // Returns 1 if any of the files failed to (de)compress.
+  return returnCode;
+}
+
+/// Construct a `ZSTD_inBuffer` that points to the data in `buffer`.
+static ZSTD_inBuffer makeZstdInBuffer(const Buffer& buffer) {
+  return ZSTD_inBuffer{buffer.data(), buffer.size(), 0};
+}
+
+/**
+ * Advance `buffer` and `inBuffer` by the amount of data read, as indicated by
+ * `inBuffer.pos`.
+ */
+void advance(Buffer& buffer, ZSTD_inBuffer& inBuffer) {
+  auto pos = inBuffer.pos;
+  inBuffer.src = static_cast(inBuffer.src) + pos;
+  inBuffer.size -= pos;
+  inBuffer.pos = 0;
+  return buffer.advance(pos);
+}
+
+/// Construct a `ZSTD_outBuffer` that points to the data in `buffer`.
+static ZSTD_outBuffer makeZstdOutBuffer(Buffer& buffer) {
+  return ZSTD_outBuffer{buffer.data(), buffer.size(), 0};
+}
+
+/**
+ * Split `buffer` and advance `outBuffer` by the amount of data written, as
+ * indicated by `outBuffer.pos`.
+ */
+Buffer split(Buffer& buffer, ZSTD_outBuffer& outBuffer) {
+  auto pos = outBuffer.pos;
+  outBuffer.dst = static_cast(outBuffer.dst) + pos;
+  outBuffer.size -= pos;
+  outBuffer.pos = 0;
+  return buffer.splitAt(pos);
+}
+
+/**
+ * Stream chunks of input from `in`, compress it, and stream it out to `out`.
+ *
+ * @param state        The shared state
+ * @param in           Queue that we `pop()` input buffers from
+ * @param out          Queue that we `push()` compressed output buffers to
+ * @param maxInputSize An upper bound on the size of the input
+ */
+static void compress(
+    SharedState& state,
+    std::shared_ptr in,
+    std::shared_ptr out,
+    size_t maxInputSize) {
+  auto& errorHolder = state.errorHolder;
+  auto guard = makeScopeGuard([&] { out->finish(); });
+  // Initialize the CCtx
+  auto ctx = state.cStreamPool->get();
+  if (!errorHolder.check(ctx != nullptr, "Failed to allocate ZSTD_CStream")) {
+    return;
+  }
+  {
+    auto err = ZSTD_resetCStream(ctx.get(), 0);
+    if (!errorHolder.check(!ZSTD_isError(err), ZSTD_getErrorName(err))) {
+      return;
+    }
+  }
+
+  // Allocate space for the result
+  auto outBuffer = Buffer(ZSTD_compressBound(maxInputSize));
+  auto zstdOutBuffer = makeZstdOutBuffer(outBuffer);
+  {
+    Buffer inBuffer;
+    // Read a buffer in from the input queue
+    while (in->pop(inBuffer) && !errorHolder.hasError()) {
+      auto zstdInBuffer = makeZstdInBuffer(inBuffer);
+      // Compress the whole buffer and send it to the output queue
+      while (!inBuffer.empty() && !errorHolder.hasError()) {
+        if (!errorHolder.check(
+                !outBuffer.empty(), "ZSTD_compressBound() was too small")) {
+          return;
+        }
+        // Compress
+        auto err =
+            ZSTD_compressStream(ctx.get(), &zstdOutBuffer, &zstdInBuffer);
+        if (!errorHolder.check(!ZSTD_isError(err), ZSTD_getErrorName(err))) {
+          return;
+        }
+        // Split the compressed data off outBuffer and pass to the output queue
+        out->push(split(outBuffer, zstdOutBuffer));
+        // Forget about the data we already compressed
+        advance(inBuffer, zstdInBuffer);
+      }
+    }
+  }
+  // Write the epilog
+  size_t bytesLeft;
+  do {
+    if (!errorHolder.check(
+            !outBuffer.empty(), "ZSTD_compressBound() was too small")) {
+      return;
+    }
+    bytesLeft = ZSTD_endStream(ctx.get(), &zstdOutBuffer);
+    if (!errorHolder.check(
+            !ZSTD_isError(bytesLeft), ZSTD_getErrorName(bytesLeft))) {
+      return;
+    }
+    out->push(split(outBuffer, zstdOutBuffer));
+  } while (bytesLeft != 0 && !errorHolder.hasError());
+}
+
+/**
+ * Calculates how large each independently compressed frame should be.
+ *
+ * @param size       The size of the source if known, 0 otherwise
+ * @param numThreads The number of threads available to run compression jobs on
+ * @param params     The zstd parameters to be used for compression
+ */
+static size_t calculateStep(
+    std::uintmax_t size,
+    size_t numThreads,
+    const ZSTD_parameters ¶ms) {
+  (void)size;
+  (void)numThreads;
+  return size_t{1} << (params.cParams.windowLog + 2);
+}
+
+namespace {
+enum class FileStatus { Continue, Done, Error };
+/// Determines the status of the file descriptor `fd`.
+FileStatus fileStatus(FILE* fd) {
+  if (std::feof(fd)) {
+    return FileStatus::Done;
+  } else if (std::ferror(fd)) {
+    return FileStatus::Error;
+  }
+  return FileStatus::Continue;
+}
+} // anonymous namespace
+
+/**
+ * Reads `size` data in chunks of `chunkSize` and puts it into `queue`.
+ * Will read less if an error or EOF occurs.
+ * Returns the status of the file after all of the reads have occurred.
+ */
+static FileStatus
+readData(BufferWorkQueue& queue, size_t chunkSize, size_t size, FILE* fd,
+         std::uint64_t *totalBytesRead) {
+  Buffer buffer(size);
+  while (!buffer.empty()) {
+    auto bytesRead =
+        std::fread(buffer.data(), 1, std::min(chunkSize, buffer.size()), fd);
+    *totalBytesRead += bytesRead;
+    queue.push(buffer.splitAt(bytesRead));
+    auto status = fileStatus(fd);
+    if (status != FileStatus::Continue) {
+      return status;
+    }
+  }
+  return FileStatus::Continue;
+}
+
+std::uint64_t asyncCompressChunks(
+    SharedState& state,
+    WorkQueue>& chunks,
+    ThreadPool& executor,
+    FILE* fd,
+    std::uintmax_t size,
+    size_t numThreads,
+    ZSTD_parameters params) {
+  auto chunksGuard = makeScopeGuard([&] { chunks.finish(); });
+  std::uint64_t bytesRead = 0;
+
+  // Break the input up into chunks of size `step` and compress each chunk
+  // independently.
+  size_t step = calculateStep(size, numThreads, params);
+  state.log(DEBUG, "Chosen frame size: %zu\n", step);
+  auto status = FileStatus::Continue;
+  while (status == FileStatus::Continue && !state.errorHolder.hasError()) {
+    // Make a new input queue that we will put the chunk's input data into.
+    auto in = std::make_shared();
+    auto inGuard = makeScopeGuard([&] { in->finish(); });
+    // Make a new output queue that compress will put the compressed data into.
+    auto out = std::make_shared();
+    // Start compression in the thread pool
+    executor.add([&state, in, out, step] {
+      return compress(
+          state, std::move(in), std::move(out), step);
+    });
+    // Pass the output queue to the writer thread.
+    chunks.push(std::move(out));
+    state.log(VERBOSE, "Starting a new frame\n");
+    // Fill the input queue for the compression job we just started
+    status = readData(*in, ZSTD_CStreamInSize(), step, fd, &bytesRead);
+  }
+  state.errorHolder.check(status != FileStatus::Error, "Error reading input");
+  return bytesRead;
+}
+
+/**
+ * Decompress a frame, whose data is streamed into `in`, and stream the output
+ * to `out`.
+ *
+ * @param state        The shared state
+ * @param in           Queue that we `pop()` input buffers from. It contains
+ *                      exactly one compressed frame.
+ * @param out          Queue that we `push()` decompressed output buffers to
+ */
+static void decompress(
+    SharedState& state,
+    std::shared_ptr in,
+    std::shared_ptr out) {
+  auto& errorHolder = state.errorHolder;
+  auto guard = makeScopeGuard([&] { out->finish(); });
+  // Initialize the DCtx
+  auto ctx = state.dStreamPool->get();
+  if (!errorHolder.check(ctx != nullptr, "Failed to allocate ZSTD_DStream")) {
+    return;
+  }
+  {
+    auto err = ZSTD_resetDStream(ctx.get());
+    if (!errorHolder.check(!ZSTD_isError(err), ZSTD_getErrorName(err))) {
+      return;
+    }
+  }
+
+  const size_t outSize = ZSTD_DStreamOutSize();
+  Buffer inBuffer;
+  size_t returnCode = 0;
+  // Read a buffer in from the input queue
+  while (in->pop(inBuffer) && !errorHolder.hasError()) {
+    auto zstdInBuffer = makeZstdInBuffer(inBuffer);
+    // Decompress the whole buffer and send it to the output queue
+    while (!inBuffer.empty() && !errorHolder.hasError()) {
+      // Allocate a buffer with at least outSize bytes.
+      Buffer outBuffer(outSize);
+      auto zstdOutBuffer = makeZstdOutBuffer(outBuffer);
+      // Decompress
+      returnCode =
+          ZSTD_decompressStream(ctx.get(), &zstdOutBuffer, &zstdInBuffer);
+      if (!errorHolder.check(
+              !ZSTD_isError(returnCode), ZSTD_getErrorName(returnCode))) {
+        return;
+      }
+      // Pass the buffer with the decompressed data to the output queue
+      out->push(split(outBuffer, zstdOutBuffer));
+      // Advance past the input we already read
+      advance(inBuffer, zstdInBuffer);
+      if (returnCode == 0) {
+        // The frame is over, prepare to (maybe) start a new frame
+        ZSTD_initDStream(ctx.get());
+      }
+    }
+  }
+  if (!errorHolder.check(returnCode <= 1, "Incomplete block")) {
+    return;
+  }
+  // We've given ZSTD_decompressStream all of our data, but there may still
+  // be data to read.
+  while (returnCode == 1) {
+    // Allocate a buffer with at least outSize bytes.
+    Buffer outBuffer(outSize);
+    auto zstdOutBuffer = makeZstdOutBuffer(outBuffer);
+    // Pass in no input.
+    ZSTD_inBuffer zstdInBuffer{nullptr, 0, 0};
+    // Decompress
+    returnCode =
+        ZSTD_decompressStream(ctx.get(), &zstdOutBuffer, &zstdInBuffer);
+    if (!errorHolder.check(
+            !ZSTD_isError(returnCode), ZSTD_getErrorName(returnCode))) {
+      return;
+    }
+    // Pass the buffer with the decompressed data to the output queue
+    out->push(split(outBuffer, zstdOutBuffer));
+  }
+}
+
+std::uint64_t asyncDecompressFrames(
+    SharedState& state,
+    WorkQueue>& frames,
+    ThreadPool& executor,
+    FILE* fd) {
+  auto framesGuard = makeScopeGuard([&] { frames.finish(); });
+  std::uint64_t totalBytesRead = 0;
+
+  // Split the source up into its component frames.
+  // If we find our recognized skippable frame we know the next frames size
+  // which means that we can decompress each standard frame in independently.
+  // Otherwise, we will decompress using only one decompression task.
+  const size_t chunkSize = ZSTD_DStreamInSize();
+  auto status = FileStatus::Continue;
+  while (status == FileStatus::Continue && !state.errorHolder.hasError()) {
+    // Make a new input queue that we will put the frames's bytes into.
+    auto in = std::make_shared();
+    auto inGuard = makeScopeGuard([&] { in->finish(); });
+    // Make a output queue that decompress will put the decompressed data into
+    auto out = std::make_shared();
+
+    size_t frameSize;
+    {
+      // Calculate the size of the next frame.
+      // frameSize is 0 if the frame info can't be decoded.
+      Buffer buffer(SkippableFrame::kSize);
+      auto bytesRead = std::fread(buffer.data(), 1, buffer.size(), fd);
+      totalBytesRead += bytesRead;
+      status = fileStatus(fd);
+      if (bytesRead == 0 && status != FileStatus::Continue) {
+        break;
+      }
+      buffer.subtract(buffer.size() - bytesRead);
+      frameSize = SkippableFrame::tryRead(buffer.range());
+      in->push(std::move(buffer));
+    }
+    if (frameSize == 0) {
+      // We hit a non SkippableFrame, so this will be the last job.
+      // Make sure that we don't use too much memory
+      in->setMaxSize(64);
+      out->setMaxSize(64);
+    }
+    // Start decompression in the thread pool
+    executor.add([&state, in, out] {
+      return decompress(state, std::move(in), std::move(out));
+    });
+    // Pass the output queue to the writer thread
+    frames.push(std::move(out));
+    if (frameSize == 0) {
+      // We hit a non SkippableFrame ==> not compressed by pzstd or corrupted
+      // Pass the rest of the source to this decompression task
+      state.log(VERBOSE,
+          "Input not in pzstd format, falling back to serial decompression\n");
+      while (status == FileStatus::Continue && !state.errorHolder.hasError()) {
+        status = readData(*in, chunkSize, chunkSize, fd, &totalBytesRead);
+      }
+      break;
+    }
+    state.log(VERBOSE, "Decompressing a frame of size %zu", frameSize);
+    // Fill the input queue for the decompression job we just started
+    status = readData(*in, chunkSize, frameSize, fd, &totalBytesRead);
+  }
+  state.errorHolder.check(status != FileStatus::Error, "Error reading input");
+  return totalBytesRead;
+}
+
+/// Write `data` to `fd`, returns true iff success.
+static bool writeData(ByteRange data, FILE* fd) {
+  while (!data.empty()) {
+    data.advance(std::fwrite(data.begin(), 1, data.size(), fd));
+    if (std::ferror(fd)) {
+      return false;
+    }
+  }
+  return true;
+}
+
+std::uint64_t writeFile(
+    SharedState& state,
+    WorkQueue>& outs,
+    FILE* outputFd,
+    bool decompress) {
+  auto& errorHolder = state.errorHolder;
+  auto lineClearGuard = makeScopeGuard([&state] {
+    state.log.clear(INFO);
+  });
+  std::uint64_t bytesWritten = 0;
+  std::shared_ptr out;
+  // Grab the output queue for each decompression job (in order).
+  while (outs.pop(out) && !errorHolder.hasError()) {
+    if (!decompress) {
+      // If we are compressing and want to write skippable frames we can't
+      // start writing before compression is done because we need to know the
+      // compressed size.
+      // Wait for the compressed size to be available and write skippable frame
+      SkippableFrame frame(out->size());
+      if (!writeData(frame.data(), outputFd)) {
+        errorHolder.setError("Failed to write output");
+        return bytesWritten;
+      }
+      bytesWritten += frame.kSize;
+    }
+    // For each chunk of the frame: Pop it from the queue and write it
+    Buffer buffer;
+    while (out->pop(buffer) && !errorHolder.hasError()) {
+      if (!writeData(buffer.range(), outputFd)) {
+        errorHolder.setError("Failed to write output");
+        return bytesWritten;
+      }
+      bytesWritten += buffer.size();
+      state.log.update(INFO, "Written: %u MB   ",
+                static_cast(bytesWritten >> 20));
+    }
+  }
+  return bytesWritten;
+}
+}
diff --git a/contrib/pzstd/Pzstd.h b/contrib/pzstd/Pzstd.h
new file mode 100644
index 0000000..dc60dd9
--- /dev/null
+++ b/contrib/pzstd/Pzstd.h
@@ -0,0 +1,150 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#pragma once
+
+#include "ErrorHolder.h"
+#include "Logging.h"
+#include "Options.h"
+#include "utils/Buffer.h"
+#include "utils/Range.h"
+#include "utils/ResourcePool.h"
+#include "utils/ThreadPool.h"
+#include "utils/WorkQueue.h"
+#define ZSTD_STATIC_LINKING_ONLY
+#include "zstd.h"
+#undef ZSTD_STATIC_LINKING_ONLY
+
+#include 
+#include 
+#include 
+
+namespace pzstd {
+/**
+ * Runs pzstd with `options` and returns the number of bytes written.
+ * An error occurred if `errorHandler.hasError()`.
+ *
+ * @param options      The pzstd options to use for (de)compression
+ * @returns            0 upon success and non-zero on failure.
+ */
+int pzstdMain(const Options& options);
+
+class SharedState {
+ public:
+  SharedState(const Options& options) : log(options.verbosity) {
+    if (!options.decompress) {
+      auto parameters = options.determineParameters();
+      cStreamPool.reset(new ResourcePool{
+          [this, parameters]() -> ZSTD_CStream* {
+            this->log(VERBOSE, "Creating new ZSTD_CStream\n");
+            auto zcs = ZSTD_createCStream();
+            if (zcs) {
+              auto err = ZSTD_initCStream_advanced(
+                  zcs, nullptr, 0, parameters, 0);
+              if (ZSTD_isError(err)) {
+                ZSTD_freeCStream(zcs);
+                return nullptr;
+              }
+            }
+            return zcs;
+          },
+          [](ZSTD_CStream *zcs) {
+            ZSTD_freeCStream(zcs);
+          }});
+    } else {
+      dStreamPool.reset(new ResourcePool{
+          [this]() -> ZSTD_DStream* {
+            this->log(VERBOSE, "Creating new ZSTD_DStream\n");
+            auto zds = ZSTD_createDStream();
+            if (zds) {
+              auto err = ZSTD_initDStream(zds);
+              if (ZSTD_isError(err)) {
+                ZSTD_freeDStream(zds);
+                return nullptr;
+              }
+            }
+            return zds;
+          },
+          [](ZSTD_DStream *zds) {
+            ZSTD_freeDStream(zds);
+          }});
+    }
+  }
+
+  ~SharedState() {
+    // The resource pools have references to this, so destroy them first.
+    cStreamPool.reset();
+    dStreamPool.reset();
+  }
+
+  Logger log;
+  ErrorHolder errorHolder;
+  std::unique_ptr> cStreamPool;
+  std::unique_ptr> dStreamPool;
+};
+
+/**
+ * Streams input from `fd`, breaks input up into chunks, and compresses each
+ * chunk independently.  Output of each chunk gets streamed to a queue, and
+ * the output queues get put into `chunks` in order.
+ *
+ * @param state        The shared state
+ * @param chunks       Each compression jobs output queue gets `pushed()` here
+ *                      as soon as it is available
+ * @param executor     The thread pool to run compression jobs in
+ * @param fd           The input file descriptor
+ * @param size         The size of the input file if known, 0 otherwise
+ * @param numThreads   The number of threads in the thread pool
+ * @param parameters   The zstd parameters to use for compression
+ * @returns            The number of bytes read from the file
+ */
+std::uint64_t asyncCompressChunks(
+    SharedState& state,
+    WorkQueue>& chunks,
+    ThreadPool& executor,
+    FILE* fd,
+    std::uintmax_t size,
+    std::size_t numThreads,
+    ZSTD_parameters parameters);
+
+/**
+ * Streams input from `fd`.  If pzstd headers are available it breaks the input
+ * up into independent frames.  It sends each frame to an independent
+ * decompression job.  Output of each frame gets streamed to a queue, and
+ * the output queues get put into `frames` in order.
+ *
+ * @param state        The shared state
+ * @param frames       Each decompression jobs output queue gets `pushed()` here
+ *                      as soon as it is available
+ * @param executor     The thread pool to run compression jobs in
+ * @param fd           The input file descriptor
+ * @returns            The number of bytes read from the file
+ */
+std::uint64_t asyncDecompressFrames(
+    SharedState& state,
+    WorkQueue>& frames,
+    ThreadPool& executor,
+    FILE* fd);
+
+/**
+ * Streams input in from each queue in `outs` in order, and writes the data to
+ * `outputFd`.
+ *
+ * @param state        The shared state
+ * @param outs         A queue of output queues, one for each
+ *                      (de)compression job.
+ * @param outputFd     The file descriptor to write to
+ * @param decompress   Are we decompressing?
+ * @returns            The number of bytes written
+ */
+std::uint64_t writeFile(
+    SharedState& state,
+    WorkQueue>& outs,
+    FILE* outputFd,
+    bool decompress);
+}
diff --git a/contrib/pzstd/README.md b/contrib/pzstd/README.md
new file mode 100644
index 0000000..84d9458
--- /dev/null
+++ b/contrib/pzstd/README.md
@@ -0,0 +1,56 @@
+# Parallel Zstandard (PZstandard)
+
+Parallel Zstandard is a Pigz-like tool for Zstandard.
+It provides Zstandard format compatible compression and decompression that is able to utilize multiple cores.
+It breaks the input up into equal sized chunks and compresses each chunk independently into a Zstandard frame.
+It then concatenates the frames together to produce the final compressed output.
+Pzstandard will write a 12 byte header for each frame that is a skippable frame in the Zstandard format, which tells PZstandard the size of the next compressed frame.
+PZstandard supports parallel decompression of files compressed with PZstandard.
+When decompressing files compressed with Zstandard, PZstandard does IO in one thread, and decompression in another.
+
+## Usage
+
+PZstandard supports the same command line interface as Zstandard, but also provides the `-p` option to specify the number of threads.
+Dictionary mode is not currently supported.
+
+Basic usage
+
+    pzstd input-file -o output-file -p num-threads -#          # Compression
+    pzstd -d input-file -o output-file -p num-threads          # Decompression
+
+PZstandard also supports piping and fifo pipes
+
+    cat input-file | pzstd -p num-threads -# -c > /dev/null
+
+For more options
+
+    pzstd --help
+
+PZstandard tries to pick a smart default number of threads if not specified (displayed in `pzstd --help`).
+If this number is not suitable, during compilation you can define `PZSTD_NUM_THREADS` to the number of threads you prefer.
+
+## Benchmarks
+
+As a reference, PZstandard and Pigz were compared on an Intel Core i7 @ 3.1 GHz, each using 4 threads, with the [Silesia compression corpus](http://sun.aei.polsl.pl/~sdeor/index.php?page=silesia).
+
+Compression Speed vs Ratio with 4 Threads | Decompression Speed with 4 Threads
+------------------------------------------|-----------------------------------
+![Compression Speed vs Ratio](images/Cspeed.png "Compression Speed vs Ratio") | ![Decompression Speed](images/Dspeed.png "Decompression Speed")
+
+The test procedure was to run each of the following commands 2 times for each compression level, and take the minimum time.
+
+    time pzstd -# -p 4    -c silesia.tar     > silesia.tar.zst
+    time pzstd -d -p 4    -c silesia.tar.zst > /dev/null
+
+    time pigz  -# -p 4 -k -c silesia.tar     > silesia.tar.gz
+    time pigz  -d -p 4 -k -c silesia.tar.gz  > /dev/null
+
+PZstandard was tested using compression levels 1-19, and Pigz was tested using compression levels 1-9.
+Pigz cannot do parallel decompression, it simply does each of reading, decompression, and writing on separate threads.
+
+## Tests
+
+Tests require that you have [gtest](https://github.com/google/googletest) installed.
+Set `GTEST_INC` and `GTEST_LIB` in `Makefile` to specify the location of the gtest headers and libraries.
+Alternatively, run `make googletest`, which will clone googletest and build it.
+Run `make tests && make check` to run tests.
diff --git a/contrib/pzstd/SkippableFrame.cpp b/contrib/pzstd/SkippableFrame.cpp
new file mode 100644
index 0000000..5dc95e5
--- /dev/null
+++ b/contrib/pzstd/SkippableFrame.cpp
@@ -0,0 +1,30 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#include "SkippableFrame.h"
+#include "mem.h"
+#include "utils/Range.h"
+
+#include 
+
+using namespace pzstd;
+
+SkippableFrame::SkippableFrame(std::uint32_t size) : frameSize_(size) {
+  MEM_writeLE32(data_.data(), kSkippableFrameMagicNumber);
+  MEM_writeLE32(data_.data() + 4, kFrameContentsSize);
+  MEM_writeLE32(data_.data() + 8, frameSize_);
+}
+
+/* static */ std::size_t SkippableFrame::tryRead(ByteRange bytes) {
+  if (bytes.size() < SkippableFrame::kSize ||
+      MEM_readLE32(bytes.begin()) != kSkippableFrameMagicNumber ||
+      MEM_readLE32(bytes.begin() + 4) != kFrameContentsSize) {
+    return 0;
+  }
+  return MEM_readLE32(bytes.begin() + 8);
+}
diff --git a/contrib/pzstd/SkippableFrame.h b/contrib/pzstd/SkippableFrame.h
new file mode 100644
index 0000000..9dc95c1
--- /dev/null
+++ b/contrib/pzstd/SkippableFrame.h
@@ -0,0 +1,64 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#pragma once
+
+#include "utils/Range.h"
+
+#include 
+#include 
+#include 
+#include 
+
+namespace pzstd {
+/**
+ * We put a skippable frame before each frame.
+ * It contains a skippable frame magic number, the size of the skippable frame,
+ * and the size of the next frame.
+ * Each skippable frame is exactly 12 bytes in little endian format.
+ * The first 8 bytes are for compatibility with the ZSTD format.
+ * If we have N threads, the output will look like
+ *
+ * [0x184D2A50|4|size1] [frame1 of size size1]
+ * [0x184D2A50|4|size2] [frame2 of size size2]
+ * ...
+ * [0x184D2A50|4|sizeN] [frameN of size sizeN]
+ *
+ * Each sizeX is 4 bytes.
+ *
+ * These skippable frames should allow us to skip through the compressed file
+ * and only load at most N pages.
+ */
+class SkippableFrame {
+ public:
+  static constexpr std::size_t kSize = 12;
+
+ private:
+  std::uint32_t frameSize_;
+  std::array data_;
+  static constexpr std::uint32_t kSkippableFrameMagicNumber = 0x184D2A50;
+  // Could be improved if the size fits in less bytes
+  static constexpr std::uint32_t kFrameContentsSize = kSize - 8;
+
+ public:
+   // Write the skippable frame to data_ in LE format.
+  explicit SkippableFrame(std::uint32_t size);
+
+  // Read the skippable frame from bytes in LE format.
+  static std::size_t tryRead(ByteRange bytes);
+
+  ByteRange data() const {
+    return {data_.data(), data_.size()};
+  }
+
+  // Size of the next frame.
+  std::size_t frameSize() const {
+    return frameSize_;
+  }
+};
+}
diff --git a/contrib/pzstd/images/Cspeed.png b/contrib/pzstd/images/Cspeed.png
new file mode 100644
index 0000000..aca4f66
Binary files /dev/null and b/contrib/pzstd/images/Cspeed.png differ
diff --git a/contrib/pzstd/images/Dspeed.png b/contrib/pzstd/images/Dspeed.png
new file mode 100644
index 0000000..e48881b
Binary files /dev/null and b/contrib/pzstd/images/Dspeed.png differ
diff --git a/contrib/pzstd/main.cpp b/contrib/pzstd/main.cpp
new file mode 100644
index 0000000..279cbfb
--- /dev/null
+++ b/contrib/pzstd/main.cpp
@@ -0,0 +1,32 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#include "ErrorHolder.h"
+#include "Options.h"
+#include "Pzstd.h"
+#include "utils/FileSystem.h"
+#include "utils/Range.h"
+#include "utils/ScopeGuard.h"
+#include "utils/ThreadPool.h"
+#include "utils/WorkQueue.h"
+
+using namespace pzstd;
+
+int main(int argc, const char** argv) {
+  Options options;
+  switch (options.parse(argc, argv)) {
+  case Options::Status::Failure:
+    return 1;
+  case Options::Status::Message:
+    return 0;
+  default:
+    break;
+  }
+
+  return pzstdMain(options);
+}
diff --git a/contrib/pzstd/test/OptionsTest.cpp b/contrib/pzstd/test/OptionsTest.cpp
new file mode 100644
index 0000000..b3efe2b
--- /dev/null
+++ b/contrib/pzstd/test/OptionsTest.cpp
@@ -0,0 +1,536 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#include "Options.h"
+
+#include 
+#include 
+
+using namespace pzstd;
+
+namespace pzstd {
+bool operator==(const Options &lhs, const Options &rhs) {
+  return lhs.numThreads == rhs.numThreads &&
+         lhs.maxWindowLog == rhs.maxWindowLog &&
+         lhs.compressionLevel == rhs.compressionLevel &&
+         lhs.decompress == rhs.decompress && lhs.inputFiles == rhs.inputFiles &&
+         lhs.outputFile == rhs.outputFile && lhs.overwrite == rhs.overwrite &&
+         lhs.keepSource == rhs.keepSource && lhs.writeMode == rhs.writeMode &&
+         lhs.checksum == rhs.checksum && lhs.verbosity == rhs.verbosity;
+}
+
+std::ostream &operator<<(std::ostream &out, const Options &opt) {
+  out << "{";
+  {
+    out << "\n\t"
+        << "numThreads: " << opt.numThreads;
+    out << ",\n\t"
+        << "maxWindowLog: " << opt.maxWindowLog;
+    out << ",\n\t"
+        << "compressionLevel: " << opt.compressionLevel;
+    out << ",\n\t"
+        << "decompress: " << opt.decompress;
+    out << ",\n\t"
+        << "inputFiles: {";
+    {
+      bool first = true;
+      for (const auto &file : opt.inputFiles) {
+        if (!first) {
+          out << ",";
+        }
+        first = false;
+        out << "\n\t\t" << file;
+      }
+    }
+    out << "\n\t}";
+    out << ",\n\t"
+        << "outputFile: " << opt.outputFile;
+    out << ",\n\t"
+        << "overwrite: " << opt.overwrite;
+    out << ",\n\t"
+        << "keepSource: " << opt.keepSource;
+    out << ",\n\t"
+        << "writeMode: " << static_cast(opt.writeMode);
+    out << ",\n\t"
+        << "checksum: " << opt.checksum;
+    out << ",\n\t"
+        << "verbosity: " << opt.verbosity;
+  }
+  out << "\n}";
+  return out;
+}
+}
+
+namespace {
+#ifdef _WIN32
+const char nullOutput[] = "nul";
+#else
+const char nullOutput[] = "/dev/null";
+#endif
+
+constexpr auto autoMode = Options::WriteMode::Auto;
+} // anonymous namespace
+
+#define EXPECT_SUCCESS(...) EXPECT_EQ(Options::Status::Success, __VA_ARGS__)
+#define EXPECT_FAILURE(...) EXPECT_EQ(Options::Status::Failure, __VA_ARGS__)
+#define EXPECT_MESSAGE(...) EXPECT_EQ(Options::Status::Message, __VA_ARGS__)
+
+template 
+std::array makeArray(Args... args) {
+  return {{nullptr, args...}};
+}
+
+TEST(Options, ValidInputs) {
+  {
+    Options options;
+    auto args = makeArray("--processes", "5", "-o", "x", "y", "-f");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    Options expected = {5,    23,   3,        false, {"y"}, "x",
+                        true, true, autoMode, true,  2};
+    EXPECT_EQ(expected, options);
+  }
+  {
+    Options options;
+    auto args = makeArray("-p", "1", "input", "-19");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    Options expected = {1,     23,   19,       false, {"input"}, "",
+                        false, true, autoMode, true,  2};
+    EXPECT_EQ(expected, options);
+  }
+  {
+    Options options;
+    auto args =
+        makeArray("--ultra", "-22", "-p", "1", "-o", "x", "-d", "x.zst", "-f");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    Options expected = {1,    0,    22,       true, {"x.zst"}, "x",
+                        true, true, autoMode, true, 2};
+    EXPECT_EQ(expected, options);
+  }
+  {
+    Options options;
+    auto args = makeArray("--processes", "100", "hello.zst", "--decompress",
+                          "--force");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    Options expected = {100,  23,       3,    true, {"hello.zst"}, "", true,
+                        true, autoMode, true, 2};
+    EXPECT_EQ(expected, options);
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "-dp", "1", "-c");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    Options expected = {1,     23,   3,        true, {"x"}, "-",
+                        false, true, autoMode, true, 2};
+    EXPECT_EQ(expected, options);
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "-dp", "1", "--stdout");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    Options expected = {1,     23,   3,        true, {"x"}, "-",
+                        false, true, autoMode, true, 2};
+    EXPECT_EQ(expected, options);
+  }
+  {
+    Options options;
+    auto args = makeArray("-p", "1", "x", "-5", "-fo", "-", "--ultra", "-d");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    Options expected = {1,    0,    5,        true, {"x"}, "-",
+                        true, true, autoMode, true, 2};
+    EXPECT_EQ(expected, options);
+  }
+  {
+    Options options;
+    auto args = makeArray("silesia.tar", "-o", "silesia.tar.pzstd", "-p", "2");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    Options expected = {2,
+                        23,
+                        3,
+                        false,
+                        {"silesia.tar"},
+                        "silesia.tar.pzstd",
+                        false,
+                        true,
+                        autoMode,
+                        true,
+                        2};
+    EXPECT_EQ(expected, options);
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "-p", "1");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "-p", "1");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+  }
+}
+
+TEST(Options, GetOutputFile) {
+  {
+    Options options;
+    auto args = makeArray("x");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ("x.zst", options.getOutputFile(options.inputFiles[0]));
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "y", "-o", nullOutput);
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(nullOutput, options.getOutputFile(options.inputFiles[0]));
+  }
+  {
+    Options options;
+    auto args = makeArray("x.zst", "-do", nullOutput);
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(nullOutput, options.getOutputFile(options.inputFiles[0]));
+  }
+  {
+    Options options;
+    auto args = makeArray("x.zst", "-d");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ("x", options.getOutputFile(options.inputFiles[0]));
+  }
+  {
+    Options options;
+    auto args = makeArray("xzst", "-d");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ("", options.getOutputFile(options.inputFiles[0]));
+  }
+  {
+    Options options;
+    auto args = makeArray("xzst", "-doxx");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ("xx", options.getOutputFile(options.inputFiles[0]));
+  }
+}
+
+TEST(Options, MultipleFiles) {
+  {
+    Options options;
+    auto args = makeArray("x", "y", "z");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    Options expected;
+    expected.inputFiles = {"x", "y", "z"};
+    expected.verbosity = 1;
+    EXPECT_EQ(expected, options);
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "y", "z", "-o", nullOutput);
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    Options expected;
+    expected.inputFiles = {"x", "y", "z"};
+    expected.outputFile = nullOutput;
+    expected.verbosity = 1;
+    EXPECT_EQ(expected, options);
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "y", "-o-");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "y", "-o", "file");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("-qqvd12qp4", "-f", "x", "--", "--rm", "-c");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    Options expected = {4,  23,   12,   true,     {"x", "--rm", "-c"},
+                        "", true, true, autoMode, true,
+                        0};
+    EXPECT_EQ(expected, options);
+  }
+}
+
+TEST(Options, NumThreads) {
+  {
+    Options options;
+    auto args = makeArray("x", "-dfo", "-");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "-p", "0", "-fo", "-");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("-f", "-p", "-o", "-");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+}
+
+TEST(Options, BadCompressionLevel) {
+  {
+    Options options;
+    auto args = makeArray("x", "-20");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "--ultra", "-23");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "--1"); // negative 1?
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+}
+
+TEST(Options, InvalidOption) {
+  {
+    Options options;
+    auto args = makeArray("x", "-x");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+}
+
+TEST(Options, BadOutputFile) {
+  {
+    Options options;
+    auto args = makeArray("notzst", "-d", "-p", "1");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ("", options.getOutputFile(options.inputFiles.front()));
+  }
+}
+
+TEST(Options, BadOptionsWithArguments) {
+  {
+    Options options;
+    auto args = makeArray("x", "-pf");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "-p", "10f");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "-p");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "-o");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "-o");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+}
+
+TEST(Options, KeepSource) {
+  {
+    Options options;
+    auto args = makeArray("x", "--rm", "-k");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(true, options.keepSource);
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "--rm", "--keep");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(true, options.keepSource);
+  }
+  {
+    Options options;
+    auto args = makeArray("x");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(true, options.keepSource);
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "--rm");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(false, options.keepSource);
+  }
+}
+
+TEST(Options, Verbosity) {
+  {
+    Options options;
+    auto args = makeArray("x");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(2, options.verbosity);
+  }
+  {
+    Options options;
+    auto args = makeArray("--quiet", "-qq", "x");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(-1, options.verbosity);
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "y");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(1, options.verbosity);
+  }
+  {
+    Options options;
+    auto args = makeArray("--", "x", "y");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(1, options.verbosity);
+  }
+  {
+    Options options;
+    auto args = makeArray("-qv", "x", "y");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(1, options.verbosity);
+  }
+  {
+    Options options;
+    auto args = makeArray("-v", "x", "y");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(3, options.verbosity);
+  }
+  {
+    Options options;
+    auto args = makeArray("-v", "x");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(3, options.verbosity);
+  }
+}
+
+TEST(Options, TestMode) {
+  {
+    Options options;
+    auto args = makeArray("x", "-t");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(true, options.keepSource);
+    EXPECT_EQ(true, options.decompress);
+    EXPECT_EQ(nullOutput, options.outputFile);
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "--test", "--rm", "-ohello");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(true, options.keepSource);
+    EXPECT_EQ(true, options.decompress);
+    EXPECT_EQ(nullOutput, options.outputFile);
+  }
+}
+
+TEST(Options, Checksum) {
+  {
+    Options options;
+    auto args = makeArray("x.zst", "--no-check", "-Cd");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(true, options.checksum);
+  }
+  {
+    Options options;
+    auto args = makeArray("x");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(true, options.checksum);
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "--no-check", "--check");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(true, options.checksum);
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "--no-check");
+    EXPECT_SUCCESS(options.parse(args.size(), args.data()));
+    EXPECT_EQ(false, options.checksum);
+  }
+}
+
+TEST(Options, InputFiles) {
+  {
+    Options options;
+    auto args = makeArray("-cd");
+    options.parse(args.size(), args.data());
+    EXPECT_EQ(1, options.inputFiles.size());
+    EXPECT_EQ("-", options.inputFiles[0]);
+    EXPECT_EQ("-", options.outputFile);
+  }
+  {
+    Options options;
+    auto args = makeArray();
+    options.parse(args.size(), args.data());
+    EXPECT_EQ(1, options.inputFiles.size());
+    EXPECT_EQ("-", options.inputFiles[0]);
+    EXPECT_EQ("-", options.outputFile);
+  }
+  {
+    Options options;
+    auto args = makeArray("-d");
+    options.parse(args.size(), args.data());
+    EXPECT_EQ(1, options.inputFiles.size());
+    EXPECT_EQ("-", options.inputFiles[0]);
+    EXPECT_EQ("-", options.outputFile);
+  }
+  {
+    Options options;
+    auto args = makeArray("x", "-");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+}
+
+TEST(Options, InvalidOptions) {
+  {
+    Options options;
+    auto args = makeArray("-ibasdf");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("- ");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("-n15");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("-0", "x");
+    EXPECT_FAILURE(options.parse(args.size(), args.data()));
+  }
+}
+
+TEST(Options, Extras) {
+  {
+    Options options;
+    auto args = makeArray("-h");
+    EXPECT_MESSAGE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("-H");
+    EXPECT_MESSAGE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("-V");
+    EXPECT_MESSAGE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("--help");
+    EXPECT_MESSAGE(options.parse(args.size(), args.data()));
+  }
+  {
+    Options options;
+    auto args = makeArray("--version");
+    EXPECT_MESSAGE(options.parse(args.size(), args.data()));
+  }
+}
diff --git a/contrib/pzstd/test/PzstdTest.cpp b/contrib/pzstd/test/PzstdTest.cpp
new file mode 100644
index 0000000..c85f73a
--- /dev/null
+++ b/contrib/pzstd/test/PzstdTest.cpp
@@ -0,0 +1,155 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#include "Pzstd.h"
+extern "C" {
+#include "datagen.h"
+}
+#include "test/RoundTrip.h"
+#include "utils/ScopeGuard.h"
+
+#include 
+#include 
+#include 
+#include 
+#include 
+
+using namespace std;
+using namespace pzstd;
+
+TEST(Pzstd, SmallSizes) {
+  unsigned seed = std::random_device{}();
+  std::fprintf(stderr, "Pzstd.SmallSizes seed: %u\n", seed);
+  std::mt19937 gen(seed);
+
+  for (unsigned len = 1; len < 256; ++len) {
+    if (len % 16 == 0) {
+      std::fprintf(stderr, "%u / 16\n", len / 16);
+    }
+    std::string inputFile = std::tmpnam(nullptr);
+    auto guard = makeScopeGuard([&] { std::remove(inputFile.c_str()); });
+    {
+      static uint8_t buf[256];
+      RDG_genBuffer(buf, len, 0.5, 0.0, gen());
+      auto fd = std::fopen(inputFile.c_str(), "wb");
+      auto written = std::fwrite(buf, 1, len, fd);
+      std::fclose(fd);
+      ASSERT_EQ(written, len);
+    }
+    for (unsigned headers = 0; headers <= 1; ++headers) {
+      for (unsigned numThreads = 1; numThreads <= 2; ++numThreads) {
+        for (unsigned level = 1; level <= 4; level *= 4) {
+          auto errorGuard = makeScopeGuard([&] {
+            std::fprintf(stderr, "pzstd headers: %u\n", headers);
+            std::fprintf(stderr, "# threads: %u\n", numThreads);
+            std::fprintf(stderr, "compression level: %u\n", level);
+          });
+          Options options;
+          options.overwrite = true;
+          options.inputFiles = {inputFile};
+          options.numThreads = numThreads;
+          options.compressionLevel = level;
+          options.verbosity = 1;
+          ASSERT_TRUE(roundTrip(options));
+          errorGuard.dismiss();
+        }
+      }
+    }
+  }
+}
+
+TEST(Pzstd, LargeSizes) {
+  unsigned seed = std::random_device{}();
+  std::fprintf(stderr, "Pzstd.LargeSizes seed: %u\n", seed);
+  std::mt19937 gen(seed);
+
+  for (unsigned len = 1 << 20; len <= (1 << 24); len *= 2) {
+    std::string inputFile = std::tmpnam(nullptr);
+    auto guard = makeScopeGuard([&] { std::remove(inputFile.c_str()); });
+    {
+      std::unique_ptr buf(new uint8_t[len]);
+      RDG_genBuffer(buf.get(), len, 0.5, 0.0, gen());
+      auto fd = std::fopen(inputFile.c_str(), "wb");
+      auto written = std::fwrite(buf.get(), 1, len, fd);
+      std::fclose(fd);
+      ASSERT_EQ(written, len);
+    }
+    for (unsigned headers = 0; headers <= 1; ++headers) {
+      for (unsigned numThreads = 1; numThreads <= 16; numThreads *= 4) {
+        for (unsigned level = 1; level <= 4; level *= 2) {
+          auto errorGuard = makeScopeGuard([&] {
+            std::fprintf(stderr, "pzstd headers: %u\n", headers);
+            std::fprintf(stderr, "# threads: %u\n", numThreads);
+            std::fprintf(stderr, "compression level: %u\n", level);
+          });
+          Options options;
+          options.overwrite = true;
+          options.inputFiles = {inputFile};
+          options.numThreads = std::min(numThreads, options.numThreads);
+          options.compressionLevel = level;
+          options.verbosity = 1;
+          ASSERT_TRUE(roundTrip(options));
+          errorGuard.dismiss();
+        }
+      }
+    }
+  }
+}
+
+TEST(Pzstd, ExtremelyLargeSize) {
+  unsigned seed = std::random_device{}();
+  std::fprintf(stderr, "Pzstd.ExtremelyLargeSize seed: %u\n", seed);
+  std::mt19937 gen(seed);
+
+  std::string inputFile = std::tmpnam(nullptr);
+  auto guard = makeScopeGuard([&] { std::remove(inputFile.c_str()); });
+
+  {
+    // Write 4GB + 64 MB
+    constexpr size_t kLength = 1 << 26;
+    std::unique_ptr buf(new uint8_t[kLength]);
+    auto fd = std::fopen(inputFile.c_str(), "wb");
+    auto closeGuard = makeScopeGuard([&] { std::fclose(fd); });
+    for (size_t i = 0; i < (1 << 6) + 1; ++i) {
+      RDG_genBuffer(buf.get(), kLength, 0.5, 0.0, gen());
+      auto written = std::fwrite(buf.get(), 1, kLength, fd);
+      if (written != kLength) {
+        std::fprintf(stderr, "Failed to write file, skipping test\n");
+        return;
+      }
+    }
+  }
+
+  Options options;
+  options.overwrite = true;
+  options.inputFiles = {inputFile};
+  options.compressionLevel = 1;
+  if (options.numThreads == 0) {
+    options.numThreads = 1;
+  }
+  ASSERT_TRUE(roundTrip(options));
+}
+
+TEST(Pzstd, ExtremelyCompressible) {
+  std::string inputFile = std::tmpnam(nullptr);
+  auto guard = makeScopeGuard([&] { std::remove(inputFile.c_str()); });
+  {
+    std::unique_ptr buf(new uint8_t[10000]);
+    std::memset(buf.get(), 'a', 10000);
+    auto fd = std::fopen(inputFile.c_str(), "wb");
+    auto written = std::fwrite(buf.get(), 1, 10000, fd);
+    std::fclose(fd);
+    ASSERT_EQ(written, 10000);
+  }
+  Options options;
+  options.overwrite = true;
+  options.inputFiles = {inputFile};
+  options.numThreads = 1;
+  options.compressionLevel = 1;
+  ASSERT_TRUE(roundTrip(options));
+}
diff --git a/contrib/pzstd/test/RoundTrip.h b/contrib/pzstd/test/RoundTrip.h
new file mode 100644
index 0000000..8b90884
--- /dev/null
+++ b/contrib/pzstd/test/RoundTrip.h
@@ -0,0 +1,86 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#pragma once
+
+#include "Options.h"
+#include "Pzstd.h"
+#include "utils/ScopeGuard.h"
+
+#include 
+#include 
+#include 
+#include 
+
+namespace pzstd {
+
+inline bool check(std::string source, std::string decompressed) {
+  std::unique_ptr sBuf(new std::uint8_t[1024]);
+  std::unique_ptr dBuf(new std::uint8_t[1024]);
+
+  auto sFd = std::fopen(source.c_str(), "rb");
+  auto dFd = std::fopen(decompressed.c_str(), "rb");
+  auto guard = makeScopeGuard([&] {
+    std::fclose(sFd);
+    std::fclose(dFd);
+  });
+
+  size_t sRead, dRead;
+
+  do {
+    sRead = std::fread(sBuf.get(), 1, 1024, sFd);
+    dRead = std::fread(dBuf.get(), 1, 1024, dFd);
+    if (std::ferror(sFd) || std::ferror(dFd)) {
+      return false;
+    }
+    if (sRead != dRead) {
+      return false;
+    }
+
+    for (size_t i = 0; i < sRead; ++i) {
+      if (sBuf.get()[i] != dBuf.get()[i]) {
+        return false;
+      }
+    }
+  } while (sRead == 1024);
+  if (!std::feof(sFd) || !std::feof(dFd)) {
+    return false;
+  }
+  return true;
+}
+
+inline bool roundTrip(Options& options) {
+  if (options.inputFiles.size() != 1) {
+    return false;
+  }
+  std::string source = options.inputFiles.front();
+  std::string compressedFile = std::tmpnam(nullptr);
+  std::string decompressedFile = std::tmpnam(nullptr);
+  auto guard = makeScopeGuard([&] {
+    std::remove(compressedFile.c_str());
+    std::remove(decompressedFile.c_str());
+  });
+
+  {
+    options.outputFile = compressedFile;
+    options.decompress = false;
+    if (pzstdMain(options) != 0) {
+      return false;
+    }
+  }
+  {
+    options.decompress = true;
+    options.inputFiles.front() = compressedFile;
+    options.outputFile = decompressedFile;
+    if (pzstdMain(options) != 0) {
+      return false;
+    }
+  }
+  return check(source, decompressedFile);
+}
+}
diff --git a/contrib/pzstd/test/RoundTripTest.cpp b/contrib/pzstd/test/RoundTripTest.cpp
new file mode 100644
index 0000000..ed2ea77
--- /dev/null
+++ b/contrib/pzstd/test/RoundTripTest.cpp
@@ -0,0 +1,86 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+extern "C" {
+#include "datagen.h"
+}
+#include "Options.h"
+#include "test/RoundTrip.h"
+#include "utils/ScopeGuard.h"
+
+#include 
+#include 
+#include 
+#include 
+#include 
+
+using namespace std;
+using namespace pzstd;
+
+namespace {
+string
+writeData(size_t size, double matchProba, double litProba, unsigned seed) {
+  std::unique_ptr buf(new uint8_t[size]);
+  RDG_genBuffer(buf.get(), size, matchProba, litProba, seed);
+  string file = tmpnam(nullptr);
+  auto fd = std::fopen(file.c_str(), "wb");
+  auto guard = makeScopeGuard([&] { std::fclose(fd); });
+  auto bytesWritten = std::fwrite(buf.get(), 1, size, fd);
+  if (bytesWritten != size) {
+    std::abort();
+  }
+  return file;
+}
+
+template 
+string generateInputFile(Generator& gen) {
+  // Use inputs ranging from 1 Byte to 2^16 Bytes
+  std::uniform_int_distribution size{1, 1 << 16};
+  std::uniform_real_distribution<> prob{0, 1};
+  return writeData(size(gen), prob(gen), prob(gen), gen());
+}
+
+template 
+Options generateOptions(Generator& gen, const string& inputFile) {
+  Options options;
+  options.inputFiles = {inputFile};
+  options.overwrite = true;
+
+  std::uniform_int_distribution numThreads{1, 32};
+  std::uniform_int_distribution compressionLevel{1, 10};
+
+  options.numThreads = numThreads(gen);
+  options.compressionLevel = compressionLevel(gen);
+
+  return options;
+}
+}
+
+int main() {
+  std::mt19937 gen(std::random_device{}());
+
+  auto newlineGuard = makeScopeGuard([] { std::fprintf(stderr, "\n"); });
+  for (unsigned i = 0; i < 10000; ++i) {
+    if (i % 100 == 0) {
+      std::fprintf(stderr, "Progress: %u%%\r", i / 100);
+    }
+    auto inputFile = generateInputFile(gen);
+    auto inputGuard = makeScopeGuard([&] { std::remove(inputFile.c_str()); });
+    for (unsigned i = 0; i < 10; ++i) {
+      auto options = generateOptions(gen, inputFile);
+      if (!roundTrip(options)) {
+        std::fprintf(stderr, "numThreads: %u\n", options.numThreads);
+        std::fprintf(stderr, "level: %u\n", options.compressionLevel);
+        std::fprintf(stderr, "decompress? %u\n", (unsigned)options.decompress);
+        std::fprintf(stderr, "file: %s\n", inputFile.c_str());
+        return 1;
+      }
+    }
+  }
+  return 0;
+}
diff --git a/contrib/pzstd/utils/Buffer.h b/contrib/pzstd/utils/Buffer.h
new file mode 100644
index 0000000..ab25bac
--- /dev/null
+++ b/contrib/pzstd/utils/Buffer.h
@@ -0,0 +1,99 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#pragma once
+
+#include "utils/Range.h"
+
+#include 
+#include 
+#include 
+
+namespace pzstd {
+
+/**
+ * A `Buffer` has a pointer to a shared buffer, and a range of the buffer that
+ * it owns.
+ * The idea is that you can allocate one buffer, and write chunks into it
+ * and break off those chunks.
+ * The underlying buffer is reference counted, and will be destroyed when all
+ * `Buffer`s that reference it are destroyed.
+ */
+class Buffer {
+  std::shared_ptr buffer_;
+  MutableByteRange range_;
+
+  static void delete_buffer(unsigned char* buffer) {
+    delete[] buffer;
+  }
+
+ public:
+  /// Construct an empty buffer that owns no data.
+  explicit Buffer() {}
+
+  /// Construct a `Buffer` that owns a new underlying buffer of size `size`.
+  explicit Buffer(std::size_t size)
+      : buffer_(new unsigned char[size], delete_buffer),
+        range_(buffer_.get(), buffer_.get() + size) {}
+
+  explicit Buffer(std::shared_ptr buffer, MutableByteRange data)
+      : buffer_(buffer), range_(data) {}
+
+  Buffer(Buffer&&) = default;
+  Buffer& operator=(Buffer&&) & = default;
+
+  /**
+   * Splits the data into two pieces: [begin, begin + n), [begin + n, end).
+   * Their data both points into the same underlying buffer.
+   * Modifies the original `Buffer` to point to only [begin + n, end).
+   *
+   * @param n  The offset to split at.
+   * @returns  A buffer that owns the data [begin, begin + n).
+   */
+  Buffer splitAt(std::size_t n) {
+    auto firstPiece = range_.subpiece(0, n);
+    range_.advance(n);
+    return Buffer(buffer_, firstPiece);
+  }
+
+  /// Modifies the buffer to point to the range [begin + n, end).
+  void advance(std::size_t n) {
+    range_.advance(n);
+  }
+
+  /// Modifies the buffer to point to the range [begin, end - n).
+  void subtract(std::size_t n) {
+    range_.subtract(n);
+  }
+
+  /// Returns a read only `Range` pointing to the `Buffer`s data.
+  ByteRange range() const {
+    return range_;
+  }
+  /// Returns a mutable `Range` pointing to the `Buffer`s data.
+  MutableByteRange range() {
+    return range_;
+  }
+
+  const unsigned char* data() const {
+    return range_.data();
+  }
+
+  unsigned char* data() {
+    return range_.data();
+  }
+
+  std::size_t size() const {
+    return range_.size();
+  }
+
+  bool empty() const {
+    return range_.empty();
+  }
+};
+}
diff --git a/contrib/pzstd/utils/FileSystem.h b/contrib/pzstd/utils/FileSystem.h
new file mode 100644
index 0000000..7d59704
--- /dev/null
+++ b/contrib/pzstd/utils/FileSystem.h
@@ -0,0 +1,94 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#pragma once
+
+#include "utils/Range.h"
+
+#include 
+#include 
+#include 
+#include 
+
+// A small subset of `std::filesystem`.
+// `std::filesystem` should be a drop in replacement.
+// See http://en.cppreference.com/w/cpp/filesystem for documentation.
+
+namespace pzstd {
+
+// using file_status = ... causes gcc to emit a false positive warning
+#if defined(_MSC_VER)
+typedef struct ::_stat64 file_status;
+#else
+typedef struct ::stat file_status;
+#endif
+
+/// http://en.cppreference.com/w/cpp/filesystem/status
+inline file_status status(StringPiece path, std::error_code& ec) noexcept {
+  file_status status;
+#if defined(_MSC_VER)
+  const auto error = ::_stat64(path.data(), &status);
+#else
+  const auto error = ::stat(path.data(), &status);
+#endif
+  if (error) {
+    ec.assign(errno, std::generic_category());
+  } else {
+    ec.clear();
+  }
+  return status;
+}
+
+/// http://en.cppreference.com/w/cpp/filesystem/is_regular_file
+inline bool is_regular_file(file_status status) noexcept {
+#if defined(S_ISREG)
+  return S_ISREG(status.st_mode);
+#elif !defined(S_ISREG) && defined(S_IFMT) && defined(S_IFREG)
+  return (status.st_mode & S_IFMT) == S_IFREG;
+#else
+  static_assert(false, "No POSIX stat() support.");
+#endif
+}
+
+/// http://en.cppreference.com/w/cpp/filesystem/is_regular_file
+inline bool is_regular_file(StringPiece path, std::error_code& ec) noexcept {
+  return is_regular_file(status(path, ec));
+}
+
+/// http://en.cppreference.com/w/cpp/filesystem/is_directory
+inline bool is_directory(file_status status) noexcept {
+#if defined(S_ISDIR)
+  return S_ISDIR(status.st_mode);
+#elif !defined(S_ISDIR) && defined(S_IFMT) && defined(S_IFDIR)
+  return (status.st_mode & S_IFMT) == S_IFDIR;
+#else
+  static_assert(false, "NO POSIX stat() support.");
+#endif
+}
+
+/// http://en.cppreference.com/w/cpp/filesystem/is_directory
+inline bool is_directory(StringPiece path, std::error_code& ec) noexcept {
+  return is_directory(status(path, ec));
+}
+
+/// http://en.cppreference.com/w/cpp/filesystem/file_size
+inline std::uintmax_t file_size(
+    StringPiece path,
+    std::error_code& ec) noexcept {
+  auto stat = status(path, ec);
+  if (ec) {
+    return -1;
+  }
+  if (!is_regular_file(stat)) {
+    ec.assign(ENOTSUP, std::generic_category());
+    return -1;
+  }
+  ec.clear();
+  return stat.st_size;
+}
+}
diff --git a/contrib/pzstd/utils/Likely.h b/contrib/pzstd/utils/Likely.h
new file mode 100644
index 0000000..c8ea102
--- /dev/null
+++ b/contrib/pzstd/utils/Likely.h
@@ -0,0 +1,28 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+ 
+/**
+ * Compiler hints to indicate the fast path of an "if" branch: whether
+ * the if condition is likely to be true or false.
+ *
+ * @author Tudor Bosman (tudorb@fb.com)
+ */
+
+#pragma once
+
+#undef LIKELY
+#undef UNLIKELY
+
+#if defined(__GNUC__) && __GNUC__ >= 4
+#define LIKELY(x) (__builtin_expect((x), 1))
+#define UNLIKELY(x) (__builtin_expect((x), 0))
+#else
+#define LIKELY(x) (x)
+#define UNLIKELY(x) (x)
+#endif
diff --git a/contrib/pzstd/utils/Range.h b/contrib/pzstd/utils/Range.h
new file mode 100644
index 0000000..111e98f
--- /dev/null
+++ b/contrib/pzstd/utils/Range.h
@@ -0,0 +1,131 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+ 
+/**
+ * A subset of `folly/Range.h`.
+ * All code copied verbatiam modulo formatting
+ */
+#pragma once
+
+#include "utils/Likely.h"
+
+#include 
+#include 
+#include 
+#include 
+#include 
+
+namespace pzstd {
+
+namespace detail {
+/*
+ *Use IsCharPointer::type to enable const char* or char*.
+ *Use IsCharPointer::const_type to enable only const char*.
+*/
+template 
+struct IsCharPointer {};
+
+template <>
+struct IsCharPointer {
+  typedef int type;
+};
+
+template <>
+struct IsCharPointer {
+  typedef int const_type;
+  typedef int type;
+};
+
+} // namespace detail
+
+template 
+class Range {
+  Iter b_;
+  Iter e_;
+
+ public:
+  using size_type = std::size_t;
+  using iterator = Iter;
+  using const_iterator = Iter;
+  using value_type = typename std::remove_reference<
+      typename std::iterator_traits::reference>::type;
+  using reference = typename std::iterator_traits::reference;
+
+  constexpr Range() : b_(), e_() {}
+  constexpr Range(Iter begin, Iter end) : b_(begin), e_(end) {}
+
+  constexpr Range(Iter begin, size_type size) : b_(begin), e_(begin + size) {}
+
+  template ::type = 0>
+  /* implicit */ Range(Iter str) : b_(str), e_(str + std::strlen(str)) {}
+
+  template ::const_type = 0>
+  /* implicit */ Range(const std::string& str)
+      : b_(str.data()), e_(b_ + str.size()) {}
+
+  // Allow implicit conversion from Range to Range if From is
+  // implicitly convertible to To.
+  template <
+      class OtherIter,
+      typename std::enable_if<
+          (!std::is_same::value &&
+           std::is_convertible::value),
+          int>::type = 0>
+  constexpr /* implicit */ Range(const Range& other)
+      : b_(other.begin()), e_(other.end()) {}
+
+  Range(const Range&) = default;
+  Range(Range&&) = default;
+
+  Range& operator=(const Range&) & = default;
+  Range& operator=(Range&&) & = default;
+
+  constexpr size_type size() const {
+    return e_ - b_;
+  }
+  bool empty() const {
+    return b_ == e_;
+  }
+  Iter data() const {
+    return b_;
+  }
+  Iter begin() const {
+    return b_;
+  }
+  Iter end() const {
+    return e_;
+  }
+
+  void advance(size_type n) {
+    if (UNLIKELY(n > size())) {
+      throw std::out_of_range("index out of range");
+    }
+    b_ += n;
+  }
+
+  void subtract(size_type n) {
+    if (UNLIKELY(n > size())) {
+      throw std::out_of_range("index out of range");
+    }
+    e_ -= n;
+  }
+
+  Range subpiece(size_type first, size_type length = std::string::npos) const {
+    if (UNLIKELY(first > size())) {
+      throw std::out_of_range("index out of range");
+    }
+
+    return Range(b_ + first, std::min(length, size() - first));
+  }
+};
+
+using ByteRange = Range;
+using MutableByteRange = Range;
+using StringPiece = Range;
+}
diff --git a/contrib/pzstd/utils/ResourcePool.h b/contrib/pzstd/utils/ResourcePool.h
new file mode 100644
index 0000000..ed01130
--- /dev/null
+++ b/contrib/pzstd/utils/ResourcePool.h
@@ -0,0 +1,96 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#pragma once
+
+#include 
+#include 
+#include 
+#include 
+#include 
+
+namespace pzstd {
+
+/**
+ * An unbounded pool of resources.
+ * A `ResourcePool` requires a factory function that takes allocates `T*` and
+ * a free function that frees a `T*`.
+ * Calling `ResourcePool::get()` will give you a new `ResourcePool::UniquePtr`
+ * to a `T`, and when it goes out of scope the resource will be returned to the
+ * pool.
+ * The `ResourcePool` *must* survive longer than any resources it hands out.
+ * Remember that `ResourcePool` hands out mutable `T`s, so make sure to clean
+ * up the resource before or after every use.
+ */
+template 
+class ResourcePool {
+ public:
+  class Deleter;
+  using Factory = std::function;
+  using Free = std::function;
+  using UniquePtr = std::unique_ptr;
+
+ private:
+  std::mutex mutex_;
+  Factory factory_;
+  Free free_;
+  std::vector resources_;
+  unsigned inUse_;
+
+ public:
+  /**
+   * Creates a `ResourcePool`.
+   *
+   * @param factory  The function to use to create new resources.
+   * @param free     The function to use to free resources created by `factory`.
+   */
+  ResourcePool(Factory factory, Free free)
+      : factory_(std::move(factory)), free_(std::move(free)), inUse_(0) {}
+
+  /**
+   * @returns  A unique pointer to a resource.  The resource is null iff
+   *           there are no avaiable resources and `factory()` returns null.
+   */
+  UniquePtr get() {
+    std::lock_guard lock(mutex_);
+    if (!resources_.empty()) {
+      UniquePtr resource{resources_.back(), Deleter{*this}};
+      resources_.pop_back();
+      ++inUse_;
+      return resource;
+    }
+    UniquePtr resource{factory_(), Deleter{*this}};
+    ++inUse_;
+    return resource;
+  }
+
+  ~ResourcePool() noexcept {
+    assert(inUse_ == 0);
+    for (const auto resource : resources_) {
+      free_(resource);
+    }
+  }
+
+  class Deleter {
+    ResourcePool *pool_;
+  public:
+    explicit Deleter(ResourcePool &pool) : pool_(&pool) {}
+
+    void operator() (T *resource) {
+      std::lock_guard lock(pool_->mutex_);
+      // Make sure we don't put null resources into the pool
+      if (resource) {
+        pool_->resources_.push_back(resource);
+      }
+      assert(pool_->inUse_ > 0);
+      --pool_->inUse_;
+    }
+  };
+};
+
+}
diff --git a/contrib/pzstd/utils/ScopeGuard.h b/contrib/pzstd/utils/ScopeGuard.h
new file mode 100644
index 0000000..5a333e0
--- /dev/null
+++ b/contrib/pzstd/utils/ScopeGuard.h
@@ -0,0 +1,50 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#pragma once
+
+#include 
+
+namespace pzstd {
+
+/**
+ * Dismissable scope guard.
+ * `Function` must be callable and take no parameters.
+ * Unless `dissmiss()` is called, the callable is executed upon destruction of
+ * `ScopeGuard`.
+ *
+ * Example:
+ *
+ *   auto guard = makeScopeGuard([&] { cleanup(); });
+ */
+template 
+class ScopeGuard {
+  Function function;
+  bool dismissed;
+
+ public:
+  explicit ScopeGuard(Function&& function)
+      : function(std::move(function)), dismissed(false) {}
+
+  void dismiss() {
+    dismissed = true;
+  }
+
+  ~ScopeGuard() noexcept {
+    if (!dismissed) {
+      function();
+    }
+  }
+};
+
+/// Creates a scope guard from `function`.
+template 
+ScopeGuard makeScopeGuard(Function&& function) {
+  return ScopeGuard(std::forward(function));
+}
+}
diff --git a/contrib/pzstd/utils/ThreadPool.h b/contrib/pzstd/utils/ThreadPool.h
new file mode 100644
index 0000000..99b3ecf
--- /dev/null
+++ b/contrib/pzstd/utils/ThreadPool.h
@@ -0,0 +1,58 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#pragma once
+
+#include "utils/WorkQueue.h"
+
+#include 
+#include 
+#include 
+#include 
+
+namespace pzstd {
+/// A simple thread pool that pulls tasks off its queue in FIFO order.
+class ThreadPool {
+  std::vector threads_;
+
+  WorkQueue> tasks_;
+
+ public:
+  /// Constructs a thread pool with `numThreads` threads.
+  explicit ThreadPool(std::size_t numThreads) {
+    threads_.reserve(numThreads);
+    for (std::size_t i = 0; i < numThreads; ++i) {
+      threads_.emplace_back([this] {
+        std::function task;
+        while (tasks_.pop(task)) {
+          task();
+        }
+      });
+    }
+  }
+
+  /// Finishes all tasks currently in the queue.
+  ~ThreadPool() {
+    tasks_.finish();
+    for (auto& thread : threads_) {
+      thread.join();
+    }
+  }
+
+  /**
+   * Adds `task` to the queue of tasks to execute. Since `task` is a
+   * `std::function<>`, it cannot be a move only type. So any lambda passed must
+   * not capture move only types (like `std::unique_ptr`).
+   *
+   * @param task  The task to execute.
+   */
+  void add(std::function task) {
+    tasks_.push(std::move(task));
+  }
+};
+}
diff --git a/contrib/pzstd/utils/WorkQueue.h b/contrib/pzstd/utils/WorkQueue.h
new file mode 100644
index 0000000..780e536
--- /dev/null
+++ b/contrib/pzstd/utils/WorkQueue.h
@@ -0,0 +1,181 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#pragma once
+
+#include "utils/Buffer.h"
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+namespace pzstd {
+
+/// Unbounded thread-safe work queue.
+template 
+class WorkQueue {
+  // Protects all member variable access
+  std::mutex mutex_;
+  std::condition_variable readerCv_;
+  std::condition_variable writerCv_;
+  std::condition_variable finishCv_;
+
+  std::queue queue_;
+  bool done_;
+  std::size_t maxSize_;
+
+  // Must have lock to call this function
+  bool full() const {
+    if (maxSize_ == 0) {
+      return false;
+    }
+    return queue_.size() >= maxSize_;
+  }
+
+ public:
+  /**
+   * Constructs an empty work queue with an optional max size.
+   * If `maxSize == 0` the queue size is unbounded.
+   *
+   * @param maxSize The maximum allowed size of the work queue.
+   */
+  WorkQueue(std::size_t maxSize = 0) : done_(false), maxSize_(maxSize) {}
+
+  /**
+   * Push an item onto the work queue.  Notify a single thread that work is
+   * available.  If `finish()` has been called, do nothing and return false.
+   * If `push()` returns false, then `item` has not been moved from.
+   *
+   * @param item  Item to push onto the queue.
+   * @returns     True upon success, false if `finish()` has been called.  An
+   *               item was pushed iff `push()` returns true.
+   */
+  bool push(T&& item) {
+    {
+      std::unique_lock lock(mutex_);
+      while (full() && !done_) {
+        writerCv_.wait(lock);
+      }
+      if (done_) {
+        return false;
+      }
+      queue_.push(std::move(item));
+    }
+    readerCv_.notify_one();
+    return true;
+  }
+
+  /**
+   * Attempts to pop an item off the work queue.  It will block until data is
+   * available or `finish()` has been called.
+   *
+   * @param[out] item  If `pop` returns `true`, it contains the popped item.
+   *                    If `pop` returns `false`, it is unmodified.
+   * @returns          True upon success.  False if the queue is empty and
+   *                    `finish()` has been called.
+   */
+  bool pop(T& item) {
+    {
+      std::unique_lock lock(mutex_);
+      while (queue_.empty() && !done_) {
+        readerCv_.wait(lock);
+      }
+      if (queue_.empty()) {
+        assert(done_);
+        return false;
+      }
+      item = std::move(queue_.front());
+      queue_.pop();
+    }
+    writerCv_.notify_one();
+    return true;
+  }
+
+  /**
+   * Sets the maximum queue size.  If `maxSize == 0` then it is unbounded.
+   *
+   * @param maxSize The new maximum queue size.
+   */
+  void setMaxSize(std::size_t maxSize) {
+    {
+      std::lock_guard lock(mutex_);
+      maxSize_ = maxSize;
+    }
+    writerCv_.notify_all();
+  }
+
+  /**
+   * Promise that `push()` won't be called again, so once the queue is empty
+   * there will never any more work.
+   */
+  void finish() {
+    {
+      std::lock_guard lock(mutex_);
+      assert(!done_);
+      done_ = true;
+    }
+    readerCv_.notify_all();
+    writerCv_.notify_all();
+    finishCv_.notify_all();
+  }
+
+  /// Blocks until `finish()` has been called (but the queue may not be empty).
+  void waitUntilFinished() {
+    std::unique_lock lock(mutex_);
+    while (!done_) {
+      finishCv_.wait(lock);
+    }
+  }
+};
+
+/// Work queue for `Buffer`s that knows the total number of bytes in the queue.
+class BufferWorkQueue {
+  WorkQueue queue_;
+  std::atomic size_;
+
+ public:
+  BufferWorkQueue(std::size_t maxSize = 0) : queue_(maxSize), size_(0) {}
+
+  void push(Buffer buffer) {
+    size_.fetch_add(buffer.size());
+    queue_.push(std::move(buffer));
+  }
+
+  bool pop(Buffer& buffer) {
+    bool result = queue_.pop(buffer);
+    if (result) {
+      size_.fetch_sub(buffer.size());
+    }
+    return result;
+  }
+
+  void setMaxSize(std::size_t maxSize) {
+    queue_.setMaxSize(maxSize);
+  }
+
+  void finish() {
+    queue_.finish();
+  }
+
+  /**
+   * Blocks until `finish()` has been called.
+   *
+   * @returns The total number of bytes of all the `Buffer`s currently in the
+   *           queue.
+   */
+  std::size_t size() {
+    queue_.waitUntilFinished();
+    return size_.load();
+  }
+};
+}
diff --git a/contrib/pzstd/utils/test/BufferTest.cpp b/contrib/pzstd/utils/test/BufferTest.cpp
new file mode 100644
index 0000000..66ec961
--- /dev/null
+++ b/contrib/pzstd/utils/test/BufferTest.cpp
@@ -0,0 +1,89 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#include "utils/Buffer.h"
+#include "utils/Range.h"
+
+#include 
+#include 
+
+using namespace pzstd;
+
+namespace {
+void deleter(const unsigned char* buf) {
+  delete[] buf;
+}
+}
+
+TEST(Buffer, Constructors) {
+  Buffer empty;
+  EXPECT_TRUE(empty.empty());
+  EXPECT_EQ(0, empty.size());
+
+  Buffer sized(5);
+  EXPECT_FALSE(sized.empty());
+  EXPECT_EQ(5, sized.size());
+
+  Buffer moved(std::move(sized));
+  EXPECT_FALSE(sized.empty());
+  EXPECT_EQ(5, sized.size());
+
+  Buffer assigned;
+  assigned = std::move(moved);
+  EXPECT_FALSE(sized.empty());
+  EXPECT_EQ(5, sized.size());
+}
+
+TEST(Buffer, BufferManagement) {
+  std::shared_ptr buf(new unsigned char[10], deleter);
+  {
+    Buffer acquired(buf, MutableByteRange(buf.get(), buf.get() + 10));
+    EXPECT_EQ(2, buf.use_count());
+    Buffer moved(std::move(acquired));
+    EXPECT_EQ(2, buf.use_count());
+    Buffer assigned;
+    assigned = std::move(moved);
+    EXPECT_EQ(2, buf.use_count());
+
+    Buffer split = assigned.splitAt(5);
+    EXPECT_EQ(3, buf.use_count());
+
+    split.advance(1);
+    assigned.subtract(1);
+    EXPECT_EQ(3, buf.use_count());
+  }
+  EXPECT_EQ(1, buf.use_count());
+}
+
+TEST(Buffer, Modifiers) {
+  Buffer buf(10);
+  {
+    unsigned char i = 0;
+    for (auto& byte : buf.range()) {
+      byte = i++;
+    }
+  }
+
+  auto prefix = buf.splitAt(2);
+
+  ASSERT_EQ(2, prefix.size());
+  EXPECT_EQ(0, *prefix.data());
+
+  ASSERT_EQ(8, buf.size());
+  EXPECT_EQ(2, *buf.data());
+
+  buf.advance(2);
+  EXPECT_EQ(4, *buf.data());
+
+  EXPECT_EQ(9, *(buf.range().end() - 1));
+
+  buf.subtract(2);
+  EXPECT_EQ(7, *(buf.range().end() - 1));
+
+  EXPECT_EQ(4, buf.size());
+}
diff --git a/contrib/pzstd/utils/test/RangeTest.cpp b/contrib/pzstd/utils/test/RangeTest.cpp
new file mode 100644
index 0000000..c761c8a
--- /dev/null
+++ b/contrib/pzstd/utils/test/RangeTest.cpp
@@ -0,0 +1,82 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#include "utils/Range.h"
+
+#include 
+#include 
+
+using namespace pzstd;
+
+// Range is directly copied from folly.
+// Just some sanity tests to make sure everything seems to work.
+
+TEST(Range, Constructors) {
+  StringPiece empty;
+  EXPECT_TRUE(empty.empty());
+  EXPECT_EQ(0, empty.size());
+
+  std::string str = "hello";
+  {
+    Range piece(str.begin(), str.end());
+    EXPECT_EQ(5, piece.size());
+    EXPECT_EQ('h', *piece.data());
+    EXPECT_EQ('o', *(piece.end() - 1));
+  }
+
+  {
+    StringPiece piece(str.data(), str.size());
+    EXPECT_EQ(5, piece.size());
+    EXPECT_EQ('h', *piece.data());
+    EXPECT_EQ('o', *(piece.end() - 1));
+  }
+
+  {
+    StringPiece piece(str);
+    EXPECT_EQ(5, piece.size());
+    EXPECT_EQ('h', *piece.data());
+    EXPECT_EQ('o', *(piece.end() - 1));
+  }
+
+  {
+    StringPiece piece(str.c_str());
+    EXPECT_EQ(5, piece.size());
+    EXPECT_EQ('h', *piece.data());
+    EXPECT_EQ('o', *(piece.end() - 1));
+  }
+}
+
+TEST(Range, Modifiers) {
+  StringPiece range("hello world");
+  ASSERT_EQ(11, range.size());
+
+  {
+    auto hello = range.subpiece(0, 5);
+    EXPECT_EQ(5, hello.size());
+    EXPECT_EQ('h', *hello.data());
+    EXPECT_EQ('o', *(hello.end() - 1));
+  }
+  {
+    auto hello = range;
+    hello.subtract(6);
+    EXPECT_EQ(5, hello.size());
+    EXPECT_EQ('h', *hello.data());
+    EXPECT_EQ('o', *(hello.end() - 1));
+  }
+  {
+    auto world = range;
+    world.advance(6);
+    EXPECT_EQ(5, world.size());
+    EXPECT_EQ('w', *world.data());
+    EXPECT_EQ('d', *(world.end() - 1));
+  }
+
+  std::string expected = "hello world";
+  EXPECT_EQ(expected, std::string(range.begin(), range.end()));
+  EXPECT_EQ(expected, std::string(range.data(), range.size()));
+}
diff --git a/contrib/pzstd/utils/test/ResourcePoolTest.cpp b/contrib/pzstd/utils/test/ResourcePoolTest.cpp
new file mode 100644
index 0000000..a6a86b3
--- /dev/null
+++ b/contrib/pzstd/utils/test/ResourcePoolTest.cpp
@@ -0,0 +1,72 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#include "utils/ResourcePool.h"
+
+#include 
+#include 
+#include 
+
+using namespace pzstd;
+
+TEST(ResourcePool, FullTest) {
+  unsigned numCreated = 0;
+  unsigned numDeleted = 0;
+  {
+    ResourcePool pool(
+      [&numCreated] { ++numCreated; return new int{5}; },
+      [&numDeleted](int *x) { ++numDeleted; delete x; });
+
+    {
+      auto i = pool.get();
+      EXPECT_EQ(5, *i);
+      *i = 6;
+    }
+    {
+      auto i = pool.get();
+      EXPECT_EQ(6, *i);
+      auto j = pool.get();
+      EXPECT_EQ(5, *j);
+      *j = 7;
+    }
+    {
+      auto i = pool.get();
+      EXPECT_EQ(6, *i);
+      auto j = pool.get();
+      EXPECT_EQ(7, *j);
+    }
+  }
+  EXPECT_EQ(2, numCreated);
+  EXPECT_EQ(numCreated, numDeleted);
+}
+
+TEST(ResourcePool, ThreadSafe) {
+  std::atomic numCreated{0};
+  std::atomic numDeleted{0};
+  {
+    ResourcePool pool(
+      [&numCreated] { ++numCreated; return new int{0}; },
+      [&numDeleted](int *x) { ++numDeleted; delete x; });
+    auto push = [&pool] {
+      for (int i = 0; i < 100; ++i) {
+        auto x = pool.get();
+        ++*x;
+      }
+    };
+    std::thread t1{push};
+    std::thread t2{push};
+    t1.join();
+    t2.join();
+
+    auto x = pool.get();
+    auto y = pool.get();
+    EXPECT_EQ(200, *x + *y);
+  }
+  EXPECT_GE(2, numCreated);
+  EXPECT_EQ(numCreated, numDeleted);
+}
diff --git a/contrib/pzstd/utils/test/ScopeGuardTest.cpp b/contrib/pzstd/utils/test/ScopeGuardTest.cpp
new file mode 100644
index 0000000..0c4dc03
--- /dev/null
+++ b/contrib/pzstd/utils/test/ScopeGuardTest.cpp
@@ -0,0 +1,28 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#include "utils/ScopeGuard.h"
+
+#include 
+
+using namespace pzstd;
+
+TEST(ScopeGuard, Dismiss) {
+  {
+    auto guard = makeScopeGuard([&] { EXPECT_TRUE(false); });
+    guard.dismiss();
+  }
+}
+
+TEST(ScopeGuard, Executes) {
+  bool executed = false;
+  {
+    auto guard = makeScopeGuard([&] { executed = true; });
+  }
+  EXPECT_TRUE(executed);
+}
diff --git a/contrib/pzstd/utils/test/ThreadPoolTest.cpp b/contrib/pzstd/utils/test/ThreadPoolTest.cpp
new file mode 100644
index 0000000..1d857aa
--- /dev/null
+++ b/contrib/pzstd/utils/test/ThreadPoolTest.cpp
@@ -0,0 +1,67 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#include "utils/ThreadPool.h"
+
+#include 
+#include 
+#include 
+#include 
+
+using namespace pzstd;
+
+TEST(ThreadPool, Ordering) {
+  std::vector results;
+
+  {
+    ThreadPool executor(1);
+    for (int i = 0; i < 10; ++i) {
+      executor.add([ &results, i ] { results.push_back(i); });
+    }
+  }
+
+  for (int i = 0; i < 10; ++i) {
+    EXPECT_EQ(i, results[i]);
+  }
+}
+
+TEST(ThreadPool, AllJobsFinished) {
+  std::atomic numFinished{0};
+  std::atomic start{false};
+  {
+    ThreadPool executor(5);
+    for (int i = 0; i < 10; ++i) {
+      executor.add([ &numFinished, &start ] {
+        while (!start.load()) {
+          // spin
+        }
+        ++numFinished;
+      });
+    }
+    start.store(true);
+  }
+  EXPECT_EQ(10, numFinished.load());
+}
+
+TEST(ThreadPool, AddJobWhileJoining) {
+  std::atomic done{false};
+  {
+    ThreadPool executor(1);
+    executor.add([&executor, &done] {
+      while (!done.load()) {
+        std::this_thread::yield();
+      }
+      // Sleep for a second to be sure that we are joining
+      std::this_thread::sleep_for(std::chrono::seconds(1));
+      executor.add([] {
+        EXPECT_TRUE(false);
+      });
+    });
+    done.store(true);
+  }
+}
diff --git a/contrib/pzstd/utils/test/WorkQueueTest.cpp b/contrib/pzstd/utils/test/WorkQueueTest.cpp
new file mode 100644
index 0000000..7f58ccb
--- /dev/null
+++ b/contrib/pzstd/utils/test/WorkQueueTest.cpp
@@ -0,0 +1,275 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#include "utils/Buffer.h"
+#include "utils/WorkQueue.h"
+
+#include 
+#include 
+#include 
+#include 
+#include 
+
+using namespace pzstd;
+
+namespace {
+struct Popper {
+  WorkQueue* queue;
+  int* results;
+  std::mutex* mutex;
+
+  void operator()() {
+    int result;
+    while (queue->pop(result)) {
+      std::lock_guard lock(*mutex);
+      results[result] = result;
+    }
+  }
+};
+}
+
+TEST(WorkQueue, SingleThreaded) {
+  WorkQueue queue;
+  int result;
+
+  queue.push(5);
+  EXPECT_TRUE(queue.pop(result));
+  EXPECT_EQ(5, result);
+
+  queue.push(1);
+  queue.push(2);
+  EXPECT_TRUE(queue.pop(result));
+  EXPECT_EQ(1, result);
+  EXPECT_TRUE(queue.pop(result));
+  EXPECT_EQ(2, result);
+
+  queue.push(1);
+  queue.push(2);
+  queue.finish();
+  EXPECT_TRUE(queue.pop(result));
+  EXPECT_EQ(1, result);
+  EXPECT_TRUE(queue.pop(result));
+  EXPECT_EQ(2, result);
+  EXPECT_FALSE(queue.pop(result));
+
+  queue.waitUntilFinished();
+}
+
+TEST(WorkQueue, SPSC) {
+  WorkQueue queue;
+  const int max = 100;
+
+  for (int i = 0; i < 10; ++i) {
+    queue.push(int{i});
+  }
+
+  std::thread thread([ &queue, max ] {
+    int result;
+    for (int i = 0;; ++i) {
+      if (!queue.pop(result)) {
+        EXPECT_EQ(i, max);
+        break;
+      }
+      EXPECT_EQ(i, result);
+    }
+  });
+
+  std::this_thread::yield();
+  for (int i = 10; i < max; ++i) {
+    queue.push(int{i});
+  }
+  queue.finish();
+
+  thread.join();
+}
+
+TEST(WorkQueue, SPMC) {
+  WorkQueue queue;
+  std::vector results(50, -1);
+  std::mutex mutex;
+  std::vector threads;
+  for (int i = 0; i < 5; ++i) {
+    threads.emplace_back(Popper{&queue, results.data(), &mutex});
+  }
+
+  for (int i = 0; i < 50; ++i) {
+    queue.push(int{i});
+  }
+  queue.finish();
+
+  for (auto& thread : threads) {
+    thread.join();
+  }
+
+  for (int i = 0; i < 50; ++i) {
+    EXPECT_EQ(i, results[i]);
+  }
+}
+
+TEST(WorkQueue, MPMC) {
+  WorkQueue queue;
+  std::vector results(100, -1);
+  std::mutex mutex;
+  std::vector popperThreads;
+  for (int i = 0; i < 4; ++i) {
+    popperThreads.emplace_back(Popper{&queue, results.data(), &mutex});
+  }
+
+  std::vector pusherThreads;
+  for (int i = 0; i < 2; ++i) {
+    auto min = i * 50;
+    auto max = (i + 1) * 50;
+    pusherThreads.emplace_back(
+        [ &queue, min, max ] {
+          for (int i = min; i < max; ++i) {
+            queue.push(int{i});
+          }
+        });
+  }
+
+  for (auto& thread : pusherThreads) {
+    thread.join();
+  }
+  queue.finish();
+
+  for (auto& thread : popperThreads) {
+    thread.join();
+  }
+
+  for (int i = 0; i < 100; ++i) {
+    EXPECT_EQ(i, results[i]);
+  }
+}
+
+TEST(WorkQueue, BoundedSizeWorks) {
+  WorkQueue queue(1);
+  int result;
+  queue.push(5);
+  queue.pop(result);
+  queue.push(5);
+  queue.pop(result);
+  queue.push(5);
+  queue.finish();
+  queue.pop(result);
+  EXPECT_EQ(5, result);
+}
+
+TEST(WorkQueue, BoundedSizePushAfterFinish) {
+  WorkQueue queue(1);
+  int result;
+  queue.push(5);
+  std::thread pusher([&queue] {
+    queue.push(6);
+  });
+  // Dirtily try and make sure that pusher has run.
+  std::this_thread::sleep_for(std::chrono::seconds(1));
+  queue.finish();
+  EXPECT_TRUE(queue.pop(result));
+  EXPECT_EQ(5, result);
+  EXPECT_FALSE(queue.pop(result));
+
+  pusher.join();
+}
+
+TEST(WorkQueue, SetMaxSize) {
+  WorkQueue queue(2);
+  int result;
+  queue.push(5);
+  queue.push(6);
+  queue.setMaxSize(1);
+  std::thread pusher([&queue] {
+    queue.push(7);
+  });
+  // Dirtily try and make sure that pusher has run.
+  std::this_thread::sleep_for(std::chrono::seconds(1));
+  queue.finish();
+  EXPECT_TRUE(queue.pop(result));
+  EXPECT_EQ(5, result);
+  EXPECT_TRUE(queue.pop(result));
+  EXPECT_EQ(6, result);
+  EXPECT_FALSE(queue.pop(result));
+
+  pusher.join();
+}
+
+TEST(WorkQueue, BoundedSizeMPMC) {
+  WorkQueue queue(10);
+  std::vector results(200, -1);
+  std::mutex mutex;
+  std::vector popperThreads;
+  for (int i = 0; i < 4; ++i) {
+    popperThreads.emplace_back(Popper{&queue, results.data(), &mutex});
+  }
+
+  std::vector pusherThreads;
+  for (int i = 0; i < 2; ++i) {
+    auto min = i * 100;
+    auto max = (i + 1) * 100;
+    pusherThreads.emplace_back(
+        [ &queue, min, max ] {
+          for (int i = min; i < max; ++i) {
+            queue.push(int{i});
+          }
+        });
+  }
+
+  for (auto& thread : pusherThreads) {
+    thread.join();
+  }
+  queue.finish();
+
+  for (auto& thread : popperThreads) {
+    thread.join();
+  }
+
+  for (int i = 0; i < 200; ++i) {
+    EXPECT_EQ(i, results[i]);
+  }
+}
+
+TEST(WorkQueue, FailedPush) {
+  WorkQueue> queue;
+  std::unique_ptr x(new int{5});
+  EXPECT_TRUE(queue.push(std::move(x)));
+  EXPECT_EQ(nullptr, x);
+  queue.finish();
+  x.reset(new int{6});
+  EXPECT_FALSE(queue.push(std::move(x)));
+  EXPECT_NE(nullptr, x);
+  EXPECT_EQ(6, *x);
+}
+
+TEST(BufferWorkQueue, SizeCalculatedCorrectly) {
+  {
+    BufferWorkQueue queue;
+    queue.finish();
+    EXPECT_EQ(0, queue.size());
+  }
+  {
+    BufferWorkQueue queue;
+    queue.push(Buffer(10));
+    queue.finish();
+    EXPECT_EQ(10, queue.size());
+  }
+  {
+    BufferWorkQueue queue;
+    queue.push(Buffer(10));
+    queue.push(Buffer(5));
+    queue.finish();
+    EXPECT_EQ(15, queue.size());
+  }
+  {
+    BufferWorkQueue queue;
+    queue.push(Buffer(10));
+    queue.push(Buffer(5));
+    queue.finish();
+    Buffer buffer;
+    queue.pop(buffer);
+    EXPECT_EQ(5, queue.size());
+  }
+}
diff --git a/doc/images/Cspeed4.png b/doc/images/Cspeed4.png
new file mode 100644
index 0000000..f0ca0ff
Binary files /dev/null and b/doc/images/Cspeed4.png differ
diff --git a/doc/images/DCspeed5.png b/doc/images/DCspeed5.png
new file mode 100644
index 0000000..900b024
Binary files /dev/null and b/doc/images/DCspeed5.png differ
diff --git a/doc/images/Dspeed4.png b/doc/images/Dspeed4.png
new file mode 100644
index 0000000..eba485d
Binary files /dev/null and b/doc/images/Dspeed4.png differ
diff --git a/doc/images/smallData.png b/doc/images/smallData.png
new file mode 100644
index 0000000..69edf58
Binary files /dev/null and b/doc/images/smallData.png differ
diff --git a/doc/zstd_compression_format.md b/doc/zstd_compression_format.md
new file mode 100644
index 0000000..b48b391
--- /dev/null
+++ b/doc/zstd_compression_format.md
@@ -0,0 +1,1366 @@
+Zstandard Compression Format
+============================
+
+### Notices
+
+Copyright (c) 2016 Yann Collet
+
+Permission is granted to copy and distribute this document
+for any purpose and without charge,
+including translations into other languages
+and incorporation into compilations,
+provided that the copyright notice and this notice are preserved,
+and that any substantive changes or deletions from the original
+are clearly marked.
+Distribution of this document is unlimited.
+
+### Version
+
+0.2.2 (14/09/16)
+
+
+Introduction
+------------
+
+The purpose of this document is to define a lossless compressed data format,
+that is independent of CPU type, operating system,
+file system and character set, suitable for
+file compression, pipe and streaming compression,
+using the [Zstandard algorithm](http://www.zstandard.org).
+
+The data can be produced or consumed,
+even for an arbitrarily long sequentially presented input data stream,
+using only an a priori bounded amount of intermediate storage,
+and hence can be used in data communications.
+The format uses the Zstandard compression method,
+and optional [xxHash-64 checksum method](http://www.xxhash.org),
+for detection of data corruption.
+
+The data format defined by this specification
+does not attempt to allow random access to compressed data.
+
+This specification is intended for use by implementers of software
+to compress data into Zstandard format and/or decompress data from Zstandard format.
+The text of the specification assumes a basic background in programming
+at the level of bits and other primitive data representations.
+
+Unless otherwise indicated below,
+a compliant compressor must produce data sets
+that conform to the specifications presented here.
+It doesn’t need to support all options though.
+
+A compliant decompressor must be able to decompress
+at least one working set of parameters
+that conforms to the specifications presented here.
+It may also ignore informative fields, such as checksum.
+Whenever it does not support a parameter defined in the compressed stream,
+it must produce a non-ambiguous error code and associated error message
+explaining which parameter is unsupported.
+
+
+Overall conventions
+-----------
+In this document:
+- square brackets i.e. `[` and `]` are used to indicate optional fields or parameters.
+- a naming convention for identifiers is `Mixed_Case_With_Underscores`
+
+Definitions
+-----------
+A content compressed by Zstandard is transformed into a Zstandard __frame__.
+Multiple frames can be appended into a single file or stream.
+A frame is totally independent, has a defined beginning and end,
+and a set of parameters which tells the decoder how to decompress it.
+
+A frame encapsulates one or multiple __blocks__.
+Each block can be compressed or not,
+and has a guaranteed maximum content size, which depends on frame parameters.
+Unlike frames, each block depends on previous blocks for proper decoding.
+However, each block can be decompressed without waiting for its successor,
+allowing streaming operations.
+
+
+Frame Concatenation
+-------------------
+
+In some circumstances, it may be required to append multiple frames,
+for example in order to add new data to an existing compressed file
+without re-framing it.
+
+In such case, each frame brings its own set of descriptor flags.
+Each frame is considered independent.
+The only relation between frames is their sequential order.
+
+The ability to decode multiple concatenated frames
+within a single stream or file is left outside of this specification.
+As an example, the reference `zstd` command line utility is able
+to decode all concatenated frames in their sequential order,
+delivering the final decompressed result as if it was a single content.
+
+
+Skippable Frames
+----------------
+
+| `Magic_Number` | `Frame_Size` | `User_Data` |
+|:--------------:|:------------:|:-----------:|
+|   4 bytes      |  4 bytes     |   n bytes   |
+
+Skippable frames allow the insertion of user-defined data
+into a flow of concatenated frames.
+Its design is pretty straightforward,
+with the sole objective to allow the decoder to quickly skip
+over user-defined data and continue decoding.
+
+Skippable frames defined in this specification are compatible with [LZ4] ones.
+
+[LZ4]:http://www.lz4.org
+
+__`Magic_Number`__
+
+4 Bytes, little-endian format.
+Value : 0x184D2A5X, which means any value from 0x184D2A50 to 0x184D2A5F.
+All 16 values are valid to identify a skippable frame.
+
+__`Frame_Size`__
+
+This is the size, in bytes, of the following `User_Data`
+(without including the magic number nor the size field itself).
+This field is represented using 4 Bytes, little-endian format, unsigned 32-bits.
+This means `User_Data` can’t be bigger than (2^32-1) bytes.
+
+__`User_Data`__
+
+The `User_Data` can be anything. Data will just be skipped by the decoder.
+
+
+
+General Structure of Zstandard Frame format
+-------------------------------------------
+The structure of a single Zstandard frame is following:
+
+| `Magic_Number` | `Frame_Header` |`Data_Block`| [More data blocks] | [`Content_Checksum`] |
+|:--------------:|:--------------:|:----------:| ------------------ |:--------------------:|
+| 4 bytes        |  2-14 bytes    | n bytes    |                    |   0-4 bytes          |
+
+__`Magic_Number`__
+
+4 Bytes, little-endian format.
+Value : 0xFD2FB528
+
+__`Frame_Header`__
+
+2 to 14 Bytes, detailed in [next part](#the-structure-of-frame_header).
+
+__`Data_Block`__
+
+Detailed in [next chapter](#the-structure-of-data_block).
+That’s where compressed data is stored.
+
+__`Content_Checksum`__
+
+An optional 32-bit checksum, only present if `Content_Checksum_flag` is set.
+The content checksum is the result
+of [xxh64() hash function](http://www.xxhash.org)
+digesting the original (decoded) data as input, and a seed of zero.
+The low 4 bytes of the checksum are stored in little endian format.
+
+
+The structure of `Frame_Header`
+-------------------------------
+The `Frame_Header` has a variable size, which uses a minimum of 2 bytes,
+and up to 14 bytes depending on optional parameters.
+The structure of `Frame_Header` is following:
+
+| `Frame_Header_Descriptor` | [`Window_Descriptor`] | [`Dictionary_ID`] | [`Frame_Content_Size`] |
+| ------------------------- | --------------------- | ----------------- | ---------------------- |
+| 1 byte                    | 0-1 byte              | 0-4 bytes         | 0-8 bytes              |
+
+### `Frame_Header_Descriptor`
+
+The first header's byte is called the `Frame_Header_Descriptor`.
+It tells which other fields are present.
+Decoding this byte is enough to tell the size of `Frame_Header`.
+
+| Bit number | Field name                |
+| ---------- | ----------                |
+| 7-6        | `Frame_Content_Size_flag` |
+| 5          | `Single_Segment_flag`     |
+| 4          | `Unused_bit`              |
+| 3          | `Reserved_bit`            |
+| 2          | `Content_Checksum_flag`   |
+| 1-0        | `Dictionary_ID_flag`      |
+
+In this table, bit 7 is highest bit, while bit 0 is lowest.
+
+__`Frame_Content_Size_flag`__
+
+This is a 2-bits flag (`= Frame_Header_Descriptor >> 6`),
+specifying if decompressed data size is provided within the header.
+The `Flag_Value` can be converted into `Field_Size`,
+which is the number of bytes used by `Frame_Content_Size`
+according to the following table:
+
+|`Flag_Value`|    0   |  1  |  2  |  3  |
+| ---------- | ------ | --- | --- | --- |
+|`Field_Size`| 0 or 1 |  2  |  4  |  8  |
+
+When `Flag_Value` is `0`, `Field_Size` depends on `Single_Segment_flag` :
+if `Single_Segment_flag` is set, `Field_Size` is 1.
+Otherwise, `Field_Size` is 0 (content size not provided).
+
+__`Single_Segment_flag`__
+
+If this flag is set,
+data must be regenerated within a single continuous memory segment.
+
+In this case, `Frame_Content_Size` is necessarily present,
+but `Window_Descriptor` byte is skipped.
+As a consequence, the decoder must allocate a memory segment
+of size equal or bigger than `Frame_Content_Size`.
+
+In order to preserve the decoder from unreasonable memory requirement,
+a decoder can reject a compressed frame
+which requests a memory size beyond decoder's authorized range.
+
+For broader compatibility, decoders are recommended to support
+memory sizes of at least 8 MB.
+This is just a recommendation,
+each decoder is free to support higher or lower limits,
+depending on local limitations.
+
+__`Unused_bit`__
+
+The value of this bit should be set to zero.
+A decoder compliant with this specification version shall not interpret it.
+It might be used in a future version,
+to signal a property which is not mandatory to properly decode the frame.
+
+__`Reserved_bit`__
+
+This bit is reserved for some future feature.
+Its value _must be zero_.
+A decoder compliant with this specification version must ensure it is not set.
+This bit may be used in a future revision,
+to signal a feature that must be interpreted to decode the frame correctly.
+
+__`Content_Checksum_flag`__
+
+If this flag is set, a 32-bits `Content_Checksum` will be present at frame's end.
+See `Content_Checksum` paragraph.
+
+__`Dictionary_ID_flag`__
+
+This is a 2-bits flag (`= FHD & 3`),
+telling if a dictionary ID is provided within the header.
+It also specifies the size of this field as `Field_Size`.
+
+|`Flag_Value`|  0  |  1  |  2  |  3  |
+| ---------- | --- | --- | --- | --- |
+|`Field_Size`|  0  |  1  |  2  |  4  |
+
+### `Window_Descriptor`
+
+Provides guarantees on maximum back-reference distance
+that will be used within compressed data.
+This information is important for decoders to allocate enough memory.
+
+The `Window_Descriptor` byte is optional. It is absent when `Single_Segment_flag` is set.
+In this case, the maximum back-reference distance is the content size itself,
+which can be any value from 1 to 2^64-1 bytes (16 EB).
+
+| Bit numbers |     7-3    |     0-2    |
+| ----------- | ---------- | ---------- |
+| Field name  | `Exponent` | `Mantissa` |
+
+Maximum distance is given by the following formulas :
+```
+windowLog = 10 + Exponent;
+windowBase = 1 << windowLog;
+windowAdd = (windowBase / 8) * Mantissa;
+Window_Size = windowBase + windowAdd;
+```
+The minimum window size is 1 KB.
+The maximum size is `15*(1<<38)` bytes, which is 1.875 TB.
+
+To properly decode compressed data,
+a decoder will need to allocate a buffer of at least `Window_Size` bytes.
+
+In order to preserve decoder from unreasonable memory requirements,
+a decoder can refuse a compressed frame
+which requests a memory size beyond decoder's authorized range.
+
+For improved interoperability,
+decoders are recommended to be compatible with window sizes of 8 MB,
+and encoders are recommended to not request more than 8 MB.
+It's merely a recommendation though,
+decoders are free to support larger or lower limits,
+depending on local limitations.
+
+### `Dictionary_ID`
+
+This is a variable size field, which contains
+the ID of the dictionary required to properly decode the frame.
+Note that this field is optional. When it's not present,
+it's up to the caller to make sure it uses the correct dictionary.
+Format is little-endian.
+
+Field size depends on `Dictionary_ID_flag`.
+1 byte can represent an ID 0-255.
+2 bytes can represent an ID 0-65535.
+4 bytes can represent an ID 0-4294967295.
+
+It's allowed to represent a small ID (for example `13`)
+with a large 4-bytes dictionary ID, losing some compacity in the process.
+
+_Reserved ranges :_
+If the frame is going to be distributed in a private environment,
+any dictionary ID can be used.
+However, for public distribution of compressed frames using a dictionary,
+the following ranges are reserved for future use and should not be used :
+- low range : 1 - 32767
+- high range : >= (2^31)
+
+
+### `Frame_Content_Size`
+
+This is the original (uncompressed) size. This information is optional.
+The `Field_Size` is provided according to value of `Frame_Content_Size_flag`.
+The `Field_Size` can be equal to 0 (not present), 1, 2, 4 or 8 bytes.
+Format is little-endian.
+
+| `Field_Size` |    Range   |
+| ------------ | ---------- |
+|      1       |   0 - 255  |
+|      2       | 256 - 65791|
+|      4       | 0 - 2^32-1 |
+|      8       | 0 - 2^64-1 |
+
+When `Field_Size` is 1, 4 or 8 bytes, the value is read directly.
+When `Field_Size` is 2, _the offset of 256 is added_.
+It's allowed to represent a small size (for example `18`) using any compatible variant.
+
+
+The structure of `Data_Block`
+-----------------------------
+The structure of `Data_Block` is following:
+
+| `Last_Block` | `Block_Type` | `Block_Size` | `Block_Content` |
+|:------------:|:------------:|:------------:|:---------------:|
+|   1 bit      |  2 bits      |  21 bits     |  n bytes        |
+
+The block header (`Last_Block`, `Block_Type`, and `Block_Size`) uses 3-bytes.
+
+__`Last_Block`__
+
+The lowest bit signals if this block is the last one.
+Frame ends right after this block.
+It may be followed by an optional `Content_Checksum` .
+
+__`Block_Type` and `Block_Size`__
+
+The next 2 bits represent the `Block_Type`,
+while the remaining 21 bits represent the `Block_Size`.
+Format is __little-endian__.
+
+There are 4 block types :
+
+|    Value     |      0      |     1       |  2                 |    3      |
+| ------------ | ----------- | ----------- | ------------------ | --------- |
+| `Block_Type` | `Raw_Block` | `RLE_Block` | `Compressed_Block` | `Reserved`|
+
+- `Raw_Block` - this is an uncompressed block.
+  `Block_Size` is the number of bytes to read and copy.
+- `RLE_Block` - this is a single byte, repeated N times.
+  In which case, `Block_Size` is the size to regenerate,
+  while the "compressed" block is just 1 byte (the byte to repeat).
+- `Compressed_Block` - this is a [Zstandard compressed block](#the-format-of-compressed_block),
+  detailed in another section of this specification.
+  `Block_Size` is the compressed size.
+  Decompressed size is unknown,
+  but its maximum possible value is guaranteed (see below)
+- `Reserved` - this is not a block.
+  This value cannot be used with current version of this specification.
+
+Block sizes must respect a few rules :
+- In compressed mode, compressed size if always strictly `< decompressed size`.
+- Block decompressed size is always <= maximum back-reference distance .
+- Block decompressed size is always <= 128 KB
+
+
+__`Block_Content`__
+
+The `Block_Content` is where the actual data to decode stands.
+It might be compressed or not, depending on previous field indications.
+A data block is not necessarily "full" :
+since an arbitrary “flush” may happen anytime,
+block decompressed content can be any size,
+up to `Block_Maximum_Decompressed_Size`, which is the smallest of :
+- Maximum back-reference distance
+- 128 KB
+
+
+
+The format of `Compressed_Block`
+--------------------------------
+The size of `Compressed_Block` must be provided using `Block_Size` field from `Data_Block`.
+The `Compressed_Block` has a guaranteed maximum regenerated size,
+in order to properly allocate destination buffer.
+See [`Data_Block`](#the-structure-of-data_block) for more details.
+
+A compressed block consists of 2 sections :
+- [`Literals_Section`](#literals_section)
+- [`Sequences_Section`](#sequences_section)
+
+### Prerequisites
+To decode a compressed block, the following elements are necessary :
+- Previous decoded blocks, up to a distance of `Window_Size`,
+  or all previous blocks when `Single_Segment_flag` is set.
+- List of "recent offsets" from previous compressed block.
+- Decoding tables of previous compressed block for each symbol type
+  (literals, literals lengths, match lengths, offsets).
+
+
+### `Literals_Section`
+
+During sequence phase, literals will be entangled with match copy operations.
+All literals are regrouped in the first part of the block.
+They can be decoded first, and then copied during sequence operations,
+or they can be decoded on the flow, as needed by sequence commands.
+
+| `Literals_Section_Header` | [`Huffman_Tree_Description`] | Stream1 | [Stream2] | [Stream3] | [Stream4] |
+| ------------------------- | ---------------------------- | ------- | --------- | --------- | --------- |
+
+Literals can be stored uncompressed or compressed using Huffman prefix codes.
+When compressed, an optional tree description can be present,
+followed by 1 or 4 streams.
+
+
+#### `Literals_Section_Header`
+
+Header is in charge of describing how literals are packed.
+It's a byte-aligned variable-size bitfield, ranging from 1 to 5 bytes,
+using little-endian convention.
+
+| `Literals_Block_Type` | `Size_Format` | `Regenerated_Size` | [`Compressed_Size`] |
+| --------------------- | ------------- | ------------------ | ----------------- |
+|   2 bits              |  1 - 2 bits   |    5 - 20 bits     |    0 - 18 bits    |
+
+In this representation, bits on the left are smallest bits.
+
+__`Literals_Block_Type`__
+
+This field uses 2 lowest bits of first byte, describing 4 different block types :
+
+| `Literals_Block_Type`         | Value |
+| ----------------------------- | ----- |
+| `Raw_Literals_Block`          |   0   |
+| `RLE_Literals_Block`          |   1   |
+| `Compressed_Literals_Block`   |   2   |
+| `Repeat_Stats_Literals_Block` |   3   |
+
+- `Raw_Literals_Block` - Literals are stored uncompressed.
+- `RLE_Literals_Block` - Literals consist of a single byte value repeated N times.
+- `Compressed_Literals_Block` - This is a standard Huffman-compressed block,
+        starting with a Huffman tree description.
+        See details below.
+- `Repeat_Stats_Literals_Block` - This is a Huffman-compressed block,
+        using Huffman tree _from previous Huffman-compressed literals block_.
+        Huffman tree description will be skipped.
+
+__`Size_Format`__
+
+`Size_Format` is divided into 2 families :
+
+- For `Compressed_Block`, it requires to decode both `Compressed_Size`
+  and `Regenerated_Size` (the decompressed size). It will also decode the number of streams.
+- For `Raw_Literals_Block` and `RLE_Literals_Block` it's enough to decode `Regenerated_Size`.
+
+For values spanning several bytes, convention is little-endian.
+
+__`Size_Format` for `Raw_Literals_Block` and `RLE_Literals_Block`__ :
+
+- Value x0 : `Regenerated_Size` uses 5 bits (0-31).
+               `Literals_Section_Header` has 1 byte.
+               `Regenerated_Size = Header[0]>>3`
+- Value 01 : `Regenerated_Size` uses 12 bits (0-4095).
+               `Literals_Section_Header` has 2 bytes.
+               `Regenerated_Size = (Header[0]>>4) + (Header[1]<<4)`
+- Value 11 : `Regenerated_Size` uses 20 bits (0-1048575).
+               `Literals_Section_Header` has 3 bytes.
+               `Regenerated_Size = (Header[0]>>4) + (Header[1]<<4) + (Header[2]<<12)`
+
+Note : it's allowed to represent a short value (for example `13`)
+using a long format, accepting the increased compressed data size.
+
+__`Size_Format` for `Compressed_Literals_Block` and `Repeat_Stats_Literals_Block`__ :
+
+- Value 00 : _A single stream_.
+               Both `Compressed_Size` and `Regenerated_Size` use 10 bits (0-1023).
+               `Literals_Section_Header` has 3 bytes.
+- Value 01 : 4 streams.
+               Both `Compressed_Size` and `Regenerated_Size` use 10 bits (0-1023).
+               `Literals_Section_Header` has 3 bytes.
+- Value 10 : 4 streams.
+               Both `Compressed_Size` and `Regenerated_Size` use 14 bits (0-16383).
+               `Literals_Section_Header` has 4 bytes.
+- Value 11 : 4 streams.
+               Both `Compressed_Size` and `Regenerated_Size` use 18 bits (0-262143).
+               `Literals_Section_Header` has 5 bytes.
+
+Both `Compressed_Size` and `Regenerated_Size` fields follow little-endian convention.
+
+
+#### `Huffman_Tree_Description`
+
+This section is only present when `Literals_Block_Type` type is `Compressed_Literals_Block` (`2`).
+
+Prefix coding represents symbols from an a priori known alphabet
+by bit sequences (codewords), one codeword for each symbol,
+in a manner such that different symbols may be represented
+by bit sequences of different lengths,
+but a parser can always parse an encoded string
+unambiguously symbol-by-symbol.
+
+Given an alphabet with known symbol frequencies,
+the Huffman algorithm allows the construction of an optimal prefix code
+using the fewest bits of any possible prefix codes for that alphabet.
+
+Prefix code must not exceed a maximum code length.
+More bits improve accuracy but cost more header size,
+and require more memory or more complex decoding operations.
+This specification limits maximum code length to 11 bits.
+
+
+##### Representation
+
+All literal values from zero (included) to last present one (excluded)
+are represented by `Weight` with values from `0` to `Max_Number_of_Bits`.
+Transformation from `Weight` to `Number_of_Bits` follows this formula :
+```
+Number_of_Bits = Weight ? (Max_Number_of_Bits + 1 - Weight) : 0
+```
+The last symbol's `Weight` is deduced from previously decoded ones,
+by completing to the nearest power of 2.
+This power of 2 gives `Max_Number_of_Bits`, the depth of the current tree.
+
+__Example__ :
+Let's presume the following Huffman tree must be described :
+
+|     literal      |  0  |  1  |  2  |  3  |  4  |  5  |
+| ---------------- | --- | --- | --- | --- | --- | --- |
+| `Number_of_Bits` |  1  |  2  |  3  |  0  |  4  |  4  |
+
+The tree depth is 4, since its smallest element uses 4 bits.
+Value `5` will not be listed, nor will values above `5`.
+Values from `0` to `4` will be listed using `Weight` instead of `Number_of_Bits`.
+Weight formula is :
+```
+Weight = Number_of_Bits ? (Max_Number_of_Bits + 1 - Number_of_Bits) : 0
+```
+It gives the following serie of weights :
+
+| `Weight` |  4  |  3  |  2  |  0  |  1  |
+| -------- | --- | --- | --- | --- | --- |
+| literal  |  0  |  1  |  2  |  3  |  4  |
+
+The decoder will do the inverse operation :
+having collected weights of literals from `0` to `4`,
+it knows the last literal, `5`, is present with a non-zero weight.
+The weight of `5` can be deducted by joining to the nearest power of 2.
+Sum of `2^(Weight-1)` (excluding 0) is :
+`8 + 4 + 2 + 0 + 1 = 15`.
+Nearest power of 2 is 16.
+Therefore, `Max_Number_of_Bits = 4` and `Weight[5] = 1`.
+
+##### Huffman Tree header
+
+This is a single byte value (0-255),
+which tells how to decode the list of weights.
+
+- if `headerByte` >= 128 : this is a direct representation,
+  where each `Weight` is written directly as a 4 bits field (0-15).
+  The full representation occupies `((Number_of_Symbols+1)/2)` bytes,
+  meaning it uses a last full byte even if `Number_of_Symbols` is odd.
+  `Number_of_Symbols = headerByte - 127`.
+  Note that maximum `Number_of_Symbols` is 255-127 = 128.
+  A larger serie must necessarily use FSE compression.
+
+- if `headerByte` < 128 :
+  the serie of weights is compressed by FSE.
+  The length of the FSE-compressed serie is equal to `headerByte` (0-127).
+
+##### Finite State Entropy (FSE) compression of Huffman weights
+
+The serie of weights is compressed using FSE compression.
+It's a single bitstream with 2 interleaved states,
+sharing a single distribution table.
+
+To decode an FSE bitstream, it is necessary to know its compressed size.
+Compressed size is provided by `headerByte`.
+It's also necessary to know its _maximum possible_ decompressed size,
+which is `255`, since literal values span from `0` to `255`,
+and last symbol value is not represented.
+
+An FSE bitstream starts by a header, describing probabilities distribution.
+It will create a Decoding Table.
+Table must be pre-allocated, which requires to support a maximum accuracy.
+For a list of Huffman weights, maximum accuracy is 7 bits.
+
+FSE header is [described in relevant chapter](#fse-distribution-table--condensed-format),
+and so is [FSE bitstream](#bitstream).
+The main difference is that Huffman header compression uses 2 states,
+which share the same FSE distribution table.
+Bitstream contains only FSE symbols (no interleaved "raw bitfields").
+The number of symbols to decode is discovered
+by tracking bitStream overflow condition.
+When both states have overflowed the bitstream, end is reached.
+
+
+##### Conversion from weights to Huffman prefix codes
+
+All present symbols shall now have a `Weight` value.
+It is possible to transform weights into Number_of_Bits, using this formula:
+```
+Number_of_Bits = Number_of_Bits ? Max_Number_of_Bits + 1 - Weight : 0
+```
+Symbols are sorted by `Weight`. Within same `Weight`, symbols keep natural order.
+Symbols with a `Weight` of zero are removed.
+Then, starting from lowest weight, prefix codes are distributed in order.
+
+__Example__ :
+Let's presume the following list of weights has been decoded :
+
+| Literal  |  0  |  1  |  2  |  3  |  4  |  5  |
+| -------- | --- | --- | --- | --- | --- | --- |
+| `Weight` |  4  |  3  |  2  |  0  |  1  |  1  |
+
+Sorted by weight and then natural order,
+it gives the following distribution :
+
+| Literal          |  3  |  4  |  5  |  2  |  1  |   0  |
+| ---------------- | --- | --- | --- | --- | --- | ---- |
+| `Weight`         |  0  |  1  |  1  |  2  |  3  |   4  |
+| `Number_of_Bits` |  0  |  4  |  4  |  3  |  2  |   1  |
+| prefix codes     | N/A | 0000| 0001| 001 | 01  |   1  |
+
+
+#### The content of Huffman-compressed literal stream
+
+##### Bitstreams sizes
+
+As seen in a previous paragraph,
+there are 2 types of Huffman-compressed literals :
+a single stream and 4 streams.
+
+Encoding using 4 streams is useful for CPU with multiple execution units and out-of-order operations.
+Since each stream can be decoded independently,
+it's possible to decode them up to 4x faster than a single stream,
+presuming the CPU has enough parallelism available.
+
+For single stream, header provides both the compressed and regenerated size.
+For 4 streams though,
+header only provides compressed and regenerated size of all 4 streams combined.
+In order to properly decode the 4 streams,
+it's necessary to know the compressed and regenerated size of each stream.
+
+Regenerated size of each stream can be calculated by `(totalSize+3)/4`,
+except for last one, which can be up to 3 bytes smaller, to reach `totalSize`.
+
+Compressed size is provided explicitly : in the 4-streams variant,
+bitstreams are preceded by 3 unsigned little-endian 16-bits values.
+Each value represents the compressed size of one stream, in order.
+The last stream size is deducted from total compressed size
+and from previously decoded stream sizes :
+
+`stream4CSize = totalCSize - 6 - stream1CSize - stream2CSize - stream3CSize`.
+
+
+##### Bitstreams read and decode
+
+Each bitstream must be read _backward_,
+that is starting from the end down to the beginning.
+Therefore it's necessary to know the size of each bitstream.
+
+It's also necessary to know exactly which _bit_ is the latest.
+This is detected by a final bit flag :
+the highest bit of latest byte is a final-bit-flag.
+Consequently, a last byte of `0` is not possible.
+And the final-bit-flag itself is not part of the useful bitstream.
+Hence, the last byte contains between 0 and 7 useful bits.
+
+Starting from the end,
+it's possible to read the bitstream in a little-endian fashion,
+keeping track of already used bits.
+
+Reading the last `Max_Number_of_Bits` bits,
+it's then possible to compare extracted value to decoding table,
+determining the symbol to decode and number of bits to discard.
+
+The process continues up to reading the required number of symbols per stream.
+If a bitstream is not entirely and exactly consumed,
+hence reaching exactly its beginning position with _all_ bits consumed,
+the decoding process is considered faulty.
+
+
+### `Sequences_Section`
+
+A compressed block is a succession of _sequences_ .
+A sequence is a literal copy command, followed by a match copy command.
+A literal copy command specifies a length.
+It is the number of bytes to be copied (or extracted) from the literal section.
+A match copy command specifies an offset and a length.
+The offset gives the position to copy from,
+which can be within a previous block.
+
+When all _sequences_ are decoded,
+if there is any literal left in the _literal section_,
+these bytes are added at the end of the block.
+
+The `Sequences_Section` regroup all symbols required to decode commands.
+There are 3 symbol types : literals lengths, offsets and match lengths.
+They are encoded together, interleaved, in a single _bitstream_.
+
+The `Sequences_Section` starts by a header,
+followed by optional probability tables for each symbol type,
+followed by the bitstream.
+
+| `Sequences_Section_Header` | [`Literals_Length_Table`] | [`Offset_Table`] | [`Match_Length_Table`] | bitStream |
+| -------------------------- | ------------------------- | ---------------- | ---------------------- | --------- |
+
+To decode the `Sequences_Section`, it's required to know its size.
+This size is deducted from `blockSize - literalSectionSize`.
+
+
+#### `Sequences_Section_Header`
+
+Consists of 2 items:
+- `Number_of_Sequences`
+- Symbol compression modes
+
+__`Number_of_Sequences`__
+
+This is a variable size field using between 1 and 3 bytes.
+Let's call its first byte `byte0`.
+- `if (byte0 == 0)` : there are no sequences.
+            The sequence section stops there.
+            Regenerated content is defined entirely by literals section.
+- `if (byte0 < 128)` : `Number_of_Sequences = byte0` . Uses 1 byte.
+- `if (byte0 < 255)` : `Number_of_Sequences = ((byte0-128) << 8) + byte1` . Uses 2 bytes.
+- `if (byte0 == 255)`: `Number_of_Sequences = byte1 + (byte2<<8) + 0x7F00` . Uses 3 bytes.
+
+__Symbol compression modes__
+
+This is a single byte, defining the compression mode of each symbol type.
+
+|Bit number|   7-6                   |   5-4          |   3-2                |     1-0    |
+| -------- | ----------------------- | -------------- | -------------------- | ---------- |
+|Field name| `Literals_Lengths_Mode` | `Offsets_Mode` | `Match_Lengths_Mode` | `Reserved` |
+
+The last field, `Reserved`, must be all-zeroes.
+
+`Literals_Lengths_Mode`, `Offsets_Mode` and `Match_Lengths_Mode` define the `Compression_Mode` of
+literals lengths, offsets, and match lengths respectively.
+
+They follow the same enumeration :
+
+|        Value       |         0         |      1     |           2           |       3       |
+| ------------------ | ----------------- | ---------- | --------------------- | ------------- |
+| `Compression_Mode` | `Predefined_Mode` | `RLE_Mode` | `FSE_Compressed_Mode` | `Repeat_Mode` |
+
+- `Predefined_Mode` : uses a predefined distribution table.
+- `RLE_Mode` : it's a single code, repeated `Number_of_Sequences` times.
+- `Repeat_Mode` : re-use distribution table from previous compressed block.
+- `FSE_Compressed_Mode` : standard FSE compression.
+          A distribution table will be present.
+          It will be described in [next part](#distribution-tables).
+
+#### The codes for literals lengths, match lengths, and offsets.
+
+Each symbol is a _code_ in its own context,
+which specifies `Baseline` and `Number_of_Bits` to add.
+_Codes_ are FSE compressed,
+and interleaved with raw additional bits in the same bitstream.
+
+##### Literals length codes
+
+Literals length codes are values ranging from `0` to `35` included.
+They define lengths from 0 to 131071 bytes.
+
+| `Literals_Length_Code` |         0-15           |
+| ---------------------- | ---------------------- |
+| length                 | `Literals_Length_Code` |
+| `Number_of_Bits`       |          0             |
+
+| `Literals_Length_Code` |  16  |  17  |  18  |  19  |  20  |  21  |  22  |  23  |
+| ---------------------- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- |
+| `Baseline`             |  16  |  18  |  20  |  22  |  24  |  28  |  32  |  40  |
+| `Number_of_Bits`       |   1  |   1  |   1  |   1  |   2  |   2  |   3  |   3  |
+
+| `Literals_Length_Code` |  24  |  25  |  26  |  27  |  28  |  29  |  30  |  31  |
+| ---------------------- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- |
+| `Baseline`             |  48  |  64  |  128 |  256 |  512 | 1024 | 2048 | 4096 |
+| `Number_of_Bits`       |   4  |   6  |   7  |   8  |   9  |  10  |  11  |  12  |
+
+| `Literals_Length_Code` |  32  |  33  |  34  |  35  |
+| ---------------------- | ---- | ---- | ---- | ---- |
+| `Baseline`             | 8192 |16384 |32768 |65536 |
+| `Number_of_Bits`       |  13  |  14  |  15  |  16  |
+
+##### Default distribution for literals length codes
+
+When `Compression_Mode` is `Predefined_Mode`,
+a predefined distribution is used for FSE compression.
+
+Below is its definition. It uses an accuracy of 6 bits (64 states).
+```
+short literalsLength_defaultDistribution[36] =
+        { 4, 3, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 1, 1,
+          2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 2, 1, 1, 1, 1, 1,
+         -1,-1,-1,-1 };
+```
+
+##### Match length codes
+
+Match length codes are values ranging from `0` to `52` included.
+They define lengths from 3 to 131074 bytes.
+
+| `Match_Length_Code` |         0-31            |
+| ------------------- | ----------------------- |
+| value               | `Match_Length_Code` + 3 |
+| `Number_of_Bits`    |          0              |
+
+| `Match_Length_Code` |  32  |  33  |  34  |  35  |  36  |  37  |  38  |  39  |
+| ------------------- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- |
+| `Baseline`          |  35  |  37  |  39  |  41  |  43  |  47  |  51  |  59  |
+| `Number_of_Bits`    |   1  |   1  |   1  |   1  |   2  |   2  |   3  |   3  |
+
+| `Match_Length_Code` |  40  |  41  |  42  |  43  |  44  |  45  |  46  |  47  |
+| ------------------- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- |
+| `Baseline`          |  67  |  83  |  99  |  131 |  258 |  514 | 1026 | 2050 |
+| `Number_of_Bits`    |   4  |   4  |   5  |   7  |   8  |   9  |  10  |  11  |
+
+| `Match_Length_Code` |  48  |  49  |  50  |  51  |  52  |
+| ------------------- | ---- | ---- | ---- | ---- | ---- |
+| `Baseline`          | 4098 | 8194 |16486 |32770 |65538 |
+| `Number_of_Bits`    |  12  |  13  |  14  |  15  |  16  |
+
+##### Default distribution for match length codes
+
+When `Compression_Mode` is defined as `Predefined_Mode`,
+a predefined distribution is used for FSE compression.
+
+Below is its definition. It uses an accuracy of 6 bits (64 states).
+```
+short matchLengths_defaultDistribution[53] =
+        { 1, 4, 3, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1,
+          1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+          1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,-1,-1,
+         -1,-1,-1,-1,-1 };
+```
+
+##### Offset codes
+
+Offset codes are values ranging from `0` to `N`.
+
+A decoder is free to limit its maximum `N` supported.
+Recommendation is to support at least up to `22`.
+For information, at the time of this writing.
+the reference decoder supports a maximum `N` value of `28` in 64-bits mode.
+
+An offset code is also the number of additional bits to read,
+and can be translated into an `Offset_Value` using the following formulas :
+
+```
+Offset_Value = (1 << offsetCode) + readNBits(offsetCode);
+if (Offset_Value > 3) offset = Offset_Value - 3;
+```
+It means that maximum `Offset_Value` is `(2^(N+1))-1` and it supports back-reference distance up to `(2^(N+1))-4`
+but is limited by [maximum back-reference distance](#window_descriptor).
+
+`Offset_Value` from 1 to 3 are special : they define "repeat codes",
+which means one of the previous offsets will be repeated.
+They are sorted in recency order, with 1 meaning the most recent one.
+See [Repeat offsets](#repeat-offsets) paragraph.
+
+
+##### Default distribution for offset codes
+
+When `Compression_Mode` is defined as `Predefined_Mode`,
+a predefined distribution is used for FSE compression.
+
+Below is its definition. It uses an accuracy of 5 bits (32 states),
+and supports a maximum `N` of 28, allowing offset values up to 536,870,908 .
+
+If any sequence in the compressed block requires an offset larger than this,
+it's not possible to use the default distribution to represent it.
+
+```
+short offsetCodes_defaultDistribution[29] =
+        { 1, 1, 1, 1, 1, 1, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1,
+          1, 1, 1, 1, 1, 1, 1, 1,-1,-1,-1,-1,-1 };
+```
+
+#### Distribution tables
+
+Following the header, up to 3 distribution tables can be described.
+When present, they are in this order :
+- Literals lengths
+- Offsets
+- Match Lengths
+
+The content to decode depends on their respective encoding mode :
+- `Predefined_Mode` : no content. Use predefined distribution table.
+- `RLE_Mode` : 1 byte. This is the only code to use across the whole compressed block.
+- `FSE_Compressed_Mode` : A distribution table is present.
+- `Repeat_Mode` : no content. Re-use distribution from previous compressed block.
+
+##### FSE distribution table : condensed format
+
+An FSE distribution table describes the probabilities of all symbols
+from `0` to the last present one (included)
+on a normalized scale of `1 << Accuracy_Log` .
+
+It's a bitstream which is read forward, in little-endian fashion.
+It's not necessary to know its exact size,
+since it will be discovered and reported by the decoding process.
+
+The bitstream starts by reporting on which scale it operates.
+`Accuracy_Log = low4bits + 5`.
+Note that maximum `Accuracy_Log` for literal and match lengths is `9`,
+and for offsets is `8`. Higher values are considered errors.
+
+Then follows each symbol value, from `0` to last present one.
+The number of bits used by each field is variable.
+It depends on :
+
+- Remaining probabilities + 1 :
+  __example__ :
+  Presuming an `Accuracy_Log` of 8,
+  and presuming 100 probabilities points have already been distributed,
+  the decoder may read any value from `0` to `255 - 100 + 1 == 156` (included).
+  Therefore, it must read `log2sup(156) == 8` bits.
+
+- Value decoded : small values use 1 less bit :
+  __example__ :
+  Presuming values from 0 to 156 (included) are possible,
+  255-156 = 99 values are remaining in an 8-bits field.
+  They are used this way :
+  first 99 values (hence from 0 to 98) use only 7 bits,
+  values from 99 to 156 use 8 bits.
+  This is achieved through this scheme :
+
+  | Value read | Value decoded | Number of bits used |
+  | ---------- | ------------- | ------------------- |
+  |   0 -  98  |   0 -  98     |  7                  |
+  |  99 - 127  |  99 - 127     |  8                  |
+  | 128 - 226  |   0 -  98     |  7                  |
+  | 227 - 255  | 128 - 156     |  8                  |
+
+Symbols probabilities are read one by one, in order.
+
+Probability is obtained from Value decoded by following formula :
+`Proba = value - 1`
+
+It means value `0` becomes negative probability `-1`.
+`-1` is a special probability, which means "less than 1".
+Its effect on distribution table is described in [next paragraph].
+For the purpose of calculating cumulated distribution, it counts as one.
+
+[next paragraph]:#fse-decoding--from-normalized-distribution-to-decoding-tables
+
+When a symbol has a probability of `zero`,
+it is followed by a 2-bits repeat flag.
+This repeat flag tells how many probabilities of zeroes follow the current one.
+It provides a number ranging from 0 to 3.
+If it is a 3, another 2-bits repeat flag follows, and so on.
+
+When last symbol reaches cumulated total of `1 << Accuracy_Log`,
+decoding is complete.
+If the last symbol makes cumulated total go above `1 << Accuracy_Log`,
+distribution is considered corrupted.
+
+Then the decoder can tell how many bytes were used in this process,
+and how many symbols are present.
+The bitstream consumes a round number of bytes.
+Any remaining bit within the last byte is just unused.
+
+##### FSE decoding : from normalized distribution to decoding tables
+
+The distribution of normalized probabilities is enough
+to create a unique decoding table.
+
+It follows the following build rule :
+
+The table has a size of `tableSize = 1 << Accuracy_Log`.
+Each cell describes the symbol decoded,
+and instructions to get the next state.
+
+Symbols are scanned in their natural order for "less than 1" probabilities.
+Symbols with this probability are being attributed a single cell,
+starting from the end of the table.
+These symbols define a full state reset, reading `Accuracy_Log` bits.
+
+All remaining symbols are sorted in their natural order.
+Starting from symbol `0` and table position `0`,
+each symbol gets attributed as many cells as its probability.
+Cell allocation is spreaded, not linear :
+each successor position follow this rule :
+
+```
+position += (tableSize>>1) + (tableSize>>3) + 3;
+position &= tableSize-1;
+```
+
+A position is skipped if already occupied,
+typically by a "less than 1" probability symbol.
+
+The result is a list of state values.
+Each state will decode the current symbol.
+
+To get the `Number_of_Bits` and `Baseline` required for next state,
+it's first necessary to sort all states in their natural order.
+The lower states will need 1 more bit than higher ones.
+
+__Example__ :
+Presuming a symbol has a probability of 5.
+It receives 5 state values. States are sorted in natural order.
+
+Next power of 2 is 8.
+Space of probabilities is divided into 8 equal parts.
+Presuming the `Accuracy_Log` is 7, it defines 128 states.
+Divided by 8, each share is 16 large.
+
+In order to reach 8, 8-5=3 lowest states will count "double",
+taking shares twice larger,
+requiring one more bit in the process.
+
+Numbering starts from higher states using less bits.
+
+| state order      |   0   |   1   |    2   |   3  |   4   |
+| ---------------- | ----- | ----- | ------ | ---- | ----- |
+| width            |  32   |  32   |   32   |  16  |  16   |
+| `Number_of_Bits` |   5   |   5   |    5   |   4  |   4   |
+| range number     |   2   |   4   |    6   |   0  |   1   |
+| `Baseline`       |  32   |  64   |   96   |   0  |  16   |
+| range            | 32-63 | 64-95 | 96-127 | 0-15 | 16-31 |
+
+Next state is determined from current state
+by reading the required `Number_of_Bits`, and adding the specified `Baseline`.
+
+
+#### Bitstream
+
+FSE bitstreams are read in reverse direction than written. In zstd,
+the compressor writes bits forward into a block and the decompressor
+must read the bitstream _backwards_.
+
+To find the start of the bitstream it is therefore necessary to
+know the offset of the last byte of the block which can be found
+by counting `Block_Size` bytes after the block header.
+
+After writing the last bit containing information, the compressor
+writes a single `1`-bit and then fills the byte with 0-7 `0` bits of
+padding. The last byte of the compressed bitstream cannot be `0` for
+that reason.
+
+When decompressing, the last byte containing the padding is the first
+byte to read. The decompressor needs to skip 0-7 initial `0`-bits and
+the first `1`-bit it occurs. Afterwards, the useful part of the bitstream
+begins.
+
+##### Starting states
+
+The bitstream starts with initial state values,
+each using the required number of bits in their respective _accuracy_,
+decoded previously from their normalized distribution.
+
+It starts by `Literals_Length_State`,
+followed by `Offset_State`,
+and finally `Match_Length_State`.
+
+Reminder : always keep in mind that all values are read _backward_.
+
+##### Decoding a sequence
+
+A state gives a code.
+A code provides `Baseline` and `Number_of_Bits` to add.
+See [Symbol Decoding] section for details on each symbol.
+
+Decoding starts by reading the `Number_of_Bits` required to decode `Offset`.
+It then does the same for `Match_Length`,
+and then for `Literals_Length`.
+
+`Offset`, `Match_Length`, and `Literals_Length` define a sequence.
+It starts by inserting the number of literals defined by `Literals_Length`,
+then continue by copying `Match_Length` bytes from `currentPos - Offset`.
+
+The next operation is to update states.
+Using rules pre-calculated in the decoding tables,
+`Literals_Length_State` is updated,
+followed by `Match_Length_State`,
+and then `Offset_State`.
+
+This operation will be repeated `Number_of_Sequences` times.
+At the end, the bitstream shall be entirely consumed,
+otherwise bitstream is considered corrupted.
+
+[Symbol Decoding]:#the-codes-for-literals-lengths-match-lengths-and-offsets
+
+##### Repeat offsets
+
+As seen in [Offset Codes], the first 3 values define a repeated offset and we will call them `Repeated_Offset1`, `Repeated_Offset2`, and `Repeated_Offset3`.
+They are sorted in recency order, with `Repeated_Offset1` meaning "most recent one".
+
+There is an exception though, when current sequence's literals length is `0`.
+In which case, repeated offsets are "pushed by one",
+so `Repeated_Offset1` becomes `Repeated_Offset2`, `Repeated_Offset2` becomes `Repeated_Offset3`,
+and `Repeated_Offset3` becomes `Repeated_Offset1 - 1_byte`.
+
+On first block, offset history is populated by the following values : 1, 4 and 8 (in order).
+
+Then each block receives its start value from previous compressed block.
+Note that non-compressed blocks are skipped,
+they do not contribute to offset history.
+
+[Offset Codes]: #offset-codes
+
+###### Offset updates rules
+
+New offset take the lead in offset history,
+up to its previous place if it was already present.
+
+It means that when `Repeated_Offset1` (most recent) is used, history is unmodified.
+When `Repeated_Offset2` is used, it's swapped with `Repeated_Offset1`.
+
+
+Dictionary format
+-----------------
+
+`zstd` is compatible with "raw content" dictionaries, free of any format restriction,
+except that they must be at least 8 bytes.
+But dictionaries created by `zstd --train` follow a format, described here.
+
+__Pre-requisites__ : a dictionary has a size,
+                     defined either by a buffer limit, or a file size.
+
+| `Magic_Number` | `Dictionary_ID` | `Entropy_Tables` | `Content` |
+| -------------- | --------------- | ---------------- | --------- |
+
+__`Magic_Number`__ : 4 bytes ID, value 0xEC30A437, little-endian format
+
+__`Dictionary_ID`__ : 4 bytes, stored in little-endian format.
+              `Dictionary_ID` can be any value, except 0 (which means no `Dictionary_ID`).
+              It's used by decoders to check if they use the correct dictionary.
+
+_Reserved ranges :_
+              If the frame is going to be distributed in a private environment,
+              any `Dictionary_ID` can be used.
+              However, for public distribution of compressed frames,
+              the following ranges are reserved for future use and should not be used :
+
+              - low range : 1 - 32767
+              - high range : >= (2^31)
+
+__`Entropy_Tables`__ : following the same format as a [compressed blocks].
+              They are stored in following order :
+              Huffman tables for literals, FSE table for offsets,
+              FSE table for match lengths, and FSE table for literals lengths.
+              It's finally followed by 3 offset values, populating recent offsets,
+              stored in order, 4-bytes little-endian each, for a total of 12 bytes.
+              Each recent offset must have a value < dictionary size.
+
+__`Content`__ : The rest of the dictionary is its content.
+              The content act as a "past" in front of data to compress or decompress.
+
+[compressed blocks]: #the-format-of-compressed_block
+
+Appendix A - Decoding tables for predefined codes
+-------------------------------------------------
+
+This appendix contains FSE decoding tables for the predefined literal length, match length, and offset
+codes. The tables have been constructed using the algorithm as given above in the
+"from normalized distribution to decoding tables" chapter. The tables here can be used as examples
+to crosscheck that an implementation implements the decoding table generation algorithm correctly.
+
+#### Literal Length Code:
+
+| State | Symbol | Number_Of_Bits | Base |
+| ----- | ------ | -------------- | ---- |
+|     0 |      0 |              4 |    0 |
+|     1 |      0 |              4 |   16 |
+|     2 |      1 |              5 |   32 |
+|     3 |      3 |              5 |    0 |
+|     4 |      4 |              5 |    0 |
+|     5 |      6 |              5 |    0 |
+|     6 |      7 |              5 |    0 |
+|     7 |      9 |              5 |    0 |
+|     8 |     10 |              5 |    0 |
+|     9 |     12 |              5 |    0 |
+|    10 |     14 |              6 |    0 |
+|    11 |     16 |              5 |    0 |
+|    12 |     18 |              5 |    0 |
+|    13 |     19 |              5 |    0 |
+|    14 |     21 |              5 |    0 |
+|    15 |     22 |              5 |    0 |
+|    16 |     24 |              5 |    0 |
+|    17 |     25 |              5 |   32 |
+|    18 |     26 |              5 |    0 |
+|    19 |     27 |              6 |    0 |
+|    20 |     29 |              6 |    0 |
+|    21 |     31 |              6 |    0 |
+|    22 |      0 |              4 |   32 |
+|    23 |      1 |              4 |    0 |
+|    24 |      2 |              5 |    0 |
+|    25 |      4 |              5 |   32 |
+|    26 |      5 |              5 |    0 |
+|    27 |      7 |              5 |   32 |
+|    28 |      8 |              5 |    0 |
+|    29 |     10 |              5 |   32 |
+|    30 |     11 |              5 |    0 |
+|    31 |     13 |              6 |    0 |
+|    32 |     16 |              5 |   32 |
+|    33 |     17 |              5 |    0 |
+|    34 |     19 |              5 |   32 |
+|    35 |     20 |              5 |    0 |
+|    36 |     22 |              5 |   32 |
+|    37 |     23 |              5 |    0 |
+|    38 |     25 |              4 |    0 |
+|    39 |     25 |              4 |   16 |
+|    40 |     26 |              5 |   32 |
+|    41 |     28 |              6 |    0 |
+|    42 |     30 |              6 |    0 |
+|    43 |      0 |              4 |   48 |
+|    44 |      1 |              4 |   16 |
+|    45 |      2 |              5 |   32 |
+|    46 |      3 |              5 |   32 |
+|    47 |      5 |              5 |   32 |
+|    48 |      6 |              5 |   32 |
+|    49 |      8 |              5 |   32 |
+|    50 |      9 |              5 |   32 |
+|    51 |     11 |              5 |   32 |
+|    52 |     12 |              5 |   32 |
+|    53 |     15 |              6 |    0 |
+|    54 |     17 |              5 |   32 |
+|    55 |     18 |              5 |   32 |
+|    56 |     20 |              5 |   32 |
+|    57 |     21 |              5 |   32 |
+|    58 |     23 |              5 |   32 |
+|    59 |     24 |              5 |   32 |
+|    60 |     35 |              6 |    0 |
+|    61 |     34 |              6 |    0 |
+|    62 |     33 |              6 |    0 |
+|    63 |     32 |              6 |    0 |
+
+#### Match Length Code:
+
+| State | Symbol | Number_Of_Bits | Base |
+| ----- | ------ | -------------- | ---- |
+|     0 |      0 |              6 |    0 |
+|     1 |      1 |              4 |    0 |
+|     2 |      2 |              5 |   32 |
+|     3 |      3 |              5 |    0 |
+|     4 |      5 |              5 |    0 |
+|     5 |      6 |              5 |    0 |
+|     6 |      8 |              5 |    0 |
+|     7 |     10 |              6 |    0 |
+|     8 |     13 |              6 |    0 |
+|     9 |     16 |              6 |    0 |
+|    10 |     19 |              6 |    0 |
+|    11 |     22 |              6 |    0 |
+|    12 |     25 |              6 |    0 |
+|    13 |     28 |              6 |    0 |
+|    14 |     31 |              6 |    0 |
+|    15 |     33 |              6 |    0 |
+|    16 |     35 |              6 |    0 |
+|    17 |     37 |              6 |    0 |
+|    18 |     39 |              6 |    0 |
+|    19 |     41 |              6 |    0 |
+|    20 |     43 |              6 |    0 |
+|    21 |     45 |              6 |    0 |
+|    22 |      1 |              4 |   16 |
+|    23 |      2 |              4 |    0 |
+|    24 |      3 |              5 |   32 |
+|    25 |      4 |              5 |    0 |
+|    26 |      6 |              5 |   32 |
+|    27 |      7 |              5 |    0 |
+|    28 |      9 |              6 |    0 |
+|    29 |     12 |              6 |    0 |
+|    30 |     15 |              6 |    0 |
+|    31 |     18 |              6 |    0 |
+|    32 |     21 |              6 |    0 |
+|    33 |     24 |              6 |    0 |
+|    34 |     27 |              6 |    0 |
+|    35 |     30 |              6 |    0 |
+|    36 |     32 |              6 |    0 |
+|    37 |     34 |              6 |    0 |
+|    38 |     36 |              6 |    0 |
+|    39 |     38 |              6 |    0 |
+|    40 |     40 |              6 |    0 |
+|    41 |     42 |              6 |    0 |
+|    42 |     44 |              6 |    0 |
+|    43 |      1 |              4 |   32 |
+|    44 |      1 |              4 |   48 |
+|    45 |      2 |              4 |   16 |
+|    46 |      4 |              5 |   32 |
+|    47 |      5 |              5 |   32 |
+|    48 |      7 |              5 |   32 |
+|    49 |      8 |              5 |   32 |
+|    50 |     11 |              6 |    0 |
+|    51 |     14 |              6 |    0 |
+|    52 |     17 |              6 |    0 |
+|    53 |     20 |              6 |    0 |
+|    54 |     23 |              6 |    0 |
+|    55 |     26 |              6 |    0 |
+|    56 |     29 |              6 |    0 |
+|    57 |     52 |              6 |    0 |
+|    58 |     51 |              6 |    0 |
+|    59 |     50 |              6 |    0 |
+|    60 |     49 |              6 |    0 |
+|    61 |     48 |              6 |    0 |
+|    62 |     47 |              6 |    0 |
+|    63 |     46 |              6 |    0 |
+
+#### Offset Code:
+
+| State | Symbol | Number_Of_Bits | Base |
+| ----- | ------ | -------------- | ---- |
+|     0 |      0 |              5 |    0 |
+|     1 |      6 |              4 |    0 |
+|     2 |      9 |              5 |    0 |
+|     3 |     15 |              5 |    0 |
+|     4 |     21 |              5 |    0 |
+|     5 |      3 |              5 |    0 |
+|     6 |      7 |              4 |    0 |
+|     7 |     12 |              5 |    0 |
+|     8 |     18 |              5 |    0 |
+|     9 |     23 |              5 |    0 |
+|    10 |      5 |              5 |    0 |
+|    11 |      8 |              4 |    0 |
+|    12 |     14 |              5 |    0 |
+|    13 |     20 |              5 |    0 |
+|    14 |      2 |              5 |    0 |
+|    15 |      7 |              4 |   16 |
+|    16 |     11 |              5 |    0 |
+|    17 |     17 |              5 |    0 |
+|    18 |     22 |              5 |    0 |
+|    19 |      4 |              5 |    0 |
+|    20 |      8 |              4 |   16 |
+|    21 |     13 |              5 |    0 |
+|    22 |     19 |              5 |    0 |
+|    23 |      1 |              5 |    0 |
+|    24 |      6 |              4 |   16 |
+|    25 |     10 |              5 |    0 |
+|    26 |     16 |              5 |    0 |
+|    27 |     28 |              5 |    0 |
+|    28 |     27 |              5 |    0 |
+|    29 |     26 |              5 |    0 |
+|    30 |     25 |              5 |    0 |
+|    31 |     24 |              5 |    0 |
+
+Version changes
+---------------
+- 0.2.2 : added predefined codes, by Johannes Rudolph
+- 0.2.1 : clarify field names, by Przemyslaw Skibinski
+- 0.2.0 : numerous format adjustments for zstd v0.8
+- 0.1.2 : limit Huffman tree depth to 11 bits
+- 0.1.1 : reserved dictID ranges
+- 0.1.0 : initial release
diff --git a/doc/zstd_manual.html b/doc/zstd_manual.html
new file mode 100644
index 0000000..1badcbd
--- /dev/null
+++ b/doc/zstd_manual.html
@@ -0,0 +1,551 @@
+
+
+
+zstd 1.1.2 Manual
+
+
+
zstd 1.1.2 Manual

+

+
Contents

+
    
+
  1. Introduction
    2. 
+
  2. Version
    3. 
+
  3. Simple API
    4. 
+
  4. Explicit memory management
    5. 
+
  5. Simple dictionary API
    6. 
+
  6. Fast dictionary API
    7. 
+
  7. Streaming
    8. 
+
  8. Streaming compression - HowTo
    9. 
+
  9. Streaming decompression - HowTo
    10. 
+
  10. START OF ADVANCED AND EXPERIMENTAL FUNCTIONS
    11. 
+
  11. Advanced types
    12. 
+
  12. Advanced compression functions
    13. 
+
  13. Advanced decompression functions
    14. 
+
  14. Advanced streaming functions
    15. 
+
  15. Buffer-less and synchronous inner streaming functions
    16. 
+
  16. Buffer-less streaming compression (synchronous mode)
    17. 
+
  17. Buffer-less streaming decompression (synchronous mode)
    18. 
+
  18. Block functions
    19. 
+

+

+
Introduction
+  zstd, short for Zstandard, is a fast lossless compression algorithm, targeting real-time compression scenarios
+  at zlib-level and better compression ratios. The zstd compression library provides in-memory compression and
+  decompression functions. The library supports compression levels from 1 up to ZSTD_maxCLevel() which is 22.
+  Levels >= 20, labelled `--ultra`, should be used with caution, as they require more memory.
+  Compression can be done in:
+    - a single step (described as Simple API)
+    - a single step, reusing a context (described as Explicit memory management)
+    - unbounded multiple steps (described as Streaming compression)
+  The compression ratio achievable on small data can be highly improved using compression with a dictionary in:
+    - a single step (described as Simple dictionary API)
+    - a single step, reusing a dictionary (described as Fast dictionary API)
+
+  Advanced experimental functions can be accessed using #define ZSTD_STATIC_LINKING_ONLY before including zstd.h.
+  These APIs shall never be used with a dynamic library.
+  They are not "stable", their definition may change in the future. Only static linking is allowed.
+

+
+
Version

+
+
unsigned ZSTD_versionNumber(void);   /**< library version number; to be used when checking dll version */
+


+
Simple API

+
+
size_t ZSTD_compress( void* dst, size_t dstCapacity,
+                            const void* src, size_t srcSize,
+                                  int compressionLevel);
+
    Compresses `src` content as a single zstd compressed frame into already allocated `dst`.
+    Hint : compression runs faster if `dstCapacity` >=  `ZSTD_compressBound(srcSize)`.
+    @return : compressed size written into `dst` (<= `dstCapacity),
+              or an error code if it fails (which can be tested using ZSTD_isError()). 
+


+
+
size_t ZSTD_decompress( void* dst, size_t dstCapacity,
+                              const void* src, size_t compressedSize);
+
    `compressedSize` : must be the _exact_ size of a single compressed frame.
+    `dstCapacity` is an upper bound of originalSize.
+    If user cannot imply a maximum upper bound, it's better to use streaming mode to decompress data.
+    @return : the number of bytes decompressed into `dst` (<= `dstCapacity`),
+              or an errorCode if it fails (which can be tested using ZSTD_isError()). 
+


+
+
unsigned long long ZSTD_getDecompressedSize(const void* src, size_t srcSize);
+
   'src' is the start of a zstd compressed frame.
+   @return : content size to be decompressed, as a 64-bits value _if known_, 0 otherwise.
+    note 1 : decompressed size is an optional field, that may not be present, especially in streaming mode.
+             When `return==0`, data to decompress could be any size.
+             In which case, it's necessary to use streaming mode to decompress data.
+             Optionally, application can still use ZSTD_decompress() while relying on implied limits.
+             (For example, data may be necessarily cut into blocks <= 16 KB).
+    note 2 : decompressed size is always present when compression is done with ZSTD_compress()
+    note 3 : decompressed size can be very large (64-bits value),
+             potentially larger than what local system can handle as a single memory segment.
+             In which case, it's necessary to use streaming mode to decompress data.
+    note 4 : If source is untrusted, decompressed size could be wrong or intentionally modified.
+             Always ensure result fits within application's authorized limits.
+             Each application can set its own limits.
+    note 5 : when `return==0`, if precise failure cause is needed, use ZSTD_getFrameParams() to know more. 
+


+
+
Helper functions
int         ZSTD_maxCLevel(void);               /*!< maximum compression level available */
+size_t      ZSTD_compressBound(size_t srcSize); /*!< maximum compressed size in worst case scenario */
+unsigned    ZSTD_isError(size_t code);          /*!< tells if a `size_t` function result is an error code */
+const char* ZSTD_getErrorName(size_t code);     /*!< provides readable string from an error code */
+


+
Explicit memory management

+
+
size_t ZSTD_compressCCtx(ZSTD_CCtx* ctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize, int compressionLevel);
+
    Same as ZSTD_compress(), requires an allocated ZSTD_CCtx (see ZSTD_createCCtx()). 
+


+
+
Decompression context
typedef struct ZSTD_DCtx_s ZSTD_DCtx;
+ZSTD_DCtx* ZSTD_createDCtx(void);
+size_t     ZSTD_freeDCtx(ZSTD_DCtx* dctx);
+


+
size_t ZSTD_decompressDCtx(ZSTD_DCtx* ctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
+
   Same as ZSTD_decompress(), requires an allocated ZSTD_DCtx (see ZSTD_createDCtx()). 
+


+
+
Simple dictionary API

+
+
size_t ZSTD_compress_usingDict(ZSTD_CCtx* ctx,
+                                           void* dst, size_t dstCapacity,
+                                     const void* src, size_t srcSize,
+                                     const void* dict,size_t dictSize,
+                                           int compressionLevel);
+
   Compression using a predefined Dictionary (see dictBuilder/zdict.h).
+   Note : This function loads the dictionary, resulting in significant startup delay.
+   Note : When `dict == NULL || dictSize < 8` no dictionary is used. 
+


+
+
size_t ZSTD_decompress_usingDict(ZSTD_DCtx* dctx,
+                                             void* dst, size_t dstCapacity,
+                                       const void* src, size_t srcSize,
+                                       const void* dict,size_t dictSize);
+
   Decompression using a predefined Dictionary (see dictBuilder/zdict.h).
+   Dictionary must be identical to the one used during compression.
+   Note : This function loads the dictionary, resulting in significant startup delay.
+   Note : When `dict == NULL || dictSize < 8` no dictionary is used. 
+


+
+
Fast dictionary API

+
+
ZSTD_CDict* ZSTD_createCDict(const void* dict, size_t dictSize, int compressionLevel);
+
   When compressing multiple messages / blocks with the same dictionary, it's recommended to load it just once.
+   ZSTD_createCDict() will create a digested dictionary, ready to start future compression operations without startup delay.
+   ZSTD_CDict can be created once and used by multiple threads concurrently, as its usage is read-only.
+   `dict` can be released after ZSTD_CDict creation. 
+


+
+
size_t      ZSTD_freeCDict(ZSTD_CDict* CDict);
+
   Function frees memory allocated by ZSTD_createCDict(). 
+


+
+
size_t ZSTD_compress_usingCDict(ZSTD_CCtx* cctx,
+                                            void* dst, size_t dstCapacity,
+                                      const void* src, size_t srcSize,
+                                      const ZSTD_CDict* cdict);
+
   Compression using a digested Dictionary.
+   Faster startup than ZSTD_compress_usingDict(), recommended when same dictionary is used multiple times.
+   Note that compression level is decided during dictionary creation. 
+


+
+
ZSTD_DDict* ZSTD_createDDict(const void* dict, size_t dictSize);
+
   Create a digested dictionary, ready to start decompression operation without startup delay.
+   `dict` can be released after creation. 
+


+
+
size_t      ZSTD_freeDDict(ZSTD_DDict* ddict);
+
   Function frees memory allocated with ZSTD_createDDict() 
+


+
+
size_t ZSTD_decompress_usingDDict(ZSTD_DCtx* dctx,
+                                              void* dst, size_t dstCapacity,
+                                        const void* src, size_t srcSize,
+                                        const ZSTD_DDict* ddict);
+
   Decompression using a digested Dictionary.
+   Faster startup than ZSTD_decompress_usingDict(), recommended when same dictionary is used multiple times. 
+


+
+
Streaming

+
+
typedef struct ZSTD_inBuffer_s {
+  const void* src;    /**< start of input buffer */
+  size_t size;        /**< size of input buffer */
+  size_t pos;         /**< position where reading stopped. Will be updated. Necessarily 0 <= pos <= size */
+} ZSTD_inBuffer;
+


+
typedef struct ZSTD_outBuffer_s {
+  void*  dst;         /**< start of output buffer */
+  size_t size;        /**< size of output buffer */
+  size_t pos;         /**< position where writing stopped. Will be updated. Necessarily 0 <= pos <= size */
+} ZSTD_outBuffer;
+


+
Streaming compression - HowTo
+  A ZSTD_CStream object is required to track streaming operation.
+  Use ZSTD_createCStream() and ZSTD_freeCStream() to create/release resources.
+  ZSTD_CStream objects can be reused multiple times on consecutive compression operations.
+  It is recommended to re-use ZSTD_CStream in situations where many streaming operations will be achieved consecutively,
+  since it will play nicer with system's memory, by re-using already allocated memory.
+  Use one separate ZSTD_CStream per thread for parallel execution.
+
+  Start a new compression by initializing ZSTD_CStream.
+  Use ZSTD_initCStream() to start a new compression operation.
+  Use ZSTD_initCStream_usingDict() or ZSTD_initCStream_usingCDict() for a compression which requires a dictionary (experimental section)
+
+  Use ZSTD_compressStream() repetitively to consume input stream.
+  The function will automatically update both `pos` fields.
+  Note that it may not consume the entire input, in which case `pos < size`,
+  and it's up to the caller to present again remaining data.
+  @return : a size hint, preferred nb of bytes to use as input for next function call
+            or an error code, which can be tested using ZSTD_isError().
+            Note 1 : it's just a hint, to help latency a little, any other value will work fine.
+            Note 2 : size hint is guaranteed to be <= ZSTD_CStreamInSize()
+
+  At any moment, it's possible to flush whatever data remains within internal buffer, using ZSTD_flushStream().
+  `output->pos` will be updated.
+  Note that some content might still be left within internal buffer if `output->size` is too small.
+  @return : nb of bytes still present within internal buffer (0 if it's empty)
+            or an error code, which can be tested using ZSTD_isError().
+
+  ZSTD_endStream() instructs to finish a frame.
+  It will perform a flush and write frame epilogue.
+  The epilogue is required for decoders to consider a frame completed.
+  Similar to ZSTD_flushStream(), it may not be able to flush the full content if `output->size` is too small.
+  In which case, call again ZSTD_endStream() to complete the flush.
+  @return : nb of bytes still present within internal buffer (0 if it's empty, hence compression completed)
+            or an error code, which can be tested using ZSTD_isError().
+
+ 
+

+
+
size_t ZSTD_CStreamInSize(void);    /**< recommended size for input buffer */
+


+
size_t ZSTD_CStreamOutSize(void);   /**< recommended size for output buffer. Guarantee to successfully flush at least one complete compressed block in all circumstances. */
+


+
Streaming decompression - HowTo
+  A ZSTD_DStream object is required to track streaming operations.
+  Use ZSTD_createDStream() and ZSTD_freeDStream() to create/release resources.
+  ZSTD_DStream objects can be re-used multiple times.
+
+  Use ZSTD_initDStream() to start a new decompression operation,
+   or ZSTD_initDStream_usingDict() if decompression requires a dictionary.
+   @return : recommended first input size
+
+  Use ZSTD_decompressStream() repetitively to consume your input.
+  The function will update both `pos` fields.
+  If `input.pos < input.size`, some input has not been consumed.
+  It's up to the caller to present again remaining data.
+  If `output.pos < output.size`, decoder has flushed everything it could.
+  @return : 0 when a frame is completely decoded and fully flushed,
+            an error code, which can be tested using ZSTD_isError(),
+            any other value > 0, which means there is still some decoding to do to complete current frame.
+            The return value is a suggested next input size (a hint to improve latency) that will never load more than the current frame.
+ 
+

+
+
size_t ZSTD_DStreamInSize(void);    /*!< recommended size for input buffer */
+


+
size_t ZSTD_DStreamOutSize(void);   /*!< recommended size for output buffer. Guarantee to successfully flush at least one complete block in all circumstances. */
+


+
START OF ADVANCED AND EXPERIMENTAL FUNCTIONS
 The definitions in this section are considered experimental.
+ They should never be used with a dynamic library, as they may change in the future.
+ They are provided for advanced usages.
+ Use them only in association with static linking.
+ 
+

+
+
Advanced types

+
+
typedef enum { ZSTD_fast, ZSTD_dfast, ZSTD_greedy, ZSTD_lazy, ZSTD_lazy2, ZSTD_btlazy2, ZSTD_btopt, ZSTD_btopt2 } ZSTD_strategy;   /* from faster to stronger */
+


+
typedef struct {
+    unsigned windowLog;      /**< largest match distance : larger == more compression, more memory needed during decompression */
+    unsigned chainLog;       /**< fully searched segment : larger == more compression, slower, more memory (useless for fast) */
+    unsigned hashLog;        /**< dispatch table : larger == faster, more memory */
+    unsigned searchLog;      /**< nb of searches : larger == more compression, slower */
+    unsigned searchLength;   /**< match length searched : larger == faster decompression, sometimes less compression */
+    unsigned targetLength;   /**< acceptable match size for optimal parser (only) : larger == more compression, slower */
+    ZSTD_strategy strategy;
+} ZSTD_compressionParameters;
+


+
typedef struct {
+    unsigned contentSizeFlag; /**< 1: content size will be in frame header (if known). */
+    unsigned checksumFlag;    /**< 1: will generate a 22-bits checksum at end of frame, to be used for error detection by decompressor */
+    unsigned noDictIDFlag;    /**< 1: no dict ID will be saved into frame header (if dictionary compression) */
+} ZSTD_frameParameters;
+


+
typedef struct {
+    ZSTD_compressionParameters cParams;
+    ZSTD_frameParameters fParams;
+} ZSTD_parameters;
+


+
Custom memory allocation functions
typedef void* (*ZSTD_allocFunction) (void* opaque, size_t size);
+typedef void  (*ZSTD_freeFunction) (void* opaque, void* address);
+typedef struct { ZSTD_allocFunction customAlloc; ZSTD_freeFunction customFree; void* opaque; } ZSTD_customMem;
+


+
Advanced compression functions

+
+
size_t ZSTD_estimateCCtxSize(ZSTD_compressionParameters cParams);
+
  Gives the amount of memory allocated for a ZSTD_CCtx given a set of compression parameters.
+  `frameContentSize` is an optional parameter, provide `0` if unknown 
+


+
+
ZSTD_CCtx* ZSTD_createCCtx_advanced(ZSTD_customMem customMem);
+
  Create a ZSTD compression context using external alloc and free functions 
+


+
+
size_t ZSTD_sizeof_CCtx(const ZSTD_CCtx* cctx);
+
  Gives the amount of memory used by a given ZSTD_CCtx 
+


+
+
ZSTD_CDict* ZSTD_createCDict_advanced(const void* dict, size_t dictSize,
+                                                  ZSTD_parameters params, ZSTD_customMem customMem);
+
  Create a ZSTD_CDict using external alloc and free, and customized compression parameters 
+


+
+
size_t ZSTD_sizeof_CDict(const ZSTD_CDict* cdict);
+
  Gives the amount of memory used by a given ZSTD_sizeof_CDict 
+


+
+
ZSTD_compressionParameters ZSTD_getCParams(int compressionLevel, unsigned long long estimatedSrcSize, size_t dictSize);
+
   @return ZSTD_compressionParameters structure for a selected compression level and estimated srcSize.
+   `estimatedSrcSize` value is optional, select 0 if not known 
+


+
+
ZSTD_parameters ZSTD_getParams(int compressionLevel, unsigned long long estimatedSrcSize, size_t dictSize);
+
   same as ZSTD_getCParams(), but @return a full `ZSTD_parameters` object instead of sub-component `ZSTD_compressionParameters`.
+   All fields of `ZSTD_frameParameters` are set to default (0) 
+


+
+
size_t ZSTD_checkCParams(ZSTD_compressionParameters params);
+
   Ensure param values remain within authorized range 
+


+
+
ZSTD_compressionParameters ZSTD_adjustCParams(ZSTD_compressionParameters cPar, unsigned long long srcSize, size_t dictSize);
+
   optimize params for a given `srcSize` and `dictSize`.
+   both values are optional, select `0` if unknown. 
+


+
+
size_t ZSTD_compress_advanced (ZSTD_CCtx* ctx,
+                                           void* dst, size_t dstCapacity,
+                                     const void* src, size_t srcSize,
+                                     const void* dict,size_t dictSize,
+                                           ZSTD_parameters params);
+
   Same as ZSTD_compress_usingDict(), with fine-tune control of each compression parameter 
+


+
+
Advanced decompression functions

+
+
unsigned ZSTD_isFrame(const void* buffer, size_t size);
+
  Tells if the content of `buffer` starts with a valid Frame Identifier.
+  Note : Frame Identifier is 4 bytes. If `size < 4`, @return will always be 0.
+  Note 2 : Legacy Frame Identifiers are considered valid only if Legacy Support is enabled.
+  Note 3 : Skippable Frame Identifiers are considered valid. 
+


+
+
size_t ZSTD_estimateDCtxSize(void);
+
  Gives the potential amount of memory allocated to create a ZSTD_DCtx 
+


+
+
ZSTD_DCtx* ZSTD_createDCtx_advanced(ZSTD_customMem customMem);
+
  Create a ZSTD decompression context using external alloc and free functions 
+


+
+
size_t ZSTD_sizeof_DCtx(const ZSTD_DCtx* dctx);
+
  Gives the amount of memory used by a given ZSTD_DCtx 
+


+
+
size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict);
+
  Gives the amount of memory used by a given ZSTD_DDict 
+


+
+
unsigned ZSTD_getDictID_fromDict(const void* dict, size_t dictSize);
+
  Provides the dictID stored within dictionary.
+  if @return == 0, the dictionary is not conformant with Zstandard specification.
+  It can still be loaded, but as a content-only dictionary. 
+


+
+
unsigned ZSTD_getDictID_fromDDict(const ZSTD_DDict* ddict);
+
  Provides the dictID of the dictionary loaded into `ddict`.
+  If @return == 0, the dictionary is not conformant to Zstandard specification, or empty.
+  Non-conformant dictionaries can still be loaded, but as content-only dictionaries. 
+


+
+
unsigned ZSTD_getDictID_fromFrame(const void* src, size_t srcSize);
+
  Provides the dictID required to decompressed the frame stored within `src`.
+  If @return == 0, the dictID could not be decoded.
+  This could for one of the following reasons :
+  - The frame does not require a dictionary to be decoded (most common case).
+  - The frame was built with dictID intentionally removed. Whatever dictionary is necessary is a hidden information.
+    Note : this use case also happens when using a non-conformant dictionary.
+  - `srcSize` is too small, and as a result, the frame header could not be decoded (only possible if `srcSize < ZSTD_FRAMEHEADERSIZE_MAX`).
+  - This is not a Zstandard frame.
+  When identifying the exact failure cause, it's possible to used ZSTD_getFrameParams(), which will provide a more precise error code. 
+


+
+
Advanced streaming functions

+
+
Advanced Streaming compression functions
ZSTD_CStream* ZSTD_createCStream_advanced(ZSTD_customMem customMem);
+size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLevel, unsigned long long pledgedSrcSize);   /**< pledgedSrcSize must be correct */
+size_t ZSTD_initCStream_usingDict(ZSTD_CStream* zcs, const void* dict, size_t dictSize, int compressionLevel);
+size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs, const void* dict, size_t dictSize,
+                                             ZSTD_parameters params, unsigned long long pledgedSrcSize);  /**< pledgedSrcSize is optional and can be zero == unknown */
+size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDict* cdict);  /**< note : cdict will just be referenced, and must outlive compression session */
+size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pledgedSrcSize);  /**< re-use compression parameters from previous init; skip dictionary loading stage; zcs must be init at least once before */
+size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs);
+


+
Advanced Streaming decompression functions
typedef enum { ZSTDdsp_maxWindowSize } ZSTD_DStreamParameter_e;
+ZSTD_DStream* ZSTD_createDStream_advanced(ZSTD_customMem customMem);
+size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize);
+size_t ZSTD_setDStreamParameter(ZSTD_DStream* zds, ZSTD_DStreamParameter_e paramType, unsigned paramValue);
+size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* zds, const ZSTD_DDict* ddict);  /**< note : ddict will just be referenced, and must outlive decompression session */
+size_t ZSTD_resetDStream(ZSTD_DStream* zds);  /**< re-use decompression parameters from previous init; saves dictionary loading */
+size_t ZSTD_sizeof_DStream(const ZSTD_DStream* zds);
+


+
Buffer-less and synchronous inner streaming functions
+  This is an advanced API, giving full control over buffer management, for users which need direct control over memory.
+  But it's also a complex one, with many restrictions (documented below).
+  Prefer using normal streaming API for an easier experience
+ 
+

+
+
Buffer-less streaming compression (synchronous mode)
+  A ZSTD_CCtx object is required to track streaming operations.
+  Use ZSTD_createCCtx() / ZSTD_freeCCtx() to manage resource.
+  ZSTD_CCtx object can be re-used multiple times within successive compression operations.
+
+  Start by initializing a context.
+  Use ZSTD_compressBegin(), or ZSTD_compressBegin_usingDict() for dictionary compression,
+  or ZSTD_compressBegin_advanced(), for finer parameter control.
+  It's also possible to duplicate a reference context which has already been initialized, using ZSTD_copyCCtx()
+
+  Then, consume your input using ZSTD_compressContinue().
+  There are some important considerations to keep in mind when using this advanced function :
+  - ZSTD_compressContinue() has no internal buffer. It uses externally provided buffer only.
+  - Interface is synchronous : input is consumed entirely and produce 1+ (or more) compressed blocks.
+  - Caller must ensure there is enough space in `dst` to store compressed data under worst case scenario.
+    Worst case evaluation is provided by ZSTD_compressBound().
+    ZSTD_compressContinue() doesn't guarantee recover after a failed compression.
+  - ZSTD_compressContinue() presumes prior input ***is still accessible and unmodified*** (up to maximum distance size, see WindowLog).
+    It remembers all previous contiguous blocks, plus one separated memory segment (which can itself consists of multiple contiguous blocks)
+  - ZSTD_compressContinue() detects that prior input has been overwritten when `src` buffer overlaps.
+    In which case, it will "discard" the relevant memory section from its history.
+
+  Finish a frame with ZSTD_compressEnd(), which will write the last block(s) and optional checksum.
+  It's possible to use a NULL,0 src content, in which case, it will write a final empty block to end the frame,
+  Without last block mark, frames will be considered unfinished (broken) by decoders.
+
+  You can then reuse `ZSTD_CCtx` (ZSTD_compressBegin()) to compress some new frame.
+

+
+
Buffer-less streaming compression functions
size_t ZSTD_compressBegin(ZSTD_CCtx* cctx, int compressionLevel);
+size_t ZSTD_compressBegin_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel);
+size_t ZSTD_compressBegin_advanced(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, ZSTD_parameters params, unsigned long long pledgedSrcSize);
+size_t ZSTD_copyCCtx(ZSTD_CCtx* cctx, const ZSTD_CCtx* preparedCCtx, unsigned long long pledgedSrcSize);
+size_t ZSTD_compressContinue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
+size_t ZSTD_compressEnd(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
+


+
Buffer-less streaming decompression (synchronous mode)
+  A ZSTD_DCtx object is required to track streaming operations.
+  Use ZSTD_createDCtx() / ZSTD_freeDCtx() to manage it.
+  A ZSTD_DCtx object can be re-used multiple times.
+
+  First typical operation is to retrieve frame parameters, using ZSTD_getFrameParams().
+  It fills a ZSTD_frameParams structure which provide important information to correctly decode the frame,
+  such as the minimum rolling buffer size to allocate to decompress data (`windowSize`),
+  and the dictionary ID used.
+  (Note : content size is optional, it may not be present. 0 means : content size unknown).
+  Note that these values could be wrong, either because of data malformation, or because an attacker is spoofing deliberate false information.
+  As a consequence, check that values remain within valid application range, especially `windowSize`, before allocation.
+  Each application can set its own limit, depending on local restrictions. For extended interoperability, it is recommended to support at least 8 MB.
+  Frame parameters are extracted from the beginning of the compressed frame.
+  Data fragment must be large enough to ensure successful decoding, typically `ZSTD_frameHeaderSize_max` bytes.
+  @result : 0 : successful decoding, the `ZSTD_frameParams` structure is correctly filled.
+           >0 : `srcSize` is too small, please provide at least @result bytes on next attempt.
+           errorCode, which can be tested using ZSTD_isError().
+
+  Start decompression, with ZSTD_decompressBegin() or ZSTD_decompressBegin_usingDict().
+  Alternatively, you can copy a prepared context, using ZSTD_copyDCtx().
+
+  Then use ZSTD_nextSrcSizeToDecompress() and ZSTD_decompressContinue() alternatively.
+  ZSTD_nextSrcSizeToDecompress() tells how many bytes to provide as 'srcSize' to ZSTD_decompressContinue().
+  ZSTD_decompressContinue() requires this _exact_ amount of bytes, or it will fail.
+
+  @result of ZSTD_decompressContinue() is the number of bytes regenerated within 'dst' (necessarily <= dstCapacity).
+  It can be zero, which is not an error; it just means ZSTD_decompressContinue() has decoded some metadata item.
+  It can also be an error code, which can be tested with ZSTD_isError().
+
+  ZSTD_decompressContinue() needs previous data blocks during decompression, up to `windowSize`.
+  They should preferably be located contiguously, prior to current block.
+  Alternatively, a round buffer of sufficient size is also possible. Sufficient size is determined by frame parameters.
+  ZSTD_decompressContinue() is very sensitive to contiguity,
+  if 2 blocks don't follow each other, make sure that either the compressor breaks contiguity at the same place,
+  or that previous contiguous segment is large enough to properly handle maximum back-reference.
+
+  A frame is fully decoded when ZSTD_nextSrcSizeToDecompress() returns zero.
+  Context can then be reset to start a new decompression.
+
+  Note : it's possible to know if next input to present is a header or a block, using ZSTD_nextInputType().
+  This information is not required to properly decode a frame.
+
+  == Special case : skippable frames ==
+
+  Skippable frames allow integration of user-defined data into a flow of concatenated frames.
+  Skippable frames will be ignored (skipped) by a decompressor. The format of skippable frames is as follows :
+  a) Skippable frame ID - 4 Bytes, Little endian format, any value from 0x184D2A50 to 0x184D2A5F
+  b) Frame Size - 4 Bytes, Little endian format, unsigned 32-bits
+  c) Frame Content - any content (User Data) of length equal to Frame Size
+  For skippable frames ZSTD_decompressContinue() always returns 0.
+  For skippable frames ZSTD_getFrameParams() returns fparamsPtr->windowLog==0 what means that a frame is skippable.
+  It also returns Frame Size as fparamsPtr->frameContentSize.
+

+
+
typedef struct {
+    unsigned long long frameContentSize;
+    unsigned windowSize;
+    unsigned dictID;
+    unsigned checksumFlag;
+} ZSTD_frameParams;
+


+
Buffer-less streaming decompression functions
size_t ZSTD_getFrameParams(ZSTD_frameParams* fparamsPtr, const void* src, size_t srcSize);   /**< doesn't consume input, see details below */
+size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx);
+size_t ZSTD_decompressBegin_usingDict(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);
+void   ZSTD_copyDCtx(ZSTD_DCtx* dctx, const ZSTD_DCtx* preparedDCtx);
+size_t ZSTD_nextSrcSizeToDecompress(ZSTD_DCtx* dctx);
+size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
+typedef enum { ZSTDnit_frameHeader, ZSTDnit_blockHeader, ZSTDnit_block, ZSTDnit_lastBlock, ZSTDnit_checksum, ZSTDnit_skippableFrame } ZSTD_nextInputType_e;
+ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx);
+


+
Block functions
+    Block functions produce and decode raw zstd blocks, without frame metadata.
+    Frame metadata cost is typically ~18 bytes, which can be non-negligible for very small blocks (< 100 bytes).
+    User will have to take in charge required information to regenerate data, such as compressed and content sizes.
+
+    A few rules to respect :
+    - Compressing and decompressing require a context structure
+      + Use ZSTD_createCCtx() and ZSTD_createDCtx()
+    - It is necessary to init context before starting
+      + compression : ZSTD_compressBegin()
+      + decompression : ZSTD_decompressBegin()
+      + variants _usingDict() are also allowed
+      + copyCCtx() and copyDCtx() work too
+    - Block size is limited, it must be <= ZSTD_getBlockSizeMax()
+      + If you need to compress more, cut data into multiple blocks
+      + Consider using the regular ZSTD_compress() instead, as frame metadata costs become negligible when source size is large.
+    - When a block is considered not compressible enough, ZSTD_compressBlock() result will be zero.
+      In which case, nothing is produced into `dst`.
+      + User must test for such outcome and deal directly with uncompressed data
+      + ZSTD_decompressBlock() doesn't accept uncompressed data as input !!!
+      + In case of multiple successive blocks, decoder must be informed of uncompressed block existence to follow proper history.
+        Use ZSTD_insertBlock() in such a case.
+

+
+
Raw zstd block functions
size_t ZSTD_getBlockSizeMax(ZSTD_CCtx* cctx);
+size_t ZSTD_compressBlock  (ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
+size_t ZSTD_decompressBlock(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
+size_t ZSTD_insertBlock(ZSTD_DCtx* dctx, const void* blockStart, size_t blockSize);  /**< insert block into `dctx` history. Useful for uncompressed blocks */
+


+
+
diff --git a/examples/.gitignore b/examples/.gitignore
new file mode 100644
index 0000000..0711813
--- /dev/null
+++ b/examples/.gitignore
@@ -0,0 +1,13 @@
+#build
+simple_compression
+simple_decompression
+dictionary_compression
+dictionary_decompression
+streaming_compression
+streaming_decompression
+multiple_streaming_compression
+
+#test artefact
+tmp*
+test*
+*.zst
diff --git a/examples/Makefile b/examples/Makefile
new file mode 100644
index 0000000..7410228
--- /dev/null
+++ b/examples/Makefile
@@ -0,0 +1,68 @@
+# ################################################################
+# Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree. An additional grant
+# of patent rights can be found in the PATENTS file in the same directory.
+# ################################################################
+
+# This Makefile presumes libzstd is installed, using `sudo make install`
+
+LDFLAGS+= -lzstd
+
+.PHONY: default all clean test
+
+default: all
+
+all: simple_compression simple_decompression \
+	dictionary_compression dictionary_decompression \
+	streaming_compression streaming_decompression \
+	multiple_streaming_compression
+
+simple_compression : simple_compression.c
+	$(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@
+
+simple_decompression : simple_decompression.c
+	$(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@
+
+dictionary_compression : dictionary_compression.c
+	$(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@
+
+dictionary_decompression : dictionary_decompression.c
+	$(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@
+
+streaming_compression : streaming_compression.c
+	$(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@
+
+multiple_streaming_compression : multiple_streaming_compression.c
+	$(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@
+
+streaming_decompression : streaming_decompression.c
+	$(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@
+
+clean:
+	@rm -f core *.o tmp* result* *.zst \
+        simple_compression simple_decompression \
+        dictionary_compression dictionary_decompression \
+        streaming_compression streaming_decompression \
+		multiple_streaming_compression
+	@echo Cleaning completed
+
+test: all
+	cp README.md tmp
+	cp Makefile tmp2
+	@echo starting simple compression
+	./simple_compression tmp
+	./simple_decompression tmp.zst
+	./streaming_decompression tmp.zst > /dev/null
+	@echo starting streaming compression
+	./streaming_compression tmp
+	./streaming_decompression tmp.zst > /dev/null
+	@echo starting multiple streaming compression
+	./multiple_streaming_compression *.c
+	@echo starting dictionary compression
+	./dictionary_compression tmp2 tmp README.md
+	./dictionary_decompression tmp2.zst tmp.zst README.md
+	$(RM) tmp* *.zst
+	@echo tests completed
diff --git a/examples/README.md b/examples/README.md
new file mode 100644
index 0000000..8a40443
--- /dev/null
+++ b/examples/README.md
@@ -0,0 +1,36 @@
+Zstandard library : usage examples
+==================================
+
+- [Simple compression](simple_compression.c) :
+  Compress a single file.
+  Introduces usage of : `ZSTD_compress()`
+
+- [Simple decompression](simple_decompression.c) :
+  Decompress a single file.
+  Only compatible with simple compression.
+  Result remains in memory.
+  Introduces usage of : `ZSTD_decompress()`
+
+- [Streaming compression](streaming_compression.c) :
+  Compress a single file.
+  Introduces usage of : `ZSTD_compressStream()`
+
+- [Multiple Streaming compression](multiple_streaming_compression.c) :
+  Compress multiple files in a single command line.
+  Introduces memory usage preservation technique,
+  reducing impact of malloc()/free() and memset() by re-using existing resources.
+
+- [Streaming decompression](streaming_decompression.c) :
+  Decompress a single file compressed by zstd.
+  Compatible with both simple and streaming compression.
+  Result is sent to stdout.
+  Introduces usage of : `ZSTD_decompressStream()`
+
+- [Dictionary compression](dictionary_compression.c) :
+  Compress multiple files using the same dictionary.
+  Introduces usage of : `ZSTD_createCDict()` and `ZSTD_compress_usingCDict()`
+
+- [Dictionary decompression](dictionary_decompression.c) :
+  Decompress multiple files using the same dictionary.
+  Result remains in memory.
+  Introduces usage of : `ZSTD_createDDict()` and `ZSTD_decompress_usingDDict()`
diff --git a/examples/dictionary_compression.c b/examples/dictionary_compression.c
new file mode 100644
index 0000000..adcc3b4
--- /dev/null
+++ b/examples/dictionary_compression.c
@@ -0,0 +1,155 @@
+/**
+ * Copyright 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the license found in the
+ * LICENSE-examples file in the root directory of this source tree.
+ */
+
+
+#include     // malloc, exit
+#include      // printf
+#include     // strerror
+#include      // errno
+#include   // stat
+#include       // presumes zstd library is installed
+
+
+static off_t fsize_orDie(const char *filename)
+{
+    struct stat st;
+    if (stat(filename, &st) == 0) return st.st_size;
+    /* error */
+    perror(filename);
+    exit(1);
+}
+
+static FILE* fopen_orDie(const char *filename, const char *instruction)
+{
+    FILE* const inFile = fopen(filename, instruction);
+    if (inFile) return inFile;
+    /* error */
+    perror(filename);
+    exit(2);
+}
+
+static void* malloc_orDie(size_t size)
+{
+    void* const buff = malloc(size);
+    if (buff) return buff;
+    /* error */
+    perror("malloc");
+    exit(3);
+}
+
+static void* loadFile_orDie(const char* fileName, size_t* size)
+{
+    off_t const buffSize = fsize_orDie(fileName);
+    FILE* const inFile = fopen_orDie(fileName, "rb");
+    void* const buffer = malloc_orDie(buffSize);
+    size_t const readSize = fread(buffer, 1, buffSize, inFile);
+    if (readSize != (size_t)buffSize) {
+        fprintf(stderr, "fread: %s : %s \n", fileName, strerror(errno));
+        exit(4);
+    }
+    fclose(inFile);
+    *size = buffSize;
+    return buffer;
+}
+
+static void saveFile_orDie(const char* fileName, const void* buff, size_t buffSize)
+{
+    FILE* const oFile = fopen_orDie(fileName, "wb");
+    size_t const wSize = fwrite(buff, 1, buffSize, oFile);
+    if (wSize != (size_t)buffSize) {
+        fprintf(stderr, "fwrite: %s : %s \n", fileName, strerror(errno));
+        exit(5);
+    }
+    if (fclose(oFile)) {
+        perror(fileName);
+        exit(6);
+    }
+}
+
+/* createDict() :
+   `dictFileName` is supposed to have been created using `zstd --train` */
+static ZSTD_CDict* createCDict_orDie(const char* dictFileName, int cLevel)
+{
+    size_t dictSize;
+    printf("loading dictionary %s \n", dictFileName);
+    void* const dictBuffer = loadFile_orDie(dictFileName, &dictSize);
+    ZSTD_CDict* const cdict = ZSTD_createCDict(dictBuffer, dictSize, cLevel);
+    if (!cdict) {
+        fprintf(stderr, "ZSTD_createCDict error \n");
+        exit(7);
+    }
+    free(dictBuffer);
+    return cdict;
+}
+
+
+static void compress(const char* fname, const char* oname, const ZSTD_CDict* cdict)
+{
+    size_t fSize;
+    void* const fBuff = loadFile_orDie(fname, &fSize);
+    size_t const cBuffSize = ZSTD_compressBound(fSize);
+    void* const cBuff = malloc_orDie(cBuffSize);
+
+    ZSTD_CCtx* const cctx = ZSTD_createCCtx();
+    if (cctx==NULL) { fprintf(stderr, "ZSTD_createCCtx() error \n"); exit(10); }
+    size_t const cSize = ZSTD_compress_usingCDict(cctx, cBuff, cBuffSize, fBuff, fSize, cdict);
+    if (ZSTD_isError(cSize)) {
+        fprintf(stderr, "error compressing %s : %s \n", fname, ZSTD_getErrorName(cSize));
+        exit(7);
+    }
+
+    saveFile_orDie(oname, cBuff, cSize);
+
+    /* success */
+    printf("%25s : %6u -> %7u - %s \n", fname, (unsigned)fSize, (unsigned)cSize, oname);
+
+    ZSTD_freeCCtx(cctx);   /* never fails */
+    free(fBuff);
+    free(cBuff);
+}
+
+
+static char* createOutFilename_orDie(const char* filename)
+{
+    size_t const inL = strlen(filename);
+    size_t const outL = inL + 5;
+    void* outSpace = malloc_orDie(outL);
+    memset(outSpace, 0, outL);
+    strcat(outSpace, filename);
+    strcat(outSpace, ".zst");
+    return (char*)outSpace;
+}
+
+int main(int argc, const char** argv)
+{
+    const char* const exeName = argv[0];
+    int const cLevel = 3;
+
+    if (argc<3) {
+        fprintf(stderr, "wrong arguments\n");
+        fprintf(stderr, "usage:\n");
+        fprintf(stderr, "%s [FILES] dictionary\n", exeName);
+        return 1;
+    }
+
+    /* load dictionary only once */
+    const char* const dictName = argv[argc-1];
+    ZSTD_CDict* const dictPtr = createCDict_orDie(dictName, cLevel);
+
+    int u;
+    for (u=1; u    // malloc, exit
+#include      // printf
+#include     // strerror
+#include      // errno
+#include   // stat
+#include       // presumes zstd library is installed
+
+
+static off_t fsize_orDie(const char *filename)
+{
+    struct stat st;
+    if (stat(filename, &st) == 0) return st.st_size;
+    /* error */
+    perror(filename);
+    exit(1);
+}
+
+static FILE* fopen_orDie(const char *filename, const char *instruction)
+{
+    FILE* const inFile = fopen(filename, instruction);
+    if (inFile) return inFile;
+    /* error */
+    perror(filename);
+    exit(2);
+}
+
+static void* malloc_orDie(size_t size)
+{
+    void* const buff = malloc(size);
+    if (buff) return buff;
+    /* error */
+    perror("malloc");
+    exit(3);
+}
+
+static void* loadFile_orDie(const char* fileName, size_t* size)
+{
+    off_t const buffSize = fsize_orDie(fileName);
+    FILE* const inFile = fopen_orDie(fileName, "rb");
+    void* const buffer = malloc_orDie(buffSize);
+    size_t const readSize = fread(buffer, 1, buffSize, inFile);
+    if (readSize != (size_t)buffSize) {
+        fprintf(stderr, "fread: %s : %s \n", fileName, strerror(errno));
+        exit(4);
+    }
+    fclose(inFile);
+    *size = buffSize;
+    return buffer;
+}
+
+/* createDict() :
+   `dictFileName` is supposed to have been created using `zstd --train` */
+static ZSTD_DDict* createDict_orDie(const char* dictFileName)
+{
+    size_t dictSize;
+    printf("loading dictionary %s \n", dictFileName);
+    void* const dictBuffer = loadFile_orDie(dictFileName, &dictSize);
+    ZSTD_DDict* const ddict = ZSTD_createDDict(dictBuffer, dictSize);
+    if (ddict==NULL) { fprintf(stderr, "ZSTD_createDDict error \n"); exit(5); }
+    free(dictBuffer);
+    return ddict;
+}
+
+
+static void decompress(const char* fname, const ZSTD_DDict* ddict)
+{
+    size_t cSize;
+    void* const cBuff = loadFile_orDie(fname, &cSize);
+    unsigned long long const rSize = ZSTD_getDecompressedSize(cBuff, cSize);
+    if (rSize==0) {
+        fprintf(stderr, "%s : original size unknown \n", fname);
+        exit(6);
+    }
+    void* const rBuff = malloc_orDie((size_t)rSize);
+
+    ZSTD_DCtx* const dctx = ZSTD_createDCtx();
+    if (dctx==NULL) { fprintf(stderr, "ZSTD_createDCtx() error \n"); exit(10); }
+    size_t const dSize = ZSTD_decompress_usingDDict(dctx, rBuff, rSize, cBuff, cSize, ddict);
+    if (dSize != rSize) {
+        fprintf(stderr, "error decoding %s : %s \n", fname, ZSTD_getErrorName(dSize));
+        exit(7);
+    }
+
+    /* success */
+    printf("%25s : %6u -> %7u \n", fname, (unsigned)cSize, (unsigned)rSize);
+
+    ZSTD_freeDCtx(dctx);
+    free(rBuff);
+    free(cBuff);
+}
+
+
+int main(int argc, const char** argv)
+{
+    const char* const exeName = argv[0];
+
+    if (argc<3) {
+        printf("wrong arguments\n");
+        printf("usage:\n");
+        printf("%s [FILES] dictionary\n", exeName);
+        return 1;
+    }
+
+    /* load dictionary only once */
+    const char* const dictName = argv[argc-1];
+    ZSTD_DDict* const dictPtr = createDict_orDie(dictName);
+
+    int u;
+    for (u=1; u    // malloc, exit
+#include      // fprintf, perror, feof
+#include     // strerror
+#include      // errno
+#define ZSTD_STATIC_LINKING_ONLY  // streaming API defined as "experimental" for the time being
+#include       // presumes zstd library is installed
+
+
+static void* malloc_orDie(size_t size)
+{
+    void* const buff = malloc(size);
+    if (buff) return buff;
+    /* error */
+    perror("malloc:");
+    exit(1);
+}
+
+static FILE* fopen_orDie(const char *filename, const char *instruction)
+{
+    FILE* const inFile = fopen(filename, instruction);
+    if (inFile) return inFile;
+    /* error */
+    perror(filename);
+    exit(3);
+}
+
+static size_t fread_orDie(void* buffer, size_t sizeToRead, FILE* file)
+{
+    size_t const readSize = fread(buffer, 1, sizeToRead, file);
+    if (readSize == sizeToRead) return readSize;   /* good */
+    if (feof(file)) return readSize;   /* good, reached end of file */
+    /* error */
+    perror("fread");
+    exit(4);
+}
+
+static size_t fwrite_orDie(const void* buffer, size_t sizeToWrite, FILE* file)
+{
+    size_t const writtenSize = fwrite(buffer, 1, sizeToWrite, file);
+    if (writtenSize == sizeToWrite) return sizeToWrite;   /* good */
+    /* error */
+    perror("fwrite");
+    exit(5);
+}
+
+static size_t fclose_orDie(FILE* file)
+{
+    if (!fclose(file)) return 0;
+    /* error */
+    perror("fclose");
+    exit(6);
+}
+
+
+typedef struct {
+    void* buffIn;
+    void* buffOut;
+    size_t buffInSize;
+    size_t buffOutSize;
+    ZSTD_CStream* cstream;
+} resources ;
+
+static resources createResources_orDie()
+{
+    resources ress;
+    ress.buffInSize = ZSTD_CStreamInSize();   /* can always read one full block */
+    ress.buffOutSize= ZSTD_CStreamOutSize();  /* can always flush a full block */
+    ress.buffIn = malloc_orDie(ress.buffInSize);
+    ress.buffOut= malloc_orDie(ress.buffOutSize);
+    ress.cstream = ZSTD_createCStream();
+    if (ress.cstream==NULL) { fprintf(stderr, "ZSTD_createCStream() error \n"); exit(10); }
+    return ress;
+}
+
+static void freeResources(resources ress)
+{
+    ZSTD_freeCStream(ress.cstream);
+    free(ress.buffIn);
+    free(ress.buffOut);
+}
+
+
+static void compressFile_orDie(resources ress, const char* fname, const char* outName, int cLevel)
+{
+    FILE* const fin  = fopen_orDie(fname, "rb");
+    FILE* const fout = fopen_orDie(outName, "wb");
+
+    size_t const initResult = ZSTD_initCStream(ress.cstream, cLevel);
+    if (ZSTD_isError(initResult)) { fprintf(stderr, "ZSTD_initCStream() error : %s \n", ZSTD_getErrorName(initResult)); exit(11); }
+
+    size_t read, toRead = ress.buffInSize;
+    while( (read = fread_orDie(ress.buffIn, toRead, fin)) ) {
+        ZSTD_inBuffer input = { ress.buffIn, read, 0 };
+        while (input.pos < input.size) {
+            ZSTD_outBuffer output = { ress.buffOut, ress.buffOutSize, 0 };
+            toRead = ZSTD_compressStream(ress.cstream, &output , &input);   /* toRead is guaranteed to be <= ZSTD_CStreamInSize() */
+            if (ZSTD_isError(toRead)) { fprintf(stderr, "ZSTD_compressStream() error : %s \n", ZSTD_getErrorName(toRead)); exit(12); }
+            if (toRead > ress.buffInSize) toRead = ress.buffInSize;   /* Safely handle when `buffInSize` is manually changed to a smaller value */
+            fwrite_orDie(ress.buffOut, output.pos, fout);
+        }
+    }
+
+    ZSTD_outBuffer output = { ress.buffOut, ress.buffOutSize, 0 };
+    size_t const remainingToFlush = ZSTD_endStream(ress.cstream, &output);   /* close frame */
+    if (remainingToFlush) { fprintf(stderr, "not fully flushed"); exit(13); }
+    fwrite_orDie(ress.buffOut, output.pos, fout);
+
+    fclose_orDie(fout);
+    fclose_orDie(fin);
+}
+
+
+int main(int argc, const char** argv)
+{
+    const char* const exeName = argv[0];
+
+    if (argc<2) {
+        printf("wrong arguments\n");
+        printf("usage:\n");
+        printf("%s FILE(s)\n", exeName);
+        return 1;
+    }
+
+    resources const ress = createResources_orDie();
+    void* ofnBuffer = NULL;
+    size_t ofnbSize = 0;
+
+    int argNb;
+    for (argNb = 1; argNb < argc; argNb++) {
+        const char* const ifn = argv[argNb];
+        size_t const ifnSize = strlen(ifn);
+        size_t const ofnSize = ifnSize + 5;
+        if (ofnbSize <= ofnSize) {
+            ofnbSize = ofnSize + 16;
+            free(ofnBuffer);
+            ofnBuffer = malloc_orDie(ofnbSize);
+        }
+        memset(ofnBuffer, 0, ofnSize);
+        strcat(ofnBuffer, ifn);
+        strcat(ofnBuffer, ".zst");
+        compressFile_orDie(ress, ifn, ofnBuffer, 7);
+    }
+
+    freeResources(ress);
+    /* success */
+    printf("compressed %i files \n", argc-1);
+
+    return 0;
+}
diff --git a/examples/simple_compression.c b/examples/simple_compression.c
new file mode 100644
index 0000000..deb0bbf
--- /dev/null
+++ b/examples/simple_compression.c
@@ -0,0 +1,132 @@
+/**
+ * Copyright 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the license found in the
+ * LICENSE-examples file in the root directory of this source tree.
+ */
+
+
+
+#include     // malloc, free, exit
+#include      // fprintf, perror, fopen, etc.
+#include     // strlen, strcat, memset, strerror
+#include      // errno
+#include   // stat
+#include       // presumes zstd library is installed
+
+
+static off_t fsize_orDie(const char *filename)
+{
+    struct stat st;
+    if (stat(filename, &st) == 0) return st.st_size;
+    /* error */
+    perror(filename);
+    exit(1);
+}
+
+static FILE* fopen_orDie(const char *filename, const char *instruction)
+{
+    FILE* const inFile = fopen(filename, instruction);
+    if (inFile) return inFile;
+    /* error */
+    perror(filename);
+    exit(2);
+}
+
+static void* malloc_orDie(size_t size)
+{
+    void* const buff = malloc(size);
+    if (buff) return buff;
+    /* error */
+    perror(NULL);
+    exit(3);
+}
+
+static void* loadFile_orDie(const char* fileName, size_t* size)
+{
+    off_t const fileSize = fsize_orDie(fileName);
+    size_t const buffSize = (size_t)fileSize;
+    if ((off_t)buffSize < fileSize) {   /* narrowcast overflow */
+        fprintf(stderr, "%s : filesize too large \n", fileName);
+        exit(4);
+    }
+    FILE* const inFile = fopen_orDie(fileName, "rb");
+    void* const buffer = malloc_orDie(buffSize);
+    size_t const readSize = fread(buffer, 1, buffSize, inFile);
+    if (readSize != (size_t)buffSize) {
+        fprintf(stderr, "fread: %s : %s \n", fileName, strerror(errno));
+        exit(5);
+    }
+    fclose(inFile);  /* can't fail, read only */
+    *size = buffSize;
+    return buffer;
+}
+
+
+static void saveFile_orDie(const char* fileName, const void* buff, size_t buffSize)
+{
+    FILE* const oFile = fopen_orDie(fileName, "wb");
+    size_t const wSize = fwrite(buff, 1, buffSize, oFile);
+    if (wSize != (size_t)buffSize) {
+        fprintf(stderr, "fwrite: %s : %s \n", fileName, strerror(errno));
+        exit(6);
+    }
+    if (fclose(oFile)) {
+        perror(fileName);
+        exit(7);
+    }
+}
+
+
+static void compress_orDie(const char* fname, const char* oname)
+{
+    size_t fSize;
+    void* const fBuff = loadFile_orDie(fname, &fSize);
+    size_t const cBuffSize = ZSTD_compressBound(fSize);
+    void* const cBuff = malloc_orDie(cBuffSize);
+
+    size_t const cSize = ZSTD_compress(cBuff, cBuffSize, fBuff, fSize, 1);
+    if (ZSTD_isError(cSize)) {
+        fprintf(stderr, "error compressing %s : %s \n", fname, ZSTD_getErrorName(cSize));
+        exit(8);
+    }
+
+    saveFile_orDie(oname, cBuff, cSize);
+
+    /* success */
+    printf("%25s : %6u -> %7u - %s \n", fname, (unsigned)fSize, (unsigned)cSize, oname);
+
+    free(fBuff);
+    free(cBuff);
+}
+
+
+static const char* createOutFilename_orDie(const char* filename)
+{
+    size_t const inL = strlen(filename);
+    size_t const outL = inL + 5;
+    void* const outSpace = malloc_orDie(outL);
+    memset(outSpace, 0, outL);
+    strcat(outSpace, filename);
+    strcat(outSpace, ".zst");
+    return (const char*)outSpace;
+}
+
+int main(int argc, const char** argv)
+{
+    const char* const exeName = argv[0];
+    const char* const inFilename = argv[1];
+
+    if (argc!=2) {
+        printf("wrong arguments\n");
+        printf("usage:\n");
+        printf("%s FILE\n", exeName);
+        return 1;
+    }
+
+    const char* const outFilename = createOutFilename_orDie(inFilename);
+    compress_orDie(inFilename, outFilename);
+
+    return 0;
+}
diff --git a/examples/simple_decompression.c b/examples/simple_decompression.c
new file mode 100644
index 0000000..62a881f
--- /dev/null
+++ b/examples/simple_decompression.c
@@ -0,0 +1,104 @@
+/**
+ * Copyright 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the license found in the
+ * LICENSE-examples file in the root directory of this source tree.
+ */
+
+
+
+#include     // malloc, exit
+#include      // printf
+#include     // strerror
+#include      // errno
+#include   // stat
+#include       // presumes zstd library is installed
+
+
+static off_t fsize_X(const char *filename)
+{
+    struct stat st;
+    if (stat(filename, &st) == 0) return st.st_size;
+    /* error */
+    printf("stat: %s : %s \n", filename, strerror(errno));
+    exit(1);
+}
+
+static FILE* fopen_X(const char *filename, const char *instruction)
+{
+    FILE* const inFile = fopen(filename, instruction);
+    if (inFile) return inFile;
+    /* error */
+    printf("fopen: %s : %s \n", filename, strerror(errno));
+    exit(2);
+}
+
+static void* malloc_X(size_t size)
+{
+    void* const buff = malloc(size);
+    if (buff) return buff;
+    /* error */
+    printf("malloc: %s \n", strerror(errno));
+    exit(3);
+}
+
+static void* loadFile_X(const char* fileName, size_t* size)
+{
+    off_t const buffSize = fsize_X(fileName);
+    FILE* const inFile = fopen_X(fileName, "rb");
+    void* const buffer = malloc_X(buffSize);
+    size_t const readSize = fread(buffer, 1, buffSize, inFile);
+    if (readSize != (size_t)buffSize) {
+        printf("fread: %s : %s \n", fileName, strerror(errno));
+        exit(4);
+    }
+    fclose(inFile);   /* can't fail (read only) */
+    *size = buffSize;
+    return buffer;
+}
+
+
+static void decompress(const char* fname)
+{
+    size_t cSize;
+    void* const cBuff = loadFile_X(fname, &cSize);
+    unsigned long long const rSize = ZSTD_getDecompressedSize(cBuff, cSize);
+    if (rSize==0) {
+        printf("%s : original size unknown. Use streaming decompression instead. \n", fname);
+        exit(5);
+    }
+    void* const rBuff = malloc_X((size_t)rSize);
+
+    size_t const dSize = ZSTD_decompress(rBuff, rSize, cBuff, cSize);
+
+    if (dSize != rSize) {
+        printf("error decoding %s : %s \n", fname, ZSTD_getErrorName(dSize));
+        exit(7);
+    }
+
+    /* success */
+    printf("%25s : %6u -> %7u \n", fname, (unsigned)cSize, (unsigned)rSize);
+
+    free(rBuff);
+    free(cBuff);
+}
+
+
+int main(int argc, const char** argv)
+{
+    const char* const exeName = argv[0];
+
+    if (argc!=2) {
+        printf("wrong arguments\n");
+        printf("usage:\n");
+        printf("%s FILE\n", exeName);
+        return 1;
+    }
+
+    decompress(argv[1]);
+
+    printf("%s correctly decoded (in memory). \n", argv[1]);
+
+    return 0;
+}
diff --git a/examples/streaming_compression.c b/examples/streaming_compression.c
new file mode 100644
index 0000000..4c2c1a1
--- /dev/null
+++ b/examples/streaming_compression.c
@@ -0,0 +1,128 @@
+/**
+ * Copyright 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the license found in the
+ * LICENSE-examples file in the root directory of this source tree.
+ */
+
+
+#include     // malloc, free, exit
+#include      // fprintf, perror, feof, fopen, etc.
+#include     // strlen, memset, strcat
+#include       // presumes zstd library is installed
+
+
+static void* malloc_orDie(size_t size)
+{
+    void* const buff = malloc(size);
+    if (buff) return buff;
+    /* error */
+    perror("malloc:");
+    exit(1);
+}
+
+static FILE* fopen_orDie(const char *filename, const char *instruction)
+{
+    FILE* const inFile = fopen(filename, instruction);
+    if (inFile) return inFile;
+    /* error */
+    perror(filename);
+    exit(3);
+}
+
+static size_t fread_orDie(void* buffer, size_t sizeToRead, FILE* file)
+{
+    size_t const readSize = fread(buffer, 1, sizeToRead, file);
+    if (readSize == sizeToRead) return readSize;   /* good */
+    if (feof(file)) return readSize;   /* good, reached end of file */
+    /* error */
+    perror("fread");
+    exit(4);
+}
+
+static size_t fwrite_orDie(const void* buffer, size_t sizeToWrite, FILE* file)
+{
+    size_t const writtenSize = fwrite(buffer, 1, sizeToWrite, file);
+    if (writtenSize == sizeToWrite) return sizeToWrite;   /* good */
+    /* error */
+    perror("fwrite");
+    exit(5);
+}
+
+static size_t fclose_orDie(FILE* file)
+{
+    if (!fclose(file)) return 0;
+    /* error */
+    perror("fclose");
+    exit(6);
+}
+
+
+static void compressFile_orDie(const char* fname, const char* outName, int cLevel)
+{
+    FILE* const fin  = fopen_orDie(fname, "rb");
+    FILE* const fout = fopen_orDie(outName, "wb");
+    size_t const buffInSize = ZSTD_CStreamInSize();    /* can always read one full block */
+    void*  const buffIn  = malloc_orDie(buffInSize);
+    size_t const buffOutSize = ZSTD_CStreamOutSize();  /* can always flush a full block */
+    void*  const buffOut = malloc_orDie(buffOutSize);
+
+    ZSTD_CStream* const cstream = ZSTD_createCStream();
+    if (cstream==NULL) { fprintf(stderr, "ZSTD_createCStream() error \n"); exit(10); }
+    size_t const initResult = ZSTD_initCStream(cstream, cLevel);
+    if (ZSTD_isError(initResult)) { fprintf(stderr, "ZSTD_initCStream() error : %s \n", ZSTD_getErrorName(initResult)); exit(11); }
+
+    size_t read, toRead = buffInSize;
+    while( (read = fread_orDie(buffIn, toRead, fin)) ) {
+        ZSTD_inBuffer input = { buffIn, read, 0 };
+        while (input.pos < input.size) {
+            ZSTD_outBuffer output = { buffOut, buffOutSize, 0 };
+            toRead = ZSTD_compressStream(cstream, &output , &input);   /* toRead is guaranteed to be <= ZSTD_CStreamInSize() */
+            if (ZSTD_isError(toRead)) { fprintf(stderr, "ZSTD_compressStream() error : %s \n", ZSTD_getErrorName(toRead)); exit(12); }
+            if (toRead > buffInSize) toRead = buffInSize;   /* Safely handle case when `buffInSize` is manually changed to a value < ZSTD_CStreamInSize()*/
+            fwrite_orDie(buffOut, output.pos, fout);
+        }
+    }
+
+    ZSTD_outBuffer output = { buffOut, buffOutSize, 0 };
+    size_t const remainingToFlush = ZSTD_endStream(cstream, &output);   /* close frame */
+    if (remainingToFlush) { fprintf(stderr, "not fully flushed"); exit(13); }
+    fwrite_orDie(buffOut, output.pos, fout);
+
+    ZSTD_freeCStream(cstream);
+    fclose_orDie(fout);
+    fclose_orDie(fin);
+    free(buffIn);
+    free(buffOut);
+}
+
+
+static const char* createOutFilename_orDie(const char* filename)
+{
+    size_t const inL = strlen(filename);
+    size_t const outL = inL + 5;
+    void* outSpace = malloc_orDie(outL);
+    memset(outSpace, 0, outL);
+    strcat(outSpace, filename);
+    strcat(outSpace, ".zst");
+    return (const char*)outSpace;
+}
+
+int main(int argc, const char** argv)
+{
+    const char* const exeName = argv[0];
+    const char* const inFilename = argv[1];
+
+    if (argc!=2) {
+        printf("wrong arguments\n");
+        printf("usage:\n");
+        printf("%s FILE\n", exeName);
+        return 1;
+    }
+
+    const char* const outFilename = createOutFilename_orDie(inFilename);
+    compressFile_orDie(inFilename, outFilename, 1);
+
+    return 0;
+}
diff --git a/examples/streaming_decompression.c b/examples/streaming_decompression.c
new file mode 100644
index 0000000..400aa67
--- /dev/null
+++ b/examples/streaming_decompression.c
@@ -0,0 +1,113 @@
+/**
+ * Copyright 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the license found in the
+ * LICENSE-examples file in the root directory of this source tree.
+ */
+
+
+#include     // malloc, exit
+#include      // fprintf, perror, feof
+#include     // strerror
+#include      // errno
+#include       // presumes zstd library is installed
+
+
+static void* malloc_orDie(size_t size)
+{
+    void* const buff = malloc(size);
+    if (buff) return buff;
+    /* error */
+    perror("malloc:");
+    exit(1);
+}
+
+static FILE* fopen_orDie(const char *filename, const char *instruction)
+{
+    FILE* const inFile = fopen(filename, instruction);
+    if (inFile) return inFile;
+    /* error */
+    perror(filename);
+    exit(3);
+}
+
+static size_t fread_orDie(void* buffer, size_t sizeToRead, FILE* file)
+{
+    size_t const readSize = fread(buffer, 1, sizeToRead, file);
+    if (readSize == sizeToRead) return readSize;   /* good */
+    if (feof(file)) return readSize;   /* good, reached end of file */
+    /* error */
+    perror("fread");
+    exit(4);
+}
+
+static size_t fwrite_orDie(const void* buffer, size_t sizeToWrite, FILE* file)
+{
+    size_t const writtenSize = fwrite(buffer, 1, sizeToWrite, file);
+    if (writtenSize == sizeToWrite) return sizeToWrite;   /* good */
+    /* error */
+    perror("fwrite");
+    exit(5);
+}
+
+static size_t fclose_orDie(FILE* file)
+{
+    if (!fclose(file)) return 0;
+    /* error */
+    perror("fclose");
+    exit(6);
+}
+
+
+static void decompressFile_orDie(const char* fname)
+{
+    FILE* const fin  = fopen_orDie(fname, "rb");
+    size_t const buffInSize = ZSTD_DStreamInSize();
+    void*  const buffIn  = malloc_orDie(buffInSize);
+    FILE* const fout = stdout;
+    size_t const buffOutSize = ZSTD_DStreamOutSize();  /* Guarantee to successfully flush at least one complete compressed block in all circumstances. */
+    void*  const buffOut = malloc_orDie(buffOutSize);
+
+    ZSTD_DStream* const dstream = ZSTD_createDStream();
+    if (dstream==NULL) { fprintf(stderr, "ZSTD_createDStream() error \n"); exit(10); }
+
+    /* In more complex scenarios, a file may consist of multiple appended frames (ex : pzstd).
+    *  The following example decompresses only the first frame.
+    *  It is compatible with other provided streaming examples */
+    size_t const initResult = ZSTD_initDStream(dstream);
+    if (ZSTD_isError(initResult)) { fprintf(stderr, "ZSTD_initDStream() error : %s \n", ZSTD_getErrorName(initResult)); exit(11); }
+    size_t read, toRead = initResult;
+    while ( (read = fread_orDie(buffIn, toRead, fin)) ) {
+        ZSTD_inBuffer input = { buffIn, read, 0 };
+        while (input.pos < input.size) {
+            ZSTD_outBuffer output = { buffOut, buffOutSize, 0 };
+            toRead = ZSTD_decompressStream(dstream, &output , &input);  /* toRead : size of next compressed block */
+            if (ZSTD_isError(toRead)) { fprintf(stderr, "ZSTD_decompressStream() error : %s \n", ZSTD_getErrorName(toRead)); exit(12); }
+            fwrite_orDie(buffOut, output.pos, fout);
+        }
+    }
+
+    ZSTD_freeDStream(dstream);
+    fclose_orDie(fin);
+    fclose_orDie(fout);
+    free(buffIn);
+    free(buffOut);
+}
+
+
+int main(int argc, const char** argv)
+{
+    const char* const exeName = argv[0];
+    const char* const inFilename = argv[1];
+
+    if (argc!=2) {
+        fprintf(stderr, "wrong arguments\n");
+        fprintf(stderr, "usage:\n");
+        fprintf(stderr, "%s FILE\n", exeName);
+        return 1;
+    }
+
+    decompressFile_orDie(inFilename);
+    return 0;
+}
diff --git a/lib/.gitignore b/lib/.gitignore
new file mode 100644
index 0000000..b43a854
--- /dev/null
+++ b/lib/.gitignore
@@ -0,0 +1,2 @@
+# make install artefact
+libzstd.pc
diff --git a/lib/Makefile b/lib/Makefile
new file mode 100644
index 0000000..fcc0d09
--- /dev/null
+++ b/lib/Makefile
@@ -0,0 +1,133 @@
+# ################################################################
+# Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree. An additional grant
+# of patent rights can be found in the PATENTS file in the same directory.
+# ################################################################
+
+# Version numbers
+LIBVER_MAJOR_SCRIPT:=`sed -n '/define ZSTD_VERSION_MAJOR/s/.*[[:blank:]]\([0-9][0-9]*\).*/\1/p' < ./zstd.h`
+LIBVER_MINOR_SCRIPT:=`sed -n '/define ZSTD_VERSION_MINOR/s/.*[[:blank:]]\([0-9][0-9]*\).*/\1/p' < ./zstd.h`
+LIBVER_PATCH_SCRIPT:=`sed -n '/define ZSTD_VERSION_RELEASE/s/.*[[:blank:]]\([0-9][0-9]*\).*/\1/p' < ./zstd.h`
+LIBVER_SCRIPT:= $(LIBVER_MAJOR_SCRIPT).$(LIBVER_MINOR_SCRIPT).$(LIBVER_PATCH_SCRIPT)
+LIBVER_MAJOR := $(shell echo $(LIBVER_MAJOR_SCRIPT))
+LIBVER_MINOR := $(shell echo $(LIBVER_MINOR_SCRIPT))
+LIBVER_PATCH := $(shell echo $(LIBVER_PATCH_SCRIPT))
+LIBVER := $(shell echo $(LIBVER_SCRIPT))
+VERSION?= $(LIBVER)
+
+DESTDIR?=
+PREFIX ?= /usr/local
+LIBDIR ?= $(PREFIX)/lib
+INCLUDEDIR=$(PREFIX)/include
+
+CPPFLAGS+= -I. -I./common -DXXH_NAMESPACE=ZSTD_
+CFLAGS  ?= -O3
+CFLAGS  += -Wall -Wextra -Wcast-qual -Wcast-align -Wshadow -Wstrict-aliasing=1 \
+           -Wswitch-enum -Wdeclaration-after-statement -Wstrict-prototypes -Wundef \
+           -Wpointer-arith
+CFLAGS  += $(MOREFLAGS)
+FLAGS    = $(CPPFLAGS) $(CFLAGS)
+
+
+ZSTD_FILES := $(wildcard common/*.c compress/*.c decompress/*.c dictBuilder/*.c deprecated/*.c)
+
+ifeq ($(ZSTD_LEGACY_SUPPORT), 0)
+CPPFLAGS  += -DZSTD_LEGACY_SUPPORT=0
+else
+CPPFLAGS  += -I./legacy -DZSTD_LEGACY_SUPPORT=1
+ZSTD_FILES+= $(wildcard legacy/*.c)
+endif
+
+# OS X linker doesn't support -soname, and use different extension
+# see : https://developer.apple.com/library/mac/documentation/DeveloperTools/Conceptual/DynamicLibraries/100-Articles/DynamicLibraryDesignGuidelines.html
+ifeq ($(shell uname), Darwin)
+	SHARED_EXT = dylib
+	SHARED_EXT_MAJOR = $(LIBVER_MAJOR).$(SHARED_EXT)
+	SHARED_EXT_VER = $(LIBVER).$(SHARED_EXT)
+	SONAME_FLAGS = -install_name $(LIBDIR)/libzstd.$(SHARED_EXT_MAJOR) -compatibility_version $(LIBVER_MAJOR) -current_version $(LIBVER)
+else
+	SONAME_FLAGS = -Wl,-soname=libzstd.$(SHARED_EXT).$(LIBVER_MAJOR)
+	SHARED_EXT = so
+	SHARED_EXT_MAJOR = $(SHARED_EXT).$(LIBVER_MAJOR)
+	SHARED_EXT_VER = $(SHARED_EXT).$(LIBVER)
+endif
+
+LIBZSTD = libzstd.$(SHARED_EXT_VER)
+
+
+.PHONY: default all clean install uninstall
+
+default: lib
+
+all: lib
+
+libzstd.a: ARFLAGS = rcs
+libzstd.a: $(ZSTD_FILES)
+	@echo compiling static library
+	@$(CC) $(FLAGS) -c $^
+	@$(AR) $(ARFLAGS) $@ *.o
+
+$(LIBZSTD): LDFLAGS += -shared -fPIC -fvisibility=hidden
+$(LIBZSTD): $(ZSTD_FILES)
+	@echo compiling dynamic library $(LIBVER)
+ifneq (,$(filter Windows%,$(OS)))
+	@$(CC) $(FLAGS) -DZSTD_DLL_EXPORT=1 -shared $^ -o dll\libzstd.dll
+	dlltool -D dll\libzstd.dll -d dll\libzstd.def -l dll\libzstd.lib
+else
+	@$(CC) $(FLAGS) $^ $(LDFLAGS) $(SONAME_FLAGS) -o $@
+	@echo creating versioned links
+	@ln -sf $@ libzstd.$(SHARED_EXT_MAJOR)
+	@ln -sf $@ libzstd.$(SHARED_EXT)
+endif
+
+libzstd : $(LIBZSTD)
+
+lib: libzstd.a libzstd
+
+clean:
+	@$(RM) core *.o *.a *.gcda *.$(SHARED_EXT) *.$(SHARED_EXT).* libzstd.pc dll/libzstd.dll dll/libzstd.lib
+	@$(RM) decompress/*.o
+	@echo Cleaning library completed
+
+#------------------------------------------------------------------------
+#make install is validated only for Linux, OSX, kFreeBSD, Hurd and some BSD targets
+ifneq (,$(filter $(shell uname),Linux Darwin GNU/kFreeBSD GNU FreeBSD DragonFly NetBSD))
+
+libzstd.pc:
+libzstd.pc: libzstd.pc.in
+	@echo creating pkgconfig
+	@sed -e 's|@PREFIX@|$(PREFIX)|' \
+             -e 's|@LIBDIR@|$(LIBDIR)|' \
+             -e 's|@INCLUDEDIR@|$(INCLUDEDIR)|' \
+             -e 's|@VERSION@|$(VERSION)|' \
+             $< >$@
+
+install: libzstd.a libzstd libzstd.pc
+	@install -d -m 755 $(DESTDIR)$(LIBDIR)/pkgconfig/ $(DESTDIR)$(INCLUDEDIR)/
+	@install -m 755 libzstd.$(SHARED_EXT_VER) $(DESTDIR)$(LIBDIR)
+	@cp -a libzstd.$(SHARED_EXT_MAJOR) $(DESTDIR)$(LIBDIR)
+	@cp -a libzstd.$(SHARED_EXT) $(DESTDIR)$(LIBDIR)
+	@cp -a libzstd.pc $(DESTDIR)$(LIBDIR)/pkgconfig/
+	@install -m 644 libzstd.a $(DESTDIR)$(LIBDIR)
+	@install -m 644 zstd.h $(DESTDIR)$(INCLUDEDIR)
+	@install -m 644 common/zstd_errors.h $(DESTDIR)$(INCLUDEDIR)
+	@install -m 644 deprecated/zbuff.h $(DESTDIR)$(INCLUDEDIR)     # prototypes generate deprecation warnings
+	@install -m 644 dictBuilder/zdict.h $(DESTDIR)$(INCLUDEDIR)
+	@echo zstd static and shared library installed
+
+uninstall:
+	@$(RM) $(DESTDIR)$(LIBDIR)/libzstd.a
+	@$(RM) $(DESTDIR)$(LIBDIR)/libzstd.$(SHARED_EXT)
+	@$(RM) $(DESTDIR)$(LIBDIR)/libzstd.$(SHARED_EXT_MAJOR)
+	@$(RM) $(DESTDIR)$(LIBDIR)/libzstd.$(SHARED_EXT_VER)
+	@$(RM) $(DESTDIR)$(LIBDIR)/pkgconfig/libzstd.pc
+	@$(RM) $(DESTDIR)$(INCLUDEDIR)/zstd.h
+	@$(RM) $(DESTDIR)$(INCLUDEDIR)/zstd_errors.h
+	@$(RM) $(DESTDIR)$(INCLUDEDIR)/zbuff.h   # Deprecated streaming functions
+	@$(RM) $(DESTDIR)$(INCLUDEDIR)/zdict.h
+	@echo zstd libraries successfully uninstalled
+
+endif
diff --git a/lib/README.md b/lib/README.md
new file mode 100644
index 0000000..3357e3d
--- /dev/null
+++ b/lib/README.md
@@ -0,0 +1,77 @@
+Zstandard library files
+================================
+
+The __lib__ directory contains several directories.
+Depending on target use case, it's enough to include only files from relevant directories.
+
+
+#### API
+
+Zstandard's stable API is exposed within [zstd.h](zstd.h),
+at the root of `lib` directory.
+
+
+#### Advanced API
+
+Some additional API may be useful if you're looking into advanced features :
+- common/error_public.h : transforms `size_t` function results into an `enum`,
+                          for precise error handling.
+- ZSTD_STATIC_LINKING_ONLY : if you define this macro _before_ including `zstd.h`,
+                          it will give access to advanced and experimental API.
+                          These APIs shall ___never be used with dynamic library___ !
+                          They are not "stable", their definition may change in the future.
+                          Only static linking is allowed.
+
+
+#### Modular build
+
+Directory `common/` is required in all circumstances.
+You can select to support compression only, by just adding files from the `compress/` directory,
+In a similar way, you can build a decompressor-only library with the `decompress/` directory.
+
+Other optional functionalities provided are :
+
+- `dictBuilder/`  : source files to create dictionaries.
+                    The API can be consulted in `dictBuilder/zdict.h`.
+                    This module also depends on `common/` and `compress/` .
+
+- `legacy/` : source code to decompress previous versions of zstd, starting from `v0.1`.
+              This module also depends on `common/` and `decompress/` .
+              Library compilation must include directive `ZSTD_LEGACY_SUPPORT = 1` .
+              The main API can be consulted in `legacy/zstd_legacy.h`.
+              Advanced API from each version can be found in their relevant header file.
+              For example, advanced API for version `v0.4` is in `legacy/zstd_v04.h` .
+
+
+#### Using MinGW+MSYS to create DLL
+
+DLL can be created using MinGW+MSYS with the `make libzstd` command.
+This command creates `dll\libzstd.dll` and the import library `dll\libzstd.lib`.
+The import library is only required with Visual C++.
+The header file `zstd.h` and the dynamic library `dll\libzstd.dll` are required to
+compile a project using gcc/MinGW.
+The dynamic library has to be added to linking options.
+It means that if a project that uses ZSTD consists of a single `test-dll.c`
+file it should be linked with `dll\libzstd.dll`. For example:
+```
+    gcc $(CFLAGS) -Iinclude/ test-dll.c -o test-dll dll\libzstd.dll
+```
+The compiled executable will require ZSTD DLL which is available at `dll\libzstd.dll`.
+
+
+#### Obsolete streaming API
+
+Streaming is now provided within `zstd.h`.
+Older streaming API is still available within `deprecated/zbuff.h`.
+It will be removed in a future version.
+Consider migrating code towards newer streaming API in `zstd.h`.
+
+
+#### Miscellaneous
+
+The other files are not source code. There are :
+
+ - LICENSE : contains the BSD license text
+ - Makefile : script to compile or install zstd library (static and dynamic)
+ - libzstd.pc.in : for pkg-config (`make install`)
+ - README.md : this file
diff --git a/lib/common/bitstream.h b/lib/common/bitstream.h
new file mode 100644
index 0000000..3a45244
--- /dev/null
+++ b/lib/common/bitstream.h
@@ -0,0 +1,414 @@
+/* ******************************************************************
+   bitstream
+   Part of FSE library
+   header file (to include)
+   Copyright (C) 2013-2016, Yann Collet.
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+   You can contact the author at :
+   - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+****************************************************************** */
+#ifndef BITSTREAM_H_MODULE
+#define BITSTREAM_H_MODULE
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+
+/*
+*  This API consists of small unitary functions, which must be inlined for best performance.
+*  Since link-time-optimization is not available for all compilers,
+*  these functions are defined into a .h to be included.
+*/
+
+/*-****************************************
+*  Dependencies
+******************************************/
+#include "mem.h"            /* unaligned access routines */
+#include "error_private.h"  /* error codes and messages */
+
+
+/*=========================================
+*  Target specific
+=========================================*/
+#if defined(__BMI__) && defined(__GNUC__)
+#  include    /* support for bextr (experimental) */
+#endif
+
+
+/*-******************************************
+*  bitStream encoding API (write forward)
+********************************************/
+/* bitStream can mix input from multiple sources.
+*  A critical property of these streams is that they encode and decode in **reverse** direction.
+*  So the first bit sequence you add will be the last to be read, like a LIFO stack.
+*/
+typedef struct
+{
+    size_t bitContainer;
+    int    bitPos;
+    char*  startPtr;
+    char*  ptr;
+    char*  endPtr;
+} BIT_CStream_t;
+
+MEM_STATIC size_t BIT_initCStream(BIT_CStream_t* bitC, void* dstBuffer, size_t dstCapacity);
+MEM_STATIC void   BIT_addBits(BIT_CStream_t* bitC, size_t value, unsigned nbBits);
+MEM_STATIC void   BIT_flushBits(BIT_CStream_t* bitC);
+MEM_STATIC size_t BIT_closeCStream(BIT_CStream_t* bitC);
+
+/* Start with initCStream, providing the size of buffer to write into.
+*  bitStream will never write outside of this buffer.
+*  `dstCapacity` must be >= sizeof(bitD->bitContainer), otherwise @return will be an error code.
+*
+*  bits are first added to a local register.
+*  Local register is size_t, hence 64-bits on 64-bits systems, or 32-bits on 32-bits systems.
+*  Writing data into memory is an explicit operation, performed by the flushBits function.
+*  Hence keep track how many bits are potentially stored into local register to avoid register overflow.
+*  After a flushBits, a maximum of 7 bits might still be stored into local register.
+*
+*  Avoid storing elements of more than 24 bits if you want compatibility with 32-bits bitstream readers.
+*
+*  Last operation is to close the bitStream.
+*  The function returns the final size of CStream in bytes.
+*  If data couldn't fit into `dstBuffer`, it will return a 0 ( == not storable)
+*/
+
+
+/*-********************************************
+*  bitStream decoding API (read backward)
+**********************************************/
+typedef struct
+{
+    size_t   bitContainer;
+    unsigned bitsConsumed;
+    const char* ptr;
+    const char* start;
+} BIT_DStream_t;
+
+typedef enum { BIT_DStream_unfinished = 0,
+               BIT_DStream_endOfBuffer = 1,
+               BIT_DStream_completed = 2,
+               BIT_DStream_overflow = 3 } BIT_DStream_status;  /* result of BIT_reloadDStream() */
+               /* 1,2,4,8 would be better for bitmap combinations, but slows down performance a bit ... :( */
+
+MEM_STATIC size_t   BIT_initDStream(BIT_DStream_t* bitD, const void* srcBuffer, size_t srcSize);
+MEM_STATIC size_t   BIT_readBits(BIT_DStream_t* bitD, unsigned nbBits);
+MEM_STATIC BIT_DStream_status BIT_reloadDStream(BIT_DStream_t* bitD);
+MEM_STATIC unsigned BIT_endOfDStream(const BIT_DStream_t* bitD);
+
+
+/* Start by invoking BIT_initDStream().
+*  A chunk of the bitStream is then stored into a local register.
+*  Local register size is 64-bits on 64-bits systems, 32-bits on 32-bits systems (size_t).
+*  You can then retrieve bitFields stored into the local register, **in reverse order**.
+*  Local register is explicitly reloaded from memory by the BIT_reloadDStream() method.
+*  A reload guarantee a minimum of ((8*sizeof(bitD->bitContainer))-7) bits when its result is BIT_DStream_unfinished.
+*  Otherwise, it can be less than that, so proceed accordingly.
+*  Checking if DStream has reached its end can be performed with BIT_endOfDStream().
+*/
+
+
+/*-****************************************
+*  unsafe API
+******************************************/
+MEM_STATIC void BIT_addBitsFast(BIT_CStream_t* bitC, size_t value, unsigned nbBits);
+/* faster, but works only if value is "clean", meaning all high bits above nbBits are 0 */
+
+MEM_STATIC void BIT_flushBitsFast(BIT_CStream_t* bitC);
+/* unsafe version; does not check buffer overflow */
+
+MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, unsigned nbBits);
+/* faster, but works only if nbBits >= 1 */
+
+
+
+/*-**************************************************************
+*  Internal functions
+****************************************************************/
+MEM_STATIC unsigned BIT_highbit32 (register U32 val)
+{
+#   if defined(_MSC_VER)   /* Visual */
+    unsigned long r=0;
+    _BitScanReverse ( &r, val );
+    return (unsigned) r;
+#   elif defined(__GNUC__) && (__GNUC__ >= 3)   /* Use GCC Intrinsic */
+    return 31 - __builtin_clz (val);
+#   else   /* Software version */
+    static const unsigned DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 };
+    U32 v = val;
+    v |= v >> 1;
+    v |= v >> 2;
+    v |= v >> 4;
+    v |= v >> 8;
+    v |= v >> 16;
+    return DeBruijnClz[ (U32) (v * 0x07C4ACDDU) >> 27];
+#   endif
+}
+
+/*=====    Local Constants   =====*/
+static const unsigned BIT_mask[] = { 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, 0x1FF, 0x3FF, 0x7FF, 0xFFF, 0x1FFF, 0x3FFF, 0x7FFF, 0xFFFF, 0x1FFFF, 0x3FFFF, 0x7FFFF, 0xFFFFF, 0x1FFFFF, 0x3FFFFF, 0x7FFFFF,  0xFFFFFF, 0x1FFFFFF, 0x3FFFFFF };   /* up to 26 bits */
+
+
+/*-**************************************************************
+*  bitStream encoding
+****************************************************************/
+/*! BIT_initCStream() :
+ *  `dstCapacity` must be > sizeof(void*)
+ *  @return : 0 if success,
+              otherwise an error code (can be tested using ERR_isError() ) */
+MEM_STATIC size_t BIT_initCStream(BIT_CStream_t* bitC, void* startPtr, size_t dstCapacity)
+{
+    bitC->bitContainer = 0;
+    bitC->bitPos = 0;
+    bitC->startPtr = (char*)startPtr;
+    bitC->ptr = bitC->startPtr;
+    bitC->endPtr = bitC->startPtr + dstCapacity - sizeof(bitC->ptr);
+    if (dstCapacity <= sizeof(bitC->ptr)) return ERROR(dstSize_tooSmall);
+    return 0;
+}
+
+/*! BIT_addBits() :
+    can add up to 26 bits into `bitC`.
+    Does not check for register overflow ! */
+MEM_STATIC void BIT_addBits(BIT_CStream_t* bitC, size_t value, unsigned nbBits)
+{
+    bitC->bitContainer |= (value & BIT_mask[nbBits]) << bitC->bitPos;
+    bitC->bitPos += nbBits;
+}
+
+/*! BIT_addBitsFast() :
+ *  works only if `value` is _clean_, meaning all high bits above nbBits are 0 */
+MEM_STATIC void BIT_addBitsFast(BIT_CStream_t* bitC, size_t value, unsigned nbBits)
+{
+    bitC->bitContainer |= value << bitC->bitPos;
+    bitC->bitPos += nbBits;
+}
+
+/*! BIT_flushBitsFast() :
+ *  unsafe version; does not check buffer overflow */
+MEM_STATIC void BIT_flushBitsFast(BIT_CStream_t* bitC)
+{
+    size_t const nbBytes = bitC->bitPos >> 3;
+    MEM_writeLEST(bitC->ptr, bitC->bitContainer);
+    bitC->ptr += nbBytes;
+    bitC->bitPos &= 7;
+    bitC->bitContainer >>= nbBytes*8;   /* if bitPos >= sizeof(bitContainer)*8 --> undefined behavior */
+}
+
+/*! BIT_flushBits() :
+ *  safe version; check for buffer overflow, and prevents it.
+ *  note : does not signal buffer overflow. This will be revealed later on using BIT_closeCStream() */
+MEM_STATIC void BIT_flushBits(BIT_CStream_t* bitC)
+{
+    size_t const nbBytes = bitC->bitPos >> 3;
+    MEM_writeLEST(bitC->ptr, bitC->bitContainer);
+    bitC->ptr += nbBytes;
+    if (bitC->ptr > bitC->endPtr) bitC->ptr = bitC->endPtr;
+    bitC->bitPos &= 7;
+    bitC->bitContainer >>= nbBytes*8;   /* if bitPos >= sizeof(bitContainer)*8 --> undefined behavior */
+}
+
+/*! BIT_closeCStream() :
+ *  @return : size of CStream, in bytes,
+              or 0 if it could not fit into dstBuffer */
+MEM_STATIC size_t BIT_closeCStream(BIT_CStream_t* bitC)
+{
+    BIT_addBitsFast(bitC, 1, 1);   /* endMark */
+    BIT_flushBits(bitC);
+
+    if (bitC->ptr >= bitC->endPtr) return 0; /* doesn't fit within authorized budget : cancel */
+
+    return (bitC->ptr - bitC->startPtr) + (bitC->bitPos > 0);
+}
+
+
+/*-********************************************************
+* bitStream decoding
+**********************************************************/
+/*! BIT_initDStream() :
+*   Initialize a BIT_DStream_t.
+*   `bitD` : a pointer to an already allocated BIT_DStream_t structure.
+*   `srcSize` must be the *exact* size of the bitStream, in bytes.
+*   @return : size of stream (== srcSize) or an errorCode if a problem is detected
+*/
+MEM_STATIC size_t BIT_initDStream(BIT_DStream_t* bitD, const void* srcBuffer, size_t srcSize)
+{
+    if (srcSize < 1) { memset(bitD, 0, sizeof(*bitD)); return ERROR(srcSize_wrong); }
+
+    if (srcSize >=  sizeof(bitD->bitContainer)) {  /* normal case */
+        bitD->start = (const char*)srcBuffer;
+        bitD->ptr   = (const char*)srcBuffer + srcSize - sizeof(bitD->bitContainer);
+        bitD->bitContainer = MEM_readLEST(bitD->ptr);
+        { BYTE const lastByte = ((const BYTE*)srcBuffer)[srcSize-1];
+          bitD->bitsConsumed = lastByte ? 8 - BIT_highbit32(lastByte) : 0;  /* ensures bitsConsumed is always set */
+          if (lastByte == 0) return ERROR(GENERIC); /* endMark not present */ }
+    } else {
+        bitD->start = (const char*)srcBuffer;
+        bitD->ptr   = bitD->start;
+        bitD->bitContainer = *(const BYTE*)(bitD->start);
+        switch(srcSize)
+        {
+            case 7: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[6]) << (sizeof(bitD->bitContainer)*8 - 16);
+            case 6: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[5]) << (sizeof(bitD->bitContainer)*8 - 24);
+            case 5: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[4]) << (sizeof(bitD->bitContainer)*8 - 32);
+            case 4: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[3]) << 24;
+            case 3: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[2]) << 16;
+            case 2: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[1]) <<  8;
+            default:;
+        }
+        { BYTE const lastByte = ((const BYTE*)srcBuffer)[srcSize-1];
+          bitD->bitsConsumed = lastByte ? 8 - BIT_highbit32(lastByte) : 0;
+          if (lastByte == 0) return ERROR(GENERIC); /* endMark not present */ }
+        bitD->bitsConsumed += (U32)(sizeof(bitD->bitContainer) - srcSize)*8;
+    }
+
+    return srcSize;
+}
+
+MEM_STATIC size_t BIT_getUpperBits(size_t bitContainer, U32 const start)
+{
+    return bitContainer >> start;
+}
+
+MEM_STATIC size_t BIT_getMiddleBits(size_t bitContainer, U32 const start, U32 const nbBits)
+{
+#if defined(__BMI__) && defined(__GNUC__) && __GNUC__*1000+__GNUC_MINOR__ >= 4008  /* experimental */
+#  if defined(__x86_64__)
+    if (sizeof(bitContainer)==8)
+        return _bextr_u64(bitContainer, start, nbBits);
+    else
+#  endif
+        return _bextr_u32(bitContainer, start, nbBits);
+#else
+    return (bitContainer >> start) & BIT_mask[nbBits];
+#endif
+}
+
+MEM_STATIC size_t BIT_getLowerBits(size_t bitContainer, U32 const nbBits)
+{
+    return bitContainer & BIT_mask[nbBits];
+}
+
+/*! BIT_lookBits() :
+ *  Provides next n bits from local register.
+ *  local register is not modified.
+ *  On 32-bits, maxNbBits==24.
+ *  On 64-bits, maxNbBits==56.
+ *  @return : value extracted
+ */
+ MEM_STATIC size_t BIT_lookBits(const BIT_DStream_t* bitD, U32 nbBits)
+{
+#if defined(__BMI__) && defined(__GNUC__)   /* experimental; fails if bitD->bitsConsumed + nbBits > sizeof(bitD->bitContainer)*8 */
+    return BIT_getMiddleBits(bitD->bitContainer, (sizeof(bitD->bitContainer)*8) - bitD->bitsConsumed - nbBits, nbBits);
+#else
+    U32 const bitMask = sizeof(bitD->bitContainer)*8 - 1;
+    return ((bitD->bitContainer << (bitD->bitsConsumed & bitMask)) >> 1) >> ((bitMask-nbBits) & bitMask);
+#endif
+}
+
+/*! BIT_lookBitsFast() :
+*   unsafe version; only works only if nbBits >= 1 */
+MEM_STATIC size_t BIT_lookBitsFast(const BIT_DStream_t* bitD, U32 nbBits)
+{
+    U32 const bitMask = sizeof(bitD->bitContainer)*8 - 1;
+    return (bitD->bitContainer << (bitD->bitsConsumed & bitMask)) >> (((bitMask+1)-nbBits) & bitMask);
+}
+
+MEM_STATIC void BIT_skipBits(BIT_DStream_t* bitD, U32 nbBits)
+{
+    bitD->bitsConsumed += nbBits;
+}
+
+/*! BIT_readBits() :
+ *  Read (consume) next n bits from local register and update.
+ *  Pay attention to not read more than nbBits contained into local register.
+ *  @return : extracted value.
+ */
+MEM_STATIC size_t BIT_readBits(BIT_DStream_t* bitD, U32 nbBits)
+{
+    size_t const value = BIT_lookBits(bitD, nbBits);
+    BIT_skipBits(bitD, nbBits);
+    return value;
+}
+
+/*! BIT_readBitsFast() :
+*   unsafe version; only works only if nbBits >= 1 */
+MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, U32 nbBits)
+{
+    size_t const value = BIT_lookBitsFast(bitD, nbBits);
+    BIT_skipBits(bitD, nbBits);
+    return value;
+}
+
+/*! BIT_reloadDStream() :
+*   Refill `bitD` from buffer previously set in BIT_initDStream() .
+*   This function is safe, it guarantees it will not read beyond src buffer.
+*   @return : status of `BIT_DStream_t` internal register.
+              if status == BIT_DStream_unfinished, internal register is filled with >= (sizeof(bitD->bitContainer)*8 - 7) bits */
+MEM_STATIC BIT_DStream_status BIT_reloadDStream(BIT_DStream_t* bitD)
+{
+	if (bitD->bitsConsumed > (sizeof(bitD->bitContainer)*8))  /* should not happen => corruption detected */
+		return BIT_DStream_overflow;
+
+    if (bitD->ptr >= bitD->start + sizeof(bitD->bitContainer)) {
+        bitD->ptr -= bitD->bitsConsumed >> 3;
+        bitD->bitsConsumed &= 7;
+        bitD->bitContainer = MEM_readLEST(bitD->ptr);
+        return BIT_DStream_unfinished;
+    }
+    if (bitD->ptr == bitD->start) {
+        if (bitD->bitsConsumed < sizeof(bitD->bitContainer)*8) return BIT_DStream_endOfBuffer;
+        return BIT_DStream_completed;
+    }
+    {   U32 nbBytes = bitD->bitsConsumed >> 3;
+        BIT_DStream_status result = BIT_DStream_unfinished;
+        if (bitD->ptr - nbBytes < bitD->start) {
+            nbBytes = (U32)(bitD->ptr - bitD->start);  /* ptr > start */
+            result = BIT_DStream_endOfBuffer;
+        }
+        bitD->ptr -= nbBytes;
+        bitD->bitsConsumed -= nbBytes*8;
+        bitD->bitContainer = MEM_readLEST(bitD->ptr);   /* reminder : srcSize > sizeof(bitD) */
+        return result;
+    }
+}
+
+/*! BIT_endOfDStream() :
+*   @return Tells if DStream has exactly reached its end (all bits consumed).
+*/
+MEM_STATIC unsigned BIT_endOfDStream(const BIT_DStream_t* DStream)
+{
+    return ((DStream->ptr == DStream->start) && (DStream->bitsConsumed == sizeof(DStream->bitContainer)*8));
+}
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* BITSTREAM_H_MODULE */
diff --git a/lib/common/entropy_common.c b/lib/common/entropy_common.c
new file mode 100644
index 0000000..83fd971
--- /dev/null
+++ b/lib/common/entropy_common.c
@@ -0,0 +1,227 @@
+/*
+   Common functions of New Generation Entropy library
+   Copyright (C) 2016, Yann Collet.
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+    You can contact the author at :
+    - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
+    - Public forum : https://groups.google.com/forum/#!forum/lz4c
+*************************************************************************** */
+
+/* *************************************
+*  Dependencies
+***************************************/
+#include "mem.h"
+#include "error_private.h"       /* ERR_*, ERROR */
+#define FSE_STATIC_LINKING_ONLY  /* FSE_MIN_TABLELOG */
+#include "fse.h"
+#define HUF_STATIC_LINKING_ONLY  /* HUF_TABLELOG_ABSOLUTEMAX */
+#include "huf.h"
+
+
+/*-****************************************
+*  FSE Error Management
+******************************************/
+unsigned FSE_isError(size_t code) { return ERR_isError(code); }
+
+const char* FSE_getErrorName(size_t code) { return ERR_getErrorName(code); }
+
+
+/* **************************************************************
+*  HUF Error Management
+****************************************************************/
+unsigned HUF_isError(size_t code) { return ERR_isError(code); }
+
+const char* HUF_getErrorName(size_t code) { return ERR_getErrorName(code); }
+
+
+/*-**************************************************************
+*  FSE NCount encoding-decoding
+****************************************************************/
+static short FSE_abs(short a) { return (short)(a<0 ? -a : a); }
+
+size_t FSE_readNCount (short* normalizedCounter, unsigned* maxSVPtr, unsigned* tableLogPtr,
+                 const void* headerBuffer, size_t hbSize)
+{
+    const BYTE* const istart = (const BYTE*) headerBuffer;
+    const BYTE* const iend = istart + hbSize;
+    const BYTE* ip = istart;
+    int nbBits;
+    int remaining;
+    int threshold;
+    U32 bitStream;
+    int bitCount;
+    unsigned charnum = 0;
+    int previous0 = 0;
+
+    if (hbSize < 4) return ERROR(srcSize_wrong);
+    bitStream = MEM_readLE32(ip);
+    nbBits = (bitStream & 0xF) + FSE_MIN_TABLELOG;   /* extract tableLog */
+    if (nbBits > FSE_TABLELOG_ABSOLUTE_MAX) return ERROR(tableLog_tooLarge);
+    bitStream >>= 4;
+    bitCount = 4;
+    *tableLogPtr = nbBits;
+    remaining = (1<1) & (charnum<=*maxSVPtr)) {
+        if (previous0) {
+            unsigned n0 = charnum;
+            while ((bitStream & 0xFFFF) == 0xFFFF) {
+                n0 += 24;
+                if (ip < iend-5) {
+                    ip += 2;
+                    bitStream = MEM_readLE32(ip) >> bitCount;
+                } else {
+                    bitStream >>= 16;
+                    bitCount   += 16;
+            }   }
+            while ((bitStream & 3) == 3) {
+                n0 += 3;
+                bitStream >>= 2;
+                bitCount += 2;
+            }
+            n0 += bitStream & 3;
+            bitCount += 2;
+            if (n0 > *maxSVPtr) return ERROR(maxSymbolValue_tooSmall);
+            while (charnum < n0) normalizedCounter[charnum++] = 0;
+            if ((ip <= iend-7) || (ip + (bitCount>>3) <= iend-4)) {
+                ip += bitCount>>3;
+                bitCount &= 7;
+                bitStream = MEM_readLE32(ip) >> bitCount;
+            } else {
+                bitStream >>= 2;
+        }   }
+        {   short const max = (short)((2*threshold-1)-remaining);
+            short count;
+
+            if ((bitStream & (threshold-1)) < (U32)max) {
+                count = (short)(bitStream & (threshold-1));
+                bitCount   += nbBits-1;
+            } else {
+                count = (short)(bitStream & (2*threshold-1));
+                if (count >= threshold) count -= max;
+                bitCount   += nbBits;
+            }
+
+            count--;   /* extra accuracy */
+            remaining -= FSE_abs(count);
+            normalizedCounter[charnum++] = count;
+            previous0 = !count;
+            while (remaining < threshold) {
+                nbBits--;
+                threshold >>= 1;
+            }
+
+            if ((ip <= iend-7) || (ip + (bitCount>>3) <= iend-4)) {
+                ip += bitCount>>3;
+                bitCount &= 7;
+            } else {
+                bitCount -= (int)(8 * (iend - 4 - ip));
+                ip = iend - 4;
+            }
+            bitStream = MEM_readLE32(ip) >> (bitCount & 31);
+    }   }   /* while ((remaining>1) & (charnum<=*maxSVPtr)) */
+    if (remaining != 1) return ERROR(corruption_detected);
+    if (bitCount > 32) return ERROR(corruption_detected);
+    *maxSVPtr = charnum-1;
+
+    ip += (bitCount+7)>>3;
+    return ip-istart;
+}
+
+
+/*! HUF_readStats() :
+    Read compact Huffman tree, saved by HUF_writeCTable().
+    `huffWeight` is destination buffer.
+    `rankStats` is assumed to be a table of at least HUF_TABLELOG_MAX U32.
+    @return : size read from `src` , or an error Code .
+    Note : Needed by HUF_readCTable() and HUF_readDTableX?() .
+*/
+size_t HUF_readStats(BYTE* huffWeight, size_t hwSize, U32* rankStats,
+                     U32* nbSymbolsPtr, U32* tableLogPtr,
+                     const void* src, size_t srcSize)
+{
+    U32 weightTotal;
+    const BYTE* ip = (const BYTE*) src;
+    size_t iSize;
+    size_t oSize;
+
+    if (!srcSize) return ERROR(srcSize_wrong);
+    iSize = ip[0];
+    /* memset(huffWeight, 0, hwSize);   *//* is not necessary, even though some analyzer complain ... */
+
+    if (iSize >= 128) {  /* special header */
+        oSize = iSize - 127;
+        iSize = ((oSize+1)/2);
+        if (iSize+1 > srcSize) return ERROR(srcSize_wrong);
+        if (oSize >= hwSize) return ERROR(corruption_detected);
+        ip += 1;
+        {   U32 n;
+            for (n=0; n> 4;
+                huffWeight[n+1] = ip[n/2] & 15;
+    }   }   }
+    else  {   /* header compressed with FSE (normal case) */
+        FSE_DTable fseWorkspace[FSE_DTABLE_SIZE_U32(6)];  /* 6 is max possible tableLog for HUF header (maybe even 5, to be tested) */
+        if (iSize+1 > srcSize) return ERROR(srcSize_wrong);
+        oSize = FSE_decompress_wksp(huffWeight, hwSize-1, ip+1, iSize, fseWorkspace, 6);   /* max (hwSize-1) values decoded, as last one is implied */
+        if (FSE_isError(oSize)) return oSize;
+    }
+
+    /* collect weight stats */
+    memset(rankStats, 0, (HUF_TABLELOG_MAX + 1) * sizeof(U32));
+    weightTotal = 0;
+    {   U32 n; for (n=0; n= HUF_TABLELOG_MAX) return ERROR(corruption_detected);
+            rankStats[huffWeight[n]]++;
+            weightTotal += (1 << huffWeight[n]) >> 1;
+    }   }
+    if (weightTotal == 0) return ERROR(corruption_detected);
+
+    /* get last non-null symbol weight (implied, total must be 2^n) */
+    {   U32 const tableLog = BIT_highbit32(weightTotal) + 1;
+        if (tableLog > HUF_TABLELOG_MAX) return ERROR(corruption_detected);
+        *tableLogPtr = tableLog;
+        /* determine last weight */
+        {   U32 const total = 1 << tableLog;
+            U32 const rest = total - weightTotal;
+            U32 const verif = 1 << BIT_highbit32(rest);
+            U32 const lastWeight = BIT_highbit32(rest) + 1;
+            if (verif != rest) return ERROR(corruption_detected);    /* last value must be a clean power of 2 */
+            huffWeight[oSize] = (BYTE)lastWeight;
+            rankStats[lastWeight]++;
+    }   }
+
+    /* check tree construction validity */
+    if ((rankStats[1] < 2) || (rankStats[1] & 1)) return ERROR(corruption_detected);   /* by construction : at least 2 elts of rank 1, must be even */
+
+    /* results */
+    *nbSymbolsPtr = (U32)(oSize+1);
+    return iSize+1;
+}
diff --git a/lib/common/error_private.c b/lib/common/error_private.c
new file mode 100644
index 0000000..a0fa172
--- /dev/null
+++ b/lib/common/error_private.c
@@ -0,0 +1,43 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+/* The purpose of this file is to have a single list of error strings embedded in binary */
+
+#include "error_private.h"
+
+const char* ERR_getErrorString(ERR_enum code)
+{
+    static const char* const notErrorCode = "Unspecified error code";
+    switch( code )
+    {
+    case PREFIX(no_error): return "No error detected";
+    case PREFIX(GENERIC):  return "Error (generic)";
+    case PREFIX(prefix_unknown): return "Unknown frame descriptor";
+    case PREFIX(version_unsupported): return "Version not supported";
+    case PREFIX(parameter_unknown): return "Unknown parameter type";
+    case PREFIX(frameParameter_unsupported): return "Unsupported frame parameter";
+    case PREFIX(frameParameter_unsupportedBy32bits): return "Frame parameter unsupported in 32-bits mode";
+    case PREFIX(frameParameter_windowTooLarge): return "Frame requires too much memory for decoding";
+    case PREFIX(compressionParameter_unsupported): return "Compression parameter is out of bound";
+    case PREFIX(init_missing): return "Context should be init first";
+    case PREFIX(memory_allocation): return "Allocation error : not enough memory";
+    case PREFIX(stage_wrong): return "Operation not authorized at current processing stage";
+    case PREFIX(dstSize_tooSmall): return "Destination buffer is too small";
+    case PREFIX(srcSize_wrong): return "Src size incorrect";
+    case PREFIX(corruption_detected): return "Corrupted block detected";
+    case PREFIX(checksum_wrong): return "Restored data doesn't match checksum";
+    case PREFIX(tableLog_tooLarge): return "tableLog requires too much memory : unsupported";
+    case PREFIX(maxSymbolValue_tooLarge): return "Unsupported max Symbol Value : too large";
+    case PREFIX(maxSymbolValue_tooSmall): return "Specified maxSymbolValue is too small";
+    case PREFIX(dictionary_corrupted): return "Dictionary is corrupted";
+    case PREFIX(dictionary_wrong): return "Dictionary mismatch";
+    case PREFIX(maxCode):
+    default: return notErrorCode;
+    }
+}
diff --git a/lib/common/error_private.h b/lib/common/error_private.h
new file mode 100644
index 0000000..1bc2e49
--- /dev/null
+++ b/lib/common/error_private.h
@@ -0,0 +1,76 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+/* Note : this module is expected to remain private, do not expose it */
+
+#ifndef ERROR_H_MODULE
+#define ERROR_H_MODULE
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+
+/* ****************************************
+*  Dependencies
+******************************************/
+#include         /* size_t */
+#include "zstd_errors.h"  /* enum list */
+
+
+/* ****************************************
+*  Compiler-specific
+******************************************/
+#if defined(__GNUC__)
+#  define ERR_STATIC static __attribute__((unused))
+#elif defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
+#  define ERR_STATIC static inline
+#elif defined(_MSC_VER)
+#  define ERR_STATIC static __inline
+#else
+#  define ERR_STATIC static  /* this version may generate warnings for unused static functions; disable the relevant warning */
+#endif
+
+
+/*-****************************************
+*  Customization (error_public.h)
+******************************************/
+typedef ZSTD_ErrorCode ERR_enum;
+#define PREFIX(name) ZSTD_error_##name
+
+
+/*-****************************************
+*  Error codes handling
+******************************************/
+#ifdef ERROR
+#  undef ERROR   /* reported already defined on VS 2015 (Rich Geldreich) */
+#endif
+#define ERROR(name) ((size_t)-PREFIX(name))
+
+ERR_STATIC unsigned ERR_isError(size_t code) { return (code > ERROR(maxCode)); }
+
+ERR_STATIC ERR_enum ERR_getErrorCode(size_t code) { if (!ERR_isError(code)) return (ERR_enum)0; return (ERR_enum) (0-code); }
+
+
+/*-****************************************
+*  Error Strings
+******************************************/
+
+const char* ERR_getErrorString(ERR_enum code);   /* error_private.c */
+
+ERR_STATIC const char* ERR_getErrorName(size_t code)
+{
+    return ERR_getErrorString(ERR_getErrorCode(code));
+}
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ERROR_H_MODULE */
diff --git a/lib/common/fse.h b/lib/common/fse.h
new file mode 100644
index 0000000..8b07d18
--- /dev/null
+++ b/lib/common/fse.h
@@ -0,0 +1,668 @@
+/* ******************************************************************
+   FSE : Finite State Entropy codec
+   Public Prototypes declaration
+   Copyright (C) 2013-2016, Yann Collet.
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+   You can contact the author at :
+   - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+****************************************************************** */
+#ifndef FSE_H
+#define FSE_H
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+
+/*-*****************************************
+*  Dependencies
+******************************************/
+#include     /* size_t, ptrdiff_t */
+
+
+/*-****************************************
+*  FSE simple functions
+******************************************/
+/*! FSE_compress() :
+    Compress content of buffer 'src', of size 'srcSize', into destination buffer 'dst'.
+    'dst' buffer must be already allocated. Compression runs faster is dstCapacity >= FSE_compressBound(srcSize).
+    @return : size of compressed data (<= dstCapacity).
+    Special values : if return == 0, srcData is not compressible => Nothing is stored within dst !!!
+                     if return == 1, srcData is a single byte symbol * srcSize times. Use RLE compression instead.
+                     if FSE_isError(return), compression failed (more details using FSE_getErrorName())
+*/
+size_t FSE_compress(void* dst, size_t dstCapacity,
+              const void* src, size_t srcSize);
+
+/*! FSE_decompress():
+    Decompress FSE data from buffer 'cSrc', of size 'cSrcSize',
+    into already allocated destination buffer 'dst', of size 'dstCapacity'.
+    @return : size of regenerated data (<= maxDstSize),
+              or an error code, which can be tested using FSE_isError() .
+
+    ** Important ** : FSE_decompress() does not decompress non-compressible nor RLE data !!!
+    Why ? : making this distinction requires a header.
+    Header management is intentionally delegated to the user layer, which can better manage special cases.
+*/
+size_t FSE_decompress(void* dst,  size_t dstCapacity,
+                const void* cSrc, size_t cSrcSize);
+
+
+/*-*****************************************
+*  Tool functions
+******************************************/
+size_t FSE_compressBound(size_t size);       /* maximum compressed size */
+
+/* Error Management */
+unsigned    FSE_isError(size_t code);        /* tells if a return value is an error code */
+const char* FSE_getErrorName(size_t code);   /* provides error code string (useful for debugging) */
+
+
+/*-*****************************************
+*  FSE advanced functions
+******************************************/
+/*! FSE_compress2() :
+    Same as FSE_compress(), but allows the selection of 'maxSymbolValue' and 'tableLog'
+    Both parameters can be defined as '0' to mean : use default value
+    @return : size of compressed data
+    Special values : if return == 0, srcData is not compressible => Nothing is stored within cSrc !!!
+                     if return == 1, srcData is a single byte symbol * srcSize times. Use RLE compression.
+                     if FSE_isError(return), it's an error code.
+*/
+size_t FSE_compress2 (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog);
+
+
+/*-*****************************************
+*  FSE detailed API
+******************************************/
+/*!
+FSE_compress() does the following:
+1. count symbol occurrence from source[] into table count[]
+2. normalize counters so that sum(count[]) == Power_of_2 (2^tableLog)
+3. save normalized counters to memory buffer using writeNCount()
+4. build encoding table 'CTable' from normalized counters
+5. encode the data stream using encoding table 'CTable'
+
+FSE_decompress() does the following:
+1. read normalized counters with readNCount()
+2. build decoding table 'DTable' from normalized counters
+3. decode the data stream using decoding table 'DTable'
+
+The following API allows targeting specific sub-functions for advanced tasks.
+For example, it's possible to compress several blocks using the same 'CTable',
+or to save and provide normalized distribution using external method.
+*/
+
+/* *** COMPRESSION *** */
+
+/*! FSE_count():
+    Provides the precise count of each byte within a table 'count'.
+    'count' is a table of unsigned int, of minimum size (*maxSymbolValuePtr+1).
+    *maxSymbolValuePtr will be updated if detected smaller than initial value.
+    @return : the count of the most frequent symbol (which is not identified).
+              if return == srcSize, there is only one symbol.
+              Can also return an error code, which can be tested with FSE_isError(). */
+size_t FSE_count(unsigned* count, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize);
+
+/*! FSE_optimalTableLog():
+    dynamically downsize 'tableLog' when conditions are met.
+    It saves CPU time, by using smaller tables, while preserving or even improving compression ratio.
+    @return : recommended tableLog (necessarily <= 'maxTableLog') */
+unsigned FSE_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue);
+
+/*! FSE_normalizeCount():
+    normalize counts so that sum(count[]) == Power_of_2 (2^tableLog)
+    'normalizedCounter' is a table of short, of minimum size (maxSymbolValue+1).
+    @return : tableLog,
+              or an errorCode, which can be tested using FSE_isError() */
+size_t FSE_normalizeCount(short* normalizedCounter, unsigned tableLog, const unsigned* count, size_t srcSize, unsigned maxSymbolValue);
+
+/*! FSE_NCountWriteBound():
+    Provides the maximum possible size of an FSE normalized table, given 'maxSymbolValue' and 'tableLog'.
+    Typically useful for allocation purpose. */
+size_t FSE_NCountWriteBound(unsigned maxSymbolValue, unsigned tableLog);
+
+/*! FSE_writeNCount():
+    Compactly save 'normalizedCounter' into 'buffer'.
+    @return : size of the compressed table,
+              or an errorCode, which can be tested using FSE_isError(). */
+size_t FSE_writeNCount (void* buffer, size_t bufferSize, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
+
+
+/*! Constructor and Destructor of FSE_CTable.
+    Note that FSE_CTable size depends on 'tableLog' and 'maxSymbolValue' */
+typedef unsigned FSE_CTable;   /* don't allocate that. It's only meant to be more restrictive than void* */
+FSE_CTable* FSE_createCTable (unsigned tableLog, unsigned maxSymbolValue);
+void        FSE_freeCTable (FSE_CTable* ct);
+
+/*! FSE_buildCTable():
+    Builds `ct`, which must be already allocated, using FSE_createCTable().
+    @return : 0, or an errorCode, which can be tested using FSE_isError() */
+size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
+
+/*! FSE_compress_usingCTable():
+    Compress `src` using `ct` into `dst` which must be already allocated.
+    @return : size of compressed data (<= `dstCapacity`),
+              or 0 if compressed data could not fit into `dst`,
+              or an errorCode, which can be tested using FSE_isError() */
+size_t FSE_compress_usingCTable (void* dst, size_t dstCapacity, const void* src, size_t srcSize, const FSE_CTable* ct);
+
+/*!
+Tutorial :
+----------
+The first step is to count all symbols. FSE_count() does this job very fast.
+Result will be saved into 'count', a table of unsigned int, which must be already allocated, and have 'maxSymbolValuePtr[0]+1' cells.
+'src' is a table of bytes of size 'srcSize'. All values within 'src' MUST be <= maxSymbolValuePtr[0]
+maxSymbolValuePtr[0] will be updated, with its real value (necessarily <= original value)
+FSE_count() will return the number of occurrence of the most frequent symbol.
+This can be used to know if there is a single symbol within 'src', and to quickly evaluate its compressibility.
+If there is an error, the function will return an ErrorCode (which can be tested using FSE_isError()).
+
+The next step is to normalize the frequencies.
+FSE_normalizeCount() will ensure that sum of frequencies is == 2 ^'tableLog'.
+It also guarantees a minimum of 1 to any Symbol with frequency >= 1.
+You can use 'tableLog'==0 to mean "use default tableLog value".
+If you are unsure of which tableLog value to use, you can ask FSE_optimalTableLog(),
+which will provide the optimal valid tableLog given sourceSize, maxSymbolValue, and a user-defined maximum (0 means "default").
+
+The result of FSE_normalizeCount() will be saved into a table,
+called 'normalizedCounter', which is a table of signed short.
+'normalizedCounter' must be already allocated, and have at least 'maxSymbolValue+1' cells.
+The return value is tableLog if everything proceeded as expected.
+It is 0 if there is a single symbol within distribution.
+If there is an error (ex: invalid tableLog value), the function will return an ErrorCode (which can be tested using FSE_isError()).
+
+'normalizedCounter' can be saved in a compact manner to a memory area using FSE_writeNCount().
+'buffer' must be already allocated.
+For guaranteed success, buffer size must be at least FSE_headerBound().
+The result of the function is the number of bytes written into 'buffer'.
+If there is an error, the function will return an ErrorCode (which can be tested using FSE_isError(); ex : buffer size too small).
+
+'normalizedCounter' can then be used to create the compression table 'CTable'.
+The space required by 'CTable' must be already allocated, using FSE_createCTable().
+You can then use FSE_buildCTable() to fill 'CTable'.
+If there is an error, both functions will return an ErrorCode (which can be tested using FSE_isError()).
+
+'CTable' can then be used to compress 'src', with FSE_compress_usingCTable().
+Similar to FSE_count(), the convention is that 'src' is assumed to be a table of char of size 'srcSize'
+The function returns the size of compressed data (without header), necessarily <= `dstCapacity`.
+If it returns '0', compressed data could not fit into 'dst'.
+If there is an error, the function will return an ErrorCode (which can be tested using FSE_isError()).
+*/
+
+
+/* *** DECOMPRESSION *** */
+
+/*! FSE_readNCount():
+    Read compactly saved 'normalizedCounter' from 'rBuffer'.
+    @return : size read from 'rBuffer',
+              or an errorCode, which can be tested using FSE_isError().
+              maxSymbolValuePtr[0] and tableLogPtr[0] will also be updated with their respective values */
+size_t FSE_readNCount (short* normalizedCounter, unsigned* maxSymbolValuePtr, unsigned* tableLogPtr, const void* rBuffer, size_t rBuffSize);
+
+/*! Constructor and Destructor of FSE_DTable.
+    Note that its size depends on 'tableLog' */
+typedef unsigned FSE_DTable;   /* don't allocate that. It's just a way to be more restrictive than void* */
+FSE_DTable* FSE_createDTable(unsigned tableLog);
+void        FSE_freeDTable(FSE_DTable* dt);
+
+/*! FSE_buildDTable():
+    Builds 'dt', which must be already allocated, using FSE_createDTable().
+    return : 0, or an errorCode, which can be tested using FSE_isError() */
+size_t FSE_buildDTable (FSE_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
+
+/*! FSE_decompress_usingDTable():
+    Decompress compressed source `cSrc` of size `cSrcSize` using `dt`
+    into `dst` which must be already allocated.
+    @return : size of regenerated data (necessarily <= `dstCapacity`),
+              or an errorCode, which can be tested using FSE_isError() */
+size_t FSE_decompress_usingDTable(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, const FSE_DTable* dt);
+
+/*!
+Tutorial :
+----------
+(Note : these functions only decompress FSE-compressed blocks.
+ If block is uncompressed, use memcpy() instead
+ If block is a single repeated byte, use memset() instead )
+
+The first step is to obtain the normalized frequencies of symbols.
+This can be performed by FSE_readNCount() if it was saved using FSE_writeNCount().
+'normalizedCounter' must be already allocated, and have at least 'maxSymbolValuePtr[0]+1' cells of signed short.
+In practice, that means it's necessary to know 'maxSymbolValue' beforehand,
+or size the table to handle worst case situations (typically 256).
+FSE_readNCount() will provide 'tableLog' and 'maxSymbolValue'.
+The result of FSE_readNCount() is the number of bytes read from 'rBuffer'.
+Note that 'rBufferSize' must be at least 4 bytes, even if useful information is less than that.
+If there is an error, the function will return an error code, which can be tested using FSE_isError().
+
+The next step is to build the decompression tables 'FSE_DTable' from 'normalizedCounter'.
+This is performed by the function FSE_buildDTable().
+The space required by 'FSE_DTable' must be already allocated using FSE_createDTable().
+If there is an error, the function will return an error code, which can be tested using FSE_isError().
+
+`FSE_DTable` can then be used to decompress `cSrc`, with FSE_decompress_usingDTable().
+`cSrcSize` must be strictly correct, otherwise decompression will fail.
+FSE_decompress_usingDTable() result will tell how many bytes were regenerated (<=`dstCapacity`).
+If there is an error, the function will return an error code, which can be tested using FSE_isError(). (ex: dst buffer too small)
+*/
+
+
+#ifdef FSE_STATIC_LINKING_ONLY
+
+/* *** Dependency *** */
+#include "bitstream.h"
+
+
+/* *****************************************
+*  Static allocation
+*******************************************/
+/* FSE buffer bounds */
+#define FSE_NCOUNTBOUND 512
+#define FSE_BLOCKBOUND(size) (size + (size>>7))
+#define FSE_COMPRESSBOUND(size) (FSE_NCOUNTBOUND + FSE_BLOCKBOUND(size))   /* Macro version, useful for static allocation */
+
+/* It is possible to statically allocate FSE CTable/DTable as a table of FSE_CTable/FSE_DTable using below macros */
+#define FSE_CTABLE_SIZE_U32(maxTableLog, maxSymbolValue)   (1 + (1<<(maxTableLog-1)) + ((maxSymbolValue+1)*2))
+#define FSE_DTABLE_SIZE_U32(maxTableLog)                   (1 + (1<= `1024` unsigned
+ */
+size_t FSE_count_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
+                 const void* source, size_t sourceSize, unsigned* workSpace);
+
+/** FSE_countFast() :
+ *  same as FSE_count(), but blindly trusts that all byte values within src are <= *maxSymbolValuePtr
+ */
+size_t FSE_countFast(unsigned* count, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize);
+
+/* FSE_countFast_wksp() :
+ * Same as FSE_countFast(), but using an externally provided scratch buffer.
+ * `workSpace` must be a table of minimum `1024` unsigned
+ */
+size_t FSE_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize, unsigned* workSpace);
+
+/*! FSE_count_simple
+ * Same as FSE_countFast(), but does not use any additional memory (not even on stack).
+ * This function is unsafe, and will segfault if any value within `src` is `> *maxSymbolValuePtr` (presuming it's also the size of `count`).
+*/
+size_t FSE_count_simple(unsigned* count, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize);
+
+
+
+unsigned FSE_optimalTableLog_internal(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue, unsigned minus);
+/**< same as FSE_optimalTableLog(), which used `minus==2` */
+
+/* FSE_compress_wksp() :
+ * Same as FSE_compress2(), but using an externally allocated scratch buffer (`workSpace`).
+ * FSE_WKSP_SIZE_U32() provides the minimum size required for `workSpace` as a table of FSE_CTable.
+ */
+#define FSE_WKSP_SIZE_U32(maxTableLog, maxSymbolValue)   ( FSE_CTABLE_SIZE_U32(maxTableLog, maxSymbolValue) + (1<<((maxTableLog>2)?(maxTableLog-2):0)) )
+size_t FSE_compress_wksp (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize);
+
+size_t FSE_buildCTable_raw (FSE_CTable* ct, unsigned nbBits);
+/**< build a fake FSE_CTable, designed for a flat distribution, where each symbol uses nbBits */
+
+size_t FSE_buildCTable_rle (FSE_CTable* ct, unsigned char symbolValue);
+/**< build a fake FSE_CTable, designed to compress always the same symbolValue */
+
+/* FSE_buildCTable_wksp() :
+ * Same as FSE_buildCTable(), but using an externally allocated scratch buffer (`workSpace`).
+ * `wkspSize` must be >= `(1<= BIT_DStream_completed
+
+When it's done, verify decompression is fully completed, by checking both DStream and the relevant states.
+Checking if DStream has reached its end is performed by :
+    BIT_endOfDStream(&DStream);
+Check also the states. There might be some symbols left there, if some high probability ones (>50%) are possible.
+    FSE_endOfDState(&DState);
+*/
+
+
+/* *****************************************
+*  FSE unsafe API
+*******************************************/
+static unsigned char FSE_decodeSymbolFast(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD);
+/* faster, but works only if nbBits is always >= 1 (otherwise, result will be corrupted) */
+
+
+/* *****************************************
+*  Implementation of inlined functions
+*******************************************/
+typedef struct {
+    int deltaFindState;
+    U32 deltaNbBits;
+} FSE_symbolCompressionTransform; /* total 8 bytes */
+
+MEM_STATIC void FSE_initCState(FSE_CState_t* statePtr, const FSE_CTable* ct)
+{
+    const void* ptr = ct;
+    const U16* u16ptr = (const U16*) ptr;
+    const U32 tableLog = MEM_read16(ptr);
+    statePtr->value = (ptrdiff_t)1<stateTable = u16ptr+2;
+    statePtr->symbolTT = ((const U32*)ct + 1 + (tableLog ? (1<<(tableLog-1)) : 1));
+    statePtr->stateLog = tableLog;
+}
+
+
+/*! FSE_initCState2() :
+*   Same as FSE_initCState(), but the first symbol to include (which will be the last to be read)
+*   uses the smallest state value possible, saving the cost of this symbol */
+MEM_STATIC void FSE_initCState2(FSE_CState_t* statePtr, const FSE_CTable* ct, U32 symbol)
+{
+    FSE_initCState(statePtr, ct);
+    {   const FSE_symbolCompressionTransform symbolTT = ((const FSE_symbolCompressionTransform*)(statePtr->symbolTT))[symbol];
+        const U16* stateTable = (const U16*)(statePtr->stateTable);
+        U32 nbBitsOut  = (U32)((symbolTT.deltaNbBits + (1<<15)) >> 16);
+        statePtr->value = (nbBitsOut << 16) - symbolTT.deltaNbBits;
+        statePtr->value = stateTable[(statePtr->value >> nbBitsOut) + symbolTT.deltaFindState];
+    }
+}
+
+MEM_STATIC void FSE_encodeSymbol(BIT_CStream_t* bitC, FSE_CState_t* statePtr, U32 symbol)
+{
+    const FSE_symbolCompressionTransform symbolTT = ((const FSE_symbolCompressionTransform*)(statePtr->symbolTT))[symbol];
+    const U16* const stateTable = (const U16*)(statePtr->stateTable);
+    U32 nbBitsOut  = (U32)((statePtr->value + symbolTT.deltaNbBits) >> 16);
+    BIT_addBits(bitC, statePtr->value, nbBitsOut);
+    statePtr->value = stateTable[ (statePtr->value >> nbBitsOut) + symbolTT.deltaFindState];
+}
+
+MEM_STATIC void FSE_flushCState(BIT_CStream_t* bitC, const FSE_CState_t* statePtr)
+{
+    BIT_addBits(bitC, statePtr->value, statePtr->stateLog);
+    BIT_flushBits(bitC);
+}
+
+
+/* ======    Decompression    ====== */
+
+typedef struct {
+    U16 tableLog;
+    U16 fastMode;
+} FSE_DTableHeader;   /* sizeof U32 */
+
+typedef struct
+{
+    unsigned short newState;
+    unsigned char  symbol;
+    unsigned char  nbBits;
+} FSE_decode_t;   /* size == U32 */
+
+MEM_STATIC void FSE_initDState(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD, const FSE_DTable* dt)
+{
+    const void* ptr = dt;
+    const FSE_DTableHeader* const DTableH = (const FSE_DTableHeader*)ptr;
+    DStatePtr->state = BIT_readBits(bitD, DTableH->tableLog);
+    BIT_reloadDStream(bitD);
+    DStatePtr->table = dt + 1;
+}
+
+MEM_STATIC BYTE FSE_peekSymbol(const FSE_DState_t* DStatePtr)
+{
+    FSE_decode_t const DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
+    return DInfo.symbol;
+}
+
+MEM_STATIC void FSE_updateState(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD)
+{
+    FSE_decode_t const DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
+    U32 const nbBits = DInfo.nbBits;
+    size_t const lowBits = BIT_readBits(bitD, nbBits);
+    DStatePtr->state = DInfo.newState + lowBits;
+}
+
+MEM_STATIC BYTE FSE_decodeSymbol(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD)
+{
+    FSE_decode_t const DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
+    U32 const nbBits = DInfo.nbBits;
+    BYTE const symbol = DInfo.symbol;
+    size_t const lowBits = BIT_readBits(bitD, nbBits);
+
+    DStatePtr->state = DInfo.newState + lowBits;
+    return symbol;
+}
+
+/*! FSE_decodeSymbolFast() :
+    unsafe, only works if no symbol has a probability > 50% */
+MEM_STATIC BYTE FSE_decodeSymbolFast(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD)
+{
+    FSE_decode_t const DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
+    U32 const nbBits = DInfo.nbBits;
+    BYTE const symbol = DInfo.symbol;
+    size_t const lowBits = BIT_readBitsFast(bitD, nbBits);
+
+    DStatePtr->state = DInfo.newState + lowBits;
+    return symbol;
+}
+
+MEM_STATIC unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr)
+{
+    return DStatePtr->state == 0;
+}
+
+
+
+#ifndef FSE_COMMONDEFS_ONLY
+
+/* **************************************************************
+*  Tuning parameters
+****************************************************************/
+/*!MEMORY_USAGE :
+*  Memory usage formula : N->2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; etc.)
+*  Increasing memory usage improves compression ratio
+*  Reduced memory usage can improve speed, due to cache effect
+*  Recommended max value is 14, for 16KB, which nicely fits into Intel x86 L1 cache */
+#ifndef FSE_MAX_MEMORY_USAGE
+#  define FSE_MAX_MEMORY_USAGE 14
+#endif
+#ifndef FSE_DEFAULT_MEMORY_USAGE
+#  define FSE_DEFAULT_MEMORY_USAGE 13
+#endif
+
+/*!FSE_MAX_SYMBOL_VALUE :
+*  Maximum symbol value authorized.
+*  Required for proper stack allocation */
+#ifndef FSE_MAX_SYMBOL_VALUE
+#  define FSE_MAX_SYMBOL_VALUE 255
+#endif
+
+/* **************************************************************
+*  template functions type & suffix
+****************************************************************/
+#define FSE_FUNCTION_TYPE BYTE
+#define FSE_FUNCTION_EXTENSION
+#define FSE_DECODE_TYPE FSE_decode_t
+
+
+#endif   /* !FSE_COMMONDEFS_ONLY */
+
+
+/* ***************************************************************
+*  Constants
+*****************************************************************/
+#define FSE_MAX_TABLELOG  (FSE_MAX_MEMORY_USAGE-2)
+#define FSE_MAX_TABLESIZE (1U< FSE_TABLELOG_ABSOLUTE_MAX
+#  error "FSE_MAX_TABLELOG > FSE_TABLELOG_ABSOLUTE_MAX is not supported"
+#endif
+
+#define FSE_TABLESTEP(tableSize) ((tableSize>>1) + (tableSize>>3) + 3)
+
+
+#endif /* FSE_STATIC_LINKING_ONLY */
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif  /* FSE_H */
diff --git a/lib/common/fse_decompress.c b/lib/common/fse_decompress.c
new file mode 100644
index 0000000..1479a5e
--- /dev/null
+++ b/lib/common/fse_decompress.c
@@ -0,0 +1,329 @@
+/* ******************************************************************
+   FSE : Finite State Entropy decoder
+   Copyright (C) 2013-2015, Yann Collet.
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+    You can contact the author at :
+    - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
+    - Public forum : https://groups.google.com/forum/#!forum/lz4c
+****************************************************************** */
+
+
+/* **************************************************************
+*  Compiler specifics
+****************************************************************/
+#ifdef _MSC_VER    /* Visual Studio */
+#  define FORCE_INLINE static __forceinline
+#  include                     /* For Visual 2005 */
+#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
+#  pragma warning(disable : 4214)        /* disable: C4214: non-int bitfields */
+#else
+#  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+#    ifdef __GNUC__
+#      define FORCE_INLINE static inline __attribute__((always_inline))
+#    else
+#      define FORCE_INLINE static inline
+#    endif
+#  else
+#    define FORCE_INLINE static
+#  endif /* __STDC_VERSION__ */
+#endif
+
+
+/* **************************************************************
+*  Includes
+****************************************************************/
+#include      /* malloc, free, qsort */
+#include      /* memcpy, memset */
+#include       /* printf (debug) */
+#include "bitstream.h"
+#define FSE_STATIC_LINKING_ONLY
+#include "fse.h"
+
+
+/* **************************************************************
+*  Error Management
+****************************************************************/
+#define FSE_isError ERR_isError
+#define FSE_STATIC_ASSERT(c) { enum { FSE_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
+
+/* check and forward error code */
+#define CHECK_F(f) { size_t const e = f; if (FSE_isError(e)) return e; }
+
+
+/* **************************************************************
+*  Templates
+****************************************************************/
+/*
+  designed to be included
+  for type-specific functions (template emulation in C)
+  Objective is to write these functions only once, for improved maintenance
+*/
+
+/* safety checks */
+#ifndef FSE_FUNCTION_EXTENSION
+#  error "FSE_FUNCTION_EXTENSION must be defined"
+#endif
+#ifndef FSE_FUNCTION_TYPE
+#  error "FSE_FUNCTION_TYPE must be defined"
+#endif
+
+/* Function names */
+#define FSE_CAT(X,Y) X##Y
+#define FSE_FUNCTION_NAME(X,Y) FSE_CAT(X,Y)
+#define FSE_TYPE_NAME(X,Y) FSE_CAT(X,Y)
+
+
+/* Function templates */
+FSE_DTable* FSE_createDTable (unsigned tableLog)
+{
+    if (tableLog > FSE_TABLELOG_ABSOLUTE_MAX) tableLog = FSE_TABLELOG_ABSOLUTE_MAX;
+    return (FSE_DTable*)malloc( FSE_DTABLE_SIZE_U32(tableLog) * sizeof (U32) );
+}
+
+void FSE_freeDTable (FSE_DTable* dt)
+{
+    free(dt);
+}
+
+size_t FSE_buildDTable(FSE_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
+{
+    void* const tdPtr = dt+1;   /* because *dt is unsigned, 32-bits aligned on 32-bits */
+    FSE_DECODE_TYPE* const tableDecode = (FSE_DECODE_TYPE*) (tdPtr);
+    U16 symbolNext[FSE_MAX_SYMBOL_VALUE+1];
+
+    U32 const maxSV1 = maxSymbolValue + 1;
+    U32 const tableSize = 1 << tableLog;
+    U32 highThreshold = tableSize-1;
+
+    /* Sanity Checks */
+    if (maxSymbolValue > FSE_MAX_SYMBOL_VALUE) return ERROR(maxSymbolValue_tooLarge);
+    if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge);
+
+    /* Init, lay down lowprob symbols */
+    {   FSE_DTableHeader DTableH;
+        DTableH.tableLog = (U16)tableLog;
+        DTableH.fastMode = 1;
+        {   S16 const largeLimit= (S16)(1 << (tableLog-1));
+            U32 s;
+            for (s=0; s= largeLimit) DTableH.fastMode=0;
+                    symbolNext[s] = normalizedCounter[s];
+        }   }   }
+        memcpy(dt, &DTableH, sizeof(DTableH));
+    }
+
+    /* Spread symbols */
+    {   U32 const tableMask = tableSize-1;
+        U32 const step = FSE_TABLESTEP(tableSize);
+        U32 s, position = 0;
+        for (s=0; s highThreshold) position = (position + step) & tableMask;   /* lowprob area */
+        }   }
+        if (position!=0) return ERROR(GENERIC);   /* position must reach all cells once, otherwise normalizedCounter is incorrect */
+    }
+
+    /* Build Decoding table */
+    {   U32 u;
+        for (u=0; utableLog = 0;
+    DTableH->fastMode = 0;
+
+    cell->newState = 0;
+    cell->symbol = symbolValue;
+    cell->nbBits = 0;
+
+    return 0;
+}
+
+
+size_t FSE_buildDTable_raw (FSE_DTable* dt, unsigned nbBits)
+{
+    void* ptr = dt;
+    FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)ptr;
+    void* dPtr = dt + 1;
+    FSE_decode_t* const dinfo = (FSE_decode_t*)dPtr;
+    const unsigned tableSize = 1 << nbBits;
+    const unsigned tableMask = tableSize - 1;
+    const unsigned maxSV1 = tableMask+1;
+    unsigned s;
+
+    /* Sanity checks */
+    if (nbBits < 1) return ERROR(GENERIC);         /* min size */
+
+    /* Build Decoding Table */
+    DTableH->tableLog = (U16)nbBits;
+    DTableH->fastMode = 1;
+    for (s=0; s sizeof(bitD.bitContainer)*8)    /* This test must be static */
+            BIT_reloadDStream(&bitD);
+
+        op[1] = FSE_GETSYMBOL(&state2);
+
+        if (FSE_MAX_TABLELOG*4+7 > sizeof(bitD.bitContainer)*8)    /* This test must be static */
+            { if (BIT_reloadDStream(&bitD) > BIT_DStream_unfinished) { op+=2; break; } }
+
+        op[2] = FSE_GETSYMBOL(&state1);
+
+        if (FSE_MAX_TABLELOG*2+7 > sizeof(bitD.bitContainer)*8)    /* This test must be static */
+            BIT_reloadDStream(&bitD);
+
+        op[3] = FSE_GETSYMBOL(&state2);
+    }
+
+    /* tail */
+    /* note : BIT_reloadDStream(&bitD) >= FSE_DStream_partiallyFilled; Ends at exactly BIT_DStream_completed */
+    while (1) {
+        if (op>(omax-2)) return ERROR(dstSize_tooSmall);
+        *op++ = FSE_GETSYMBOL(&state1);
+        if (BIT_reloadDStream(&bitD)==BIT_DStream_overflow) {
+            *op++ = FSE_GETSYMBOL(&state2);
+            break;
+        }
+
+        if (op>(omax-2)) return ERROR(dstSize_tooSmall);
+        *op++ = FSE_GETSYMBOL(&state2);
+        if (BIT_reloadDStream(&bitD)==BIT_DStream_overflow) {
+            *op++ = FSE_GETSYMBOL(&state1);
+            break;
+    }   }
+
+    return op-ostart;
+}
+
+
+size_t FSE_decompress_usingDTable(void* dst, size_t originalSize,
+                            const void* cSrc, size_t cSrcSize,
+                            const FSE_DTable* dt)
+{
+    const void* ptr = dt;
+    const FSE_DTableHeader* DTableH = (const FSE_DTableHeader*)ptr;
+    const U32 fastMode = DTableH->fastMode;
+
+    /* select fast mode (static) */
+    if (fastMode) return FSE_decompress_usingDTable_generic(dst, originalSize, cSrc, cSrcSize, dt, 1);
+    return FSE_decompress_usingDTable_generic(dst, originalSize, cSrc, cSrcSize, dt, 0);
+}
+
+
+size_t FSE_decompress_wksp(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, FSE_DTable* workSpace, unsigned maxLog)
+{
+    const BYTE* const istart = (const BYTE*)cSrc;
+    const BYTE* ip = istart;
+    short counting[FSE_MAX_SYMBOL_VALUE+1];
+    unsigned tableLog;
+    unsigned maxSymbolValue = FSE_MAX_SYMBOL_VALUE;
+
+    /* normal FSE decoding mode */
+    size_t const NCountLength = FSE_readNCount (counting, &maxSymbolValue, &tableLog, istart, cSrcSize);
+    if (FSE_isError(NCountLength)) return NCountLength;
+    //if (NCountLength >= cSrcSize) return ERROR(srcSize_wrong);   /* too small input size; supposed to be already checked in NCountLength, only remaining case : NCountLength==cSrcSize */
+    if (tableLog > maxLog) return ERROR(tableLog_tooLarge);
+    ip += NCountLength;
+    cSrcSize -= NCountLength;
+
+    CHECK_F( FSE_buildDTable (workSpace, counting, maxSymbolValue, tableLog) );
+
+    return FSE_decompress_usingDTable (dst, dstCapacity, ip, cSrcSize, workSpace);   /* always return, even if it is an error code */
+}
+
+
+typedef FSE_DTable DTable_max_t[FSE_DTABLE_SIZE_U32(FSE_MAX_TABLELOG)];
+
+size_t FSE_decompress(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize)
+{
+    DTable_max_t dt;   /* Static analyzer seems unable to understand this table will be properly initialized later */
+    return FSE_decompress_wksp(dst, dstCapacity, cSrc, cSrcSize, dt, FSE_MAX_TABLELOG);
+}
+
+
+
+#endif   /* FSE_COMMONDEFS_ONLY */
diff --git a/lib/common/huf.h b/lib/common/huf.h
new file mode 100644
index 0000000..9427ae8
--- /dev/null
+++ b/lib/common/huf.h
@@ -0,0 +1,238 @@
+/* ******************************************************************
+   Huffman coder, part of New Generation Entropy library
+   header file
+   Copyright (C) 2013-2016, Yann Collet.
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+   You can contact the author at :
+   - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+****************************************************************** */
+#ifndef HUF_H_298734234
+#define HUF_H_298734234
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+
+/* *** Dependencies *** */
+#include     /* size_t */
+
+
+/* *** simple functions *** */
+/**
+HUF_compress() :
+    Compress content from buffer 'src', of size 'srcSize', into buffer 'dst'.
+    'dst' buffer must be already allocated.
+    Compression runs faster if `dstCapacity` >= HUF_compressBound(srcSize).
+    `srcSize` must be <= `HUF_BLOCKSIZE_MAX` == 128 KB.
+    @return : size of compressed data (<= `dstCapacity`).
+    Special values : if return == 0, srcData is not compressible => Nothing is stored within dst !!!
+                     if return == 1, srcData is a single repeated byte symbol (RLE compression).
+                     if HUF_isError(return), compression failed (more details using HUF_getErrorName())
+*/
+size_t HUF_compress(void* dst, size_t dstCapacity,
+              const void* src, size_t srcSize);
+
+/**
+HUF_decompress() :
+    Decompress HUF data from buffer 'cSrc', of size 'cSrcSize',
+    into already allocated buffer 'dst', of minimum size 'dstSize'.
+    `originalSize` : **must** be the ***exact*** size of original (uncompressed) data.
+    Note : in contrast with FSE, HUF_decompress can regenerate
+           RLE (cSrcSize==1) and uncompressed (cSrcSize==dstSize) data,
+           because it knows size to regenerate.
+    @return : size of regenerated data (== originalSize),
+              or an error code, which can be tested using HUF_isError()
+*/
+size_t HUF_decompress(void* dst,  size_t originalSize,
+                const void* cSrc, size_t cSrcSize);
+
+
+/* ***   Tool functions *** */
+#define HUF_BLOCKSIZE_MAX (128 * 1024)       /**< maximum input size for a single block compressed with HUF_compress */
+size_t HUF_compressBound(size_t size);       /**< maximum compressed size (worst case) */
+
+/* Error Management */
+unsigned    HUF_isError(size_t code);        /**< tells if a return value is an error code */
+const char* HUF_getErrorName(size_t code);   /**< provides error code string (useful for debugging) */
+
+
+/* ***   Advanced function   *** */
+
+/** HUF_compress2() :
+ *   Same as HUF_compress(), but offers direct control over `maxSymbolValue` and `tableLog` .
+ *   `tableLog` must be `<= HUF_TABLELOG_MAX` . */
+size_t HUF_compress2 (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog);
+
+/** HUF_compress4X_wksp() :
+*   Same as HUF_compress2(), but uses externally allocated `workSpace`, which must be a table of >= 1024 unsigned */
+size_t HUF_compress4X_wksp (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize);  /**< `workSpace` must be a table of at least 1024 unsigned */
+
+
+
+#ifdef HUF_STATIC_LINKING_ONLY
+
+/* *** Dependencies *** */
+#include "mem.h"   /* U32 */
+
+
+/* *** Constants *** */
+#define HUF_TABLELOG_ABSOLUTEMAX  15   /* absolute limit of HUF_MAX_TABLELOG. Beyond that value, code does not work */
+#define HUF_TABLELOG_MAX  12           /* max configured tableLog (for static allocation); can be modified up to HUF_ABSOLUTEMAX_TABLELOG */
+#define HUF_TABLELOG_DEFAULT  11       /* tableLog by default, when not specified */
+#define HUF_SYMBOLVALUE_MAX 255
+#if (HUF_TABLELOG_MAX > HUF_TABLELOG_ABSOLUTEMAX)
+#  error "HUF_TABLELOG_MAX is too large !"
+#endif
+
+
+/* ****************************************
+*  Static allocation
+******************************************/
+/* HUF buffer bounds */
+#define HUF_CTABLEBOUND 129
+#define HUF_BLOCKBOUND(size) (size + (size>>8) + 8)   /* only true if incompressible pre-filtered with fast heuristic */
+#define HUF_COMPRESSBOUND(size) (HUF_CTABLEBOUND + HUF_BLOCKBOUND(size))   /* Macro version, useful for static allocation */
+
+/* static allocation of HUF's Compression Table */
+#define HUF_CREATE_STATIC_CTABLE(name, maxSymbolValue) \
+    U32 name##hb[maxSymbolValue+1]; \
+    void* name##hv = &(name##hb); \
+    HUF_CElt* name = (HUF_CElt*)(name##hv)   /* no final ; */
+
+/* static allocation of HUF's DTable */
+typedef U32 HUF_DTable;
+#define HUF_DTABLE_SIZE(maxTableLog)   (1 + (1<<(maxTableLog)))
+#define HUF_CREATE_STATIC_DTABLEX2(DTable, maxTableLog) \
+        HUF_DTable DTable[HUF_DTABLE_SIZE((maxTableLog)-1)] = { ((U32)((maxTableLog)-1) * 0x01000001) }
+#define HUF_CREATE_STATIC_DTABLEX4(DTable, maxTableLog) \
+        HUF_DTable DTable[HUF_DTABLE_SIZE(maxTableLog)] = { ((U32)(maxTableLog) * 0x01000001) }
+
+
+/* ****************************************
+*  Advanced decompression functions
+******************************************/
+size_t HUF_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< single-symbol decoder */
+size_t HUF_decompress4X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< double-symbols decoder */
+
+size_t HUF_decompress4X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< decodes RLE and uncompressed */
+size_t HUF_decompress4X_hufOnly(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize); /**< considers RLE and uncompressed as errors */
+size_t HUF_decompress4X2_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< single-symbol decoder */
+size_t HUF_decompress4X4_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< double-symbols decoder */
+
+
+/* ****************************************
+*  HUF detailed API
+******************************************/
+/*!
+HUF_compress() does the following:
+1. count symbol occurrence from source[] into table count[] using FSE_count()
+2. (optional) refine tableLog using HUF_optimalTableLog()
+3. build Huffman table from count using HUF_buildCTable()
+4. save Huffman table to memory buffer using HUF_writeCTable()
+5. encode the data stream using HUF_compress4X_usingCTable()
+
+The following API allows targeting specific sub-functions for advanced tasks.
+For example, it's possible to compress several blocks using the same 'CTable',
+or to save and regenerate 'CTable' using external methods.
+*/
+/* FSE_count() : find it within "fse.h" */
+unsigned HUF_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue);
+typedef struct HUF_CElt_s HUF_CElt;   /* incomplete type */
+size_t HUF_buildCTable (HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue, unsigned maxNbBits);
+size_t HUF_writeCTable (void* dst, size_t maxDstSize, const HUF_CElt* CTable, unsigned maxSymbolValue, unsigned huffLog);
+size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable);
+
+
+/** HUF_buildCTable_wksp() :
+ *  Same as HUF_buildCTable(), but using externally allocated scratch buffer.
+ *  `workSpace` must be aligned on 4-bytes boundaries, and be at least as large as a table of 1024 unsigned.
+ */
+size_t HUF_buildCTable_wksp (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U32 maxNbBits, void* workSpace, size_t wkspSize);
+
+/*! HUF_readStats() :
+    Read compact Huffman tree, saved by HUF_writeCTable().
+    `huffWeight` is destination buffer.
+    @return : size read from `src` , or an error Code .
+    Note : Needed by HUF_readCTable() and HUF_readDTableXn() . */
+size_t HUF_readStats(BYTE* huffWeight, size_t hwSize, U32* rankStats,
+                     U32* nbSymbolsPtr, U32* tableLogPtr,
+                     const void* src, size_t srcSize);
+
+/** HUF_readCTable() :
+*   Loading a CTable saved with HUF_writeCTable() */
+size_t HUF_readCTable (HUF_CElt* CTable, unsigned maxSymbolValue, const void* src, size_t srcSize);
+
+
+/*
+HUF_decompress() does the following:
+1. select the decompression algorithm (X2, X4) based on pre-computed heuristics
+2. build Huffman table from save, using HUF_readDTableXn()
+3. decode 1 or 4 segments in parallel using HUF_decompressSXn_usingDTable
+*/
+
+/** HUF_selectDecoder() :
+*   Tells which decoder is likely to decode faster,
+*   based on a set of pre-determined metrics.
+*   @return : 0==HUF_decompress4X2, 1==HUF_decompress4X4 .
+*   Assumption : 0 < cSrcSize < dstSize <= 128 KB */
+U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize);
+
+size_t HUF_readDTableX2 (HUF_DTable* DTable, const void* src, size_t srcSize);
+size_t HUF_readDTableX4 (HUF_DTable* DTable, const void* src, size_t srcSize);
+
+size_t HUF_decompress4X_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
+size_t HUF_decompress4X2_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
+size_t HUF_decompress4X4_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
+
+
+/* single stream variants */
+
+size_t HUF_compress1X (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog);
+size_t HUF_compress1X_wksp (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize);  /**< `workSpace` must be a table of at least 1024 unsigned */
+size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable);
+
+size_t HUF_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* single-symbol decoder */
+size_t HUF_decompress1X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* double-symbol decoder */
+
+size_t HUF_decompress1X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);
+size_t HUF_decompress1X2_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< single-symbol decoder */
+size_t HUF_decompress1X4_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< double-symbols decoder */
+
+size_t HUF_decompress1X_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);   /**< automatic selection of sing or double symbol decoder, based on DTable */
+size_t HUF_decompress1X2_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
+size_t HUF_decompress1X4_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
+
+#endif /* HUF_STATIC_LINKING_ONLY */
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif   /* HUF_H_298734234 */
diff --git a/lib/common/mem.h b/lib/common/mem.h
new file mode 100644
index 0000000..32c63dd
--- /dev/null
+++ b/lib/common/mem.h
@@ -0,0 +1,372 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+#ifndef MEM_H_MODULE
+#define MEM_H_MODULE
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/*-****************************************
+*  Dependencies
+******************************************/
+#include      /* size_t, ptrdiff_t */
+#include      /* memcpy */
+
+
+/*-****************************************
+*  Compiler specifics
+******************************************/
+#if defined(_MSC_VER)   /* Visual Studio */
+#   include   /* _byteswap_ulong */
+#   include   /* _byteswap_* */
+#endif
+#if defined(__GNUC__)
+#  define MEM_STATIC static __inline __attribute__((unused))
+#elif defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
+#  define MEM_STATIC static inline
+#elif defined(_MSC_VER)
+#  define MEM_STATIC static __inline
+#else
+#  define MEM_STATIC static  /* this version may generate warnings for unused static functions; disable the relevant warning */
+#endif
+
+/* code only tested on 32 and 64 bits systems */
+#define MEM_STATIC_ASSERT(c)   { enum { XXH_static_assert = 1/(int)(!!(c)) }; }
+MEM_STATIC void MEM_check(void) { MEM_STATIC_ASSERT((sizeof(size_t)==4) || (sizeof(size_t)==8)); }
+
+
+/*-**************************************************************
+*  Basic Types
+*****************************************************************/
+#if  !defined (__VMS) && (defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) )
+# include 
+  typedef  uint8_t BYTE;
+  typedef uint16_t U16;
+  typedef  int16_t S16;
+  typedef uint32_t U32;
+  typedef  int32_t S32;
+  typedef uint64_t U64;
+  typedef  int64_t S64;
+  typedef intptr_t iPtrDiff;
+#else
+  typedef unsigned char      BYTE;
+  typedef unsigned short      U16;
+  typedef   signed short      S16;
+  typedef unsigned int        U32;
+  typedef   signed int        S32;
+  typedef unsigned long long  U64;
+  typedef   signed long long  S64;
+  typedef ptrdiff_t      iPtrDiff;
+#endif
+
+
+/*-**************************************************************
+*  Memory I/O
+*****************************************************************/
+/* MEM_FORCE_MEMORY_ACCESS :
+ * By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
+ * Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
+ * The below switch allow to select different access method for improved performance.
+ * Method 0 (default) : use `memcpy()`. Safe and portable.
+ * Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
+ *            This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
+ * Method 2 : direct access. This method is portable but violate C standard.
+ *            It can generate buggy code on targets depending on alignment.
+ *            In some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
+ * See http://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
+ * Prefer these methods in priority order (0 > 1 > 2)
+ */
+#ifndef MEM_FORCE_MEMORY_ACCESS   /* can be defined externally, on command line for example */
+#  if defined(__GNUC__) && ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) )
+#    define MEM_FORCE_MEMORY_ACCESS 2
+#  elif defined(__INTEL_COMPILER) /*|| defined(_MSC_VER)*/ || \
+  (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7S__) ))
+#    define MEM_FORCE_MEMORY_ACCESS 1
+#  endif
+#endif
+
+MEM_STATIC unsigned MEM_32bits(void) { return sizeof(size_t)==4; }
+MEM_STATIC unsigned MEM_64bits(void) { return sizeof(size_t)==8; }
+
+MEM_STATIC unsigned MEM_isLittleEndian(void)
+{
+    const union { U32 u; BYTE c[4]; } one = { 1 };   /* don't use static : performance detrimental  */
+    return one.c[0];
+}
+
+#if defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==2)
+
+/* violates C standard, by lying on structure alignment.
+Only use if no other choice to achieve best performance on target platform */
+MEM_STATIC U16 MEM_read16(const void* memPtr) { return *(const U16*) memPtr; }
+MEM_STATIC U32 MEM_read32(const void* memPtr) { return *(const U32*) memPtr; }
+MEM_STATIC U64 MEM_read64(const void* memPtr) { return *(const U64*) memPtr; }
+MEM_STATIC U64 MEM_readST(const void* memPtr) { return *(const size_t*) memPtr; }
+
+MEM_STATIC void MEM_write16(void* memPtr, U16 value) { *(U16*)memPtr = value; }
+MEM_STATIC void MEM_write32(void* memPtr, U32 value) { *(U32*)memPtr = value; }
+MEM_STATIC void MEM_write64(void* memPtr, U64 value) { *(U64*)memPtr = value; }
+
+#elif defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==1)
+
+/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
+/* currently only defined for gcc and icc */
+#if defined(_MSC_VER) || (defined(__INTEL_COMPILER) && defined(WIN32))
+	__pragma( pack(push, 1) )
+    typedef union { U16 u16; U32 u32; U64 u64; size_t st; } unalign;
+    __pragma( pack(pop) )
+#else
+    typedef union { U16 u16; U32 u32; U64 u64; size_t st; } __attribute__((packed)) unalign;
+#endif
+
+MEM_STATIC U16 MEM_read16(const void* ptr) { return ((const unalign*)ptr)->u16; }
+MEM_STATIC U32 MEM_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
+MEM_STATIC U64 MEM_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
+MEM_STATIC U64 MEM_readST(const void* ptr) { return ((const unalign*)ptr)->st; }
+
+MEM_STATIC void MEM_write16(void* memPtr, U16 value) { ((unalign*)memPtr)->u16 = value; }
+MEM_STATIC void MEM_write32(void* memPtr, U32 value) { ((unalign*)memPtr)->u32 = value; }
+MEM_STATIC void MEM_write64(void* memPtr, U64 value) { ((unalign*)memPtr)->u64 = value; }
+
+#else
+
+/* default method, safe and standard.
+   can sometimes prove slower */
+
+MEM_STATIC U16 MEM_read16(const void* memPtr)
+{
+    U16 val; memcpy(&val, memPtr, sizeof(val)); return val;
+}
+
+MEM_STATIC U32 MEM_read32(const void* memPtr)
+{
+    U32 val; memcpy(&val, memPtr, sizeof(val)); return val;
+}
+
+MEM_STATIC U64 MEM_read64(const void* memPtr)
+{
+    U64 val; memcpy(&val, memPtr, sizeof(val)); return val;
+}
+
+MEM_STATIC size_t MEM_readST(const void* memPtr)
+{
+    size_t val; memcpy(&val, memPtr, sizeof(val)); return val;
+}
+
+MEM_STATIC void MEM_write16(void* memPtr, U16 value)
+{
+    memcpy(memPtr, &value, sizeof(value));
+}
+
+MEM_STATIC void MEM_write32(void* memPtr, U32 value)
+{
+    memcpy(memPtr, &value, sizeof(value));
+}
+
+MEM_STATIC void MEM_write64(void* memPtr, U64 value)
+{
+    memcpy(memPtr, &value, sizeof(value));
+}
+
+#endif /* MEM_FORCE_MEMORY_ACCESS */
+
+MEM_STATIC U32 MEM_swap32(U32 in)
+{
+#if defined(_MSC_VER)     /* Visual Studio */
+    return _byteswap_ulong(in);
+#elif defined (__GNUC__)
+    return __builtin_bswap32(in);
+#else
+    return  ((in << 24) & 0xff000000 ) |
+            ((in <<  8) & 0x00ff0000 ) |
+            ((in >>  8) & 0x0000ff00 ) |
+            ((in >> 24) & 0x000000ff );
+#endif
+}
+
+MEM_STATIC U64 MEM_swap64(U64 in)
+{
+#if defined(_MSC_VER)     /* Visual Studio */
+    return _byteswap_uint64(in);
+#elif defined (__GNUC__)
+    return __builtin_bswap64(in);
+#else
+    return  ((in << 56) & 0xff00000000000000ULL) |
+            ((in << 40) & 0x00ff000000000000ULL) |
+            ((in << 24) & 0x0000ff0000000000ULL) |
+            ((in << 8)  & 0x000000ff00000000ULL) |
+            ((in >> 8)  & 0x00000000ff000000ULL) |
+            ((in >> 24) & 0x0000000000ff0000ULL) |
+            ((in >> 40) & 0x000000000000ff00ULL) |
+            ((in >> 56) & 0x00000000000000ffULL);
+#endif
+}
+
+MEM_STATIC size_t MEM_swapST(size_t in)
+{
+    if (MEM_32bits())
+        return (size_t)MEM_swap32((U32)in);
+    else
+        return (size_t)MEM_swap64((U64)in);
+}
+
+/*=== Little endian r/w ===*/
+
+MEM_STATIC U16 MEM_readLE16(const void* memPtr)
+{
+    if (MEM_isLittleEndian())
+        return MEM_read16(memPtr);
+    else {
+        const BYTE* p = (const BYTE*)memPtr;
+        return (U16)(p[0] + (p[1]<<8));
+    }
+}
+
+MEM_STATIC void MEM_writeLE16(void* memPtr, U16 val)
+{
+    if (MEM_isLittleEndian()) {
+        MEM_write16(memPtr, val);
+    } else {
+        BYTE* p = (BYTE*)memPtr;
+        p[0] = (BYTE)val;
+        p[1] = (BYTE)(val>>8);
+    }
+}
+
+MEM_STATIC U32 MEM_readLE24(const void* memPtr)
+{
+    return MEM_readLE16(memPtr) + (((const BYTE*)memPtr)[2] << 16);
+}
+
+MEM_STATIC void MEM_writeLE24(void* memPtr, U32 val)
+{
+    MEM_writeLE16(memPtr, (U16)val);
+    ((BYTE*)memPtr)[2] = (BYTE)(val>>16);
+}
+
+MEM_STATIC U32 MEM_readLE32(const void* memPtr)
+{
+    if (MEM_isLittleEndian())
+        return MEM_read32(memPtr);
+    else
+        return MEM_swap32(MEM_read32(memPtr));
+}
+
+MEM_STATIC void MEM_writeLE32(void* memPtr, U32 val32)
+{
+    if (MEM_isLittleEndian())
+        MEM_write32(memPtr, val32);
+    else
+        MEM_write32(memPtr, MEM_swap32(val32));
+}
+
+MEM_STATIC U64 MEM_readLE64(const void* memPtr)
+{
+    if (MEM_isLittleEndian())
+        return MEM_read64(memPtr);
+    else
+        return MEM_swap64(MEM_read64(memPtr));
+}
+
+MEM_STATIC void MEM_writeLE64(void* memPtr, U64 val64)
+{
+    if (MEM_isLittleEndian())
+        MEM_write64(memPtr, val64);
+    else
+        MEM_write64(memPtr, MEM_swap64(val64));
+}
+
+MEM_STATIC size_t MEM_readLEST(const void* memPtr)
+{
+    if (MEM_32bits())
+        return (size_t)MEM_readLE32(memPtr);
+    else
+        return (size_t)MEM_readLE64(memPtr);
+}
+
+MEM_STATIC void MEM_writeLEST(void* memPtr, size_t val)
+{
+    if (MEM_32bits())
+        MEM_writeLE32(memPtr, (U32)val);
+    else
+        MEM_writeLE64(memPtr, (U64)val);
+}
+
+/*=== Big endian r/w ===*/
+
+MEM_STATIC U32 MEM_readBE32(const void* memPtr)
+{
+    if (MEM_isLittleEndian())
+        return MEM_swap32(MEM_read32(memPtr));
+    else
+        return MEM_read32(memPtr);
+}
+
+MEM_STATIC void MEM_writeBE32(void* memPtr, U32 val32)
+{
+    if (MEM_isLittleEndian())
+        MEM_write32(memPtr, MEM_swap32(val32));
+    else
+        MEM_write32(memPtr, val32);
+}
+
+MEM_STATIC U64 MEM_readBE64(const void* memPtr)
+{
+    if (MEM_isLittleEndian())
+        return MEM_swap64(MEM_read64(memPtr));
+    else
+        return MEM_read64(memPtr);
+}
+
+MEM_STATIC void MEM_writeBE64(void* memPtr, U64 val64)
+{
+    if (MEM_isLittleEndian())
+        MEM_write64(memPtr, MEM_swap64(val64));
+    else
+        MEM_write64(memPtr, val64);
+}
+
+MEM_STATIC size_t MEM_readBEST(const void* memPtr)
+{
+    if (MEM_32bits())
+        return (size_t)MEM_readBE32(memPtr);
+    else
+        return (size_t)MEM_readBE64(memPtr);
+}
+
+MEM_STATIC void MEM_writeBEST(void* memPtr, size_t val)
+{
+    if (MEM_32bits())
+        MEM_writeBE32(memPtr, (U32)val);
+    else
+        MEM_writeBE64(memPtr, (U64)val);
+}
+
+
+/* function safe only for comparisons */
+MEM_STATIC U32 MEM_readMINMATCH(const void* memPtr, U32 length)
+{
+    switch (length)
+    {
+    default :
+    case 4 : return MEM_read32(memPtr);
+    case 3 : if (MEM_isLittleEndian())
+                return MEM_read32(memPtr)<<8;
+             else
+                return MEM_read32(memPtr)>>8;
+    }
+}
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* MEM_H_MODULE */
diff --git a/lib/common/xxhash.c b/lib/common/xxhash.c
new file mode 100644
index 0000000..29e4fa6
--- /dev/null
+++ b/lib/common/xxhash.c
@@ -0,0 +1,867 @@
+/*
+*  xxHash - Fast Hash algorithm
+*  Copyright (C) 2012-2016, Yann Collet
+*
+*  BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+*
+*  Redistribution and use in source and binary forms, with or without
+*  modification, are permitted provided that the following conditions are
+*  met:
+*
+*  * Redistributions of source code must retain the above copyright
+*  notice, this list of conditions and the following disclaimer.
+*  * Redistributions in binary form must reproduce the above
+*  copyright notice, this list of conditions and the following disclaimer
+*  in the documentation and/or other materials provided with the
+*  distribution.
+*
+*  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+*  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+*  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+*  A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+*  OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+*  SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+*  LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+*  DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+*  THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+*  (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+*  OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*
+*  You can contact the author at :
+*  - xxHash homepage: http://www.xxhash.com
+*  - xxHash source repository : https://github.com/Cyan4973/xxHash
+*/
+
+
+/* *************************************
+*  Tuning parameters
+***************************************/
+/*!XXH_FORCE_MEMORY_ACCESS :
+ * By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
+ * Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
+ * The below switch allow to select different access method for improved performance.
+ * Method 0 (default) : use `memcpy()`. Safe and portable.
+ * Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
+ *            This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
+ * Method 2 : direct access. This method doesn't depend on compiler but violate C standard.
+ *            It can generate buggy code on targets which do not support unaligned memory accesses.
+ *            But in some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
+ * See http://stackoverflow.com/a/32095106/646947 for details.
+ * Prefer these methods in priority order (0 > 1 > 2)
+ */
+#ifndef XXH_FORCE_MEMORY_ACCESS   /* can be defined externally, on command line for example */
+#  if defined(__GNUC__) && ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) )
+#    define XXH_FORCE_MEMORY_ACCESS 2
+#  elif (defined(__INTEL_COMPILER) && !defined(WIN32)) || \
+  (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7S__) ))
+#    define XXH_FORCE_MEMORY_ACCESS 1
+#  endif
+#endif
+
+/*!XXH_ACCEPT_NULL_INPUT_POINTER :
+ * If the input pointer is a null pointer, xxHash default behavior is to trigger a memory access error, since it is a bad pointer.
+ * When this option is enabled, xxHash output for null input pointers will be the same as a null-length input.
+ * By default, this option is disabled. To enable it, uncomment below define :
+ */
+/* #define XXH_ACCEPT_NULL_INPUT_POINTER 1 */
+
+/*!XXH_FORCE_NATIVE_FORMAT :
+ * By default, xxHash library provides endian-independant Hash values, based on little-endian convention.
+ * Results are therefore identical for little-endian and big-endian CPU.
+ * This comes at a performance cost for big-endian CPU, since some swapping is required to emulate little-endian format.
+ * Should endian-independance be of no importance for your application, you may set the #define below to 1,
+ * to improve speed for Big-endian CPU.
+ * This option has no impact on Little_Endian CPU.
+ */
+#ifndef XXH_FORCE_NATIVE_FORMAT   /* can be defined externally */
+#  define XXH_FORCE_NATIVE_FORMAT 0
+#endif
+
+/*!XXH_FORCE_ALIGN_CHECK :
+ * This is a minor performance trick, only useful with lots of very small keys.
+ * It means : check for aligned/unaligned input.
+ * The check costs one initial branch per hash; set to 0 when the input data
+ * is guaranteed to be aligned.
+ */
+#ifndef XXH_FORCE_ALIGN_CHECK /* can be defined externally */
+#  if defined(__i386) || defined(_M_IX86) || defined(__x86_64__) || defined(_M_X64)
+#    define XXH_FORCE_ALIGN_CHECK 0
+#  else
+#    define XXH_FORCE_ALIGN_CHECK 1
+#  endif
+#endif
+
+
+/* *************************************
+*  Includes & Memory related functions
+***************************************/
+/* Modify the local functions below should you wish to use some other memory routines */
+/* for malloc(), free() */
+#include 
+static void* XXH_malloc(size_t s) { return malloc(s); }
+static void  XXH_free  (void* p)  { free(p); }
+/* for memcpy() */
+#include 
+static void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcpy(dest,src,size); }
+
+#define XXH_STATIC_LINKING_ONLY
+#include "xxhash.h"
+
+
+/* *************************************
+*  Compiler Specific Options
+***************************************/
+#ifdef _MSC_VER    /* Visual Studio */
+#  pragma warning(disable : 4127)      /* disable: C4127: conditional expression is constant */
+#  define FORCE_INLINE static __forceinline
+#else
+#  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+#    ifdef __GNUC__
+#      define FORCE_INLINE static inline __attribute__((always_inline))
+#    else
+#      define FORCE_INLINE static inline
+#    endif
+#  else
+#    define FORCE_INLINE static
+#  endif /* __STDC_VERSION__ */
+#endif
+
+
+/* *************************************
+*  Basic Types
+***************************************/
+#ifndef MEM_MODULE
+# define MEM_MODULE
+# if !defined (__VMS) && (defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) )
+#   include 
+    typedef uint8_t  BYTE;
+    typedef uint16_t U16;
+    typedef uint32_t U32;
+    typedef  int32_t S32;
+    typedef uint64_t U64;
+#  else
+    typedef unsigned char      BYTE;
+    typedef unsigned short     U16;
+    typedef unsigned int       U32;
+    typedef   signed int       S32;
+    typedef unsigned long long U64;   /* if your compiler doesn't support unsigned long long, replace by another 64-bit type here. Note that xxhash.h will also need to be updated. */
+#  endif
+#endif
+
+
+#if (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==2))
+
+/* Force direct memory access. Only works on CPU which support unaligned memory access in hardware */
+static U32 XXH_read32(const void* memPtr) { return *(const U32*) memPtr; }
+static U64 XXH_read64(const void* memPtr) { return *(const U64*) memPtr; }
+
+#elif (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==1))
+
+/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
+/* currently only defined for gcc and icc */
+typedef union { U32 u32; U64 u64; } __attribute__((packed)) unalign;
+
+static U32 XXH_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
+static U64 XXH_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
+
+#else
+
+/* portable and safe solution. Generally efficient.
+ * see : http://stackoverflow.com/a/32095106/646947
+ */
+
+static U32 XXH_read32(const void* memPtr)
+{
+    U32 val;
+    memcpy(&val, memPtr, sizeof(val));
+    return val;
+}
+
+static U64 XXH_read64(const void* memPtr)
+{
+    U64 val;
+    memcpy(&val, memPtr, sizeof(val));
+    return val;
+}
+
+#endif   /* XXH_FORCE_DIRECT_MEMORY_ACCESS */
+
+
+/* ****************************************
+*  Compiler-specific Functions and Macros
+******************************************/
+#define GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
+
+/* Note : although _rotl exists for minGW (GCC under windows), performance seems poor */
+#if defined(_MSC_VER)
+#  define XXH_rotl32(x,r) _rotl(x,r)
+#  define XXH_rotl64(x,r) _rotl64(x,r)
+#else
+#  define XXH_rotl32(x,r) ((x << r) | (x >> (32 - r)))
+#  define XXH_rotl64(x,r) ((x << r) | (x >> (64 - r)))
+#endif
+
+#if defined(_MSC_VER)     /* Visual Studio */
+#  define XXH_swap32 _byteswap_ulong
+#  define XXH_swap64 _byteswap_uint64
+#elif GCC_VERSION >= 403
+#  define XXH_swap32 __builtin_bswap32
+#  define XXH_swap64 __builtin_bswap64
+#else
+static U32 XXH_swap32 (U32 x)
+{
+    return  ((x << 24) & 0xff000000 ) |
+            ((x <<  8) & 0x00ff0000 ) |
+            ((x >>  8) & 0x0000ff00 ) |
+            ((x >> 24) & 0x000000ff );
+}
+static U64 XXH_swap64 (U64 x)
+{
+    return  ((x << 56) & 0xff00000000000000ULL) |
+            ((x << 40) & 0x00ff000000000000ULL) |
+            ((x << 24) & 0x0000ff0000000000ULL) |
+            ((x << 8)  & 0x000000ff00000000ULL) |
+            ((x >> 8)  & 0x00000000ff000000ULL) |
+            ((x >> 24) & 0x0000000000ff0000ULL) |
+            ((x >> 40) & 0x000000000000ff00ULL) |
+            ((x >> 56) & 0x00000000000000ffULL);
+}
+#endif
+
+
+/* *************************************
+*  Architecture Macros
+***************************************/
+typedef enum { XXH_bigEndian=0, XXH_littleEndian=1 } XXH_endianess;
+
+/* XXH_CPU_LITTLE_ENDIAN can be defined externally, for example on the compiler command line */
+#ifndef XXH_CPU_LITTLE_ENDIAN
+    static const int g_one = 1;
+#   define XXH_CPU_LITTLE_ENDIAN   (*(const char*)(&g_one))
+#endif
+
+
+/* ***************************
+*  Memory reads
+*****************************/
+typedef enum { XXH_aligned, XXH_unaligned } XXH_alignment;
+
+FORCE_INLINE U32 XXH_readLE32_align(const void* ptr, XXH_endianess endian, XXH_alignment align)
+{
+    if (align==XXH_unaligned)
+        return endian==XXH_littleEndian ? XXH_read32(ptr) : XXH_swap32(XXH_read32(ptr));
+    else
+        return endian==XXH_littleEndian ? *(const U32*)ptr : XXH_swap32(*(const U32*)ptr);
+}
+
+FORCE_INLINE U32 XXH_readLE32(const void* ptr, XXH_endianess endian)
+{
+    return XXH_readLE32_align(ptr, endian, XXH_unaligned);
+}
+
+static U32 XXH_readBE32(const void* ptr)
+{
+    return XXH_CPU_LITTLE_ENDIAN ? XXH_swap32(XXH_read32(ptr)) : XXH_read32(ptr);
+}
+
+FORCE_INLINE U64 XXH_readLE64_align(const void* ptr, XXH_endianess endian, XXH_alignment align)
+{
+    if (align==XXH_unaligned)
+        return endian==XXH_littleEndian ? XXH_read64(ptr) : XXH_swap64(XXH_read64(ptr));
+    else
+        return endian==XXH_littleEndian ? *(const U64*)ptr : XXH_swap64(*(const U64*)ptr);
+}
+
+FORCE_INLINE U64 XXH_readLE64(const void* ptr, XXH_endianess endian)
+{
+    return XXH_readLE64_align(ptr, endian, XXH_unaligned);
+}
+
+static U64 XXH_readBE64(const void* ptr)
+{
+    return XXH_CPU_LITTLE_ENDIAN ? XXH_swap64(XXH_read64(ptr)) : XXH_read64(ptr);
+}
+
+
+/* *************************************
+*  Macros
+***************************************/
+#define XXH_STATIC_ASSERT(c)   { enum { XXH_static_assert = 1/(int)(!!(c)) }; }    /* use only *after* variable declarations */
+
+
+/* *************************************
+*  Constants
+***************************************/
+static const U32 PRIME32_1 = 2654435761U;
+static const U32 PRIME32_2 = 2246822519U;
+static const U32 PRIME32_3 = 3266489917U;
+static const U32 PRIME32_4 =  668265263U;
+static const U32 PRIME32_5 =  374761393U;
+
+static const U64 PRIME64_1 = 11400714785074694791ULL;
+static const U64 PRIME64_2 = 14029467366897019727ULL;
+static const U64 PRIME64_3 =  1609587929392839161ULL;
+static const U64 PRIME64_4 =  9650029242287828579ULL;
+static const U64 PRIME64_5 =  2870177450012600261ULL;
+
+XXH_PUBLIC_API unsigned XXH_versionNumber (void) { return XXH_VERSION_NUMBER; }
+
+
+/* **************************
+*  Utils
+****************************/
+XXH_PUBLIC_API void XXH32_copyState(XXH32_state_t* restrict dstState, const XXH32_state_t* restrict srcState)
+{
+    memcpy(dstState, srcState, sizeof(*dstState));
+}
+
+XXH_PUBLIC_API void XXH64_copyState(XXH64_state_t* restrict dstState, const XXH64_state_t* restrict srcState)
+{
+    memcpy(dstState, srcState, sizeof(*dstState));
+}
+
+
+/* ***************************
+*  Simple Hash Functions
+*****************************/
+
+static U32 XXH32_round(U32 seed, U32 input)
+{
+    seed += input * PRIME32_2;
+    seed  = XXH_rotl32(seed, 13);
+    seed *= PRIME32_1;
+    return seed;
+}
+
+FORCE_INLINE U32 XXH32_endian_align(const void* input, size_t len, U32 seed, XXH_endianess endian, XXH_alignment align)
+{
+    const BYTE* p = (const BYTE*)input;
+    const BYTE* bEnd = p + len;
+    U32 h32;
+#define XXH_get32bits(p) XXH_readLE32_align(p, endian, align)
+
+#ifdef XXH_ACCEPT_NULL_INPUT_POINTER
+    if (p==NULL) {
+        len=0;
+        bEnd=p=(const BYTE*)(size_t)16;
+    }
+#endif
+
+    if (len>=16) {
+        const BYTE* const limit = bEnd - 16;
+        U32 v1 = seed + PRIME32_1 + PRIME32_2;
+        U32 v2 = seed + PRIME32_2;
+        U32 v3 = seed + 0;
+        U32 v4 = seed - PRIME32_1;
+
+        do {
+            v1 = XXH32_round(v1, XXH_get32bits(p)); p+=4;
+            v2 = XXH32_round(v2, XXH_get32bits(p)); p+=4;
+            v3 = XXH32_round(v3, XXH_get32bits(p)); p+=4;
+            v4 = XXH32_round(v4, XXH_get32bits(p)); p+=4;
+        } while (p<=limit);
+
+        h32 = XXH_rotl32(v1, 1) + XXH_rotl32(v2, 7) + XXH_rotl32(v3, 12) + XXH_rotl32(v4, 18);
+    } else {
+        h32  = seed + PRIME32_5;
+    }
+
+    h32 += (U32) len;
+
+    while (p+4<=bEnd) {
+        h32 += XXH_get32bits(p) * PRIME32_3;
+        h32  = XXH_rotl32(h32, 17) * PRIME32_4 ;
+        p+=4;
+    }
+
+    while (p> 15;
+    h32 *= PRIME32_2;
+    h32 ^= h32 >> 13;
+    h32 *= PRIME32_3;
+    h32 ^= h32 >> 16;
+
+    return h32;
+}
+
+
+XXH_PUBLIC_API unsigned int XXH32 (const void* input, size_t len, unsigned int seed)
+{
+#if 0
+    /* Simple version, good for code maintenance, but unfortunately slow for small inputs */
+    XXH32_CREATESTATE_STATIC(state);
+    XXH32_reset(state, seed);
+    XXH32_update(state, input, len);
+    return XXH32_digest(state);
+#else
+    XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
+
+    if (XXH_FORCE_ALIGN_CHECK) {
+        if ((((size_t)input) & 3) == 0) {   /* Input is 4-bytes aligned, leverage the speed benefit */
+            if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
+                return XXH32_endian_align(input, len, seed, XXH_littleEndian, XXH_aligned);
+            else
+                return XXH32_endian_align(input, len, seed, XXH_bigEndian, XXH_aligned);
+    }   }
+
+    if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
+        return XXH32_endian_align(input, len, seed, XXH_littleEndian, XXH_unaligned);
+    else
+        return XXH32_endian_align(input, len, seed, XXH_bigEndian, XXH_unaligned);
+#endif
+}
+
+
+static U64 XXH64_round(U64 acc, U64 input)
+{
+    acc += input * PRIME64_2;
+    acc  = XXH_rotl64(acc, 31);
+    acc *= PRIME64_1;
+    return acc;
+}
+
+static U64 XXH64_mergeRound(U64 acc, U64 val)
+{
+    val  = XXH64_round(0, val);
+    acc ^= val;
+    acc  = acc * PRIME64_1 + PRIME64_4;
+    return acc;
+}
+
+FORCE_INLINE U64 XXH64_endian_align(const void* input, size_t len, U64 seed, XXH_endianess endian, XXH_alignment align)
+{
+    const BYTE* p = (const BYTE*)input;
+    const BYTE* const bEnd = p + len;
+    U64 h64;
+#define XXH_get64bits(p) XXH_readLE64_align(p, endian, align)
+
+#ifdef XXH_ACCEPT_NULL_INPUT_POINTER
+    if (p==NULL) {
+        len=0;
+        bEnd=p=(const BYTE*)(size_t)32;
+    }
+#endif
+
+    if (len>=32) {
+        const BYTE* const limit = bEnd - 32;
+        U64 v1 = seed + PRIME64_1 + PRIME64_2;
+        U64 v2 = seed + PRIME64_2;
+        U64 v3 = seed + 0;
+        U64 v4 = seed - PRIME64_1;
+
+        do {
+            v1 = XXH64_round(v1, XXH_get64bits(p)); p+=8;
+            v2 = XXH64_round(v2, XXH_get64bits(p)); p+=8;
+            v3 = XXH64_round(v3, XXH_get64bits(p)); p+=8;
+            v4 = XXH64_round(v4, XXH_get64bits(p)); p+=8;
+        } while (p<=limit);
+
+        h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18);
+        h64 = XXH64_mergeRound(h64, v1);
+        h64 = XXH64_mergeRound(h64, v2);
+        h64 = XXH64_mergeRound(h64, v3);
+        h64 = XXH64_mergeRound(h64, v4);
+
+    } else {
+        h64  = seed + PRIME64_5;
+    }
+
+    h64 += (U64) len;
+
+    while (p+8<=bEnd) {
+        U64 const k1 = XXH64_round(0, XXH_get64bits(p));
+        h64 ^= k1;
+        h64  = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4;
+        p+=8;
+    }
+
+    if (p+4<=bEnd) {
+        h64 ^= (U64)(XXH_get32bits(p)) * PRIME64_1;
+        h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
+        p+=4;
+    }
+
+    while (p> 33;
+    h64 *= PRIME64_2;
+    h64 ^= h64 >> 29;
+    h64 *= PRIME64_3;
+    h64 ^= h64 >> 32;
+
+    return h64;
+}
+
+
+XXH_PUBLIC_API unsigned long long XXH64 (const void* input, size_t len, unsigned long long seed)
+{
+#if 0
+    /* Simple version, good for code maintenance, but unfortunately slow for small inputs */
+    XXH64_CREATESTATE_STATIC(state);
+    XXH64_reset(state, seed);
+    XXH64_update(state, input, len);
+    return XXH64_digest(state);
+#else
+    XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
+
+    if (XXH_FORCE_ALIGN_CHECK) {
+        if ((((size_t)input) & 7)==0) {  /* Input is aligned, let's leverage the speed advantage */
+            if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
+                return XXH64_endian_align(input, len, seed, XXH_littleEndian, XXH_aligned);
+            else
+                return XXH64_endian_align(input, len, seed, XXH_bigEndian, XXH_aligned);
+    }   }
+
+    if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
+        return XXH64_endian_align(input, len, seed, XXH_littleEndian, XXH_unaligned);
+    else
+        return XXH64_endian_align(input, len, seed, XXH_bigEndian, XXH_unaligned);
+#endif
+}
+
+
+/* **************************************************
+*  Advanced Hash Functions
+****************************************************/
+
+XXH_PUBLIC_API XXH32_state_t* XXH32_createState(void)
+{
+    return (XXH32_state_t*)XXH_malloc(sizeof(XXH32_state_t));
+}
+XXH_PUBLIC_API XXH_errorcode XXH32_freeState(XXH32_state_t* statePtr)
+{
+    XXH_free(statePtr);
+    return XXH_OK;
+}
+
+XXH_PUBLIC_API XXH64_state_t* XXH64_createState(void)
+{
+    return (XXH64_state_t*)XXH_malloc(sizeof(XXH64_state_t));
+}
+XXH_PUBLIC_API XXH_errorcode XXH64_freeState(XXH64_state_t* statePtr)
+{
+    XXH_free(statePtr);
+    return XXH_OK;
+}
+
+
+/*** Hash feed ***/
+
+XXH_PUBLIC_API XXH_errorcode XXH32_reset(XXH32_state_t* statePtr, unsigned int seed)
+{
+    XXH32_state_t state;   /* using a local state to memcpy() in order to avoid strict-aliasing warnings */
+    memset(&state, 0, sizeof(state)-4);   /* do not write into reserved, for future removal */
+    state.v1 = seed + PRIME32_1 + PRIME32_2;
+    state.v2 = seed + PRIME32_2;
+    state.v3 = seed + 0;
+    state.v4 = seed - PRIME32_1;
+    memcpy(statePtr, &state, sizeof(state));
+    return XXH_OK;
+}
+
+
+XXH_PUBLIC_API XXH_errorcode XXH64_reset(XXH64_state_t* statePtr, unsigned long long seed)
+{
+    XXH64_state_t state;   /* using a local state to memcpy() in order to avoid strict-aliasing warnings */
+    memset(&state, 0, sizeof(state)-8);   /* do not write into reserved, for future removal */
+    state.v1 = seed + PRIME64_1 + PRIME64_2;
+    state.v2 = seed + PRIME64_2;
+    state.v3 = seed + 0;
+    state.v4 = seed - PRIME64_1;
+    memcpy(statePtr, &state, sizeof(state));
+    return XXH_OK;
+}
+
+
+FORCE_INLINE XXH_errorcode XXH32_update_endian (XXH32_state_t* state, const void* input, size_t len, XXH_endianess endian)
+{
+    const BYTE* p = (const BYTE*)input;
+    const BYTE* const bEnd = p + len;
+
+#ifdef XXH_ACCEPT_NULL_INPUT_POINTER
+    if (input==NULL) return XXH_ERROR;
+#endif
+
+    state->total_len_32 += (unsigned)len;
+    state->large_len |= (len>=16) | (state->total_len_32>=16);
+
+    if (state->memsize + len < 16)  {   /* fill in tmp buffer */
+        XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, len);
+        state->memsize += (unsigned)len;
+        return XXH_OK;
+    }
+
+    if (state->memsize) {   /* some data left from previous update */
+        XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, 16-state->memsize);
+        {   const U32* p32 = state->mem32;
+            state->v1 = XXH32_round(state->v1, XXH_readLE32(p32, endian)); p32++;
+            state->v2 = XXH32_round(state->v2, XXH_readLE32(p32, endian)); p32++;
+            state->v3 = XXH32_round(state->v3, XXH_readLE32(p32, endian)); p32++;
+            state->v4 = XXH32_round(state->v4, XXH_readLE32(p32, endian)); p32++;
+        }
+        p += 16-state->memsize;
+        state->memsize = 0;
+    }
+
+    if (p <= bEnd-16) {
+        const BYTE* const limit = bEnd - 16;
+        U32 v1 = state->v1;
+        U32 v2 = state->v2;
+        U32 v3 = state->v3;
+        U32 v4 = state->v4;
+
+        do {
+            v1 = XXH32_round(v1, XXH_readLE32(p, endian)); p+=4;
+            v2 = XXH32_round(v2, XXH_readLE32(p, endian)); p+=4;
+            v3 = XXH32_round(v3, XXH_readLE32(p, endian)); p+=4;
+            v4 = XXH32_round(v4, XXH_readLE32(p, endian)); p+=4;
+        } while (p<=limit);
+
+        state->v1 = v1;
+        state->v2 = v2;
+        state->v3 = v3;
+        state->v4 = v4;
+    }
+
+    if (p < bEnd) {
+        XXH_memcpy(state->mem32, p, (size_t)(bEnd-p));
+        state->memsize = (unsigned)(bEnd-p);
+    }
+
+    return XXH_OK;
+}
+
+XXH_PUBLIC_API XXH_errorcode XXH32_update (XXH32_state_t* state_in, const void* input, size_t len)
+{
+    XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
+
+    if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
+        return XXH32_update_endian(state_in, input, len, XXH_littleEndian);
+    else
+        return XXH32_update_endian(state_in, input, len, XXH_bigEndian);
+}
+
+
+
+FORCE_INLINE U32 XXH32_digest_endian (const XXH32_state_t* state, XXH_endianess endian)
+{
+    const BYTE * p = (const BYTE*)state->mem32;
+    const BYTE* const bEnd = (const BYTE*)(state->mem32) + state->memsize;
+    U32 h32;
+
+    if (state->large_len) {
+        h32 = XXH_rotl32(state->v1, 1) + XXH_rotl32(state->v2, 7) + XXH_rotl32(state->v3, 12) + XXH_rotl32(state->v4, 18);
+    } else {
+        h32 = state->v3 /* == seed */ + PRIME32_5;
+    }
+
+    h32 += state->total_len_32;
+
+    while (p+4<=bEnd) {
+        h32 += XXH_readLE32(p, endian) * PRIME32_3;
+        h32  = XXH_rotl32(h32, 17) * PRIME32_4;
+        p+=4;
+    }
+
+    while (p> 15;
+    h32 *= PRIME32_2;
+    h32 ^= h32 >> 13;
+    h32 *= PRIME32_3;
+    h32 ^= h32 >> 16;
+
+    return h32;
+}
+
+
+XXH_PUBLIC_API unsigned int XXH32_digest (const XXH32_state_t* state_in)
+{
+    XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
+
+    if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
+        return XXH32_digest_endian(state_in, XXH_littleEndian);
+    else
+        return XXH32_digest_endian(state_in, XXH_bigEndian);
+}
+
+
+
+/* **** XXH64 **** */
+
+FORCE_INLINE XXH_errorcode XXH64_update_endian (XXH64_state_t* state, const void* input, size_t len, XXH_endianess endian)
+{
+    const BYTE* p = (const BYTE*)input;
+    const BYTE* const bEnd = p + len;
+
+#ifdef XXH_ACCEPT_NULL_INPUT_POINTER
+    if (input==NULL) return XXH_ERROR;
+#endif
+
+    state->total_len += len;
+
+    if (state->memsize + len < 32) {  /* fill in tmp buffer */
+        XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, len);
+        state->memsize += (U32)len;
+        return XXH_OK;
+    }
+
+    if (state->memsize) {   /* tmp buffer is full */
+        XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, 32-state->memsize);
+        state->v1 = XXH64_round(state->v1, XXH_readLE64(state->mem64+0, endian));
+        state->v2 = XXH64_round(state->v2, XXH_readLE64(state->mem64+1, endian));
+        state->v3 = XXH64_round(state->v3, XXH_readLE64(state->mem64+2, endian));
+        state->v4 = XXH64_round(state->v4, XXH_readLE64(state->mem64+3, endian));
+        p += 32-state->memsize;
+        state->memsize = 0;
+    }
+
+    if (p+32 <= bEnd) {
+        const BYTE* const limit = bEnd - 32;
+        U64 v1 = state->v1;
+        U64 v2 = state->v2;
+        U64 v3 = state->v3;
+        U64 v4 = state->v4;
+
+        do {
+            v1 = XXH64_round(v1, XXH_readLE64(p, endian)); p+=8;
+            v2 = XXH64_round(v2, XXH_readLE64(p, endian)); p+=8;
+            v3 = XXH64_round(v3, XXH_readLE64(p, endian)); p+=8;
+            v4 = XXH64_round(v4, XXH_readLE64(p, endian)); p+=8;
+        } while (p<=limit);
+
+        state->v1 = v1;
+        state->v2 = v2;
+        state->v3 = v3;
+        state->v4 = v4;
+    }
+
+    if (p < bEnd) {
+        XXH_memcpy(state->mem64, p, (size_t)(bEnd-p));
+        state->memsize = (unsigned)(bEnd-p);
+    }
+
+    return XXH_OK;
+}
+
+XXH_PUBLIC_API XXH_errorcode XXH64_update (XXH64_state_t* state_in, const void* input, size_t len)
+{
+    XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
+
+    if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
+        return XXH64_update_endian(state_in, input, len, XXH_littleEndian);
+    else
+        return XXH64_update_endian(state_in, input, len, XXH_bigEndian);
+}
+
+
+
+FORCE_INLINE U64 XXH64_digest_endian (const XXH64_state_t* state, XXH_endianess endian)
+{
+    const BYTE * p = (const BYTE*)state->mem64;
+    const BYTE* const bEnd = (const BYTE*)state->mem64 + state->memsize;
+    U64 h64;
+
+    if (state->total_len >= 32) {
+        U64 const v1 = state->v1;
+        U64 const v2 = state->v2;
+        U64 const v3 = state->v3;
+        U64 const v4 = state->v4;
+
+        h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18);
+        h64 = XXH64_mergeRound(h64, v1);
+        h64 = XXH64_mergeRound(h64, v2);
+        h64 = XXH64_mergeRound(h64, v3);
+        h64 = XXH64_mergeRound(h64, v4);
+    } else {
+        h64  = state->v3 + PRIME64_5;
+    }
+
+    h64 += (U64) state->total_len;
+
+    while (p+8<=bEnd) {
+        U64 const k1 = XXH64_round(0, XXH_readLE64(p, endian));
+        h64 ^= k1;
+        h64  = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4;
+        p+=8;
+    }
+
+    if (p+4<=bEnd) {
+        h64 ^= (U64)(XXH_readLE32(p, endian)) * PRIME64_1;
+        h64  = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
+        p+=4;
+    }
+
+    while (p> 33;
+    h64 *= PRIME64_2;
+    h64 ^= h64 >> 29;
+    h64 *= PRIME64_3;
+    h64 ^= h64 >> 32;
+
+    return h64;
+}
+
+
+XXH_PUBLIC_API unsigned long long XXH64_digest (const XXH64_state_t* state_in)
+{
+    XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
+
+    if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
+        return XXH64_digest_endian(state_in, XXH_littleEndian);
+    else
+        return XXH64_digest_endian(state_in, XXH_bigEndian);
+}
+
+
+/* **************************
+*  Canonical representation
+****************************/
+
+/*! Default XXH result types are basic unsigned 32 and 64 bits.
+*   The canonical representation follows human-readable write convention, aka big-endian (large digits first).
+*   These functions allow transformation of hash result into and from its canonical format.
+*   This way, hash values can be written into a file or buffer, and remain comparable across different systems and programs.
+*/
+
+XXH_PUBLIC_API void XXH32_canonicalFromHash(XXH32_canonical_t* dst, XXH32_hash_t hash)
+{
+    XXH_STATIC_ASSERT(sizeof(XXH32_canonical_t) == sizeof(XXH32_hash_t));
+    if (XXH_CPU_LITTLE_ENDIAN) hash = XXH_swap32(hash);
+    memcpy(dst, &hash, sizeof(*dst));
+}
+
+XXH_PUBLIC_API void XXH64_canonicalFromHash(XXH64_canonical_t* dst, XXH64_hash_t hash)
+{
+    XXH_STATIC_ASSERT(sizeof(XXH64_canonical_t) == sizeof(XXH64_hash_t));
+    if (XXH_CPU_LITTLE_ENDIAN) hash = XXH_swap64(hash);
+    memcpy(dst, &hash, sizeof(*dst));
+}
+
+XXH_PUBLIC_API XXH32_hash_t XXH32_hashFromCanonical(const XXH32_canonical_t* src)
+{
+    return XXH_readBE32(src);
+}
+
+XXH_PUBLIC_API XXH64_hash_t XXH64_hashFromCanonical(const XXH64_canonical_t* src)
+{
+    return XXH_readBE64(src);
+}
diff --git a/lib/common/xxhash.h b/lib/common/xxhash.h
new file mode 100644
index 0000000..2c9b7c6
--- /dev/null
+++ b/lib/common/xxhash.h
@@ -0,0 +1,309 @@
+/*
+   xxHash - Extremely Fast Hash algorithm
+   Header File
+   Copyright (C) 2012-2016, Yann Collet.
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+   You can contact the author at :
+   - xxHash source repository : https://github.com/Cyan4973/xxHash
+*/
+
+/* Notice extracted from xxHash homepage :
+
+xxHash is an extremely fast Hash algorithm, running at RAM speed limits.
+It also successfully passes all tests from the SMHasher suite.
+
+Comparison (single thread, Windows Seven 32 bits, using SMHasher on a Core 2 Duo @3GHz)
+
+Name            Speed       Q.Score   Author
+xxHash          5.4 GB/s     10
+CrapWow         3.2 GB/s      2       Andrew
+MumurHash 3a    2.7 GB/s     10       Austin Appleby
+SpookyHash      2.0 GB/s     10       Bob Jenkins
+SBox            1.4 GB/s      9       Bret Mulvey
+Lookup3         1.2 GB/s      9       Bob Jenkins
+SuperFastHash   1.2 GB/s      1       Paul Hsieh
+CityHash64      1.05 GB/s    10       Pike & Alakuijala
+FNV             0.55 GB/s     5       Fowler, Noll, Vo
+CRC32           0.43 GB/s     9
+MD5-32          0.33 GB/s    10       Ronald L. Rivest
+SHA1-32         0.28 GB/s    10
+
+Q.Score is a measure of quality of the hash function.
+It depends on successfully passing SMHasher test set.
+10 is a perfect score.
+
+A 64-bits version, named XXH64, is available since r35.
+It offers much better speed, but for 64-bits applications only.
+Name     Speed on 64 bits    Speed on 32 bits
+XXH64       13.8 GB/s            1.9 GB/s
+XXH32        6.8 GB/s            6.0 GB/s
+*/
+
+#ifndef XXHASH_H_5627135585666179
+#define XXHASH_H_5627135585666179 1
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+#ifndef XXH_NAMESPACE
+#  define XXH_NAMESPACE ZSTD_  /* Zstandard specific */
+#endif
+
+
+/* ****************************
+*  Definitions
+******************************/
+#include    /* size_t */
+typedef enum { XXH_OK=0, XXH_ERROR } XXH_errorcode;
+
+
+/* ****************************
+*  API modifier
+******************************/
+/** XXH_PRIVATE_API
+*   This is useful if you want to include xxhash functions in `static` mode
+*   in order to inline them, and remove their symbol from the public list.
+*   Methodology :
+*     #define XXH_PRIVATE_API
+*     #include "xxhash.h"
+*   `xxhash.c` is automatically included.
+*   It's not useful to compile and link it as a separate module anymore.
+*/
+#ifdef XXH_PRIVATE_API
+#  ifndef XXH_STATIC_LINKING_ONLY
+#    define XXH_STATIC_LINKING_ONLY
+#  endif
+#  if defined(__GNUC__)
+#    define XXH_PUBLIC_API static __inline __attribute__((unused))
+#  elif defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
+#    define XXH_PUBLIC_API static inline
+#  elif defined(_MSC_VER)
+#    define XXH_PUBLIC_API static __inline
+#  else
+#    define XXH_PUBLIC_API static   /* this version may generate warnings for unused static functions; disable the relevant warning */
+#  endif
+#else
+#  define XXH_PUBLIC_API   /* do nothing */
+#endif /* XXH_PRIVATE_API */
+
+/*!XXH_NAMESPACE, aka Namespace Emulation :
+
+If you want to include _and expose_ xxHash functions from within your own library,
+but also want to avoid symbol collisions with another library which also includes xxHash,
+
+you can use XXH_NAMESPACE, to automatically prefix any public symbol from xxhash library
+with the value of XXH_NAMESPACE (so avoid to keep it NULL and avoid numeric values).
+
+Note that no change is required within the calling program as long as it includes `xxhash.h` :
+regular symbol name will be automatically translated by this header.
+*/
+#ifdef XXH_NAMESPACE
+#  define XXH_CAT(A,B) A##B
+#  define XXH_NAME2(A,B) XXH_CAT(A,B)
+#  define XXH32 XXH_NAME2(XXH_NAMESPACE, XXH32)
+#  define XXH64 XXH_NAME2(XXH_NAMESPACE, XXH64)
+#  define XXH_versionNumber XXH_NAME2(XXH_NAMESPACE, XXH_versionNumber)
+#  define XXH32_createState XXH_NAME2(XXH_NAMESPACE, XXH32_createState)
+#  define XXH64_createState XXH_NAME2(XXH_NAMESPACE, XXH64_createState)
+#  define XXH32_freeState XXH_NAME2(XXH_NAMESPACE, XXH32_freeState)
+#  define XXH64_freeState XXH_NAME2(XXH_NAMESPACE, XXH64_freeState)
+#  define XXH32_reset XXH_NAME2(XXH_NAMESPACE, XXH32_reset)
+#  define XXH64_reset XXH_NAME2(XXH_NAMESPACE, XXH64_reset)
+#  define XXH32_update XXH_NAME2(XXH_NAMESPACE, XXH32_update)
+#  define XXH64_update XXH_NAME2(XXH_NAMESPACE, XXH64_update)
+#  define XXH32_digest XXH_NAME2(XXH_NAMESPACE, XXH32_digest)
+#  define XXH64_digest XXH_NAME2(XXH_NAMESPACE, XXH64_digest)
+#  define XXH32_copyState XXH_NAME2(XXH_NAMESPACE, XXH32_copyState)
+#  define XXH64_copyState XXH_NAME2(XXH_NAMESPACE, XXH64_copyState)
+#  define XXH32_canonicalFromHash XXH_NAME2(XXH_NAMESPACE, XXH32_canonicalFromHash)
+#  define XXH64_canonicalFromHash XXH_NAME2(XXH_NAMESPACE, XXH64_canonicalFromHash)
+#  define XXH32_hashFromCanonical XXH_NAME2(XXH_NAMESPACE, XXH32_hashFromCanonical)
+#  define XXH64_hashFromCanonical XXH_NAME2(XXH_NAMESPACE, XXH64_hashFromCanonical)
+#endif
+
+
+/* *************************************
+*  Version
+***************************************/
+#define XXH_VERSION_MAJOR    0
+#define XXH_VERSION_MINOR    6
+#define XXH_VERSION_RELEASE  2
+#define XXH_VERSION_NUMBER  (XXH_VERSION_MAJOR *100*100 + XXH_VERSION_MINOR *100 + XXH_VERSION_RELEASE)
+XXH_PUBLIC_API unsigned XXH_versionNumber (void);
+
+
+/* ****************************
+*  Simple Hash Functions
+******************************/
+typedef unsigned int       XXH32_hash_t;
+typedef unsigned long long XXH64_hash_t;
+
+XXH_PUBLIC_API XXH32_hash_t XXH32 (const void* input, size_t length, unsigned int seed);
+XXH_PUBLIC_API XXH64_hash_t XXH64 (const void* input, size_t length, unsigned long long seed);
+
+/*!
+XXH32() :
+    Calculate the 32-bits hash of sequence "length" bytes stored at memory address "input".
+    The memory between input & input+length must be valid (allocated and read-accessible).
+    "seed" can be used to alter the result predictably.
+    Speed on Core 2 Duo @ 3 GHz (single thread, SMHasher benchmark) : 5.4 GB/s
+XXH64() :
+    Calculate the 64-bits hash of sequence of length "len" stored at memory address "input".
+    "seed" can be used to alter the result predictably.
+    This function runs 2x faster on 64-bits systems, but slower on 32-bits systems (see benchmark).
+*/
+
+
+/* ****************************
+*  Streaming Hash Functions
+******************************/
+typedef struct XXH32_state_s XXH32_state_t;   /* incomplete type */
+typedef struct XXH64_state_s XXH64_state_t;   /* incomplete type */
+
+/*! State allocation, compatible with dynamic libraries */
+
+XXH_PUBLIC_API XXH32_state_t* XXH32_createState(void);
+XXH_PUBLIC_API XXH_errorcode  XXH32_freeState(XXH32_state_t* statePtr);
+
+XXH_PUBLIC_API XXH64_state_t* XXH64_createState(void);
+XXH_PUBLIC_API XXH_errorcode  XXH64_freeState(XXH64_state_t* statePtr);
+
+
+/* hash streaming */
+
+XXH_PUBLIC_API XXH_errorcode XXH32_reset  (XXH32_state_t* statePtr, unsigned int seed);
+XXH_PUBLIC_API XXH_errorcode XXH32_update (XXH32_state_t* statePtr, const void* input, size_t length);
+XXH_PUBLIC_API XXH32_hash_t  XXH32_digest (const XXH32_state_t* statePtr);
+
+XXH_PUBLIC_API XXH_errorcode XXH64_reset  (XXH64_state_t* statePtr, unsigned long long seed);
+XXH_PUBLIC_API XXH_errorcode XXH64_update (XXH64_state_t* statePtr, const void* input, size_t length);
+XXH_PUBLIC_API XXH64_hash_t  XXH64_digest (const XXH64_state_t* statePtr);
+
+/*
+These functions generate the xxHash of an input provided in multiple segments.
+Note that, for small input, they are slower than single-call functions, due to state management.
+For small input, prefer `XXH32()` and `XXH64()` .
+
+XXH state must first be allocated, using XXH*_createState() .
+
+Start a new hash by initializing state with a seed, using XXH*_reset().
+
+Then, feed the hash state by calling XXH*_update() as many times as necessary.
+Obviously, input must be allocated and read accessible.
+The function returns an error code, with 0 meaning OK, and any other value meaning there is an error.
+
+Finally, a hash value can be produced anytime, by using XXH*_digest().
+This function returns the nn-bits hash as an int or long long.
+
+It's still possible to continue inserting input into the hash state after a digest,
+and generate some new hashes later on, by calling again XXH*_digest().
+
+When done, free XXH state space if it was allocated dynamically.
+*/
+
+
+/* **************************
+*  Utils
+****************************/
+#if !(defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L))   /* ! C99 */
+#  define restrict   /* disable restrict */
+#endif
+
+XXH_PUBLIC_API void XXH32_copyState(XXH32_state_t* restrict dst_state, const XXH32_state_t* restrict src_state);
+XXH_PUBLIC_API void XXH64_copyState(XXH64_state_t* restrict dst_state, const XXH64_state_t* restrict src_state);
+
+
+/* **************************
+*  Canonical representation
+****************************/
+typedef struct { unsigned char digest[4]; } XXH32_canonical_t;
+typedef struct { unsigned char digest[8]; } XXH64_canonical_t;
+
+XXH_PUBLIC_API void XXH32_canonicalFromHash(XXH32_canonical_t* dst, XXH32_hash_t hash);
+XXH_PUBLIC_API void XXH64_canonicalFromHash(XXH64_canonical_t* dst, XXH64_hash_t hash);
+
+XXH_PUBLIC_API XXH32_hash_t XXH32_hashFromCanonical(const XXH32_canonical_t* src);
+XXH_PUBLIC_API XXH64_hash_t XXH64_hashFromCanonical(const XXH64_canonical_t* src);
+
+/* Default result type for XXH functions are primitive unsigned 32 and 64 bits.
+*  The canonical representation uses human-readable write convention, aka big-endian (large digits first).
+*  These functions allow transformation of hash result into and from its canonical format.
+*  This way, hash values can be written into a file / memory, and remain comparable on different systems and programs.
+*/
+
+
+#ifdef XXH_STATIC_LINKING_ONLY
+
+/* ================================================================================================
+   This section contains definitions which are not guaranteed to remain stable.
+   They may change in future versions, becoming incompatible with a different version of the library.
+   They shall only be used with static linking.
+   Never use these definitions in association with dynamic linking !
+=================================================================================================== */
+
+/* These definitions are only meant to allow allocation of XXH state
+   statically, on stack, or in a struct for example.
+   Do not use members directly. */
+
+   struct XXH32_state_s {
+       unsigned total_len_32;
+       unsigned large_len;
+       unsigned v1;
+       unsigned v2;
+       unsigned v3;
+       unsigned v4;
+       unsigned mem32[4];   /* buffer defined as U32 for alignment */
+       unsigned memsize;
+       unsigned reserved;   /* never read nor write, will be removed in a future version */
+   };   /* typedef'd to XXH32_state_t */
+
+   struct XXH64_state_s {
+       unsigned long long total_len;
+       unsigned long long v1;
+       unsigned long long v2;
+       unsigned long long v3;
+       unsigned long long v4;
+       unsigned long long mem64[4];   /* buffer defined as U64 for alignment */
+       unsigned memsize;
+       unsigned reserved[2];          /* never read nor write, will be removed in a future version */
+   };   /* typedef'd to XXH64_state_t */
+
+
+#  ifdef XXH_PRIVATE_API
+#    include "xxhash.c"   /* include xxhash functions as `static`, for inlining */
+#  endif
+
+#endif /* XXH_STATIC_LINKING_ONLY */
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* XXHASH_H_5627135585666179 */
diff --git a/lib/common/zstd_common.c b/lib/common/zstd_common.c
new file mode 100644
index 0000000..f30128c
--- /dev/null
+++ b/lib/common/zstd_common.c
@@ -0,0 +1,77 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+
+
+/*-*************************************
+*  Dependencies
+***************************************/
+#include          /* malloc */
+#include "error_private.h"
+#define ZSTD_STATIC_LINKING_ONLY
+#include "zstd.h"           /* declaration of ZSTD_isError, ZSTD_getErrorName, ZSTD_getErrorCode, ZSTD_getErrorString, ZSTD_versionNumber */
+
+
+/*-****************************************
+*  Version
+******************************************/
+unsigned ZSTD_versionNumber (void) { return ZSTD_VERSION_NUMBER; }
+
+
+/*-****************************************
+*  ZSTD Error Management
+******************************************/
+/*! ZSTD_isError() :
+*   tells if a return value is an error code */
+unsigned ZSTD_isError(size_t code) { return ERR_isError(code); }
+
+/*! ZSTD_getErrorName() :
+*   provides error code string from function result (useful for debugging) */
+const char* ZSTD_getErrorName(size_t code) { return ERR_getErrorName(code); }
+
+/*! ZSTD_getError() :
+*   convert a `size_t` function result into a proper ZSTD_errorCode enum */
+ZSTD_ErrorCode ZSTD_getErrorCode(size_t code) { return ERR_getErrorCode(code); }
+
+/*! ZSTD_getErrorString() :
+*   provides error code string from enum */
+const char* ZSTD_getErrorString(ZSTD_ErrorCode code) { return ERR_getErrorName(code); }
+
+/* ---   ZBUFF Error Management  (deprecated)   --- */
+unsigned ZBUFF_isError(size_t errorCode) { return ERR_isError(errorCode); }
+const char* ZBUFF_getErrorName(size_t errorCode) { return ERR_getErrorName(errorCode); }
+
+
+/*=**************************************************************
+*  Custom allocator
+****************************************************************/
+/* default uses stdlib */
+void* ZSTD_defaultAllocFunction(void* opaque, size_t size)
+{
+    void* address = malloc(size);
+    (void)opaque;
+    return address;
+}
+
+void ZSTD_defaultFreeFunction(void* opaque, void* address)
+{
+    (void)opaque;
+    free(address);
+}
+
+void* ZSTD_malloc(size_t size, ZSTD_customMem customMem)
+{
+    return customMem.customAlloc(customMem.opaque, size);
+}
+
+void ZSTD_free(void* ptr, ZSTD_customMem customMem)
+{
+    if (ptr!=NULL)
+        customMem.customFree(customMem.opaque, ptr);
+}
diff --git a/lib/common/zstd_errors.h b/lib/common/zstd_errors.h
new file mode 100644
index 0000000..50dc4f7
--- /dev/null
+++ b/lib/common/zstd_errors.h
@@ -0,0 +1,60 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+#ifndef ZSTD_ERRORS_H_398273423
+#define ZSTD_ERRORS_H_398273423
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/*===== dependency =====*/
+#include    /* size_t */
+
+
+/*-****************************************
+*  error codes list
+******************************************/
+typedef enum {
+  ZSTD_error_no_error,
+  ZSTD_error_GENERIC,
+  ZSTD_error_prefix_unknown,
+  ZSTD_error_version_unsupported,
+  ZSTD_error_parameter_unknown,
+  ZSTD_error_frameParameter_unsupported,
+  ZSTD_error_frameParameter_unsupportedBy32bits,
+  ZSTD_error_frameParameter_windowTooLarge,
+  ZSTD_error_compressionParameter_unsupported,
+  ZSTD_error_init_missing,
+  ZSTD_error_memory_allocation,
+  ZSTD_error_stage_wrong,
+  ZSTD_error_dstSize_tooSmall,
+  ZSTD_error_srcSize_wrong,
+  ZSTD_error_corruption_detected,
+  ZSTD_error_checksum_wrong,
+  ZSTD_error_tableLog_tooLarge,
+  ZSTD_error_maxSymbolValue_tooLarge,
+  ZSTD_error_maxSymbolValue_tooSmall,
+  ZSTD_error_dictionary_corrupted,
+  ZSTD_error_dictionary_wrong,
+  ZSTD_error_maxCode
+} ZSTD_ErrorCode;
+
+/*! ZSTD_getErrorCode() :
+    convert a `size_t` function result into a `ZSTD_ErrorCode` enum type,
+    which can be used to compare directly with enum list published into "error_public.h" */
+ZSTD_ErrorCode ZSTD_getErrorCode(size_t functionResult);
+const char* ZSTD_getErrorString(ZSTD_ErrorCode code);
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_ERRORS_H_398273423 */
diff --git a/lib/common/zstd_internal.h b/lib/common/zstd_internal.h
new file mode 100644
index 0000000..96e0577
--- /dev/null
+++ b/lib/common/zstd_internal.h
@@ -0,0 +1,270 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+#ifndef ZSTD_CCOMMON_H_MODULE
+#define ZSTD_CCOMMON_H_MODULE
+
+/*-*******************************************************
+*  Compiler specifics
+*********************************************************/
+#ifdef _MSC_VER    /* Visual Studio */
+#  define FORCE_INLINE static __forceinline
+#  include                     /* For Visual 2005 */
+#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
+#  pragma warning(disable : 4324)        /* disable: C4324: padded structure */
+#  pragma warning(disable : 4100)        /* disable: C4100: unreferenced formal parameter */
+#else
+#  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+#    ifdef __GNUC__
+#      define FORCE_INLINE static inline __attribute__((always_inline))
+#    else
+#      define FORCE_INLINE static inline
+#    endif
+#  else
+#    define FORCE_INLINE static
+#  endif /* __STDC_VERSION__ */
+#endif
+
+#ifdef _MSC_VER
+#  define FORCE_NOINLINE static __declspec(noinline)
+#else
+#  ifdef __GNUC__
+#    define FORCE_NOINLINE static __attribute__((__noinline__))
+#  else
+#    define FORCE_NOINLINE static
+#  endif
+#endif
+
+
+/*-*************************************
+*  Dependencies
+***************************************/
+#include "mem.h"
+#include "error_private.h"
+#define ZSTD_STATIC_LINKING_ONLY
+#include "zstd.h"
+
+
+/*-*************************************
+*  shared macros
+***************************************/
+#define MIN(a,b) ((a)<(b) ? (a) : (b))
+#define MAX(a,b) ((a)>(b) ? (a) : (b))
+#define CHECK_F(f) { size_t const errcod = f; if (ERR_isError(errcod)) return errcod; }  /* check and Forward error code */
+#define CHECK_E(f, e) { size_t const errcod = f; if (ERR_isError(errcod)) return ERROR(e); }  /* check and send Error code */
+
+
+/*-*************************************
+*  Common constants
+***************************************/
+#define ZSTD_OPT_NUM    (1<<12)
+#define ZSTD_DICT_MAGIC  0xEC30A437   /* v0.7+ */
+
+#define ZSTD_REP_NUM      3                 /* number of repcodes */
+#define ZSTD_REP_CHECK    (ZSTD_REP_NUM)    /* number of repcodes to check by the optimal parser */
+#define ZSTD_REP_MOVE     (ZSTD_REP_NUM-1)
+#define ZSTD_REP_MOVE_OPT (ZSTD_REP_NUM)
+static const U32 repStartValue[ZSTD_REP_NUM] = { 1, 4, 8 };
+
+#define KB *(1 <<10)
+#define MB *(1 <<20)
+#define GB *(1U<<30)
+
+#define BIT7 128
+#define BIT6  64
+#define BIT5  32
+#define BIT4  16
+#define BIT1   2
+#define BIT0   1
+
+#define ZSTD_WINDOWLOG_ABSOLUTEMIN 10
+static const size_t ZSTD_fcs_fieldSize[4] = { 0, 2, 4, 8 };
+static const size_t ZSTD_did_fieldSize[4] = { 0, 1, 2, 4 };
+
+#define ZSTD_BLOCKHEADERSIZE 3   /* C standard doesn't allow `static const` variable to be init using another `static const` variable */
+static const size_t ZSTD_blockHeaderSize = ZSTD_BLOCKHEADERSIZE;
+typedef enum { bt_raw, bt_rle, bt_compressed, bt_reserved } blockType_e;
+
+#define MIN_SEQUENCES_SIZE 1 /* nbSeq==0 */
+#define MIN_CBLOCK_SIZE (1 /*litCSize*/ + 1 /* RLE or RAW */ + MIN_SEQUENCES_SIZE /* nbSeq==0 */)   /* for a non-null block */
+
+#define HufLog 12
+typedef enum { set_basic, set_rle, set_compressed, set_repeat } symbolEncodingType_e;
+
+#define LONGNBSEQ 0x7F00
+
+#define MINMATCH 3
+#define EQUAL_READ32 4
+
+#define Litbits  8
+#define MaxLit ((1<= 3)   /* GCC Intrinsic */
+    return 31 - __builtin_clz(val);
+#   else   /* Software version */
+    static const int DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 };
+    U32 v = val;
+    int r;
+    v |= v >> 1;
+    v |= v >> 2;
+    v |= v >> 4;
+    v |= v >> 8;
+    v |= v >> 16;
+    r = DeBruijnClz[(U32)(v * 0x07C4ACDDU) >> 27];
+    return r;
+#   endif
+}
+
+
+#endif   /* ZSTD_CCOMMON_H_MODULE */
diff --git a/lib/compress/fse_compress.c b/lib/compress/fse_compress.c
new file mode 100644
index 0000000..6627fac
--- /dev/null
+++ b/lib/compress/fse_compress.c
@@ -0,0 +1,850 @@
+/* ******************************************************************
+   FSE : Finite State Entropy encoder
+   Copyright (C) 2013-2015, Yann Collet.
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+    You can contact the author at :
+    - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
+    - Public forum : https://groups.google.com/forum/#!forum/lz4c
+****************************************************************** */
+
+/* **************************************************************
+*  Compiler specifics
+****************************************************************/
+#ifdef _MSC_VER    /* Visual Studio */
+#  define FORCE_INLINE static __forceinline
+#  include                     /* For Visual 2005 */
+#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
+#  pragma warning(disable : 4214)        /* disable: C4214: non-int bitfields */
+#else
+#  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+#    ifdef __GNUC__
+#      define FORCE_INLINE static inline __attribute__((always_inline))
+#    else
+#      define FORCE_INLINE static inline
+#    endif
+#  else
+#    define FORCE_INLINE static
+#  endif /* __STDC_VERSION__ */
+#endif
+
+
+/* **************************************************************
+*  Includes
+****************************************************************/
+#include      /* malloc, free, qsort */
+#include      /* memcpy, memset */
+#include       /* printf (debug) */
+#include "bitstream.h"
+#define FSE_STATIC_LINKING_ONLY
+#include "fse.h"
+
+
+/* **************************************************************
+*  Error Management
+****************************************************************/
+#define FSE_STATIC_ASSERT(c) { enum { FSE_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
+
+
+/* **************************************************************
+*  Templates
+****************************************************************/
+/*
+  designed to be included
+  for type-specific functions (template emulation in C)
+  Objective is to write these functions only once, for improved maintenance
+*/
+
+/* safety checks */
+#ifndef FSE_FUNCTION_EXTENSION
+#  error "FSE_FUNCTION_EXTENSION must be defined"
+#endif
+#ifndef FSE_FUNCTION_TYPE
+#  error "FSE_FUNCTION_TYPE must be defined"
+#endif
+
+/* Function names */
+#define FSE_CAT(X,Y) X##Y
+#define FSE_FUNCTION_NAME(X,Y) FSE_CAT(X,Y)
+#define FSE_TYPE_NAME(X,Y) FSE_CAT(X,Y)
+
+
+/* Function templates */
+
+/* FSE_buildCTable_wksp() :
+ * Same as FSE_buildCTable(), but using an externally allocated scratch buffer (`workSpace`).
+ * wkspSize should be sized to handle worst case situation, which is `1<>1 : 1) ;
+    FSE_symbolCompressionTransform* const symbolTT = (FSE_symbolCompressionTransform*) (FSCT);
+    U32 const step = FSE_TABLESTEP(tableSize);
+    U32 cumul[FSE_MAX_SYMBOL_VALUE+2];
+
+    FSE_FUNCTION_TYPE* const tableSymbol = (FSE_FUNCTION_TYPE*)workSpace;
+    U32 highThreshold = tableSize-1;
+
+    /* CTable header */
+    if (((size_t)1 << tableLog) * sizeof(FSE_FUNCTION_TYPE) > wkspSize) return ERROR(tableLog_tooLarge);
+    tableU16[-2] = (U16) tableLog;
+    tableU16[-1] = (U16) maxSymbolValue;
+
+    /* For explanations on how to distribute symbol values over the table :
+    *  http://fastcompression.blogspot.fr/2014/02/fse-distributing-symbol-values.html */
+
+    /* symbol start positions */
+    {   U32 u;
+        cumul[0] = 0;
+        for (u=1; u<=maxSymbolValue+1; u++) {
+            if (normalizedCounter[u-1]==-1) {  /* Low proba symbol */
+                cumul[u] = cumul[u-1] + 1;
+                tableSymbol[highThreshold--] = (FSE_FUNCTION_TYPE)(u-1);
+            } else {
+                cumul[u] = cumul[u-1] + normalizedCounter[u-1];
+        }   }
+        cumul[maxSymbolValue+1] = tableSize+1;
+    }
+
+    /* Spread symbols */
+    {   U32 position = 0;
+        U32 symbol;
+        for (symbol=0; symbol<=maxSymbolValue; symbol++) {
+            int nbOccurences;
+            for (nbOccurences=0; nbOccurences highThreshold) position = (position + step) & tableMask;   /* Low proba area */
+        }   }
+
+        if (position!=0) return ERROR(GENERIC);   /* Must have gone through all positions */
+    }
+
+    /* Build table */
+    {   U32 u; for (u=0; u> 3) + 3;
+    return maxSymbolValue ? maxHeaderSize : FSE_NCOUNTBOUND;  /* maxSymbolValue==0 ? use default */
+}
+
+static short FSE_abs(short a) { return (short)(a<0 ? -a : a); }
+
+static size_t FSE_writeNCount_generic (void* header, size_t headerBufferSize,
+                                       const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog,
+                                       unsigned writeIsSafe)
+{
+    BYTE* const ostart = (BYTE*) header;
+    BYTE* out = ostart;
+    BYTE* const oend = ostart + headerBufferSize;
+    int nbBits;
+    const int tableSize = 1 << tableLog;
+    int remaining;
+    int threshold;
+    U32 bitStream;
+    int bitCount;
+    unsigned charnum = 0;
+    int previous0 = 0;
+
+    bitStream = 0;
+    bitCount  = 0;
+    /* Table Size */
+    bitStream += (tableLog-FSE_MIN_TABLELOG) << bitCount;
+    bitCount  += 4;
+
+    /* Init */
+    remaining = tableSize+1;   /* +1 for extra accuracy */
+    threshold = tableSize;
+    nbBits = tableLog+1;
+
+    while (remaining>1) {  /* stops at 1 */
+        if (previous0) {
+            unsigned start = charnum;
+            while (!normalizedCounter[charnum]) charnum++;
+            while (charnum >= start+24) {
+                start+=24;
+                bitStream += 0xFFFFU << bitCount;
+                if ((!writeIsSafe) && (out > oend-2)) return ERROR(dstSize_tooSmall);   /* Buffer overflow */
+                out[0] = (BYTE) bitStream;
+                out[1] = (BYTE)(bitStream>>8);
+                out+=2;
+                bitStream>>=16;
+            }
+            while (charnum >= start+3) {
+                start+=3;
+                bitStream += 3 << bitCount;
+                bitCount += 2;
+            }
+            bitStream += (charnum-start) << bitCount;
+            bitCount += 2;
+            if (bitCount>16) {
+                if ((!writeIsSafe) && (out > oend - 2)) return ERROR(dstSize_tooSmall);   /* Buffer overflow */
+                out[0] = (BYTE)bitStream;
+                out[1] = (BYTE)(bitStream>>8);
+                out += 2;
+                bitStream >>= 16;
+                bitCount -= 16;
+        }   }
+        {   short count = normalizedCounter[charnum++];
+            const short max = (short)((2*threshold-1)-remaining);
+            remaining -= FSE_abs(count);
+            if (remaining<1) return ERROR(GENERIC);
+            count++;   /* +1 for extra accuracy */
+            if (count>=threshold) count += max;   /* [0..max[ [max..threshold[ (...) [threshold+max 2*threshold[ */
+            bitStream += count << bitCount;
+            bitCount  += nbBits;
+            bitCount  -= (count>=1;
+        }
+        if (bitCount>16) {
+            if ((!writeIsSafe) && (out > oend - 2)) return ERROR(dstSize_tooSmall);   /* Buffer overflow */
+            out[0] = (BYTE)bitStream;
+            out[1] = (BYTE)(bitStream>>8);
+            out += 2;
+            bitStream >>= 16;
+            bitCount -= 16;
+    }   }
+
+    /* flush remaining bitStream */
+    if ((!writeIsSafe) && (out > oend - 2)) return ERROR(dstSize_tooSmall);   /* Buffer overflow */
+    out[0] = (BYTE)bitStream;
+    out[1] = (BYTE)(bitStream>>8);
+    out+= (bitCount+7) /8;
+
+    if (charnum > maxSymbolValue + 1) return ERROR(GENERIC);
+
+    return (out-ostart);
+}
+
+
+size_t FSE_writeNCount (void* buffer, size_t bufferSize, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
+{
+    if (tableLog > FSE_MAX_TABLELOG) return ERROR(GENERIC);   /* Unsupported */
+    if (tableLog < FSE_MIN_TABLELOG) return ERROR(GENERIC);   /* Unsupported */
+
+    if (bufferSize < FSE_NCountWriteBound(maxSymbolValue, tableLog))
+        return FSE_writeNCount_generic(buffer, bufferSize, normalizedCounter, maxSymbolValue, tableLog, 0);
+
+    return FSE_writeNCount_generic(buffer, bufferSize, normalizedCounter, maxSymbolValue, tableLog, 1);
+}
+
+
+
+/*-**************************************************************
+*  Counting histogram
+****************************************************************/
+/*! FSE_count_simple
+    This function counts byte values within `src`, and store the histogram into table `count`.
+    It doesn't use any additional memory.
+    But this function is unsafe : it doesn't check that all values within `src` can fit into `count`.
+    For this reason, prefer using a table `count` with 256 elements.
+    @return : count of most numerous element
+*/
+size_t FSE_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
+                        const void* src, size_t srcSize)
+{
+    const BYTE* ip = (const BYTE*)src;
+    const BYTE* const end = ip + srcSize;
+    unsigned maxSymbolValue = *maxSymbolValuePtr;
+    unsigned max=0;
+
+    memset(count, 0, (maxSymbolValue+1)*sizeof(*count));
+    if (srcSize==0) { *maxSymbolValuePtr = 0; return 0; }
+
+    while (ip max) max = count[s]; }
+
+    return (size_t)max;
+}
+
+
+/* FSE_count_parallel_wksp() :
+ * Same as FSE_count_parallel(), but using an externally provided scratch buffer.
+ * `workSpace` size must be a minimum of `1024 * sizeof(unsigned)`` */
+static size_t FSE_count_parallel_wksp(
+                                unsigned* count, unsigned* maxSymbolValuePtr,
+                                const void* source, size_t sourceSize,
+                                unsigned checkMax, unsigned* const workSpace)
+{
+    const BYTE* ip = (const BYTE*)source;
+    const BYTE* const iend = ip+sourceSize;
+    unsigned maxSymbolValue = *maxSymbolValuePtr;
+    unsigned max=0;
+    U32* const Counting1 = workSpace;
+    U32* const Counting2 = Counting1 + 256;
+    U32* const Counting3 = Counting2 + 256;
+    U32* const Counting4 = Counting3 + 256;
+
+    memset(Counting1, 0, 4*256*sizeof(unsigned));
+
+    /* safety checks */
+    if (!sourceSize) {
+        memset(count, 0, maxSymbolValue + 1);
+        *maxSymbolValuePtr = 0;
+        return 0;
+    }
+    if (!maxSymbolValue) maxSymbolValue = 255;            /* 0 == default */
+
+    /* by stripes of 16 bytes */
+    {   U32 cached = MEM_read32(ip); ip += 4;
+        while (ip < iend-15) {
+            U32 c = cached; cached = MEM_read32(ip); ip += 4;
+            Counting1[(BYTE) c     ]++;
+            Counting2[(BYTE)(c>>8) ]++;
+            Counting3[(BYTE)(c>>16)]++;
+            Counting4[       c>>24 ]++;
+            c = cached; cached = MEM_read32(ip); ip += 4;
+            Counting1[(BYTE) c     ]++;
+            Counting2[(BYTE)(c>>8) ]++;
+            Counting3[(BYTE)(c>>16)]++;
+            Counting4[       c>>24 ]++;
+            c = cached; cached = MEM_read32(ip); ip += 4;
+            Counting1[(BYTE) c     ]++;
+            Counting2[(BYTE)(c>>8) ]++;
+            Counting3[(BYTE)(c>>16)]++;
+            Counting4[       c>>24 ]++;
+            c = cached; cached = MEM_read32(ip); ip += 4;
+            Counting1[(BYTE) c     ]++;
+            Counting2[(BYTE)(c>>8) ]++;
+            Counting3[(BYTE)(c>>16)]++;
+            Counting4[       c>>24 ]++;
+        }
+        ip-=4;
+    }
+
+    /* finish last symbols */
+    while (ipmaxSymbolValue; s--) {
+            Counting1[s] += Counting2[s] + Counting3[s] + Counting4[s];
+            if (Counting1[s]) return ERROR(maxSymbolValue_tooSmall);
+    }   }
+
+    {   U32 s; for (s=0; s<=maxSymbolValue; s++) {
+            count[s] = Counting1[s] + Counting2[s] + Counting3[s] + Counting4[s];
+            if (count[s] > max) max = count[s];
+    }   }
+
+    while (!count[maxSymbolValue]) maxSymbolValue--;
+    *maxSymbolValuePtr = maxSymbolValue;
+    return (size_t)max;
+}
+
+/* FSE_countFast_wksp() :
+ * Same as FSE_countFast(), but using an externally provided scratch buffer.
+ * `workSpace` size must be table of >= `1024` unsigned */
+size_t FSE_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
+                     const void* source, size_t sourceSize, unsigned* workSpace)
+{
+    if (sourceSize < 1500) return FSE_count_simple(count, maxSymbolValuePtr, source, sourceSize);
+    return FSE_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, 0, workSpace);
+}
+
+/* fast variant (unsafe : won't check if src contains values beyond count[] limit) */
+size_t FSE_countFast(unsigned* count, unsigned* maxSymbolValuePtr,
+                     const void* source, size_t sourceSize)
+{
+    unsigned tmpCounters[1024];
+    return FSE_countFast_wksp(count, maxSymbolValuePtr, source, sourceSize, tmpCounters);
+}
+
+/* FSE_count_wksp() :
+ * Same as FSE_count(), but using an externally provided scratch buffer.
+ * `workSpace` size must be table of >= `1024` unsigned */
+size_t FSE_count_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
+                 const void* source, size_t sourceSize, unsigned* workSpace)
+{
+    if (*maxSymbolValuePtr < 255)
+        return FSE_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, 1, workSpace);
+    *maxSymbolValuePtr = 255;
+    return FSE_countFast_wksp(count, maxSymbolValuePtr, source, sourceSize, workSpace);
+}
+
+size_t FSE_count(unsigned* count, unsigned* maxSymbolValuePtr,
+                 const void* src, size_t srcSize)
+{
+    unsigned tmpCounters[1024];
+    return FSE_count_wksp(count, maxSymbolValuePtr, src, srcSize, tmpCounters);
+}
+
+
+
+/*-**************************************************************
+*  FSE Compression Code
+****************************************************************/
+/*! FSE_sizeof_CTable() :
+    FSE_CTable is a variable size structure which contains :
+    `U16 tableLog;`
+    `U16 maxSymbolValue;`
+    `U16 nextStateNumber[1 << tableLog];`                         // This size is variable
+    `FSE_symbolCompressionTransform symbolTT[maxSymbolValue+1];`  // This size is variable
+Allocation is manual (C standard does not support variable-size structures).
+*/
+size_t FSE_sizeof_CTable (unsigned maxSymbolValue, unsigned tableLog)
+{
+    if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge);
+    return FSE_CTABLE_SIZE_U32 (tableLog, maxSymbolValue) * sizeof(U32);
+}
+
+FSE_CTable* FSE_createCTable (unsigned maxSymbolValue, unsigned tableLog)
+{
+    size_t size;
+    if (tableLog > FSE_TABLELOG_ABSOLUTE_MAX) tableLog = FSE_TABLELOG_ABSOLUTE_MAX;
+    size = FSE_CTABLE_SIZE_U32 (tableLog, maxSymbolValue) * sizeof(U32);
+    return (FSE_CTable*)malloc(size);
+}
+
+void FSE_freeCTable (FSE_CTable* ct) { free(ct); }
+
+/* provides the minimum logSize to safely represent a distribution */
+static unsigned FSE_minTableLog(size_t srcSize, unsigned maxSymbolValue)
+{
+	U32 minBitsSrc = BIT_highbit32((U32)(srcSize - 1)) + 1;
+	U32 minBitsSymbols = BIT_highbit32(maxSymbolValue) + 2;
+	U32 minBits = minBitsSrc < minBitsSymbols ? minBitsSrc : minBitsSymbols;
+	return minBits;
+}
+
+unsigned FSE_optimalTableLog_internal(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue, unsigned minus)
+{
+	U32 maxBitsSrc = BIT_highbit32((U32)(srcSize - 1)) - minus;
+    U32 tableLog = maxTableLog;
+	U32 minBits = FSE_minTableLog(srcSize, maxSymbolValue);
+    if (tableLog==0) tableLog = FSE_DEFAULT_TABLELOG;
+	if (maxBitsSrc < tableLog) tableLog = maxBitsSrc;   /* Accuracy can be reduced */
+	if (minBits > tableLog) tableLog = minBits;   /* Need a minimum to safely represent all symbol values */
+    if (tableLog < FSE_MIN_TABLELOG) tableLog = FSE_MIN_TABLELOG;
+    if (tableLog > FSE_MAX_TABLELOG) tableLog = FSE_MAX_TABLELOG;
+    return tableLog;
+}
+
+unsigned FSE_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue)
+{
+    return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 2);
+}
+
+
+/* Secondary normalization method.
+   To be used when primary method fails. */
+
+static size_t FSE_normalizeM2(short* norm, U32 tableLog, const unsigned* count, size_t total, U32 maxSymbolValue)
+{
+    U32 s;
+    U32 distributed = 0;
+    U32 ToDistribute;
+
+    /* Init */
+    U32 const lowThreshold = (U32)(total >> tableLog);
+    U32 lowOne = (U32)((total * 3) >> (tableLog + 1));
+
+    for (s=0; s<=maxSymbolValue; s++) {
+        if (count[s] == 0) {
+            norm[s]=0;
+            continue;
+        }
+        if (count[s] <= lowThreshold) {
+            norm[s] = -1;
+            distributed++;
+            total -= count[s];
+            continue;
+        }
+        if (count[s] <= lowOne) {
+            norm[s] = 1;
+            distributed++;
+            total -= count[s];
+            continue;
+        }
+        norm[s]=-2;
+    }
+    ToDistribute = (1 << tableLog) - distributed;
+
+    if ((total / ToDistribute) > lowOne) {
+        /* risk of rounding to zero */
+        lowOne = (U32)((total * 3) / (ToDistribute * 2));
+        for (s=0; s<=maxSymbolValue; s++) {
+            if ((norm[s] == -2) && (count[s] <= lowOne)) {
+                norm[s] = 1;
+                distributed++;
+                total -= count[s];
+                continue;
+        }   }
+        ToDistribute = (1 << tableLog) - distributed;
+    }
+
+    if (distributed == maxSymbolValue+1) {
+        /* all values are pretty poor;
+           probably incompressible data (should have already been detected);
+           find max, then give all remaining points to max */
+        U32 maxV = 0, maxC = 0;
+        for (s=0; s<=maxSymbolValue; s++)
+            if (count[s] > maxC) maxV=s, maxC=count[s];
+        norm[maxV] += (short)ToDistribute;
+        return 0;
+    }
+
+    {   U64 const vStepLog = 62 - tableLog;
+        U64 const mid = (1ULL << (vStepLog-1)) - 1;
+        U64 const rStep = ((((U64)1<> vStepLog);
+                U32 const sEnd = (U32)(end >> vStepLog);
+                U32 const weight = sEnd - sStart;
+                if (weight < 1)
+                    return ERROR(GENERIC);
+                norm[s] = (short)weight;
+                tmpTotal = end;
+    }   }   }
+
+    return 0;
+}
+
+
+size_t FSE_normalizeCount (short* normalizedCounter, unsigned tableLog,
+                           const unsigned* count, size_t total,
+                           unsigned maxSymbolValue)
+{
+    /* Sanity checks */
+    if (tableLog==0) tableLog = FSE_DEFAULT_TABLELOG;
+    if (tableLog < FSE_MIN_TABLELOG) return ERROR(GENERIC);   /* Unsupported size */
+    if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge);   /* Unsupported size */
+    if (tableLog < FSE_minTableLog(total, maxSymbolValue)) return ERROR(GENERIC);   /* Too small tableLog, compression potentially impossible */
+
+    {   U32 const rtbTable[] = {     0, 473195, 504333, 520860, 550000, 700000, 750000, 830000 };
+        U64 const scale = 62 - tableLog;
+        U64 const step = ((U64)1<<62) / total;   /* <== here, one division ! */
+        U64 const vStep = 1ULL<<(scale-20);
+        int stillToDistribute = 1<> tableLog);
+
+        for (s=0; s<=maxSymbolValue; s++) {
+            if (count[s] == total) return 0;   /* rle special case */
+            if (count[s] == 0) { normalizedCounter[s]=0; continue; }
+            if (count[s] <= lowThreshold) {
+                normalizedCounter[s] = -1;
+                stillToDistribute--;
+            } else {
+                short proba = (short)((count[s]*step) >> scale);
+                if (proba<8) {
+                    U64 restToBeat = vStep * rtbTable[proba];
+                    proba += (count[s]*step) - ((U64)proba< restToBeat;
+                }
+                if (proba > largestP) largestP=proba, largest=s;
+                normalizedCounter[s] = proba;
+                stillToDistribute -= proba;
+        }   }
+        if (-stillToDistribute >= (normalizedCounter[largest] >> 1)) {
+            /* corner case, need another normalization method */
+            size_t const errorCode = FSE_normalizeM2(normalizedCounter, tableLog, count, total, maxSymbolValue);
+            if (FSE_isError(errorCode)) return errorCode;
+        }
+        else normalizedCounter[largest] += (short)stillToDistribute;
+    }
+
+#if 0
+    {   /* Print Table (debug) */
+        U32 s;
+        U32 nTotal = 0;
+        for (s=0; s<=maxSymbolValue; s++)
+            printf("%3i: %4i \n", s, normalizedCounter[s]);
+        for (s=0; s<=maxSymbolValue; s++)
+            nTotal += abs(normalizedCounter[s]);
+        if (nTotal != (1U<>1);   /* assumption : tableLog >= 1 */
+    FSE_symbolCompressionTransform* const symbolTT = (FSE_symbolCompressionTransform*) (FSCT);
+    unsigned s;
+
+    /* Sanity checks */
+    if (nbBits < 1) return ERROR(GENERIC);             /* min size */
+
+    /* header */
+    tableU16[-2] = (U16) nbBits;
+    tableU16[-1] = (U16) maxSymbolValue;
+
+    /* Build table */
+    for (s=0; s FSE_MAX_TABLELOG*4+7 ) && (srcSize & 2)) {  /* test bit 2 */
+        FSE_encodeSymbol(&bitC, &CState2, *--ip);
+        FSE_encodeSymbol(&bitC, &CState1, *--ip);
+        FSE_FLUSHBITS(&bitC);
+    }
+
+    /* 2 or 4 encoding per loop */
+    while ( ip>istart ) {
+
+        FSE_encodeSymbol(&bitC, &CState2, *--ip);
+
+        if (sizeof(bitC.bitContainer)*8 < FSE_MAX_TABLELOG*2+7 )   /* this test must be static */
+            FSE_FLUSHBITS(&bitC);
+
+        FSE_encodeSymbol(&bitC, &CState1, *--ip);
+
+        if (sizeof(bitC.bitContainer)*8 > FSE_MAX_TABLELOG*4+7 ) {  /* this test must be static */
+            FSE_encodeSymbol(&bitC, &CState2, *--ip);
+            FSE_encodeSymbol(&bitC, &CState1, *--ip);
+        }
+
+        FSE_FLUSHBITS(&bitC);
+    }
+
+    FSE_flushCState(&bitC, &CState2);
+    FSE_flushCState(&bitC, &CState1);
+    return BIT_closeCStream(&bitC);
+}
+
+size_t FSE_compress_usingCTable (void* dst, size_t dstSize,
+                           const void* src, size_t srcSize,
+                           const FSE_CTable* ct)
+{
+    unsigned const fast = (dstSize >= FSE_BLOCKBOUND(srcSize));
+
+    if (fast)
+        return FSE_compress_usingCTable_generic(dst, dstSize, src, srcSize, ct, 1);
+    else
+        return FSE_compress_usingCTable_generic(dst, dstSize, src, srcSize, ct, 0);
+}
+
+
+size_t FSE_compressBound(size_t size) { return FSE_COMPRESSBOUND(size); }
+
+#define CHECK_V_F(e, f) size_t const e = f; if (ERR_isError(e)) return f
+#define CHECK_F(f)   { CHECK_V_F(_var_err__, f); }
+
+/* FSE_compress_wksp() :
+ * Same as FSE_compress2(), but using an externally allocated scratch buffer (`workSpace`).
+ * `wkspSize` size must be `(1< not compressible */
+        if (maxCount < (srcSize >> 7)) return 0;   /* Heuristic : not compressible enough */
+    }
+
+    tableLog = FSE_optimalTableLog(tableLog, srcSize, maxSymbolValue);
+    CHECK_F( FSE_normalizeCount(norm, tableLog, count, srcSize, maxSymbolValue) );
+
+    /* Write table description header */
+    {   CHECK_V_F(nc_err, FSE_writeNCount(op, oend-op, norm, maxSymbolValue, tableLog) );
+        op += nc_err;
+    }
+
+    /* Compress */
+    CHECK_F( FSE_buildCTable_wksp(CTable, norm, maxSymbolValue, tableLog, scratchBuffer, scratchBufferSize) );
+    {   CHECK_V_F(cSize, FSE_compress_usingCTable(op, oend - op, src, srcSize, CTable) );
+        if (cSize == 0) return 0;   /* not enough space for compressed data */
+        op += cSize;
+    }
+
+    /* check compressibility */
+    if ( (size_t)(op-ostart) >= srcSize-1 ) return 0;
+
+    return op-ostart;
+}
+
+typedef struct {
+    FSE_CTable CTable_max[FSE_CTABLE_SIZE_U32(FSE_MAX_TABLELOG, FSE_MAX_SYMBOL_VALUE)];
+    BYTE scratchBuffer[1 << FSE_MAX_TABLELOG];
+} fseWkspMax_t;
+
+size_t FSE_compress2 (void* dst, size_t dstCapacity, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog)
+{
+    fseWkspMax_t scratchBuffer;
+    FSE_STATIC_ASSERT(sizeof(scratchBuffer) >= FSE_WKSP_SIZE_U32(FSE_MAX_TABLELOG, FSE_MAX_SYMBOL_VALUE));   /* compilation failures here means scratchBuffer is not large enough */
+    if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge);
+    return FSE_compress_wksp(dst, dstCapacity, src, srcSize, maxSymbolValue, tableLog, &scratchBuffer, sizeof(scratchBuffer));
+}
+
+size_t FSE_compress (void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+    return FSE_compress2(dst, dstCapacity, src, srcSize, FSE_MAX_SYMBOL_VALUE, FSE_DEFAULT_TABLELOG);
+}
+
+
+#endif   /* FSE_COMMONDEFS_ONLY */
diff --git a/lib/compress/huf_compress.c b/lib/compress/huf_compress.c
new file mode 100644
index 0000000..bf464da
--- /dev/null
+++ b/lib/compress/huf_compress.c
@@ -0,0 +1,609 @@
+/* ******************************************************************
+   Huffman encoder, part of New Generation Entropy library
+   Copyright (C) 2013-2016, Yann Collet.
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+    You can contact the author at :
+    - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
+    - Public forum : https://groups.google.com/forum/#!forum/lz4c
+****************************************************************** */
+
+/* **************************************************************
+*  Compiler specifics
+****************************************************************/
+#ifdef _MSC_VER    /* Visual Studio */
+#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
+#endif
+
+
+/* **************************************************************
+*  Includes
+****************************************************************/
+#include      /* memcpy, memset */
+#include       /* printf (debug) */
+#include "bitstream.h"
+#define FSE_STATIC_LINKING_ONLY   /* FSE_optimalTableLog_internal */
+#include "fse.h"        /* header compression */
+#define HUF_STATIC_LINKING_ONLY
+#include "huf.h"
+
+
+/* **************************************************************
+*  Error Management
+****************************************************************/
+#define HUF_STATIC_ASSERT(c) { enum { HUF_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
+#define CHECK_V_F(e, f) size_t const e = f; if (ERR_isError(e)) return f
+#define CHECK_F(f)   { CHECK_V_F(_var_err__, f); }
+
+
+/* **************************************************************
+*  Utils
+****************************************************************/
+unsigned HUF_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue)
+{
+    return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 1);
+}
+
+
+/* *******************************************************
+*  HUF : Huffman block compression
+*********************************************************/
+/* HUF_compressWeights() :
+ * Same as FSE_compress(), but dedicated to huff0's weights compression.
+ * The use case needs much less stack memory.
+ * Note : all elements within weightTable are supposed to be <= HUF_TABLELOG_MAX.
+ */
+#define MAX_FSE_TABLELOG_FOR_HUFF_HEADER 6
+size_t HUF_compressWeights (void* dst, size_t dstSize, const void* weightTable, size_t wtSize)
+{
+    BYTE* const ostart = (BYTE*) dst;
+    BYTE* op = ostart;
+    BYTE* const oend = ostart + dstSize;
+
+    U32 maxSymbolValue = HUF_TABLELOG_MAX;
+    U32 tableLog = MAX_FSE_TABLELOG_FOR_HUFF_HEADER;
+
+    FSE_CTable CTable[FSE_CTABLE_SIZE_U32(MAX_FSE_TABLELOG_FOR_HUFF_HEADER, HUF_TABLELOG_MAX)];
+    BYTE scratchBuffer[1< not compressible */
+    }
+
+    tableLog = FSE_optimalTableLog(tableLog, wtSize, maxSymbolValue);
+    CHECK_F( FSE_normalizeCount(norm, tableLog, count, wtSize, maxSymbolValue) );
+
+    /* Write table description header */
+    {   CHECK_V_F(hSize, FSE_writeNCount(op, oend-op, norm, maxSymbolValue, tableLog) );
+        op += hSize;
+    }
+
+    /* Compress */
+    CHECK_F( FSE_buildCTable_wksp(CTable, norm, maxSymbolValue, tableLog, scratchBuffer, sizeof(scratchBuffer)) );
+    {   CHECK_V_F(cSize, FSE_compress_usingCTable(op, oend - op, weightTable, wtSize, CTable) );
+        if (cSize == 0) return 0;   /* not enough space for compressed data */
+        op += cSize;
+    }
+
+    return op-ostart;
+}
+
+
+struct HUF_CElt_s {
+  U16  val;
+  BYTE nbBits;
+};   /* typedef'd to HUF_CElt within "huf.h" */
+
+/*! HUF_writeCTable() :
+    `CTable` : huffman tree to save, using huf representation.
+    @return : size of saved CTable */
+size_t HUF_writeCTable (void* dst, size_t maxDstSize,
+                        const HUF_CElt* CTable, U32 maxSymbolValue, U32 huffLog)
+{
+    BYTE bitsToWeight[HUF_TABLELOG_MAX + 1];   /* precomputed conversion table */
+    BYTE huffWeight[HUF_SYMBOLVALUE_MAX];
+    BYTE* op = (BYTE*)dst;
+    U32 n;
+
+     /* check conditions */
+    if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge);
+
+    /* convert to weight */
+    bitsToWeight[0] = 0;
+    for (n=1; n1) & (hSize < maxSymbolValue/2)) {   /* FSE compressed */
+            op[0] = (BYTE)hSize;
+            return hSize+1;
+    }   }
+
+    /* write raw values as 4-bits (max : 15) */
+    if (maxSymbolValue > (256-128)) return ERROR(GENERIC);   /* should not happen : likely means source cannot be compressed */
+    if (((maxSymbolValue+1)/2) + 1 > maxDstSize) return ERROR(dstSize_tooSmall);   /* not enough space within dst buffer */
+    op[0] = (BYTE)(128 /*special case*/ + (maxSymbolValue-1));
+    huffWeight[maxSymbolValue] = 0;   /* to be sure it doesn't cause msan issue in final combination */
+    for (n=0; n HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
+    if (nbSymbols > maxSymbolValue+1) return ERROR(maxSymbolValue_tooSmall);
+
+    /* Prepare base value per rank */
+    {   U32 n, nextRankStart = 0;
+        for (n=1; n<=tableLog; n++) {
+            U32 current = nextRankStart;
+            nextRankStart += (rankVal[n] << (n-1));
+            rankVal[n] = current;
+    }   }
+
+    /* fill nbBits */
+    {   U32 n; for (n=0; nn=tableLog+1 */
+        U16 valPerRank[HUF_TABLELOG_MAX+2] = {0};
+        { U32 n; for (n=0; n0; n--) {  /* start at n=tablelog <-> w=1 */
+                valPerRank[n] = min;     /* get starting value within each rank */
+                min += nbPerRank[n];
+                min >>= 1;
+        }   }
+        /* assign value within rank, symbol order */
+        { U32 n; for (n=0; n<=maxSymbolValue; n++) CTable[n].val = valPerRank[CTable[n].nbBits]++; }
+    }
+
+    return readSize;
+}
+
+
+typedef struct nodeElt_s {
+    U32 count;
+    U16 parent;
+    BYTE byte;
+    BYTE nbBits;
+} nodeElt;
+
+static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
+{
+    const U32 largestBits = huffNode[lastNonNull].nbBits;
+    if (largestBits <= maxNbBits) return largestBits;   /* early exit : no elt > maxNbBits */
+
+    /* there are several too large elements (at least >= 2) */
+    {   int totalCost = 0;
+        const U32 baseCost = 1 << (largestBits - maxNbBits);
+        U32 n = lastNonNull;
+
+        while (huffNode[n].nbBits > maxNbBits) {
+            totalCost += baseCost - (1 << (largestBits - huffNode[n].nbBits));
+            huffNode[n].nbBits = (BYTE)maxNbBits;
+            n --;
+        }  /* n stops at huffNode[n].nbBits <= maxNbBits */
+        while (huffNode[n].nbBits == maxNbBits) n--;   /* n end at index of smallest symbol using < maxNbBits */
+
+        /* renorm totalCost */
+        totalCost >>= (largestBits - maxNbBits);  /* note : totalCost is necessarily a multiple of baseCost */
+
+        /* repay normalized cost */
+        {   U32 const noSymbol = 0xF0F0F0F0;
+            U32 rankLast[HUF_TABLELOG_MAX+2];
+            int pos;
+
+            /* Get pos of last (smallest) symbol per rank */
+            memset(rankLast, 0xF0, sizeof(rankLast));
+            {   U32 currentNbBits = maxNbBits;
+                for (pos=n ; pos >= 0; pos--) {
+                    if (huffNode[pos].nbBits >= currentNbBits) continue;
+                    currentNbBits = huffNode[pos].nbBits;   /* < maxNbBits */
+                    rankLast[maxNbBits-currentNbBits] = pos;
+            }   }
+
+            while (totalCost > 0) {
+                U32 nBitsToDecrease = BIT_highbit32(totalCost) + 1;
+                for ( ; nBitsToDecrease > 1; nBitsToDecrease--) {
+                    U32 highPos = rankLast[nBitsToDecrease];
+                    U32 lowPos = rankLast[nBitsToDecrease-1];
+                    if (highPos == noSymbol) continue;
+                    if (lowPos == noSymbol) break;
+                    {   U32 const highTotal = huffNode[highPos].count;
+                        U32 const lowTotal = 2 * huffNode[lowPos].count;
+                        if (highTotal <= lowTotal) break;
+                }   }
+                /* only triggered when no more rank 1 symbol left => find closest one (note : there is necessarily at least one !) */
+                while ((nBitsToDecrease<=HUF_TABLELOG_MAX) && (rankLast[nBitsToDecrease] == noSymbol))  /* HUF_MAX_TABLELOG test just to please gcc 5+; but it should not be necessary */
+                    nBitsToDecrease ++;
+                totalCost -= 1 << (nBitsToDecrease-1);
+                if (rankLast[nBitsToDecrease-1] == noSymbol)
+                    rankLast[nBitsToDecrease-1] = rankLast[nBitsToDecrease];   /* this rank is no longer empty */
+                huffNode[rankLast[nBitsToDecrease]].nbBits ++;
+                if (rankLast[nBitsToDecrease] == 0)    /* special case, reached largest symbol */
+                    rankLast[nBitsToDecrease] = noSymbol;
+                else {
+                    rankLast[nBitsToDecrease]--;
+                    if (huffNode[rankLast[nBitsToDecrease]].nbBits != maxNbBits-nBitsToDecrease)
+                        rankLast[nBitsToDecrease] = noSymbol;   /* this rank is now empty */
+            }   }   /* while (totalCost > 0) */
+
+            while (totalCost < 0) {  /* Sometimes, cost correction overshoot */
+                if (rankLast[1] == noSymbol) {  /* special case : no rank 1 symbol (using maxNbBits-1); let's create one from largest rank 0 (using maxNbBits) */
+                    while (huffNode[n].nbBits == maxNbBits) n--;
+                    huffNode[n+1].nbBits--;
+                    rankLast[1] = n+1;
+                    totalCost++;
+                    continue;
+                }
+                huffNode[ rankLast[1] + 1 ].nbBits--;
+                rankLast[1]++;
+                totalCost ++;
+    }   }   }   /* there are several too large elements (at least >= 2) */
+
+    return maxNbBits;
+}
+
+
+typedef struct {
+    U32 base;
+    U32 current;
+} rankPos;
+
+static void HUF_sort(nodeElt* huffNode, const U32* count, U32 maxSymbolValue)
+{
+    rankPos rank[32];
+    U32 n;
+
+    memset(rank, 0, sizeof(rank));
+    for (n=0; n<=maxSymbolValue; n++) {
+        U32 r = BIT_highbit32(count[n] + 1);
+        rank[r].base ++;
+    }
+    for (n=30; n>0; n--) rank[n-1].base += rank[n].base;
+    for (n=0; n<32; n++) rank[n].current = rank[n].base;
+    for (n=0; n<=maxSymbolValue; n++) {
+        U32 const c = count[n];
+        U32 const r = BIT_highbit32(c+1) + 1;
+        U32 pos = rank[r].current++;
+        while ((pos > rank[r].base) && (c > huffNode[pos-1].count)) huffNode[pos]=huffNode[pos-1], pos--;
+        huffNode[pos].count = c;
+        huffNode[pos].byte  = (BYTE)n;
+    }
+}
+
+
+/** HUF_buildCTable_wksp() :
+ *  Same as HUF_buildCTable(), but using externally allocated scratch buffer.
+ *  `workSpace` must be aligned on 4-bytes boundaries, and be at least as large as a table of 1024 unsigned.
+ */
+#define STARTNODE (HUF_SYMBOLVALUE_MAX+1)
+typedef nodeElt huffNodeTable[2*HUF_SYMBOLVALUE_MAX+1 +1];
+size_t HUF_buildCTable_wksp (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U32 maxNbBits, void* workSpace, size_t wkspSize)
+{
+    nodeElt* const huffNode0 = (nodeElt*)workSpace;
+    nodeElt* const huffNode = huffNode0+1;
+    U32 n, nonNullRank;
+    int lowS, lowN;
+    U16 nodeNb = STARTNODE;
+    U32 nodeRoot;
+
+    /* safety checks */
+    if (wkspSize < sizeof(huffNodeTable)) return ERROR(GENERIC);   /* workSpace is not large enough */
+    if (maxNbBits == 0) maxNbBits = HUF_TABLELOG_DEFAULT;
+    if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(GENERIC);
+    memset(huffNode0, 0, sizeof(huffNodeTable));
+
+    /* sort, decreasing order */
+    HUF_sort(huffNode, count, maxSymbolValue);
+
+    /* init for parents */
+    nonNullRank = maxSymbolValue;
+    while(huffNode[nonNullRank].count == 0) nonNullRank--;
+    lowS = nonNullRank; nodeRoot = nodeNb + lowS - 1; lowN = nodeNb;
+    huffNode[nodeNb].count = huffNode[lowS].count + huffNode[lowS-1].count;
+    huffNode[lowS].parent = huffNode[lowS-1].parent = nodeNb;
+    nodeNb++; lowS-=2;
+    for (n=nodeNb; n<=nodeRoot; n++) huffNode[n].count = (U32)(1U<<30);
+    huffNode0[0].count = (U32)(1U<<31);  /* fake entry, strong barrier */
+
+    /* create parents */
+    while (nodeNb <= nodeRoot) {
+        U32 n1 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++;
+        U32 n2 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++;
+        huffNode[nodeNb].count = huffNode[n1].count + huffNode[n2].count;
+        huffNode[n1].parent = huffNode[n2].parent = nodeNb;
+        nodeNb++;
+    }
+
+    /* distribute weights (unlimited tree height) */
+    huffNode[nodeRoot].nbBits = 0;
+    for (n=nodeRoot-1; n>=STARTNODE; n--)
+        huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1;
+    for (n=0; n<=nonNullRank; n++)
+        huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1;
+
+    /* enforce maxTableLog */
+    maxNbBits = HUF_setMaxHeight(huffNode, nonNullRank, maxNbBits);
+
+    /* fill result into tree (val, nbBits) */
+    {   U16 nbPerRank[HUF_TABLELOG_MAX+1] = {0};
+        U16 valPerRank[HUF_TABLELOG_MAX+1] = {0};
+        if (maxNbBits > HUF_TABLELOG_MAX) return ERROR(GENERIC);   /* check fit into table */
+        for (n=0; n<=nonNullRank; n++)
+            nbPerRank[huffNode[n].nbBits]++;
+        /* determine stating value per rank */
+        {   U16 min = 0;
+            for (n=maxNbBits; n>0; n--) {
+                valPerRank[n] = min;      /* get starting value within each rank */
+                min += nbPerRank[n];
+                min >>= 1;
+        }   }
+        for (n=0; n<=maxSymbolValue; n++)
+            tree[huffNode[n].byte].nbBits = huffNode[n].nbBits;   /* push nbBits per symbol, symbol order */
+        for (n=0; n<=maxSymbolValue; n++)
+            tree[n].val = valPerRank[tree[n].nbBits]++;   /* assign value within rank, symbol order */
+    }
+
+    return maxNbBits;
+}
+
+/** HUF_buildCTable() :
+ *  Note : count is used before tree is written, so they can safely overlap
+ */
+size_t HUF_buildCTable (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U32 maxNbBits)
+{
+    huffNodeTable nodeTable;
+    return HUF_buildCTable_wksp(tree, count, maxSymbolValue, maxNbBits, nodeTable, sizeof(nodeTable));
+}
+
+static void HUF_encodeSymbol(BIT_CStream_t* bitCPtr, U32 symbol, const HUF_CElt* CTable)
+{
+    BIT_addBitsFast(bitCPtr, CTable[symbol].val, CTable[symbol].nbBits);
+}
+
+size_t HUF_compressBound(size_t size) { return HUF_COMPRESSBOUND(size); }
+
+#define HUF_FLUSHBITS(s)  (fast ? BIT_flushBitsFast(s) : BIT_flushBits(s))
+
+#define HUF_FLUSHBITS_1(stream) \
+    if (sizeof((stream)->bitContainer)*8 < HUF_TABLELOG_MAX*2+7) HUF_FLUSHBITS(stream)
+
+#define HUF_FLUSHBITS_2(stream) \
+    if (sizeof((stream)->bitContainer)*8 < HUF_TABLELOG_MAX*4+7) HUF_FLUSHBITS(stream)
+
+size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
+{
+    const BYTE* ip = (const BYTE*) src;
+    BYTE* const ostart = (BYTE*)dst;
+    BYTE* const oend = ostart + dstSize;
+    BYTE* op = ostart;
+    size_t n;
+    const unsigned fast = (dstSize >= HUF_BLOCKBOUND(srcSize));
+    BIT_CStream_t bitC;
+
+    /* init */
+    if (dstSize < 8) return 0;   /* not enough space to compress */
+    { size_t const initErr = BIT_initCStream(&bitC, op, oend-op);
+      if (HUF_isError(initErr)) return 0; }
+
+    n = srcSize & ~3;  /* join to mod 4 */
+    switch (srcSize & 3)
+    {
+        case 3 : HUF_encodeSymbol(&bitC, ip[n+ 2], CTable);
+                 HUF_FLUSHBITS_2(&bitC);
+        case 2 : HUF_encodeSymbol(&bitC, ip[n+ 1], CTable);
+                 HUF_FLUSHBITS_1(&bitC);
+        case 1 : HUF_encodeSymbol(&bitC, ip[n+ 0], CTable);
+                 HUF_FLUSHBITS(&bitC);
+        case 0 :
+        default: ;
+    }
+
+    for (; n>0; n-=4) {  /* note : n&3==0 at this stage */
+        HUF_encodeSymbol(&bitC, ip[n- 1], CTable);
+        HUF_FLUSHBITS_1(&bitC);
+        HUF_encodeSymbol(&bitC, ip[n- 2], CTable);
+        HUF_FLUSHBITS_2(&bitC);
+        HUF_encodeSymbol(&bitC, ip[n- 3], CTable);
+        HUF_FLUSHBITS_1(&bitC);
+        HUF_encodeSymbol(&bitC, ip[n- 4], CTable);
+        HUF_FLUSHBITS(&bitC);
+    }
+
+    return BIT_closeCStream(&bitC);
+}
+
+
+size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
+{
+    size_t const segmentSize = (srcSize+3)/4;   /* first 3 segments */
+    const BYTE* ip = (const BYTE*) src;
+    const BYTE* const iend = ip + srcSize;
+    BYTE* const ostart = (BYTE*) dst;
+    BYTE* const oend = ostart + dstSize;
+    BYTE* op = ostart;
+
+    if (dstSize < 6 + 1 + 1 + 1 + 8) return 0;   /* minimum space to compress successfully */
+    if (srcSize < 12) return 0;   /* no saving possible : too small input */
+    op += 6;   /* jumpTable */
+
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable) );
+        if (cSize==0) return 0;
+        MEM_writeLE16(ostart, (U16)cSize);
+        op += cSize;
+    }
+
+    ip += segmentSize;
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable) );
+        if (cSize==0) return 0;
+        MEM_writeLE16(ostart+2, (U16)cSize);
+        op += cSize;
+    }
+
+    ip += segmentSize;
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable) );
+        if (cSize==0) return 0;
+        MEM_writeLE16(ostart+4, (U16)cSize);
+        op += cSize;
+    }
+
+    ip += segmentSize;
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, iend-ip, CTable) );
+        if (cSize==0) return 0;
+        op += cSize;
+    }
+
+    return op-ostart;
+}
+
+
+/* `workSpace` must a table of at least 1024 unsigned */
+static size_t HUF_compress_internal (
+                void* dst, size_t dstSize,
+                const void* src, size_t srcSize,
+                unsigned maxSymbolValue, unsigned huffLog,
+                unsigned singleStream,
+                void* workSpace, size_t wkspSize)
+{
+    BYTE* const ostart = (BYTE*)dst;
+    BYTE* const oend = ostart + dstSize;
+    BYTE* op = ostart;
+
+    union {
+        U32 count[HUF_SYMBOLVALUE_MAX+1];
+        HUF_CElt CTable[HUF_SYMBOLVALUE_MAX+1];
+    } table;   /* `count` can overlap with `CTable`; saves 1 KB */
+
+    /* checks & inits */
+    if (wkspSize < sizeof(huffNodeTable)) return ERROR(GENERIC);
+    if (!srcSize) return 0;  /* Uncompressed (note : 1 means rle, so first byte must be correct) */
+    if (!dstSize) return 0;  /* cannot fit within dst budget */
+    if (srcSize > HUF_BLOCKSIZE_MAX) return ERROR(srcSize_wrong);   /* current block size limit */
+    if (huffLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
+    if (!maxSymbolValue) maxSymbolValue = HUF_SYMBOLVALUE_MAX;
+    if (!huffLog) huffLog = HUF_TABLELOG_DEFAULT;
+
+    /* Scan input and build symbol stats */
+    {   CHECK_V_F(largest, FSE_count_wksp (table.count, &maxSymbolValue, (const BYTE*)src, srcSize, (U32*)workSpace) );
+        if (largest == srcSize) { *ostart = ((const BYTE*)src)[0]; return 1; }   /* single symbol, rle */
+        if (largest <= (srcSize >> 7)+1) return 0;   /* Fast heuristic : not compressible enough */
+    }
+
+    /* Build Huffman Tree */
+    huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
+    {   CHECK_V_F(maxBits, HUF_buildCTable_wksp (table.CTable, table.count, maxSymbolValue, huffLog, workSpace, wkspSize) );
+        huffLog = (U32)maxBits;
+    }
+
+    /* Write table description header */
+    {   CHECK_V_F(hSize, HUF_writeCTable (op, dstSize, table.CTable, maxSymbolValue, huffLog) );
+        if (hSize + 12 >= srcSize) return 0;   /* not useful to try compression */
+        op += hSize;
+    }
+
+    /* Compress */
+    {   size_t const cSize = (singleStream) ?
+                            HUF_compress1X_usingCTable(op, oend - op, src, srcSize, table.CTable) :   /* single segment */
+                            HUF_compress4X_usingCTable(op, oend - op, src, srcSize, table.CTable);
+        if (HUF_isError(cSize)) return cSize;
+        if (cSize==0) return 0;   /* uncompressible */
+        op += cSize;
+    }
+
+    /* check compressibility */
+    if ((size_t)(op-ostart) >= srcSize-1)
+        return 0;
+
+    return op-ostart;
+}
+
+
+size_t HUF_compress1X_wksp (void* dst, size_t dstSize,
+                      const void* src, size_t srcSize,
+                      unsigned maxSymbolValue, unsigned huffLog,
+                      void* workSpace, size_t wkspSize)
+{
+    return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 1 /* single stream */, workSpace, wkspSize);
+}
+
+size_t HUF_compress1X (void* dst, size_t dstSize,
+                 const void* src, size_t srcSize,
+                 unsigned maxSymbolValue, unsigned huffLog)
+{
+    unsigned workSpace[1024];
+    return HUF_compress1X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
+}
+
+size_t HUF_compress4X_wksp (void* dst, size_t dstSize,
+                      const void* src, size_t srcSize,
+                      unsigned maxSymbolValue, unsigned huffLog,
+                      void* workSpace, size_t wkspSize)
+{
+    return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 0 /* 4 streams */, workSpace, wkspSize);
+}
+
+size_t HUF_compress2 (void* dst, size_t dstSize,
+                const void* src, size_t srcSize,
+                unsigned maxSymbolValue, unsigned huffLog)
+{
+    unsigned workSpace[1024];
+    return HUF_compress4X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
+}
+
+size_t HUF_compress (void* dst, size_t maxDstSize, const void* src, size_t srcSize)
+{
+    return HUF_compress2(dst, maxDstSize, src, (U32)srcSize, 255, HUF_TABLELOG_DEFAULT);
+}
diff --git a/lib/compress/zstd_compress.c b/lib/compress/zstd_compress.c
new file mode 100644
index 0000000..3d10fbd
--- /dev/null
+++ b/lib/compress/zstd_compress.c
@@ -0,0 +1,3291 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+
+/*-*************************************
+*  Dependencies
+***************************************/
+#include          /* memset */
+#include "mem.h"
+#define XXH_STATIC_LINKING_ONLY   /* XXH64_state_t */
+#include "xxhash.h"               /* XXH_reset, update, digest */
+#define FSE_STATIC_LINKING_ONLY   /* FSE_encodeSymbol */
+#include "fse.h"
+#define HUF_STATIC_LINKING_ONLY
+#include "huf.h"
+#include "zstd_internal.h"  /* includes zstd.h */
+
+
+/*-*************************************
+*  Constants
+***************************************/
+static const U32 g_searchStrength = 8;   /* control skip over incompressible data */
+#define HASH_READ_SIZE 8
+typedef enum { ZSTDcs_created=0, ZSTDcs_init, ZSTDcs_ongoing, ZSTDcs_ending } ZSTD_compressionStage_e;
+
+
+/*-*************************************
+*  Helper functions
+***************************************/
+#define ZSTD_STATIC_ASSERT(c) { enum { ZSTD_static_assert = 1/(int)(!!(c)) }; }
+size_t ZSTD_compressBound(size_t srcSize) { return FSE_compressBound(srcSize) + 12; }
+
+
+/*-*************************************
+*  Sequence storage
+***************************************/
+static void ZSTD_resetSeqStore(seqStore_t* ssPtr)
+{
+    ssPtr->lit = ssPtr->litStart;
+    ssPtr->sequences = ssPtr->sequencesStart;
+    ssPtr->longLengthID = 0;
+}
+
+
+/*-*************************************
+*  Context memory management
+***************************************/
+struct ZSTD_CCtx_s
+{
+    const BYTE* nextSrc;    /* next block here to continue on current prefix */
+    const BYTE* base;       /* All regular indexes relative to this position */
+    const BYTE* dictBase;   /* extDict indexes relative to this position */
+    U32   dictLimit;        /* below that point, need extDict */
+    U32   lowLimit;         /* below that point, no more data */
+    U32   nextToUpdate;     /* index from which to continue dictionary update */
+    U32   nextToUpdate3;    /* index from which to continue dictionary update */
+    U32   hashLog3;         /* dispatch table : larger == faster, more memory */
+    U32   loadedDictEnd;
+    ZSTD_compressionStage_e stage;
+    U32   rep[ZSTD_REP_NUM];
+    U32   savedRep[ZSTD_REP_NUM];
+    U32   dictID;
+    ZSTD_parameters params;
+    void* workSpace;
+    size_t workSpaceSize;
+    size_t blockSize;
+    U64 frameContentSize;
+    XXH64_state_t xxhState;
+    ZSTD_customMem customMem;
+
+    seqStore_t seqStore;    /* sequences storage ptrs */
+    U32* hashTable;
+    U32* hashTable3;
+    U32* chainTable;
+    HUF_CElt* hufTable;
+    U32 flagStaticTables;
+    FSE_CTable offcodeCTable  [FSE_CTABLE_SIZE_U32(OffFSELog, MaxOff)];
+    FSE_CTable matchlengthCTable[FSE_CTABLE_SIZE_U32(MLFSELog, MaxML)];
+    FSE_CTable litlengthCTable  [FSE_CTABLE_SIZE_U32(LLFSELog, MaxLL)];
+    unsigned tmpCounters[1024];
+};
+
+ZSTD_CCtx* ZSTD_createCCtx(void)
+{
+    return ZSTD_createCCtx_advanced(defaultCustomMem);
+}
+
+ZSTD_CCtx* ZSTD_createCCtx_advanced(ZSTD_customMem customMem)
+{
+    ZSTD_CCtx* cctx;
+
+    if (!customMem.customAlloc && !customMem.customFree) customMem = defaultCustomMem;
+    if (!customMem.customAlloc || !customMem.customFree) return NULL;
+
+    cctx = (ZSTD_CCtx*) ZSTD_malloc(sizeof(ZSTD_CCtx), customMem);
+    if (!cctx) return NULL;
+    memset(cctx, 0, sizeof(ZSTD_CCtx));
+    memcpy(&(cctx->customMem), &customMem, sizeof(customMem));
+    return cctx;
+}
+
+size_t ZSTD_freeCCtx(ZSTD_CCtx* cctx)
+{
+    if (cctx==NULL) return 0;   /* support free on NULL */
+    ZSTD_free(cctx->workSpace, cctx->customMem);
+    ZSTD_free(cctx, cctx->customMem);
+    return 0;   /* reserved as a potential error code in the future */
+}
+
+size_t ZSTD_sizeof_CCtx(const ZSTD_CCtx* cctx)
+{
+    if (cctx==NULL) return 0;   /* support sizeof on NULL */
+    return sizeof(*cctx) + cctx->workSpaceSize;
+}
+
+const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx)   /* hidden interface */
+{
+    return &(ctx->seqStore);
+}
+
+static ZSTD_parameters ZSTD_getParamsFromCCtx(const ZSTD_CCtx* cctx)
+{
+    return cctx->params;
+}
+
+
+/** ZSTD_checkParams() :
+    ensure param values remain within authorized range.
+    @return : 0, or an error code if one value is beyond authorized range */
+size_t ZSTD_checkCParams(ZSTD_compressionParameters cParams)
+{
+#   define CLAMPCHECK(val,min,max) { if ((valmax)) return ERROR(compressionParameter_unsupported); }
+    CLAMPCHECK(cParams.windowLog, ZSTD_WINDOWLOG_MIN, ZSTD_WINDOWLOG_MAX);
+    CLAMPCHECK(cParams.chainLog, ZSTD_CHAINLOG_MIN, ZSTD_CHAINLOG_MAX);
+    CLAMPCHECK(cParams.hashLog, ZSTD_HASHLOG_MIN, ZSTD_HASHLOG_MAX);
+    CLAMPCHECK(cParams.searchLog, ZSTD_SEARCHLOG_MIN, ZSTD_SEARCHLOG_MAX);
+    { U32 const searchLengthMin = ((cParams.strategy == ZSTD_fast) | (cParams.strategy == ZSTD_greedy)) ? ZSTD_SEARCHLENGTH_MIN+1 : ZSTD_SEARCHLENGTH_MIN;
+      U32 const searchLengthMax = (cParams.strategy == ZSTD_fast) ? ZSTD_SEARCHLENGTH_MAX : ZSTD_SEARCHLENGTH_MAX-1;
+      CLAMPCHECK(cParams.searchLength, searchLengthMin, searchLengthMax); }
+    CLAMPCHECK(cParams.targetLength, ZSTD_TARGETLENGTH_MIN, ZSTD_TARGETLENGTH_MAX);
+    if ((U32)(cParams.strategy) > (U32)ZSTD_btopt2) return ERROR(compressionParameter_unsupported);
+    return 0;
+}
+
+
+/** ZSTD_cycleLog() :
+ *  condition for correct operation : hashLog > 1 */
+static U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat)
+{
+    U32 const btScale = ((U32)strat >= (U32)ZSTD_btlazy2);
+    return hashLog - btScale;
+}
+
+/** ZSTD_adjustCParams() :
+    optimize `cPar` for a given input (`srcSize` and `dictSize`).
+    mostly downsizing to reduce memory consumption and initialization.
+    Both `srcSize` and `dictSize` are optional (use 0 if unknown),
+    but if both are 0, no optimization can be done.
+    Note : cPar is considered validated at this stage. Use ZSTD_checkParams() to ensure that. */
+ZSTD_compressionParameters ZSTD_adjustCParams(ZSTD_compressionParameters cPar, unsigned long long srcSize, size_t dictSize)
+{
+    if (srcSize+dictSize == 0) return cPar;   /* no size information available : no adjustment */
+
+    /* resize params, to use less memory when necessary */
+    {   U32 const minSrcSize = (srcSize==0) ? 500 : 0;
+        U64 const rSize = srcSize + dictSize + minSrcSize;
+        if (rSize < ((U64)1< srcLog) cPar.windowLog = srcLog;
+    }   }
+    if (cPar.hashLog > cPar.windowLog) cPar.hashLog = cPar.windowLog;
+    {   U32 const cycleLog = ZSTD_cycleLog(cPar.chainLog, cPar.strategy);
+        if (cycleLog > cPar.windowLog) cPar.chainLog -= (cycleLog - cPar.windowLog);
+    }
+
+    if (cPar.windowLog < ZSTD_WINDOWLOG_ABSOLUTEMIN) cPar.windowLog = ZSTD_WINDOWLOG_ABSOLUTEMIN;  /* required for frame header */
+
+    return cPar;
+}
+
+
+size_t ZSTD_estimateCCtxSize(ZSTD_compressionParameters cParams)
+{
+    size_t const blockSize = MIN(ZSTD_BLOCKSIZE_ABSOLUTEMAX, (size_t)1 << cParams.windowLog);
+    U32    const divider = (cParams.searchLength==3) ? 3 : 4;
+    size_t const maxNbSeq = blockSize / divider;
+    size_t const tokenSpace = blockSize + 11*maxNbSeq;
+
+    size_t const chainSize = (cParams.strategy == ZSTD_fast) ? 0 : (1 << cParams.chainLog);
+    size_t const hSize = ((size_t)1) << cParams.hashLog;
+    U32    const hashLog3 = (cParams.searchLength>3) ? 0 : MIN(ZSTD_HASHLOG3_MAX, cParams.windowLog);
+    size_t const h3Size = ((size_t)1) << hashLog3;
+    size_t const tableSpace = (chainSize + hSize + h3Size) * sizeof(U32);
+
+    size_t const optSpace = ((MaxML+1) + (MaxLL+1) + (MaxOff+1) + (1<nextSrc - cctx->base);
+    cctx->params = params;
+    cctx->frameContentSize = frameContentSize;
+    cctx->lowLimit = end;
+    cctx->dictLimit = end;
+    cctx->nextToUpdate = end+1;
+    cctx->stage = ZSTDcs_init;
+    cctx->dictID = 0;
+    cctx->loadedDictEnd = 0;
+    { int i; for (i=0; irep[i] = repStartValue[i]; }
+    cctx->seqStore.litLengthSum = 0;  /* force reset of btopt stats */
+    XXH64_reset(&cctx->xxhState, 0);
+    return 0;
+}
+
+typedef enum { ZSTDcrp_continue, ZSTDcrp_noMemset, ZSTDcrp_fullReset } ZSTD_compResetPolicy_e;
+
+/*! ZSTD_resetCCtx_advanced() :
+    note : 'params' must be validated */
+static size_t ZSTD_resetCCtx_advanced (ZSTD_CCtx* zc,
+                                       ZSTD_parameters params, U64 frameContentSize,
+                                       ZSTD_compResetPolicy_e const crp)
+{
+    if (crp == ZSTDcrp_continue)
+        if (ZSTD_equivalentParams(params, zc->params))
+            return ZSTD_continueCCtx(zc, params, frameContentSize);
+
+    {   size_t const blockSize = MIN(ZSTD_BLOCKSIZE_ABSOLUTEMAX, (size_t)1 << params.cParams.windowLog);
+        U32    const divider = (params.cParams.searchLength==3) ? 3 : 4;
+        size_t const maxNbSeq = blockSize / divider;
+        size_t const tokenSpace = blockSize + 11*maxNbSeq;
+        size_t const chainSize = (params.cParams.strategy == ZSTD_fast) ? 0 : (1 << params.cParams.chainLog);
+        size_t const hSize = ((size_t)1) << params.cParams.hashLog;
+        U32    const hashLog3 = (params.cParams.searchLength>3) ? 0 : MIN(ZSTD_HASHLOG3_MAX, params.cParams.windowLog);
+        size_t const h3Size = ((size_t)1) << hashLog3;
+        size_t const tableSpace = (chainSize + hSize + h3Size) * sizeof(U32);
+        void* ptr;
+
+        /* Check if workSpace is large enough, alloc a new one if needed */
+        {   size_t const optSpace = ((MaxML+1) + (MaxLL+1) + (MaxOff+1) + (1<workSpaceSize < neededSpace) {
+                ZSTD_free(zc->workSpace, zc->customMem);
+                zc->workSpace = ZSTD_malloc(neededSpace, zc->customMem);
+                if (zc->workSpace == NULL) return ERROR(memory_allocation);
+                zc->workSpaceSize = neededSpace;
+        }   }
+
+        if (crp!=ZSTDcrp_noMemset) memset(zc->workSpace, 0, tableSpace);   /* reset tables only */
+        XXH64_reset(&zc->xxhState, 0);
+        zc->hashLog3 = hashLog3;
+        zc->hashTable = (U32*)(zc->workSpace);
+        zc->chainTable = zc->hashTable + hSize;
+        zc->hashTable3 = zc->chainTable + chainSize;
+        ptr = zc->hashTable3 + h3Size;
+        zc->hufTable = (HUF_CElt*)ptr;
+        zc->flagStaticTables = 0;
+        ptr = ((U32*)ptr) + 256;  /* note : HUF_CElt* is incomplete type, size is simulated using U32 */
+
+        zc->nextToUpdate = 1;
+        zc->nextSrc = NULL;
+        zc->base = NULL;
+        zc->dictBase = NULL;
+        zc->dictLimit = 0;
+        zc->lowLimit = 0;
+        zc->params = params;
+        zc->blockSize = blockSize;
+        zc->frameContentSize = frameContentSize;
+        { int i; for (i=0; irep[i] = repStartValue[i]; }
+
+        if ((params.cParams.strategy == ZSTD_btopt) || (params.cParams.strategy == ZSTD_btopt2)) {
+            zc->seqStore.litFreq = (U32*)ptr;
+            zc->seqStore.litLengthFreq = zc->seqStore.litFreq + (1<seqStore.matchLengthFreq = zc->seqStore.litLengthFreq + (MaxLL+1);
+            zc->seqStore.offCodeFreq = zc->seqStore.matchLengthFreq + (MaxML+1);
+            ptr = zc->seqStore.offCodeFreq + (MaxOff+1);
+            zc->seqStore.matchTable = (ZSTD_match_t*)ptr;
+            ptr = zc->seqStore.matchTable + ZSTD_OPT_NUM+1;
+            zc->seqStore.priceTable = (ZSTD_optimal_t*)ptr;
+            ptr = zc->seqStore.priceTable + ZSTD_OPT_NUM+1;
+            zc->seqStore.litLengthSum = 0;
+        }
+        zc->seqStore.sequencesStart = (seqDef*)ptr;
+        ptr = zc->seqStore.sequencesStart + maxNbSeq;
+        zc->seqStore.llCode = (BYTE*) ptr;
+        zc->seqStore.mlCode = zc->seqStore.llCode + maxNbSeq;
+        zc->seqStore.ofCode = zc->seqStore.mlCode + maxNbSeq;
+        zc->seqStore.litStart = zc->seqStore.ofCode + maxNbSeq;
+
+        zc->stage = ZSTDcs_init;
+        zc->dictID = 0;
+        zc->loadedDictEnd = 0;
+
+        return 0;
+    }
+}
+
+
+/*! ZSTD_copyCCtx() :
+*   Duplicate an existing context `srcCCtx` into another one `dstCCtx`.
+*   Only works during stage ZSTDcs_init (i.e. after creation, but before first call to ZSTD_compressContinue()).
+*   @return : 0, or an error code */
+size_t ZSTD_copyCCtx(ZSTD_CCtx* dstCCtx, const ZSTD_CCtx* srcCCtx, unsigned long long pledgedSrcSize)
+{
+    if (srcCCtx->stage!=ZSTDcs_init) return ERROR(stage_wrong);
+
+    memcpy(&dstCCtx->customMem, &srcCCtx->customMem, sizeof(ZSTD_customMem));
+    ZSTD_resetCCtx_advanced(dstCCtx, srcCCtx->params, pledgedSrcSize, ZSTDcrp_noMemset);
+
+    /* copy tables */
+    {   size_t const chainSize = (srcCCtx->params.cParams.strategy == ZSTD_fast) ? 0 : (1 << srcCCtx->params.cParams.chainLog);
+        size_t const hSize = ((size_t)1) << srcCCtx->params.cParams.hashLog;
+        size_t const h3Size = (size_t)1 << srcCCtx->hashLog3;
+        size_t const tableSpace = (chainSize + hSize + h3Size) * sizeof(U32);
+        memcpy(dstCCtx->workSpace, srcCCtx->workSpace, tableSpace);
+    }
+
+    /* copy dictionary offsets */
+    dstCCtx->nextToUpdate = srcCCtx->nextToUpdate;
+    dstCCtx->nextToUpdate3= srcCCtx->nextToUpdate3;
+    dstCCtx->nextSrc      = srcCCtx->nextSrc;
+    dstCCtx->base         = srcCCtx->base;
+    dstCCtx->dictBase     = srcCCtx->dictBase;
+    dstCCtx->dictLimit    = srcCCtx->dictLimit;
+    dstCCtx->lowLimit     = srcCCtx->lowLimit;
+    dstCCtx->loadedDictEnd= srcCCtx->loadedDictEnd;
+    dstCCtx->dictID       = srcCCtx->dictID;
+
+    /* copy entropy tables */
+    dstCCtx->flagStaticTables = srcCCtx->flagStaticTables;
+    if (srcCCtx->flagStaticTables) {
+        memcpy(dstCCtx->hufTable, srcCCtx->hufTable, 256*4);
+        memcpy(dstCCtx->litlengthCTable, srcCCtx->litlengthCTable, sizeof(dstCCtx->litlengthCTable));
+        memcpy(dstCCtx->matchlengthCTable, srcCCtx->matchlengthCTable, sizeof(dstCCtx->matchlengthCTable));
+        memcpy(dstCCtx->offcodeCTable, srcCCtx->offcodeCTable, sizeof(dstCCtx->offcodeCTable));
+    }
+
+    return 0;
+}
+
+
+/*! ZSTD_reduceTable() :
+*   reduce table indexes by `reducerValue` */
+static void ZSTD_reduceTable (U32* const table, U32 const size, U32 const reducerValue)
+{
+    U32 u;
+    for (u=0 ; u < size ; u++) {
+        if (table[u] < reducerValue) table[u] = 0;
+        else table[u] -= reducerValue;
+    }
+}
+
+/*! ZSTD_reduceIndex() :
+*   rescale all indexes to avoid future overflow (indexes are U32) */
+static void ZSTD_reduceIndex (ZSTD_CCtx* zc, const U32 reducerValue)
+{
+    { U32 const hSize = 1 << zc->params.cParams.hashLog;
+      ZSTD_reduceTable(zc->hashTable, hSize, reducerValue); }
+
+    { U32 const chainSize = (zc->params.cParams.strategy == ZSTD_fast) ? 0 : (1 << zc->params.cParams.chainLog);
+      ZSTD_reduceTable(zc->chainTable, chainSize, reducerValue); }
+
+    { U32 const h3Size = (zc->hashLog3) ? 1 << zc->hashLog3 : 0;
+      ZSTD_reduceTable(zc->hashTable3, h3Size, reducerValue); }
+}
+
+
+/*-*******************************************************
+*  Block entropic compression
+*********************************************************/
+
+/* See doc/zstd_compression_format.md for detailed format description */
+
+size_t ZSTD_noCompressBlock (void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+    if (srcSize + ZSTD_blockHeaderSize > dstCapacity) return ERROR(dstSize_tooSmall);
+    memcpy((BYTE*)dst + ZSTD_blockHeaderSize, src, srcSize);
+    MEM_writeLE24(dst, (U32)(srcSize << 2) + (U32)bt_raw);
+    return ZSTD_blockHeaderSize+srcSize;
+}
+
+
+static size_t ZSTD_noCompressLiterals (void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+    BYTE* const ostart = (BYTE* const)dst;
+    U32   const flSize = 1 + (srcSize>31) + (srcSize>4095);
+
+    if (srcSize + flSize > dstCapacity) return ERROR(dstSize_tooSmall);
+
+    switch(flSize)
+    {
+        case 1: /* 2 - 1 - 5 */
+            ostart[0] = (BYTE)((U32)set_basic + (srcSize<<3));
+            break;
+        case 2: /* 2 - 2 - 12 */
+            MEM_writeLE16(ostart, (U16)((U32)set_basic + (1<<2) + (srcSize<<4)));
+            break;
+        default:   /*note : should not be necessary : flSize is within {1,2,3} */
+        case 3: /* 2 - 2 - 20 */
+            MEM_writeLE32(ostart, (U32)((U32)set_basic + (3<<2) + (srcSize<<4)));
+            break;
+    }
+
+    memcpy(ostart + flSize, src, srcSize);
+    return srcSize + flSize;
+}
+
+static size_t ZSTD_compressRleLiteralsBlock (void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+    BYTE* const ostart = (BYTE* const)dst;
+    U32   const flSize = 1 + (srcSize>31) + (srcSize>4095);
+
+    (void)dstCapacity;  /* dstCapacity already guaranteed to be >=4, hence large enough */
+
+    switch(flSize)
+    {
+        case 1: /* 2 - 1 - 5 */
+            ostart[0] = (BYTE)((U32)set_rle + (srcSize<<3));
+            break;
+        case 2: /* 2 - 2 - 12 */
+            MEM_writeLE16(ostart, (U16)((U32)set_rle + (1<<2) + (srcSize<<4)));
+            break;
+        default:   /*note : should not be necessary : flSize is necessarily within {1,2,3} */
+        case 3: /* 2 - 2 - 20 */
+            MEM_writeLE32(ostart, (U32)((U32)set_rle + (3<<2) + (srcSize<<4)));
+            break;
+    }
+
+    ostart[flSize] = *(const BYTE*)src;
+    return flSize+1;
+}
+
+
+static size_t ZSTD_minGain(size_t srcSize) { return (srcSize >> 6) + 2; }
+
+static size_t ZSTD_compressLiterals (ZSTD_CCtx* zc,
+                                     void* dst, size_t dstCapacity,
+                               const void* src, size_t srcSize)
+{
+    size_t const minGain = ZSTD_minGain(srcSize);
+    size_t const lhSize = 3 + (srcSize >= 1 KB) + (srcSize >= 16 KB);
+    BYTE*  const ostart = (BYTE*)dst;
+    U32 singleStream = srcSize < 256;
+    symbolEncodingType_e hType = set_compressed;
+    size_t cLitSize;
+
+
+    /* small ? don't even attempt compression (speed opt) */
+#   define LITERAL_NOENTROPY 63
+    {   size_t const minLitSize = zc->flagStaticTables ? 6 : LITERAL_NOENTROPY;
+        if (srcSize <= minLitSize) return ZSTD_noCompressLiterals(dst, dstCapacity, src, srcSize);
+    }
+
+    if (dstCapacity < lhSize+1) return ERROR(dstSize_tooSmall);   /* not enough space for compression */
+    if (zc->flagStaticTables && (lhSize==3)) {
+        hType = set_repeat;
+        singleStream = 1;
+        cLitSize = HUF_compress1X_usingCTable(ostart+lhSize, dstCapacity-lhSize, src, srcSize, zc->hufTable);
+    } else {
+        cLitSize = singleStream ? HUF_compress1X_wksp(ostart+lhSize, dstCapacity-lhSize, src, srcSize, 255, 11, zc->tmpCounters, sizeof(zc->tmpCounters))
+                                : HUF_compress4X_wksp(ostart+lhSize, dstCapacity-lhSize, src, srcSize, 255, 11, zc->tmpCounters, sizeof(zc->tmpCounters));
+    }
+
+    if ((cLitSize==0) | (cLitSize >= srcSize - minGain))
+        return ZSTD_noCompressLiterals(dst, dstCapacity, src, srcSize);
+    if (cLitSize==1)
+        return ZSTD_compressRleLiteralsBlock(dst, dstCapacity, src, srcSize);
+
+    /* Build header */
+    switch(lhSize)
+    {
+    case 3: /* 2 - 2 - 10 - 10 */
+        {   U32 const lhc = hType + ((!singleStream) << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<14);
+            MEM_writeLE24(ostart, lhc);
+            break;
+        }
+    case 4: /* 2 - 2 - 14 - 14 */
+        {   U32 const lhc = hType + (2 << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<18);
+            MEM_writeLE32(ostart, lhc);
+            break;
+        }
+    default:   /* should not be necessary, lhSize is only {3,4,5} */
+    case 5: /* 2 - 2 - 18 - 18 */
+        {   U32 const lhc = hType + (3 << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<22);
+            MEM_writeLE32(ostart, lhc);
+            ostart[4] = (BYTE)(cLitSize >> 10);
+            break;
+        }
+    }
+    return lhSize+cLitSize;
+}
+
+static const BYTE LL_Code[64] = {  0,  1,  2,  3,  4,  5,  6,  7,
+                                   8,  9, 10, 11, 12, 13, 14, 15,
+                                  16, 16, 17, 17, 18, 18, 19, 19,
+                                  20, 20, 20, 20, 21, 21, 21, 21,
+                                  22, 22, 22, 22, 22, 22, 22, 22,
+                                  23, 23, 23, 23, 23, 23, 23, 23,
+                                  24, 24, 24, 24, 24, 24, 24, 24,
+                                  24, 24, 24, 24, 24, 24, 24, 24 };
+
+static const BYTE ML_Code[128] = { 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
+                                  16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
+                                  32, 32, 33, 33, 34, 34, 35, 35, 36, 36, 36, 36, 37, 37, 37, 37,
+                                  38, 38, 38, 38, 38, 38, 38, 38, 39, 39, 39, 39, 39, 39, 39, 39,
+                                  40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40,
+                                  41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,
+                                  42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,
+                                  42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42 };
+
+
+void ZSTD_seqToCodes(const seqStore_t* seqStorePtr)
+{
+    BYTE const LL_deltaCode = 19;
+    BYTE const ML_deltaCode = 36;
+    const seqDef* const sequences = seqStorePtr->sequencesStart;
+    BYTE* const llCodeTable = seqStorePtr->llCode;
+    BYTE* const ofCodeTable = seqStorePtr->ofCode;
+    BYTE* const mlCodeTable = seqStorePtr->mlCode;
+    U32 const nbSeq = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
+    U32 u;
+    for (u=0; u 63) ? (BYTE)ZSTD_highbit32(llv) + LL_deltaCode : LL_Code[llv];
+        ofCodeTable[u] = (BYTE)ZSTD_highbit32(sequences[u].offset);
+        mlCodeTable[u] = (mlv>127) ? (BYTE)ZSTD_highbit32(mlv) + ML_deltaCode : ML_Code[mlv];
+    }
+    if (seqStorePtr->longLengthID==1)
+        llCodeTable[seqStorePtr->longLengthPos] = MaxLL;
+    if (seqStorePtr->longLengthID==2)
+        mlCodeTable[seqStorePtr->longLengthPos] = MaxML;
+}
+
+
+size_t ZSTD_compressSequences(ZSTD_CCtx* zc,
+                              void* dst, size_t dstCapacity,
+                              size_t srcSize)
+{
+    const seqStore_t* seqStorePtr = &(zc->seqStore);
+    U32 count[MaxSeq+1];
+    S16 norm[MaxSeq+1];
+    FSE_CTable* CTable_LitLength = zc->litlengthCTable;
+    FSE_CTable* CTable_OffsetBits = zc->offcodeCTable;
+    FSE_CTable* CTable_MatchLength = zc->matchlengthCTable;
+    U32 LLtype, Offtype, MLtype;   /* compressed, raw or rle */
+    const seqDef* const sequences = seqStorePtr->sequencesStart;
+    const BYTE* const ofCodeTable = seqStorePtr->ofCode;
+    const BYTE* const llCodeTable = seqStorePtr->llCode;
+    const BYTE* const mlCodeTable = seqStorePtr->mlCode;
+    BYTE* const ostart = (BYTE*)dst;
+    BYTE* const oend = ostart + dstCapacity;
+    BYTE* op = ostart;
+    size_t const nbSeq = seqStorePtr->sequences - seqStorePtr->sequencesStart;
+    BYTE* seqHead;
+    BYTE scratchBuffer[1<litStart;
+        size_t const litSize = seqStorePtr->lit - literals;
+        size_t const cSize = ZSTD_compressLiterals(zc, op, dstCapacity, literals, litSize);
+        if (ZSTD_isError(cSize)) return cSize;
+        op += cSize;
+    }
+
+    /* Sequences Header */
+    if ((oend-op) < 3 /*max nbSeq Size*/ + 1 /*seqHead */) return ERROR(dstSize_tooSmall);
+    if (nbSeq < 0x7F) *op++ = (BYTE)nbSeq;
+    else if (nbSeq < LONGNBSEQ) op[0] = (BYTE)((nbSeq>>8) + 0x80), op[1] = (BYTE)nbSeq, op+=2;
+    else op[0]=0xFF, MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ)), op+=3;
+    if (nbSeq==0) goto _check_compressibility;
+
+    /* seqHead : flags for FSE encoding type */
+    seqHead = op++;
+
+#define MIN_SEQ_FOR_DYNAMIC_FSE   64
+#define MAX_SEQ_FOR_STATIC_FSE  1000
+
+    /* convert length/distances into codes */
+    ZSTD_seqToCodes(seqStorePtr);
+
+    /* CTable for Literal Lengths */
+    {   U32 max = MaxLL;
+        size_t const mostFrequent = FSE_countFast_wksp(count, &max, llCodeTable, nbSeq, zc->tmpCounters);
+        if ((mostFrequent == nbSeq) && (nbSeq > 2)) {
+            *op++ = llCodeTable[0];
+            FSE_buildCTable_rle(CTable_LitLength, (BYTE)max);
+            LLtype = set_rle;
+        } else if ((zc->flagStaticTables) && (nbSeq < MAX_SEQ_FOR_STATIC_FSE)) {
+            LLtype = set_repeat;
+        } else if ((nbSeq < MIN_SEQ_FOR_DYNAMIC_FSE) || (mostFrequent < (nbSeq >> (LL_defaultNormLog-1)))) {
+            FSE_buildCTable_wksp(CTable_LitLength, LL_defaultNorm, MaxLL, LL_defaultNormLog, scratchBuffer, sizeof(scratchBuffer));
+            LLtype = set_basic;
+        } else {
+            size_t nbSeq_1 = nbSeq;
+            const U32 tableLog = FSE_optimalTableLog(LLFSELog, nbSeq, max);
+            if (count[llCodeTable[nbSeq-1]]>1) { count[llCodeTable[nbSeq-1]]--; nbSeq_1--; }
+            FSE_normalizeCount(norm, tableLog, count, nbSeq_1, max);
+            { size_t const NCountSize = FSE_writeNCount(op, oend-op, norm, max, tableLog);   /* overflow protected */
+              if (FSE_isError(NCountSize)) return ERROR(GENERIC);
+              op += NCountSize; }
+            FSE_buildCTable_wksp(CTable_LitLength, norm, max, tableLog, scratchBuffer, sizeof(scratchBuffer));
+            LLtype = set_compressed;
+    }   }
+
+    /* CTable for Offsets */
+    {   U32 max = MaxOff;
+        size_t const mostFrequent = FSE_countFast_wksp(count, &max, ofCodeTable, nbSeq, zc->tmpCounters);
+        if ((mostFrequent == nbSeq) && (nbSeq > 2)) {
+            *op++ = ofCodeTable[0];
+            FSE_buildCTable_rle(CTable_OffsetBits, (BYTE)max);
+            Offtype = set_rle;
+        } else if ((zc->flagStaticTables) && (nbSeq < MAX_SEQ_FOR_STATIC_FSE)) {
+            Offtype = set_repeat;
+        } else if ((nbSeq < MIN_SEQ_FOR_DYNAMIC_FSE) || (mostFrequent < (nbSeq >> (OF_defaultNormLog-1)))) {
+            FSE_buildCTable_wksp(CTable_OffsetBits, OF_defaultNorm, MaxOff, OF_defaultNormLog, scratchBuffer, sizeof(scratchBuffer));
+            Offtype = set_basic;
+        } else {
+            size_t nbSeq_1 = nbSeq;
+            const U32 tableLog = FSE_optimalTableLog(OffFSELog, nbSeq, max);
+            if (count[ofCodeTable[nbSeq-1]]>1) { count[ofCodeTable[nbSeq-1]]--; nbSeq_1--; }
+            FSE_normalizeCount(norm, tableLog, count, nbSeq_1, max);
+            { size_t const NCountSize = FSE_writeNCount(op, oend-op, norm, max, tableLog);   /* overflow protected */
+              if (FSE_isError(NCountSize)) return ERROR(GENERIC);
+              op += NCountSize; }
+            FSE_buildCTable_wksp(CTable_OffsetBits, norm, max, tableLog, scratchBuffer, sizeof(scratchBuffer));
+            Offtype = set_compressed;
+    }   }
+
+    /* CTable for MatchLengths */
+    {   U32 max = MaxML;
+        size_t const mostFrequent = FSE_countFast_wksp(count, &max, mlCodeTable, nbSeq, zc->tmpCounters);
+        if ((mostFrequent == nbSeq) && (nbSeq > 2)) {
+            *op++ = *mlCodeTable;
+            FSE_buildCTable_rle(CTable_MatchLength, (BYTE)max);
+            MLtype = set_rle;
+        } else if ((zc->flagStaticTables) && (nbSeq < MAX_SEQ_FOR_STATIC_FSE)) {
+            MLtype = set_repeat;
+        } else if ((nbSeq < MIN_SEQ_FOR_DYNAMIC_FSE) || (mostFrequent < (nbSeq >> (ML_defaultNormLog-1)))) {
+            FSE_buildCTable_wksp(CTable_MatchLength, ML_defaultNorm, MaxML, ML_defaultNormLog, scratchBuffer, sizeof(scratchBuffer));
+            MLtype = set_basic;
+        } else {
+            size_t nbSeq_1 = nbSeq;
+            const U32 tableLog = FSE_optimalTableLog(MLFSELog, nbSeq, max);
+            if (count[mlCodeTable[nbSeq-1]]>1) { count[mlCodeTable[nbSeq-1]]--; nbSeq_1--; }
+            FSE_normalizeCount(norm, tableLog, count, nbSeq_1, max);
+            { size_t const NCountSize = FSE_writeNCount(op, oend-op, norm, max, tableLog);   /* overflow protected */
+              if (FSE_isError(NCountSize)) return ERROR(GENERIC);
+              op += NCountSize; }
+            FSE_buildCTable_wksp(CTable_MatchLength, norm, max, tableLog, scratchBuffer, sizeof(scratchBuffer));
+            MLtype = set_compressed;
+    }   }
+
+    *seqHead = (BYTE)((LLtype<<6) + (Offtype<<4) + (MLtype<<2));
+    zc->flagStaticTables = 0;
+
+    /* Encoding Sequences */
+    {   BIT_CStream_t blockStream;
+        FSE_CState_t  stateMatchLength;
+        FSE_CState_t  stateOffsetBits;
+        FSE_CState_t  stateLitLength;
+
+        CHECK_E(BIT_initCStream(&blockStream, op, oend-op), dstSize_tooSmall); /* not enough space remaining */
+
+        /* first symbols */
+        FSE_initCState2(&stateMatchLength, CTable_MatchLength, mlCodeTable[nbSeq-1]);
+        FSE_initCState2(&stateOffsetBits,  CTable_OffsetBits,  ofCodeTable[nbSeq-1]);
+        FSE_initCState2(&stateLitLength,   CTable_LitLength,   llCodeTable[nbSeq-1]);
+        BIT_addBits(&blockStream, sequences[nbSeq-1].litLength, LL_bits[llCodeTable[nbSeq-1]]);
+        if (MEM_32bits()) BIT_flushBits(&blockStream);
+        BIT_addBits(&blockStream, sequences[nbSeq-1].matchLength, ML_bits[mlCodeTable[nbSeq-1]]);
+        if (MEM_32bits()) BIT_flushBits(&blockStream);
+        BIT_addBits(&blockStream, sequences[nbSeq-1].offset, ofCodeTable[nbSeq-1]);
+        BIT_flushBits(&blockStream);
+
+        {   size_t n;
+            for (n=nbSeq-2 ; n= 64-7-(LLFSELog+MLFSELog+OffFSELog)))
+                    BIT_flushBits(&blockStream);                                /* (7)*/
+                BIT_addBits(&blockStream, sequences[n].litLength, llBits);
+                if (MEM_32bits() && ((llBits+mlBits)>24)) BIT_flushBits(&blockStream);
+                BIT_addBits(&blockStream, sequences[n].matchLength, mlBits);
+                if (MEM_32bits()) BIT_flushBits(&blockStream);                  /* (7)*/
+                BIT_addBits(&blockStream, sequences[n].offset, ofBits);         /* 31 */
+                BIT_flushBits(&blockStream);                                    /* (7)*/
+        }   }
+
+        FSE_flushCState(&blockStream, &stateMatchLength);
+        FSE_flushCState(&blockStream, &stateOffsetBits);
+        FSE_flushCState(&blockStream, &stateLitLength);
+
+        {   size_t const streamSize = BIT_closeCStream(&blockStream);
+            if (streamSize==0) return ERROR(dstSize_tooSmall);   /* not enough space */
+            op += streamSize;
+    }   }
+
+    /* check compressibility */
+_check_compressibility:
+    { size_t const minGain = ZSTD_minGain(srcSize);
+      size_t const maxCSize = srcSize - minGain;
+      if ((size_t)(op-ostart) >= maxCSize) return 0; }
+
+    /* confirm repcodes */
+    { int i; for (i=0; irep[i] = zc->savedRep[i]; }
+
+    return op - ostart;
+}
+
+
+/*! ZSTD_storeSeq() :
+    Store a sequence (literal length, literals, offset code and match length code) into seqStore_t.
+    `offsetCode` : distance to match, or 0 == repCode.
+    `matchCode` : matchLength - MINMATCH
+*/
+MEM_STATIC void ZSTD_storeSeq(seqStore_t* seqStorePtr, size_t litLength, const void* literals, U32 offsetCode, size_t matchCode)
+{
+#if 0  /* for debug */
+    static const BYTE* g_start = NULL;
+    const U32 pos = (U32)((const BYTE*)literals - g_start);
+    if (g_start==NULL) g_start = (const BYTE*)literals;
+    //if ((pos > 1) && (pos < 50000))
+        printf("Cpos %6u :%5u literals & match %3u bytes at distance %6u \n",
+               pos, (U32)litLength, (U32)matchCode+MINMATCH, (U32)offsetCode);
+#endif
+    /* copy Literals */
+    ZSTD_wildcopy(seqStorePtr->lit, literals, litLength);
+    seqStorePtr->lit += litLength;
+
+    /* literal Length */
+    if (litLength>0xFFFF) { seqStorePtr->longLengthID = 1; seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart); }
+    seqStorePtr->sequences[0].litLength = (U16)litLength;
+
+    /* match offset */
+    seqStorePtr->sequences[0].offset = offsetCode + 1;
+
+    /* match Length */
+    if (matchCode>0xFFFF) { seqStorePtr->longLengthID = 2; seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart); }
+    seqStorePtr->sequences[0].matchLength = (U16)matchCode;
+
+    seqStorePtr->sequences++;
+}
+
+
+/*-*************************************
+*  Match length counter
+***************************************/
+static unsigned ZSTD_NbCommonBytes (register size_t val)
+{
+    if (MEM_isLittleEndian()) {
+        if (MEM_64bits()) {
+#       if defined(_MSC_VER) && defined(_WIN64)
+            unsigned long r = 0;
+            _BitScanForward64( &r, (U64)val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 3)
+            return (__builtin_ctzll((U64)val) >> 3);
+#       else
+            static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2, 0, 3, 1, 3, 1, 4, 2, 7, 0, 2, 3, 6, 1, 5, 3, 5, 1, 3, 4, 4, 2, 5, 6, 7, 7, 0, 1, 2, 3, 3, 4, 6, 2, 6, 5, 5, 3, 4, 5, 6, 7, 1, 2, 4, 6, 4, 4, 5, 7, 2, 6, 5, 7, 6, 7, 7 };
+            return DeBruijnBytePos[((U64)((val & -(long long)val) * 0x0218A392CDABBD3FULL)) >> 58];
+#       endif
+        } else { /* 32 bits */
+#       if defined(_MSC_VER)
+            unsigned long r=0;
+            _BitScanForward( &r, (U32)val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 3)
+            return (__builtin_ctz((U32)val) >> 3);
+#       else
+            static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0, 3, 2, 2, 1, 3, 2, 0, 1, 3, 3, 1, 2, 2, 2, 2, 0, 3, 1, 2, 0, 1, 0, 1, 1 };
+            return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27];
+#       endif
+        }
+    } else {  /* Big Endian CPU */
+        if (MEM_64bits()) {
+#       if defined(_MSC_VER) && defined(_WIN64)
+            unsigned long r = 0;
+            _BitScanReverse64( &r, val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 3)
+            return (__builtin_clzll(val) >> 3);
+#       else
+            unsigned r;
+            const unsigned n32 = sizeof(size_t)*4;   /* calculate this way due to compiler complaining in 32-bits mode */
+            if (!(val>>n32)) { r=4; } else { r=0; val>>=n32; }
+            if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; }
+            r += (!val);
+            return r;
+#       endif
+        } else { /* 32 bits */
+#       if defined(_MSC_VER)
+            unsigned long r = 0;
+            _BitScanReverse( &r, (unsigned long)val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 3)
+            return (__builtin_clz((U32)val) >> 3);
+#       else
+            unsigned r;
+            if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; }
+            r += (!val);
+            return r;
+#       endif
+    }   }
+}
+
+
+static size_t ZSTD_count(const BYTE* pIn, const BYTE* pMatch, const BYTE* const pInLimit)
+{
+    const BYTE* const pStart = pIn;
+    const BYTE* const pInLoopLimit = pInLimit - (sizeof(size_t)-1);
+
+    while (pIn < pInLoopLimit) {
+        size_t const diff = MEM_readST(pMatch) ^ MEM_readST(pIn);
+        if (!diff) { pIn+=sizeof(size_t); pMatch+=sizeof(size_t); continue; }
+        pIn += ZSTD_NbCommonBytes(diff);
+        return (size_t)(pIn - pStart);
+    }
+    if (MEM_64bits()) if ((pIn<(pInLimit-3)) && (MEM_read32(pMatch) == MEM_read32(pIn))) { pIn+=4; pMatch+=4; }
+    if ((pIn<(pInLimit-1)) && (MEM_read16(pMatch) == MEM_read16(pIn))) { pIn+=2; pMatch+=2; }
+    if ((pIn> (32-h) ; }
+MEM_STATIC size_t ZSTD_hash3Ptr(const void* ptr, U32 h) { return ZSTD_hash3(MEM_readLE32(ptr), h); }   /* only in zstd_opt.h */
+
+static const U32 prime4bytes = 2654435761U;
+static U32    ZSTD_hash4(U32 u, U32 h) { return (u * prime4bytes) >> (32-h) ; }
+static size_t ZSTD_hash4Ptr(const void* ptr, U32 h) { return ZSTD_hash4(MEM_read32(ptr), h); }
+
+static const U64 prime5bytes = 889523592379ULL;
+static size_t ZSTD_hash5(U64 u, U32 h) { return (size_t)(((u  << (64-40)) * prime5bytes) >> (64-h)) ; }
+static size_t ZSTD_hash5Ptr(const void* p, U32 h) { return ZSTD_hash5(MEM_readLE64(p), h); }
+
+static const U64 prime6bytes = 227718039650203ULL;
+static size_t ZSTD_hash6(U64 u, U32 h) { return (size_t)(((u  << (64-48)) * prime6bytes) >> (64-h)) ; }
+static size_t ZSTD_hash6Ptr(const void* p, U32 h) { return ZSTD_hash6(MEM_readLE64(p), h); }
+
+static const U64 prime7bytes = 58295818150454627ULL;
+static size_t ZSTD_hash7(U64 u, U32 h) { return (size_t)(((u  << (64-56)) * prime7bytes) >> (64-h)) ; }
+static size_t ZSTD_hash7Ptr(const void* p, U32 h) { return ZSTD_hash7(MEM_readLE64(p), h); }
+
+static const U64 prime8bytes = 0xCF1BBCDCB7A56463ULL;
+static size_t ZSTD_hash8(U64 u, U32 h) { return (size_t)(((u) * prime8bytes) >> (64-h)) ; }
+static size_t ZSTD_hash8Ptr(const void* p, U32 h) { return ZSTD_hash8(MEM_readLE64(p), h); }
+
+static size_t ZSTD_hashPtr(const void* p, U32 hBits, U32 mls)
+{
+    switch(mls)
+    {
+    default:
+    case 4: return ZSTD_hash4Ptr(p, hBits);
+    case 5: return ZSTD_hash5Ptr(p, hBits);
+    case 6: return ZSTD_hash6Ptr(p, hBits);
+    case 7: return ZSTD_hash7Ptr(p, hBits);
+    case 8: return ZSTD_hash8Ptr(p, hBits);
+    }
+}
+
+
+/*-*************************************
+*  Fast Scan
+***************************************/
+static void ZSTD_fillHashTable (ZSTD_CCtx* zc, const void* end, const U32 mls)
+{
+    U32* const hashTable = zc->hashTable;
+    U32  const hBits = zc->params.cParams.hashLog;
+    const BYTE* const base = zc->base;
+    const BYTE* ip = base + zc->nextToUpdate;
+    const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
+    const size_t fastHashFillStep = 3;
+
+    while(ip <= iend) {
+        hashTable[ZSTD_hashPtr(ip, hBits, mls)] = (U32)(ip - base);
+        ip += fastHashFillStep;
+    }
+}
+
+
+FORCE_INLINE
+void ZSTD_compressBlock_fast_generic(ZSTD_CCtx* cctx,
+                               const void* src, size_t srcSize,
+                               const U32 mls)
+{
+    U32* const hashTable = cctx->hashTable;
+    U32  const hBits = cctx->params.cParams.hashLog;
+    seqStore_t* seqStorePtr = &(cctx->seqStore);
+    const BYTE* const base = cctx->base;
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const U32   lowestIndex = cctx->dictLimit;
+    const BYTE* const lowest = base + lowestIndex;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - HASH_READ_SIZE;
+    U32 offset_1=cctx->rep[0], offset_2=cctx->rep[1];
+    U32 offsetSaved = 0;
+
+    /* init */
+    ip += (ip==lowest);
+    {   U32 const maxRep = (U32)(ip-lowest);
+        if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
+        if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
+    }
+
+    /* Main Search Loop */
+    while (ip < ilimit) {   /* < instead of <=, because repcode check at (ip+1) */
+        size_t mLength;
+        size_t const h = ZSTD_hashPtr(ip, hBits, mls);
+        U32 const current = (U32)(ip-base);
+        U32 const matchIndex = hashTable[h];
+        const BYTE* match = base + matchIndex;
+        hashTable[h] = current;   /* update hash table */
+
+        if ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1))) {
+            mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
+            ip++;
+            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
+        } else {
+            U32 offset;
+            if ( (matchIndex <= lowestIndex) || (MEM_read32(match) != MEM_read32(ip)) ) {
+                ip += ((ip-anchor) >> g_searchStrength) + 1;
+                continue;
+            }
+            mLength = ZSTD_count(ip+4, match+4, iend) + 4;
+            offset = (U32)(ip-match);
+            while (((ip>anchor) & (match>lowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
+            offset_2 = offset_1;
+            offset_1 = offset;
+
+            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+        }
+
+        /* match found */
+        ip += mLength;
+        anchor = ip;
+
+        if (ip <= ilimit) {
+            /* Fill Table */
+            hashTable[ZSTD_hashPtr(base+current+2, hBits, mls)] = current+2;  /* here because current+2 could be > iend-8 */
+            hashTable[ZSTD_hashPtr(ip-2, hBits, mls)] = (U32)(ip-2-base);
+            /* check immediate repcode */
+            while ( (ip <= ilimit)
+                 && ( (offset_2>0)
+                 & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
+                /* store sequence */
+                size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
+                { U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; }  /* swap offset_2 <=> offset_1 */
+                hashTable[ZSTD_hashPtr(ip, hBits, mls)] = (U32)(ip-base);
+                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
+                ip += rLength;
+                anchor = ip;
+                continue;   /* faster when present ... (?) */
+    }   }   }
+
+    /* save reps for next block */
+    cctx->savedRep[0] = offset_1 ? offset_1 : offsetSaved;
+    cctx->savedRep[1] = offset_2 ? offset_2 : offsetSaved;
+
+    /* Last Literals */
+    {   size_t const lastLLSize = iend - anchor;
+        memcpy(seqStorePtr->lit, anchor, lastLLSize);
+        seqStorePtr->lit += lastLLSize;
+    }
+}
+
+
+static void ZSTD_compressBlock_fast(ZSTD_CCtx* ctx,
+                       const void* src, size_t srcSize)
+{
+    const U32 mls = ctx->params.cParams.searchLength;
+    switch(mls)
+    {
+    default:
+    case 4 :
+        ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 4); return;
+    case 5 :
+        ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 5); return;
+    case 6 :
+        ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 6); return;
+    case 7 :
+        ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 7); return;
+    }
+}
+
+
+static void ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
+                                 const void* src, size_t srcSize,
+                                 const U32 mls)
+{
+    U32* hashTable = ctx->hashTable;
+    const U32 hBits = ctx->params.cParams.hashLog;
+    seqStore_t* seqStorePtr = &(ctx->seqStore);
+    const BYTE* const base = ctx->base;
+    const BYTE* const dictBase = ctx->dictBase;
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const U32   lowestIndex = ctx->lowLimit;
+    const BYTE* const dictStart = dictBase + lowestIndex;
+    const U32   dictLimit = ctx->dictLimit;
+    const BYTE* const lowPrefixPtr = base + dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - 8;
+    U32 offset_1=ctx->rep[0], offset_2=ctx->rep[1];
+
+    /* Search Loop */
+    while (ip < ilimit) {  /* < instead of <=, because (ip+1) */
+        const size_t h = ZSTD_hashPtr(ip, hBits, mls);
+        const U32 matchIndex = hashTable[h];
+        const BYTE* matchBase = matchIndex < dictLimit ? dictBase : base;
+        const BYTE* match = matchBase + matchIndex;
+        const U32 current = (U32)(ip-base);
+        const U32 repIndex = current + 1 - offset_1;   /* offset_1 expected <= current +1 */
+        const BYTE* repBase = repIndex < dictLimit ? dictBase : base;
+        const BYTE* repMatch = repBase + repIndex;
+        size_t mLength;
+        hashTable[h] = current;   /* update hash table */
+
+        if ( (((U32)((dictLimit-1) - repIndex) >= 3) /* intentional underflow */ & (repIndex > lowestIndex))
+           && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
+            const BYTE* repMatchEnd = repIndex < dictLimit ? dictEnd : iend;
+            mLength = ZSTD_count_2segments(ip+1+EQUAL_READ32, repMatch+EQUAL_READ32, iend, repMatchEnd, lowPrefixPtr) + EQUAL_READ32;
+            ip++;
+            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
+        } else {
+            if ( (matchIndex < lowestIndex) ||
+                 (MEM_read32(match) != MEM_read32(ip)) ) {
+                ip += ((ip-anchor) >> g_searchStrength) + 1;
+                continue;
+            }
+            {   const BYTE* matchEnd = matchIndex < dictLimit ? dictEnd : iend;
+                const BYTE* lowMatchPtr = matchIndex < dictLimit ? dictStart : lowPrefixPtr;
+                U32 offset;
+                mLength = ZSTD_count_2segments(ip+EQUAL_READ32, match+EQUAL_READ32, iend, matchEnd, lowPrefixPtr) + EQUAL_READ32;
+                while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; }   /* catch up */
+                offset = current - matchIndex;
+                offset_2 = offset_1;
+                offset_1 = offset;
+                ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+        }   }
+
+        /* found a match : store it */
+        ip += mLength;
+        anchor = ip;
+
+        if (ip <= ilimit) {
+            /* Fill Table */
+            hashTable[ZSTD_hashPtr(base+current+2, hBits, mls)] = current+2;
+            hashTable[ZSTD_hashPtr(ip-2, hBits, mls)] = (U32)(ip-2-base);
+            /* check immediate repcode */
+            while (ip <= ilimit) {
+                U32 const current2 = (U32)(ip-base);
+                U32 const repIndex2 = current2 - offset_2;
+                const BYTE* repMatch2 = repIndex2 < dictLimit ? dictBase + repIndex2 : base + repIndex2;
+                if ( (((U32)((dictLimit-1) - repIndex2) >= 3) & (repIndex2 > lowestIndex))  /* intentional overflow */
+                   && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
+                    const BYTE* const repEnd2 = repIndex2 < dictLimit ? dictEnd : iend;
+                    size_t repLength2 = ZSTD_count_2segments(ip+EQUAL_READ32, repMatch2+EQUAL_READ32, iend, repEnd2, lowPrefixPtr) + EQUAL_READ32;
+                    U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset;   /* swap offset_2 <=> offset_1 */
+                    ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
+                    hashTable[ZSTD_hashPtr(ip, hBits, mls)] = current2;
+                    ip += repLength2;
+                    anchor = ip;
+                    continue;
+                }
+                break;
+    }   }   }
+
+    /* save reps for next block */
+    ctx->savedRep[0] = offset_1; ctx->savedRep[1] = offset_2;
+
+    /* Last Literals */
+    {   size_t const lastLLSize = iend - anchor;
+        memcpy(seqStorePtr->lit, anchor, lastLLSize);
+        seqStorePtr->lit += lastLLSize;
+    }
+}
+
+
+static void ZSTD_compressBlock_fast_extDict(ZSTD_CCtx* ctx,
+                         const void* src, size_t srcSize)
+{
+    U32 const mls = ctx->params.cParams.searchLength;
+    switch(mls)
+    {
+    default:
+    case 4 :
+        ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 4); return;
+    case 5 :
+        ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 5); return;
+    case 6 :
+        ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 6); return;
+    case 7 :
+        ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 7); return;
+    }
+}
+
+
+/*-*************************************
+*  Double Fast
+***************************************/
+static void ZSTD_fillDoubleHashTable (ZSTD_CCtx* cctx, const void* end, const U32 mls)
+{
+    U32* const hashLarge = cctx->hashTable;
+    U32  const hBitsL = cctx->params.cParams.hashLog;
+    U32* const hashSmall = cctx->chainTable;
+    U32  const hBitsS = cctx->params.cParams.chainLog;
+    const BYTE* const base = cctx->base;
+    const BYTE* ip = base + cctx->nextToUpdate;
+    const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
+    const size_t fastHashFillStep = 3;
+
+    while(ip <= iend) {
+        hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip - base);
+        hashLarge[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip - base);
+        ip += fastHashFillStep;
+    }
+}
+
+
+FORCE_INLINE
+void ZSTD_compressBlock_doubleFast_generic(ZSTD_CCtx* cctx,
+                                 const void* src, size_t srcSize,
+                                 const U32 mls)
+{
+    U32* const hashLong = cctx->hashTable;
+    const U32 hBitsL = cctx->params.cParams.hashLog;
+    U32* const hashSmall = cctx->chainTable;
+    const U32 hBitsS = cctx->params.cParams.chainLog;
+    seqStore_t* seqStorePtr = &(cctx->seqStore);
+    const BYTE* const base = cctx->base;
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const U32 lowestIndex = cctx->dictLimit;
+    const BYTE* const lowest = base + lowestIndex;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - HASH_READ_SIZE;
+    U32 offset_1=cctx->rep[0], offset_2=cctx->rep[1];
+    U32 offsetSaved = 0;
+
+    /* init */
+    ip += (ip==lowest);
+    {   U32 const maxRep = (U32)(ip-lowest);
+        if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
+        if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
+    }
+
+    /* Main Search Loop */
+    while (ip < ilimit) {   /* < instead of <=, because repcode check at (ip+1) */
+        size_t mLength;
+        size_t const h2 = ZSTD_hashPtr(ip, hBitsL, 8);
+        size_t const h = ZSTD_hashPtr(ip, hBitsS, mls);
+        U32 const current = (U32)(ip-base);
+        U32 const matchIndexL = hashLong[h2];
+        U32 const matchIndexS = hashSmall[h];
+        const BYTE* matchLong = base + matchIndexL;
+        const BYTE* match = base + matchIndexS;
+        hashLong[h2] = hashSmall[h] = current;   /* update hash tables */
+
+        if ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1))) { /* note : by construction, offset_1 <= current */
+            mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
+            ip++;
+            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
+        } else {
+            U32 offset;
+            if ( (matchIndexL > lowestIndex) && (MEM_read64(matchLong) == MEM_read64(ip)) ) {
+                mLength = ZSTD_count(ip+8, matchLong+8, iend) + 8;
+                offset = (U32)(ip-matchLong);
+                while (((ip>anchor) & (matchLong>lowest)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */
+            } else if ( (matchIndexS > lowestIndex) && (MEM_read32(match) == MEM_read32(ip)) ) {
+                size_t const h3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
+                U32 const matchIndex3 = hashLong[h3];
+                const BYTE* match3 = base + matchIndex3;
+                hashLong[h3] = current + 1;
+                if ( (matchIndex3 > lowestIndex) && (MEM_read64(match3) == MEM_read64(ip+1)) ) {
+                    mLength = ZSTD_count(ip+9, match3+8, iend) + 8;
+                    ip++;
+                    offset = (U32)(ip-match3);
+                    while (((ip>anchor) & (match3>lowest)) && (ip[-1] == match3[-1])) { ip--; match3--; mLength++; } /* catch up */
+                } else {
+                    mLength = ZSTD_count(ip+4, match+4, iend) + 4;
+                    offset = (U32)(ip-match);
+                    while (((ip>anchor) & (match>lowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
+                }
+            } else {
+                ip += ((ip-anchor) >> g_searchStrength) + 1;
+                continue;
+            }
+
+            offset_2 = offset_1;
+            offset_1 = offset;
+
+            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+        }
+
+        /* match found */
+        ip += mLength;
+        anchor = ip;
+
+        if (ip <= ilimit) {
+            /* Fill Table */
+            hashLong[ZSTD_hashPtr(base+current+2, hBitsL, 8)] =
+                hashSmall[ZSTD_hashPtr(base+current+2, hBitsS, mls)] = current+2;  /* here because current+2 could be > iend-8 */
+            hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] =
+                hashSmall[ZSTD_hashPtr(ip-2, hBitsS, mls)] = (U32)(ip-2-base);
+
+            /* check immediate repcode */
+            while ( (ip <= ilimit)
+                 && ( (offset_2>0)
+                 & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
+                /* store sequence */
+                size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
+                { U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; } /* swap offset_2 <=> offset_1 */
+                hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip-base);
+                hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip-base);
+                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
+                ip += rLength;
+                anchor = ip;
+                continue;   /* faster when present ... (?) */
+    }   }   }
+
+    /* save reps for next block */
+    cctx->savedRep[0] = offset_1 ? offset_1 : offsetSaved;
+    cctx->savedRep[1] = offset_2 ? offset_2 : offsetSaved;
+
+    /* Last Literals */
+    {   size_t const lastLLSize = iend - anchor;
+        memcpy(seqStorePtr->lit, anchor, lastLLSize);
+        seqStorePtr->lit += lastLLSize;
+    }
+}
+
+
+static void ZSTD_compressBlock_doubleFast(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    const U32 mls = ctx->params.cParams.searchLength;
+    switch(mls)
+    {
+    default:
+    case 4 :
+        ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 4); return;
+    case 5 :
+        ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 5); return;
+    case 6 :
+        ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 6); return;
+    case 7 :
+        ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 7); return;
+    }
+}
+
+
+static void ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
+                                 const void* src, size_t srcSize,
+                                 const U32 mls)
+{
+    U32* const hashLong = ctx->hashTable;
+    U32  const hBitsL = ctx->params.cParams.hashLog;
+    U32* const hashSmall = ctx->chainTable;
+    U32  const hBitsS = ctx->params.cParams.chainLog;
+    seqStore_t* seqStorePtr = &(ctx->seqStore);
+    const BYTE* const base = ctx->base;
+    const BYTE* const dictBase = ctx->dictBase;
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const U32   lowestIndex = ctx->lowLimit;
+    const BYTE* const dictStart = dictBase + lowestIndex;
+    const U32   dictLimit = ctx->dictLimit;
+    const BYTE* const lowPrefixPtr = base + dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - 8;
+    U32 offset_1=ctx->rep[0], offset_2=ctx->rep[1];
+
+    /* Search Loop */
+    while (ip < ilimit) {  /* < instead of <=, because (ip+1) */
+        const size_t hSmall = ZSTD_hashPtr(ip, hBitsS, mls);
+        const U32 matchIndex = hashSmall[hSmall];
+        const BYTE* matchBase = matchIndex < dictLimit ? dictBase : base;
+        const BYTE* match = matchBase + matchIndex;
+
+        const size_t hLong = ZSTD_hashPtr(ip, hBitsL, 8);
+        const U32 matchLongIndex = hashLong[hLong];
+        const BYTE* matchLongBase = matchLongIndex < dictLimit ? dictBase : base;
+        const BYTE* matchLong = matchLongBase + matchLongIndex;
+
+        const U32 current = (U32)(ip-base);
+        const U32 repIndex = current + 1 - offset_1;   /* offset_1 expected <= current +1 */
+        const BYTE* repBase = repIndex < dictLimit ? dictBase : base;
+        const BYTE* repMatch = repBase + repIndex;
+        size_t mLength;
+        hashSmall[hSmall] = hashLong[hLong] = current;   /* update hash table */
+
+        if ( (((U32)((dictLimit-1) - repIndex) >= 3) /* intentional underflow */ & (repIndex > lowestIndex))
+           && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
+            const BYTE* repMatchEnd = repIndex < dictLimit ? dictEnd : iend;
+            mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, lowPrefixPtr) + 4;
+            ip++;
+            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
+        } else {
+            if ((matchLongIndex > lowestIndex) && (MEM_read64(matchLong) == MEM_read64(ip))) {
+                const BYTE* matchEnd = matchLongIndex < dictLimit ? dictEnd : iend;
+                const BYTE* lowMatchPtr = matchLongIndex < dictLimit ? dictStart : lowPrefixPtr;
+                U32 offset;
+                mLength = ZSTD_count_2segments(ip+8, matchLong+8, iend, matchEnd, lowPrefixPtr) + 8;
+                offset = current - matchLongIndex;
+                while (((ip>anchor) & (matchLong>lowMatchPtr)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; }   /* catch up */
+                offset_2 = offset_1;
+                offset_1 = offset;
+                ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+
+            } else if ((matchIndex > lowestIndex) && (MEM_read32(match) == MEM_read32(ip))) {
+                size_t const h3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
+                U32 const matchIndex3 = hashLong[h3];
+                const BYTE* const match3Base = matchIndex3 < dictLimit ? dictBase : base;
+                const BYTE* match3 = match3Base + matchIndex3;
+                U32 offset;
+                hashLong[h3] = current + 1;
+                if ( (matchIndex3 > lowestIndex) && (MEM_read64(match3) == MEM_read64(ip+1)) ) {
+                    const BYTE* matchEnd = matchIndex3 < dictLimit ? dictEnd : iend;
+                    const BYTE* lowMatchPtr = matchIndex3 < dictLimit ? dictStart : lowPrefixPtr;
+                    mLength = ZSTD_count_2segments(ip+9, match3+8, iend, matchEnd, lowPrefixPtr) + 8;
+                    ip++;
+                    offset = current+1 - matchIndex3;
+                    while (((ip>anchor) & (match3>lowMatchPtr)) && (ip[-1] == match3[-1])) { ip--; match3--; mLength++; } /* catch up */
+                } else {
+                    const BYTE* matchEnd = matchIndex < dictLimit ? dictEnd : iend;
+                    const BYTE* lowMatchPtr = matchIndex < dictLimit ? dictStart : lowPrefixPtr;
+                    mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, lowPrefixPtr) + 4;
+                    offset = current - matchIndex;
+                    while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; }   /* catch up */
+                }
+                offset_2 = offset_1;
+                offset_1 = offset;
+                ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+
+            } else {
+                ip += ((ip-anchor) >> g_searchStrength) + 1;
+                continue;
+        }   }
+
+        /* found a match : store it */
+        ip += mLength;
+        anchor = ip;
+
+        if (ip <= ilimit) {
+            /* Fill Table */
+			hashSmall[ZSTD_hashPtr(base+current+2, hBitsS, mls)] = current+2;
+			hashLong[ZSTD_hashPtr(base+current+2, hBitsL, 8)] = current+2;
+            hashSmall[ZSTD_hashPtr(ip-2, hBitsS, mls)] = (U32)(ip-2-base);
+            hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base);
+            /* check immediate repcode */
+            while (ip <= ilimit) {
+                U32 const current2 = (U32)(ip-base);
+                U32 const repIndex2 = current2 - offset_2;
+                const BYTE* repMatch2 = repIndex2 < dictLimit ? dictBase + repIndex2 : base + repIndex2;
+                if ( (((U32)((dictLimit-1) - repIndex2) >= 3) & (repIndex2 > lowestIndex))  /* intentional overflow */
+                   && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
+                    const BYTE* const repEnd2 = repIndex2 < dictLimit ? dictEnd : iend;
+                    size_t const repLength2 = ZSTD_count_2segments(ip+EQUAL_READ32, repMatch2+EQUAL_READ32, iend, repEnd2, lowPrefixPtr) + EQUAL_READ32;
+                    U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset;   /* swap offset_2 <=> offset_1 */
+                    ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
+                    hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
+                    hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
+                    ip += repLength2;
+                    anchor = ip;
+                    continue;
+                }
+                break;
+    }   }   }
+
+    /* save reps for next block */
+    ctx->savedRep[0] = offset_1; ctx->savedRep[1] = offset_2;
+
+    /* Last Literals */
+    {   size_t const lastLLSize = iend - anchor;
+        memcpy(seqStorePtr->lit, anchor, lastLLSize);
+        seqStorePtr->lit += lastLLSize;
+    }
+}
+
+
+static void ZSTD_compressBlock_doubleFast_extDict(ZSTD_CCtx* ctx,
+                         const void* src, size_t srcSize)
+{
+    U32 const mls = ctx->params.cParams.searchLength;
+    switch(mls)
+    {
+    default:
+    case 4 :
+        ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 4); return;
+    case 5 :
+        ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 5); return;
+    case 6 :
+        ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 6); return;
+    case 7 :
+        ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 7); return;
+    }
+}
+
+
+/*-*************************************
+*  Binary Tree search
+***************************************/
+/** ZSTD_insertBt1() : add one or multiple positions to tree.
+*   ip : assumed <= iend-8 .
+*   @return : nb of positions added */
+static U32 ZSTD_insertBt1(ZSTD_CCtx* zc, const BYTE* const ip, const U32 mls, const BYTE* const iend, U32 nbCompares,
+                          U32 extDict)
+{
+    U32*   const hashTable = zc->hashTable;
+    U32    const hashLog = zc->params.cParams.hashLog;
+    size_t const h  = ZSTD_hashPtr(ip, hashLog, mls);
+    U32*   const bt = zc->chainTable;
+    U32    const btLog  = zc->params.cParams.chainLog - 1;
+    U32    const btMask = (1 << btLog) - 1;
+    U32 matchIndex = hashTable[h];
+    size_t commonLengthSmaller=0, commonLengthLarger=0;
+    const BYTE* const base = zc->base;
+    const BYTE* const dictBase = zc->dictBase;
+    const U32 dictLimit = zc->dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const BYTE* const prefixStart = base + dictLimit;
+    const BYTE* match;
+    const U32 current = (U32)(ip-base);
+    const U32 btLow = btMask >= current ? 0 : current - btMask;
+    U32* smallerPtr = bt + 2*(current&btMask);
+    U32* largerPtr  = smallerPtr + 1;
+    U32 dummy32;   /* to be nullified at the end */
+    U32 const windowLow = zc->lowLimit;
+    U32 matchEndIdx = current+8;
+    size_t bestLength = 8;
+#ifdef ZSTD_C_PREDICT
+    U32 predictedSmall = *(bt + 2*((current-1)&btMask) + 0);
+    U32 predictedLarge = *(bt + 2*((current-1)&btMask) + 1);
+    predictedSmall += (predictedSmall>0);
+    predictedLarge += (predictedLarge>0);
+#endif /* ZSTD_C_PREDICT */
+
+    hashTable[h] = current;   /* Update Hash Table */
+
+    while (nbCompares-- && (matchIndex > windowLow)) {
+        U32* const nextPtr = bt + 2*(matchIndex & btMask);
+        size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
+
+#ifdef ZSTD_C_PREDICT   /* note : can create issues when hlog small <= 11 */
+        const U32* predictPtr = bt + 2*((matchIndex-1) & btMask);   /* written this way, as bt is a roll buffer */
+        if (matchIndex == predictedSmall) {
+            /* no need to check length, result known */
+            *smallerPtr = matchIndex;
+            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
+            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+            predictedSmall = predictPtr[1] + (predictPtr[1]>0);
+            continue;
+        }
+        if (matchIndex == predictedLarge) {
+            *largerPtr = matchIndex;
+            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            largerPtr = nextPtr;
+            matchIndex = nextPtr[0];
+            predictedLarge = predictPtr[0] + (predictPtr[0]>0);
+            continue;
+        }
+#endif
+        if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
+            match = base + matchIndex;
+            if (match[matchLength] == ip[matchLength])
+                matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iend) +1;
+        } else {
+            match = dictBase + matchIndex;
+            matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
+            if (matchIndex+matchLength >= dictLimit)
+				match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
+        }
+
+        if (matchLength > bestLength) {
+            bestLength = matchLength;
+            if (matchLength > matchEndIdx - matchIndex)
+                matchEndIdx = matchIndex + (U32)matchLength;
+        }
+
+        if (ip+matchLength == iend)   /* equal : no way to know if inf or sup */
+            break;   /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt the tree */
+
+        if (match[matchLength] < ip[matchLength]) {  /* necessarily within correct buffer */
+            /* match is smaller than current */
+            *smallerPtr = matchIndex;             /* update smaller idx */
+            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
+            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
+            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+        } else {
+            /* match is larger than current */
+            *largerPtr = matchIndex;
+            commonLengthLarger = matchLength;
+            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            largerPtr = nextPtr;
+            matchIndex = nextPtr[0];
+    }   }
+
+    *smallerPtr = *largerPtr = 0;
+    if (bestLength > 384) return MIN(192, (U32)(bestLength - 384));   /* speed optimization */
+    if (matchEndIdx > current + 8) return matchEndIdx - current - 8;
+    return 1;
+}
+
+
+static size_t ZSTD_insertBtAndFindBestMatch (
+                        ZSTD_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iend,
+                        size_t* offsetPtr,
+                        U32 nbCompares, const U32 mls,
+                        U32 extDict)
+{
+    U32*   const hashTable = zc->hashTable;
+    U32    const hashLog = zc->params.cParams.hashLog;
+    size_t const h  = ZSTD_hashPtr(ip, hashLog, mls);
+    U32*   const bt = zc->chainTable;
+    U32    const btLog  = zc->params.cParams.chainLog - 1;
+    U32    const btMask = (1 << btLog) - 1;
+    U32 matchIndex  = hashTable[h];
+    size_t commonLengthSmaller=0, commonLengthLarger=0;
+    const BYTE* const base = zc->base;
+    const BYTE* const dictBase = zc->dictBase;
+    const U32 dictLimit = zc->dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const BYTE* const prefixStart = base + dictLimit;
+    const U32 current = (U32)(ip-base);
+    const U32 btLow = btMask >= current ? 0 : current - btMask;
+    const U32 windowLow = zc->lowLimit;
+    U32* smallerPtr = bt + 2*(current&btMask);
+    U32* largerPtr  = bt + 2*(current&btMask) + 1;
+    U32 matchEndIdx = current+8;
+    U32 dummy32;   /* to be nullified at the end */
+    size_t bestLength = 0;
+
+    hashTable[h] = current;   /* Update Hash Table */
+
+    while (nbCompares-- && (matchIndex > windowLow)) {
+        U32* const nextPtr = bt + 2*(matchIndex & btMask);
+        size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
+        const BYTE* match;
+
+        if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
+            match = base + matchIndex;
+            if (match[matchLength] == ip[matchLength])
+                matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iend) +1;
+        } else {
+            match = dictBase + matchIndex;
+            matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
+            if (matchIndex+matchLength >= dictLimit)
+				match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
+        }
+
+        if (matchLength > bestLength) {
+            if (matchLength > matchEndIdx - matchIndex)
+                matchEndIdx = matchIndex + (U32)matchLength;
+            if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
+                bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - matchIndex;
+            if (ip+matchLength == iend)   /* equal : no way to know if inf or sup */
+                break;   /* drop, to guarantee consistency (miss a little bit of compression) */
+        }
+
+        if (match[matchLength] < ip[matchLength]) {
+            /* match is smaller than current */
+            *smallerPtr = matchIndex;             /* update smaller idx */
+            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
+            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
+            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+        } else {
+            /* match is larger than current */
+            *largerPtr = matchIndex;
+            commonLengthLarger = matchLength;
+            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            largerPtr = nextPtr;
+            matchIndex = nextPtr[0];
+    }   }
+
+    *smallerPtr = *largerPtr = 0;
+
+    zc->nextToUpdate = (matchEndIdx > current + 8) ? matchEndIdx - 8 : current+1;
+    return bestLength;
+}
+
+
+static void ZSTD_updateTree(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls)
+{
+    const BYTE* const base = zc->base;
+    const U32 target = (U32)(ip - base);
+    U32 idx = zc->nextToUpdate;
+
+    while(idx < target)
+        idx += ZSTD_insertBt1(zc, base+idx, mls, iend, nbCompares, 0);
+}
+
+/** ZSTD_BtFindBestMatch() : Tree updater, providing best match */
+static size_t ZSTD_BtFindBestMatch (
+                        ZSTD_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 mls)
+{
+    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
+    ZSTD_updateTree(zc, ip, iLimit, maxNbAttempts, mls);
+    return ZSTD_insertBtAndFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, mls, 0);
+}
+
+
+static size_t ZSTD_BtFindBestMatch_selectMLS (
+                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        const BYTE* ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 matchLengthSearch)
+{
+    switch(matchLengthSearch)
+    {
+    default :
+    case 4 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4);
+    case 5 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5);
+    case 6 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6);
+    }
+}
+
+
+static void ZSTD_updateTree_extDict(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls)
+{
+    const BYTE* const base = zc->base;
+    const U32 target = (U32)(ip - base);
+    U32 idx = zc->nextToUpdate;
+
+    while (idx < target) idx += ZSTD_insertBt1(zc, base+idx, mls, iend, nbCompares, 1);
+}
+
+
+/** Tree updater, providing best match */
+static size_t ZSTD_BtFindBestMatch_extDict (
+                        ZSTD_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 mls)
+{
+    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
+    ZSTD_updateTree_extDict(zc, ip, iLimit, maxNbAttempts, mls);
+    return ZSTD_insertBtAndFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, mls, 1);
+}
+
+
+static size_t ZSTD_BtFindBestMatch_selectMLS_extDict (
+                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        const BYTE* ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 matchLengthSearch)
+{
+    switch(matchLengthSearch)
+    {
+    default :
+    case 4 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4);
+    case 5 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5);
+    case 6 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6);
+    }
+}
+
+
+
+/* *********************************
+*  Hash Chain
+***********************************/
+#define NEXT_IN_CHAIN(d, mask)   chainTable[(d) & mask]
+
+/* Update chains up to ip (excluded)
+   Assumption : always within prefix (ie. not within extDict) */
+FORCE_INLINE
+U32 ZSTD_insertAndFindFirstIndex (ZSTD_CCtx* zc, const BYTE* ip, U32 mls)
+{
+    U32* const hashTable  = zc->hashTable;
+    const U32 hashLog = zc->params.cParams.hashLog;
+    U32* const chainTable = zc->chainTable;
+    const U32 chainMask = (1 << zc->params.cParams.chainLog) - 1;
+    const BYTE* const base = zc->base;
+    const U32 target = (U32)(ip - base);
+    U32 idx = zc->nextToUpdate;
+
+    while(idx < target) { /* catch up */
+        size_t const h = ZSTD_hashPtr(base+idx, hashLog, mls);
+        NEXT_IN_CHAIN(idx, chainMask) = hashTable[h];
+        hashTable[h] = idx;
+        idx++;
+    }
+
+    zc->nextToUpdate = target;
+    return hashTable[ZSTD_hashPtr(ip, hashLog, mls)];
+}
+
+
+
+FORCE_INLINE /* inlining is important to hardwire a hot branch (template emulation) */
+size_t ZSTD_HcFindBestMatch_generic (
+                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        const BYTE* const ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 mls, const U32 extDict)
+{
+    U32* const chainTable = zc->chainTable;
+    const U32 chainSize = (1 << zc->params.cParams.chainLog);
+    const U32 chainMask = chainSize-1;
+    const BYTE* const base = zc->base;
+    const BYTE* const dictBase = zc->dictBase;
+    const U32 dictLimit = zc->dictLimit;
+    const BYTE* const prefixStart = base + dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const U32 lowLimit = zc->lowLimit;
+    const U32 current = (U32)(ip-base);
+    const U32 minChain = current > chainSize ? current - chainSize : 0;
+    int nbAttempts=maxNbAttempts;
+    size_t ml=EQUAL_READ32-1;
+
+    /* HC4 match finder */
+    U32 matchIndex = ZSTD_insertAndFindFirstIndex (zc, ip, mls);
+
+    for ( ; (matchIndex>lowLimit) & (nbAttempts>0) ; nbAttempts--) {
+        const BYTE* match;
+        size_t currentMl=0;
+        if ((!extDict) || matchIndex >= dictLimit) {
+            match = base + matchIndex;
+            if (match[ml] == ip[ml])   /* potentially better */
+                currentMl = ZSTD_count(ip, match, iLimit);
+        } else {
+            match = dictBase + matchIndex;
+            if (MEM_read32(match) == MEM_read32(ip))   /* assumption : matchIndex <= dictLimit-4 (by table construction) */
+                currentMl = ZSTD_count_2segments(ip+EQUAL_READ32, match+EQUAL_READ32, iLimit, dictEnd, prefixStart) + EQUAL_READ32;
+        }
+
+        /* save best solution */
+        if (currentMl > ml) { ml = currentMl; *offsetPtr = current - matchIndex + ZSTD_REP_MOVE; if (ip+currentMl == iLimit) break; /* best possible, and avoid read overflow*/ }
+
+        if (matchIndex <= minChain) break;
+        matchIndex = NEXT_IN_CHAIN(matchIndex, chainMask);
+    }
+
+    return ml;
+}
+
+
+FORCE_INLINE size_t ZSTD_HcFindBestMatch_selectMLS (
+                        ZSTD_CCtx* zc,
+                        const BYTE* ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 matchLengthSearch)
+{
+    switch(matchLengthSearch)
+    {
+    default :
+    case 4 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4, 0);
+    case 5 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5, 0);
+    case 6 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6, 0);
+    }
+}
+
+
+FORCE_INLINE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
+                        ZSTD_CCtx* zc,
+                        const BYTE* ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 matchLengthSearch)
+{
+    switch(matchLengthSearch)
+    {
+    default :
+    case 4 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4, 1);
+    case 5 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5, 1);
+    case 6 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6, 1);
+    }
+}
+
+
+/* *******************************
+*  Common parser - lazy strategy
+*********************************/
+FORCE_INLINE
+void ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
+                                     const void* src, size_t srcSize,
+                                     const U32 searchMethod, const U32 depth)
+{
+    seqStore_t* seqStorePtr = &(ctx->seqStore);
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - 8;
+    const BYTE* const base = ctx->base + ctx->dictLimit;
+
+    U32 const maxSearches = 1 << ctx->params.cParams.searchLog;
+    U32 const mls = ctx->params.cParams.searchLength;
+
+    typedef size_t (*searchMax_f)(ZSTD_CCtx* zc, const BYTE* ip, const BYTE* iLimit,
+                        size_t* offsetPtr,
+                        U32 maxNbAttempts, U32 matchLengthSearch);
+    searchMax_f const searchMax = searchMethod ? ZSTD_BtFindBestMatch_selectMLS : ZSTD_HcFindBestMatch_selectMLS;
+    U32 offset_1 = ctx->rep[0], offset_2 = ctx->rep[1], savedOffset=0;
+
+    /* init */
+    ip += (ip==base);
+    ctx->nextToUpdate3 = ctx->nextToUpdate;
+    {   U32 const maxRep = (U32)(ip-base);
+        if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0;
+        if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0;
+    }
+
+    /* Match Loop */
+    while (ip < ilimit) {
+        size_t matchLength=0;
+        size_t offset=0;
+        const BYTE* start=ip+1;
+
+        /* check repCode */
+        if ((offset_1>0) & (MEM_read32(ip+1) == MEM_read32(ip+1 - offset_1))) {
+            /* repcode : we take it */
+            matchLength = ZSTD_count(ip+1+EQUAL_READ32, ip+1+EQUAL_READ32-offset_1, iend) + EQUAL_READ32;
+            if (depth==0) goto _storeSequence;
+        }
+
+        /* first search (depth 0) */
+        {   size_t offsetFound = 99999999;
+            size_t const ml2 = searchMax(ctx, ip, iend, &offsetFound, maxSearches, mls);
+            if (ml2 > matchLength)
+                matchLength = ml2, start = ip, offset=offsetFound;
+        }
+
+        if (matchLength < EQUAL_READ32) {
+            ip += ((ip-anchor) >> g_searchStrength) + 1;   /* jump faster over incompressible sections */
+            continue;
+        }
+
+        /* let's try to find a better solution */
+        if (depth>=1)
+        while (ip0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
+                size_t const mlRep = ZSTD_count(ip+EQUAL_READ32, ip+EQUAL_READ32-offset_1, iend) + EQUAL_READ32;
+                int const gain2 = (int)(mlRep * 3);
+                int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
+                if ((mlRep >= EQUAL_READ32) && (gain2 > gain1))
+                    matchLength = mlRep, offset = 0, start = ip;
+            }
+            {   size_t offset2=99999999;
+                size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
+                int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
+                int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
+                if ((ml2 >= EQUAL_READ32) && (gain2 > gain1)) {
+                    matchLength = ml2, offset = offset2, start = ip;
+                    continue;   /* search a better one */
+            }   }
+
+            /* let's find an even better one */
+            if ((depth==2) && (ip0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
+                    size_t const ml2 = ZSTD_count(ip+EQUAL_READ32, ip+EQUAL_READ32-offset_1, iend) + EQUAL_READ32;
+                    int const gain2 = (int)(ml2 * 4);
+                    int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
+                    if ((ml2 >= EQUAL_READ32) && (gain2 > gain1))
+                        matchLength = ml2, offset = 0, start = ip;
+                }
+                {   size_t offset2=99999999;
+                    size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
+                    int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
+                    int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
+                    if ((ml2 >= EQUAL_READ32) && (gain2 > gain1)) {
+                        matchLength = ml2, offset = offset2, start = ip;
+                        continue;
+            }   }   }
+            break;  /* nothing found : store previous solution */
+        }
+
+        /* catch up */
+        if (offset) {
+            while ((start>anchor) && (start>base+offset-ZSTD_REP_MOVE) && (start[-1] == start[-1-offset+ZSTD_REP_MOVE]))   /* only search for offset within prefix */
+                { start--; matchLength++; }
+            offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
+        }
+
+        /* store sequence */
+_storeSequence:
+        {   size_t const litLength = start - anchor;
+            ZSTD_storeSeq(seqStorePtr, litLength, anchor, (U32)offset, matchLength-MINMATCH);
+            anchor = ip = start + matchLength;
+        }
+
+        /* check immediate repcode */
+        while ( (ip <= ilimit)
+             && ((offset_2>0)
+             & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
+            /* store sequence */
+            matchLength = ZSTD_count(ip+EQUAL_READ32, ip+EQUAL_READ32-offset_2, iend) + EQUAL_READ32;
+            offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */
+            ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, matchLength-MINMATCH);
+            ip += matchLength;
+            anchor = ip;
+            continue;   /* faster when present ... (?) */
+    }   }
+
+    /* Save reps for next block */
+    ctx->savedRep[0] = offset_1 ? offset_1 : savedOffset;
+    ctx->savedRep[1] = offset_2 ? offset_2 : savedOffset;
+
+    /* Last Literals */
+    {   size_t const lastLLSize = iend - anchor;
+        memcpy(seqStorePtr->lit, anchor, lastLLSize);
+        seqStorePtr->lit += lastLLSize;
+    }
+}
+
+
+static void ZSTD_compressBlock_btlazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 1, 2);
+}
+
+static void ZSTD_compressBlock_lazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 2);
+}
+
+static void ZSTD_compressBlock_lazy(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 1);
+}
+
+static void ZSTD_compressBlock_greedy(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 0);
+}
+
+
+FORCE_INLINE
+void ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
+                                     const void* src, size_t srcSize,
+                                     const U32 searchMethod, const U32 depth)
+{
+    seqStore_t* seqStorePtr = &(ctx->seqStore);
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - 8;
+    const BYTE* const base = ctx->base;
+    const U32 dictLimit = ctx->dictLimit;
+    const U32 lowestIndex = ctx->lowLimit;
+    const BYTE* const prefixStart = base + dictLimit;
+    const BYTE* const dictBase = ctx->dictBase;
+    const BYTE* const dictEnd  = dictBase + dictLimit;
+    const BYTE* const dictStart  = dictBase + ctx->lowLimit;
+
+    const U32 maxSearches = 1 << ctx->params.cParams.searchLog;
+    const U32 mls = ctx->params.cParams.searchLength;
+
+    typedef size_t (*searchMax_f)(ZSTD_CCtx* zc, const BYTE* ip, const BYTE* iLimit,
+                        size_t* offsetPtr,
+                        U32 maxNbAttempts, U32 matchLengthSearch);
+    searchMax_f searchMax = searchMethod ? ZSTD_BtFindBestMatch_selectMLS_extDict : ZSTD_HcFindBestMatch_extDict_selectMLS;
+
+    U32 offset_1 = ctx->rep[0], offset_2 = ctx->rep[1];
+
+    /* init */
+    ctx->nextToUpdate3 = ctx->nextToUpdate;
+    ip += (ip == prefixStart);
+
+    /* Match Loop */
+    while (ip < ilimit) {
+        size_t matchLength=0;
+        size_t offset=0;
+        const BYTE* start=ip+1;
+        U32 current = (U32)(ip-base);
+
+        /* check repCode */
+        {   const U32 repIndex = (U32)(current+1 - offset_1);
+            const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+            const BYTE* const repMatch = repBase + repIndex;
+            if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))   /* intentional overflow */
+            if (MEM_read32(ip+1) == MEM_read32(repMatch)) {
+                /* repcode detected we should take it */
+                const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+                matchLength = ZSTD_count_2segments(ip+1+EQUAL_READ32, repMatch+EQUAL_READ32, iend, repEnd, prefixStart) + EQUAL_READ32;
+                if (depth==0) goto _storeSequence;
+        }   }
+
+        /* first search (depth 0) */
+        {   size_t offsetFound = 99999999;
+            size_t const ml2 = searchMax(ctx, ip, iend, &offsetFound, maxSearches, mls);
+            if (ml2 > matchLength)
+                matchLength = ml2, start = ip, offset=offsetFound;
+        }
+
+         if (matchLength < EQUAL_READ32) {
+            ip += ((ip-anchor) >> g_searchStrength) + 1;   /* jump faster over incompressible sections */
+            continue;
+        }
+
+        /* let's try to find a better solution */
+        if (depth>=1)
+        while (ip= 3) & (repIndex > lowestIndex))  /* intentional overflow */
+                if (MEM_read32(ip) == MEM_read32(repMatch)) {
+                    /* repcode detected */
+                    const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+                    size_t const repLength = ZSTD_count_2segments(ip+EQUAL_READ32, repMatch+EQUAL_READ32, iend, repEnd, prefixStart) + EQUAL_READ32;
+                    int const gain2 = (int)(repLength * 3);
+                    int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
+                    if ((repLength >= EQUAL_READ32) && (gain2 > gain1))
+                        matchLength = repLength, offset = 0, start = ip;
+            }   }
+
+            /* search match, depth 1 */
+            {   size_t offset2=99999999;
+                size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
+                int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
+                int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
+                if ((ml2 >= EQUAL_READ32) && (gain2 > gain1)) {
+                    matchLength = ml2, offset = offset2, start = ip;
+                    continue;   /* search a better one */
+            }   }
+
+            /* let's find an even better one */
+            if ((depth==2) && (ip= 3) & (repIndex > lowestIndex))  /* intentional overflow */
+                    if (MEM_read32(ip) == MEM_read32(repMatch)) {
+                        /* repcode detected */
+                        const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+                        size_t repLength = ZSTD_count_2segments(ip+EQUAL_READ32, repMatch+EQUAL_READ32, iend, repEnd, prefixStart) + EQUAL_READ32;
+                        int gain2 = (int)(repLength * 4);
+                        int gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
+                        if ((repLength >= EQUAL_READ32) && (gain2 > gain1))
+                            matchLength = repLength, offset = 0, start = ip;
+                }   }
+
+                /* search match, depth 2 */
+                {   size_t offset2=99999999;
+                    size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
+                    int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
+                    int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
+                    if ((ml2 >= EQUAL_READ32) && (gain2 > gain1)) {
+                        matchLength = ml2, offset = offset2, start = ip;
+                        continue;
+            }   }   }
+            break;  /* nothing found : store previous solution */
+        }
+
+        /* catch up */
+        if (offset) {
+            U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE));
+            const BYTE* match = (matchIndex < dictLimit) ? dictBase + matchIndex : base + matchIndex;
+            const BYTE* const mStart = (matchIndex < dictLimit) ? dictStart : prefixStart;
+            while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; }  /* catch up */
+            offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
+        }
+
+        /* store sequence */
+_storeSequence:
+        {   size_t const litLength = start - anchor;
+            ZSTD_storeSeq(seqStorePtr, litLength, anchor, (U32)offset, matchLength-MINMATCH);
+            anchor = ip = start + matchLength;
+        }
+
+        /* check immediate repcode */
+        while (ip <= ilimit) {
+            const U32 repIndex = (U32)((ip-base) - offset_2);
+            const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+            const BYTE* const repMatch = repBase + repIndex;
+            if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))  /* intentional overflow */
+            if (MEM_read32(ip) == MEM_read32(repMatch)) {
+                /* repcode detected we should take it */
+                const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+                matchLength = ZSTD_count_2segments(ip+EQUAL_READ32, repMatch+EQUAL_READ32, iend, repEnd, prefixStart) + EQUAL_READ32;
+                offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset;   /* swap offset history */
+                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, matchLength-MINMATCH);
+                ip += matchLength;
+                anchor = ip;
+                continue;   /* faster when present ... (?) */
+            }
+            break;
+    }   }
+
+    /* Save reps for next block */
+    ctx->savedRep[0] = offset_1; ctx->savedRep[1] = offset_2;
+
+    /* Last Literals */
+    {   size_t const lastLLSize = iend - anchor;
+        memcpy(seqStorePtr->lit, anchor, lastLLSize);
+        seqStorePtr->lit += lastLLSize;
+    }
+}
+
+
+void ZSTD_compressBlock_greedy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 0);
+}
+
+static void ZSTD_compressBlock_lazy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 1);
+}
+
+static void ZSTD_compressBlock_lazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 2);
+}
+
+static void ZSTD_compressBlock_btlazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 1, 2);
+}
+
+
+/* The optimal parser */
+#include "zstd_opt.h"
+
+static void ZSTD_compressBlock_btopt(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+#ifdef ZSTD_OPT_H_91842398743
+    ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 0);
+#else
+    (void)ctx; (void)src; (void)srcSize;
+    return;
+#endif
+}
+
+static void ZSTD_compressBlock_btopt2(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+#ifdef ZSTD_OPT_H_91842398743
+    ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 1);
+#else
+    (void)ctx; (void)src; (void)srcSize;
+    return;
+#endif
+}
+
+static void ZSTD_compressBlock_btopt_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+#ifdef ZSTD_OPT_H_91842398743
+    ZSTD_compressBlock_opt_extDict_generic(ctx, src, srcSize, 0);
+#else
+    (void)ctx; (void)src; (void)srcSize;
+    return;
+#endif
+}
+
+static void ZSTD_compressBlock_btopt2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+#ifdef ZSTD_OPT_H_91842398743
+    ZSTD_compressBlock_opt_extDict_generic(ctx, src, srcSize, 1);
+#else
+    (void)ctx; (void)src; (void)srcSize;
+    return;
+#endif
+}
+
+
+typedef void (*ZSTD_blockCompressor) (ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+
+static ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, int extDict)
+{
+    static const ZSTD_blockCompressor blockCompressor[2][8] = {
+        { ZSTD_compressBlock_fast, ZSTD_compressBlock_doubleFast, ZSTD_compressBlock_greedy, ZSTD_compressBlock_lazy, ZSTD_compressBlock_lazy2, ZSTD_compressBlock_btlazy2, ZSTD_compressBlock_btopt, ZSTD_compressBlock_btopt2 },
+        { ZSTD_compressBlock_fast_extDict, ZSTD_compressBlock_doubleFast_extDict, ZSTD_compressBlock_greedy_extDict, ZSTD_compressBlock_lazy_extDict,ZSTD_compressBlock_lazy2_extDict, ZSTD_compressBlock_btlazy2_extDict, ZSTD_compressBlock_btopt_extDict, ZSTD_compressBlock_btopt2_extDict }
+    };
+
+    return blockCompressor[extDict][(U32)strat];
+}
+
+
+static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+    ZSTD_blockCompressor const blockCompressor = ZSTD_selectBlockCompressor(zc->params.cParams.strategy, zc->lowLimit < zc->dictLimit);
+    const BYTE* const base = zc->base;
+    const BYTE* const istart = (const BYTE*)src;
+    const U32 current = (U32)(istart-base);
+    if (srcSize < MIN_CBLOCK_SIZE+ZSTD_blockHeaderSize+1) return 0;   /* don't even attempt compression below a certain srcSize */
+    ZSTD_resetSeqStore(&(zc->seqStore));
+    if (current > zc->nextToUpdate + 384)
+        zc->nextToUpdate = current - MIN(192, (U32)(current - zc->nextToUpdate - 384));   /* update tree not updated after finding very long rep matches */
+    blockCompressor(zc, src, srcSize);
+    return ZSTD_compressSequences(zc, dst, dstCapacity, srcSize);
+}
+
+
+/*! ZSTD_compress_generic() :
+*   Compress a chunk of data into one or multiple blocks.
+*   All blocks will be terminated, all input will be consumed.
+*   Function will issue an error if there is not enough `dstCapacity` to hold the compressed content.
+*   Frame is supposed already started (header already produced)
+*   @return : compressed size, or an error code
+*/
+static size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
+                                     void* dst, size_t dstCapacity,
+                               const void* src, size_t srcSize,
+                                     U32 lastFrameChunk)
+{
+    size_t blockSize = cctx->blockSize;
+    size_t remaining = srcSize;
+    const BYTE* ip = (const BYTE*)src;
+    BYTE* const ostart = (BYTE*)dst;
+    BYTE* op = ostart;
+    U32 const maxDist = 1 << cctx->params.cParams.windowLog;
+
+    if (cctx->params.fParams.checksumFlag && srcSize)
+        XXH64_update(&cctx->xxhState, src, srcSize);
+
+    while (remaining) {
+        U32 const lastBlock = lastFrameChunk & (blockSize >= remaining);
+        size_t cSize;
+
+        if (dstCapacity < ZSTD_blockHeaderSize + MIN_CBLOCK_SIZE) return ERROR(dstSize_tooSmall);   /* not enough space to store compressed block */
+        if (remaining < blockSize) blockSize = remaining;
+
+        /* preemptive overflow correction */
+        if (cctx->lowLimit > (2U<<30)) {
+            U32 const cycleMask = (1 << ZSTD_cycleLog(cctx->params.cParams.hashLog, cctx->params.cParams.strategy)) - 1;
+            U32 const current = (U32)(ip - cctx->base);
+            U32 const newCurrent = (current & cycleMask) + (1 << cctx->params.cParams.windowLog);
+            U32 const correction = current - newCurrent;
+            ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX_64 <= 30);
+            ZSTD_reduceIndex(cctx, correction);
+            cctx->base += correction;
+            cctx->dictBase += correction;
+            cctx->lowLimit -= correction;
+            cctx->dictLimit -= correction;
+            if (cctx->nextToUpdate < correction) cctx->nextToUpdate = 0;
+            else cctx->nextToUpdate -= correction;
+        }
+
+        if ((U32)(ip+blockSize - cctx->base) > cctx->loadedDictEnd + maxDist) {
+            /* enforce maxDist */
+            U32 const newLowLimit = (U32)(ip+blockSize - cctx->base) - maxDist;
+            if (cctx->lowLimit < newLowLimit) cctx->lowLimit = newLowLimit;
+            if (cctx->dictLimit < cctx->lowLimit) cctx->dictLimit = cctx->lowLimit;
+        }
+
+        cSize = ZSTD_compressBlock_internal(cctx, op+ZSTD_blockHeaderSize, dstCapacity-ZSTD_blockHeaderSize, ip, blockSize);
+        if (ZSTD_isError(cSize)) return cSize;
+
+        if (cSize == 0) {  /* block is not compressible */
+            U32 const cBlockHeader24 = lastBlock + (((U32)bt_raw)<<1) + (U32)(blockSize << 3);
+            if (blockSize + ZSTD_blockHeaderSize > dstCapacity) return ERROR(dstSize_tooSmall);
+            MEM_writeLE32(op, cBlockHeader24);   /* no pb, 4th byte will be overwritten */
+            memcpy(op + ZSTD_blockHeaderSize, ip, blockSize);
+            cSize = ZSTD_blockHeaderSize+blockSize;
+        } else {
+            U32 const cBlockHeader24 = lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3);
+            MEM_writeLE24(op, cBlockHeader24);
+            cSize += ZSTD_blockHeaderSize;
+        }
+
+        remaining -= blockSize;
+        dstCapacity -= cSize;
+        ip += blockSize;
+        op += cSize;
+    }
+
+    if (lastFrameChunk && (op>ostart)) cctx->stage = ZSTDcs_ending;
+    return op-ostart;
+}
+
+
+static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity,
+                                    ZSTD_parameters params, U64 pledgedSrcSize, U32 dictID)
+{   BYTE* const op = (BYTE*)dst;
+    U32   const dictIDSizeCode = (dictID>0) + (dictID>=256) + (dictID>=65536);   /* 0-3 */
+    U32   const checksumFlag = params.fParams.checksumFlag>0;
+    U32   const windowSize = 1U << params.cParams.windowLog;
+    U32   const singleSegment = params.fParams.contentSizeFlag && (windowSize > (pledgedSrcSize-1));
+    BYTE  const windowLogByte = (BYTE)((params.cParams.windowLog - ZSTD_WINDOWLOG_ABSOLUTEMIN) << 3);
+    U32   const fcsCode = params.fParams.contentSizeFlag ?
+                     (pledgedSrcSize>=256) + (pledgedSrcSize>=65536+256) + (pledgedSrcSize>=0xFFFFFFFFU) :   /* 0-3 */
+                      0;
+    BYTE  const frameHeaderDecriptionByte = (BYTE)(dictIDSizeCode + (checksumFlag<<2) + (singleSegment<<5) + (fcsCode<<6) );
+    size_t pos;
+
+    if (dstCapacity < ZSTD_frameHeaderSize_max) return ERROR(dstSize_tooSmall);
+
+    MEM_writeLE32(dst, ZSTD_MAGICNUMBER);
+    op[4] = frameHeaderDecriptionByte; pos=5;
+    if (!singleSegment) op[pos++] = windowLogByte;
+    switch(dictIDSizeCode)
+    {
+        default:   /* impossible */
+        case 0 : break;
+        case 1 : op[pos] = (BYTE)(dictID); pos++; break;
+        case 2 : MEM_writeLE16(op+pos, (U16)dictID); pos+=2; break;
+        case 3 : MEM_writeLE32(op+pos, dictID); pos+=4; break;
+    }
+    switch(fcsCode)
+    {
+        default:   /* impossible */
+        case 0 : if (singleSegment) op[pos++] = (BYTE)(pledgedSrcSize); break;
+        case 1 : MEM_writeLE16(op+pos, (U16)(pledgedSrcSize-256)); pos+=2; break;
+        case 2 : MEM_writeLE32(op+pos, (U32)(pledgedSrcSize)); pos+=4; break;
+        case 3 : MEM_writeLE64(op+pos, (U64)(pledgedSrcSize)); pos+=8; break;
+    }
+    return pos;
+}
+
+
+static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,
+                              void* dst, size_t dstCapacity,
+                        const void* src, size_t srcSize,
+                               U32 frame, U32 lastFrameChunk)
+{
+    const BYTE* const ip = (const BYTE*) src;
+    size_t fhSize = 0;
+
+    if (cctx->stage==ZSTDcs_created) return ERROR(stage_wrong);   /* missing init (ZSTD_compressBegin) */
+
+    if (frame && (cctx->stage==ZSTDcs_init)) {
+        fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, cctx->params, cctx->frameContentSize, cctx->dictID);
+        if (ZSTD_isError(fhSize)) return fhSize;
+        dstCapacity -= fhSize;
+        dst = (char*)dst + fhSize;
+        cctx->stage = ZSTDcs_ongoing;
+    }
+
+    /* Check if blocks follow each other */
+    if (src != cctx->nextSrc) {
+        /* not contiguous */
+        ptrdiff_t const delta = cctx->nextSrc - ip;
+        cctx->lowLimit = cctx->dictLimit;
+        cctx->dictLimit = (U32)(cctx->nextSrc - cctx->base);
+        cctx->dictBase = cctx->base;
+        cctx->base -= delta;
+        cctx->nextToUpdate = cctx->dictLimit;
+        if (cctx->dictLimit - cctx->lowLimit < HASH_READ_SIZE) cctx->lowLimit = cctx->dictLimit;   /* too small extDict */
+    }
+
+    /* if input and dictionary overlap : reduce dictionary (area presumed modified by input) */
+    if ((ip+srcSize > cctx->dictBase + cctx->lowLimit) & (ip < cctx->dictBase + cctx->dictLimit)) {
+        ptrdiff_t const highInputIdx = (ip + srcSize) - cctx->dictBase;
+        U32 const lowLimitMax = (highInputIdx > (ptrdiff_t)cctx->dictLimit) ? cctx->dictLimit : (U32)highInputIdx;
+        cctx->lowLimit = lowLimitMax;
+    }
+
+    cctx->nextSrc = ip + srcSize;
+
+    {   size_t const cSize = frame ?
+                             ZSTD_compress_generic (cctx, dst, dstCapacity, src, srcSize, lastFrameChunk) :
+                             ZSTD_compressBlock_internal (cctx, dst, dstCapacity, src, srcSize);
+        if (ZSTD_isError(cSize)) return cSize;
+        return cSize + fhSize;
+    }
+}
+
+
+size_t ZSTD_compressContinue (ZSTD_CCtx* cctx,
+                              void* dst, size_t dstCapacity,
+                        const void* src, size_t srcSize)
+{
+    return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 1, 0);
+}
+
+
+size_t ZSTD_getBlockSizeMax(ZSTD_CCtx* cctx)
+{
+    return MIN (ZSTD_BLOCKSIZE_ABSOLUTEMAX, 1 << cctx->params.cParams.windowLog);
+}
+
+size_t ZSTD_compressBlock(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+    size_t const blockSizeMax = ZSTD_getBlockSizeMax(cctx);
+    if (srcSize > blockSizeMax) return ERROR(srcSize_wrong);
+    return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 0, 0);
+}
+
+
+static size_t ZSTD_loadDictionaryContent(ZSTD_CCtx* zc, const void* src, size_t srcSize)
+{
+    const BYTE* const ip = (const BYTE*) src;
+    const BYTE* const iend = ip + srcSize;
+
+    /* input becomes current prefix */
+    zc->lowLimit = zc->dictLimit;
+    zc->dictLimit = (U32)(zc->nextSrc - zc->base);
+    zc->dictBase = zc->base;
+    zc->base += ip - zc->nextSrc;
+    zc->nextToUpdate = zc->dictLimit;
+    zc->loadedDictEnd = (U32)(iend - zc->base);
+
+    zc->nextSrc = iend;
+    if (srcSize <= HASH_READ_SIZE) return 0;
+
+    switch(zc->params.cParams.strategy)
+    {
+    case ZSTD_fast:
+        ZSTD_fillHashTable (zc, iend, zc->params.cParams.searchLength);
+        break;
+
+    case ZSTD_dfast:
+        ZSTD_fillDoubleHashTable (zc, iend, zc->params.cParams.searchLength);
+        break;
+
+    case ZSTD_greedy:
+    case ZSTD_lazy:
+    case ZSTD_lazy2:
+        ZSTD_insertAndFindFirstIndex (zc, iend-HASH_READ_SIZE, zc->params.cParams.searchLength);
+        break;
+
+    case ZSTD_btlazy2:
+    case ZSTD_btopt:
+    case ZSTD_btopt2:
+        ZSTD_updateTree(zc, iend-HASH_READ_SIZE, iend, 1 << zc->params.cParams.searchLog, zc->params.cParams.searchLength);
+        break;
+
+    default:
+        return ERROR(GENERIC);   /* strategy doesn't exist; impossible */
+    }
+
+    zc->nextToUpdate = zc->loadedDictEnd;
+    return 0;
+}
+
+
+/* Dictionaries that assign zero probability to symbols that show up causes problems
+   when FSE encoding.  Refuse dictionaries that assign zero probability to symbols
+   that we may encounter during compression.
+   NOTE: This behavior is not standard and could be improved in the future. */
+static size_t ZSTD_checkDictNCount(short* normalizedCounter, unsigned dictMaxSymbolValue, unsigned maxSymbolValue) {
+    U32 s;
+    if (dictMaxSymbolValue < maxSymbolValue) return ERROR(dictionary_corrupted);
+    for (s = 0; s <= maxSymbolValue; ++s) {
+        if (normalizedCounter[s] == 0) return ERROR(dictionary_corrupted);
+    }
+    return 0;
+}
+
+
+/* Dictionary format :
+    Magic == ZSTD_DICT_MAGIC (4 bytes)
+    HUF_writeCTable(256)
+    FSE_writeNCount(off)
+    FSE_writeNCount(ml)
+    FSE_writeNCount(ll)
+    RepOffsets
+    Dictionary content
+*/
+/*! ZSTD_loadDictEntropyStats() :
+    @return : size read from dictionary
+    note : magic number supposed already checked */
+static size_t ZSTD_loadDictEntropyStats(ZSTD_CCtx* cctx, const void* dict, size_t dictSize)
+{
+    const BYTE* dictPtr = (const BYTE*)dict;
+    const BYTE* const dictEnd = dictPtr + dictSize;
+    short offcodeNCount[MaxOff+1];
+    unsigned offcodeMaxValue = MaxOff;
+    BYTE scratchBuffer[1<hufTable, 255, dict, dictSize);
+        if (HUF_isError(hufHeaderSize)) return ERROR(dictionary_corrupted);
+        dictPtr += hufHeaderSize;
+    }
+
+    {   unsigned offcodeLog;
+        size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr);
+        if (FSE_isError(offcodeHeaderSize)) return ERROR(dictionary_corrupted);
+        if (offcodeLog > OffFSELog) return ERROR(dictionary_corrupted);
+        /* Defer checking offcodeMaxValue because we need to know the size of the dictionary content */
+        CHECK_E (FSE_buildCTable_wksp(cctx->offcodeCTable, offcodeNCount, offcodeMaxValue, offcodeLog, scratchBuffer, sizeof(scratchBuffer)), dictionary_corrupted);
+        dictPtr += offcodeHeaderSize;
+    }
+
+    {   short matchlengthNCount[MaxML+1];
+        unsigned matchlengthMaxValue = MaxML, matchlengthLog;
+        size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, dictEnd-dictPtr);
+        if (FSE_isError(matchlengthHeaderSize)) return ERROR(dictionary_corrupted);
+        if (matchlengthLog > MLFSELog) return ERROR(dictionary_corrupted);
+        /* Every match length code must have non-zero probability */
+        CHECK_F (ZSTD_checkDictNCount(matchlengthNCount, matchlengthMaxValue, MaxML));
+        CHECK_E (FSE_buildCTable_wksp(cctx->matchlengthCTable, matchlengthNCount, matchlengthMaxValue, matchlengthLog, scratchBuffer, sizeof(scratchBuffer)), dictionary_corrupted);
+        dictPtr += matchlengthHeaderSize;
+    }
+
+    {   short litlengthNCount[MaxLL+1];
+        unsigned litlengthMaxValue = MaxLL, litlengthLog;
+        size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, dictEnd-dictPtr);
+        if (FSE_isError(litlengthHeaderSize)) return ERROR(dictionary_corrupted);
+        if (litlengthLog > LLFSELog) return ERROR(dictionary_corrupted);
+        /* Every literal length code must have non-zero probability */
+        CHECK_F (ZSTD_checkDictNCount(litlengthNCount, litlengthMaxValue, MaxLL));
+        CHECK_E(FSE_buildCTable_wksp(cctx->litlengthCTable, litlengthNCount, litlengthMaxValue, litlengthLog, scratchBuffer, sizeof(scratchBuffer)), dictionary_corrupted);
+        dictPtr += litlengthHeaderSize;
+    }
+
+    if (dictPtr+12 > dictEnd) return ERROR(dictionary_corrupted);
+    cctx->rep[0] = MEM_readLE32(dictPtr+0); if (cctx->rep[0] >= dictSize) return ERROR(dictionary_corrupted);
+    cctx->rep[1] = MEM_readLE32(dictPtr+4); if (cctx->rep[1] >= dictSize) return ERROR(dictionary_corrupted);
+    cctx->rep[2] = MEM_readLE32(dictPtr+8); if (cctx->rep[2] >= dictSize) return ERROR(dictionary_corrupted);
+    dictPtr += 12;
+
+    {   U32 offcodeMax = MaxOff;
+        if ((size_t)(dictEnd - dictPtr) <= ((U32)-1) - 128 KB) {
+            U32 const maxOffset = (U32)(dictEnd - dictPtr) + 128 KB; /* The maximum offset that must be supported */
+            /* Calculate minimum offset code required to represent maxOffset */
+            offcodeMax = ZSTD_highbit32(maxOffset);
+        }
+        /* Every possible supported offset <= dictContentSize + 128 KB must be representable */
+        CHECK_F (ZSTD_checkDictNCount(offcodeNCount, offcodeMaxValue, MIN(offcodeMax, MaxOff)));
+    }
+
+    cctx->flagStaticTables = 1;
+    return dictPtr - (const BYTE*)dict;
+}
+
+/** ZSTD_compress_insertDictionary() :
+*   @return : 0, or an error code */
+static size_t ZSTD_compress_insertDictionary(ZSTD_CCtx* zc, const void* dict, size_t dictSize)
+{
+    if ((dict==NULL) || (dictSize<=8)) return 0;
+
+    /* default : dict is pure content */
+    if (MEM_readLE32(dict) != ZSTD_DICT_MAGIC) return ZSTD_loadDictionaryContent(zc, dict, dictSize);
+    zc->dictID = zc->params.fParams.noDictIDFlag ? 0 :  MEM_readLE32((const char*)dict+4);
+
+    /* known magic number : dict is parsed for entropy stats and content */
+    {   size_t const loadError = ZSTD_loadDictEntropyStats(zc, (const char*)dict+8 /* skip dictHeader */, dictSize-8);
+        size_t const eSize = loadError + 8;
+        if (ZSTD_isError(loadError)) return loadError;
+        return ZSTD_loadDictionaryContent(zc, (const char*)dict+eSize, dictSize-eSize);
+    }
+}
+
+
+/*! ZSTD_compressBegin_internal() :
+*   @return : 0, or an error code */
+static size_t ZSTD_compressBegin_internal(ZSTD_CCtx* cctx,
+                             const void* dict, size_t dictSize,
+                                   ZSTD_parameters params, U64 pledgedSrcSize)
+{
+    ZSTD_compResetPolicy_e const crp = dictSize ? ZSTDcrp_fullReset : ZSTDcrp_continue;
+    CHECK_F(ZSTD_resetCCtx_advanced(cctx, params, pledgedSrcSize, crp));
+    return ZSTD_compress_insertDictionary(cctx, dict, dictSize);
+}
+
+
+/*! ZSTD_compressBegin_advanced() :
+*   @return : 0, or an error code */
+size_t ZSTD_compressBegin_advanced(ZSTD_CCtx* cctx,
+                             const void* dict, size_t dictSize,
+                                   ZSTD_parameters params, unsigned long long pledgedSrcSize)
+{
+    /* compression parameters verification and optimization */
+    CHECK_F(ZSTD_checkCParams(params.cParams));
+    return ZSTD_compressBegin_internal(cctx, dict, dictSize, params, pledgedSrcSize);
+}
+
+
+size_t ZSTD_compressBegin_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel)
+{
+    ZSTD_parameters const params = ZSTD_getParams(compressionLevel, 0, dictSize);
+    return ZSTD_compressBegin_internal(cctx, dict, dictSize, params, 0);
+}
+
+
+size_t ZSTD_compressBegin(ZSTD_CCtx* zc, int compressionLevel)
+{
+    return ZSTD_compressBegin_usingDict(zc, NULL, 0, compressionLevel);
+}
+
+
+/*! ZSTD_writeEpilogue() :
+*   Ends a frame.
+*   @return : nb of bytes written into dst (or an error code) */
+static size_t ZSTD_writeEpilogue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)
+{
+    BYTE* const ostart = (BYTE*)dst;
+    BYTE* op = ostart;
+    size_t fhSize = 0;
+
+    if (cctx->stage == ZSTDcs_created) return ERROR(stage_wrong);  /* init missing */
+
+    /* special case : empty frame */
+    if (cctx->stage == ZSTDcs_init) {
+        fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, cctx->params, 0, 0);
+        if (ZSTD_isError(fhSize)) return fhSize;
+        dstCapacity -= fhSize;
+        op += fhSize;
+        cctx->stage = ZSTDcs_ongoing;
+    }
+
+    if (cctx->stage != ZSTDcs_ending) {
+        /* write one last empty block, make it the "last" block */
+        U32 const cBlockHeader24 = 1 /* last block */ + (((U32)bt_raw)<<1) + 0;
+        if (dstCapacity<4) return ERROR(dstSize_tooSmall);
+        MEM_writeLE32(op, cBlockHeader24);
+        op += ZSTD_blockHeaderSize;
+        dstCapacity -= ZSTD_blockHeaderSize;
+    }
+
+    if (cctx->params.fParams.checksumFlag) {
+        U32 const checksum = (U32) XXH64_digest(&cctx->xxhState);
+        if (dstCapacity<4) return ERROR(dstSize_tooSmall);
+        MEM_writeLE32(op, checksum);
+        op += 4;
+    }
+
+    cctx->stage = ZSTDcs_created;  /* return to "created but no init" status */
+    return op-ostart;
+}
+
+
+size_t ZSTD_compressEnd (ZSTD_CCtx* cctx,
+                         void* dst, size_t dstCapacity,
+                   const void* src, size_t srcSize)
+{
+    size_t endResult;
+    size_t const cSize = ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 1, 1);
+    if (ZSTD_isError(cSize)) return cSize;
+    endResult = ZSTD_writeEpilogue(cctx, (char*)dst + cSize, dstCapacity-cSize);
+    if (ZSTD_isError(endResult)) return endResult;
+    return cSize + endResult;
+}
+
+
+static size_t ZSTD_compress_internal (ZSTD_CCtx* cctx,
+                               void* dst, size_t dstCapacity,
+                         const void* src, size_t srcSize,
+                         const void* dict,size_t dictSize,
+                               ZSTD_parameters params)
+{
+    CHECK_F(ZSTD_compressBegin_internal(cctx, dict, dictSize, params, srcSize));
+    return ZSTD_compressEnd(cctx, dst,  dstCapacity, src, srcSize);
+}
+
+size_t ZSTD_compress_advanced (ZSTD_CCtx* ctx,
+                               void* dst, size_t dstCapacity,
+                         const void* src, size_t srcSize,
+                         const void* dict,size_t dictSize,
+                               ZSTD_parameters params)
+{
+    CHECK_F(ZSTD_checkCParams(params.cParams));
+    return ZSTD_compress_internal(ctx, dst, dstCapacity, src, srcSize, dict, dictSize, params);
+}
+
+size_t ZSTD_compress_usingDict(ZSTD_CCtx* ctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize, const void* dict, size_t dictSize, int compressionLevel)
+{
+    ZSTD_parameters params = ZSTD_getParams(compressionLevel, srcSize, dict ? dictSize : 0);
+    params.fParams.contentSizeFlag = 1;
+    return ZSTD_compress_internal(ctx, dst, dstCapacity, src, srcSize, dict, dictSize, params);
+}
+
+size_t ZSTD_compressCCtx (ZSTD_CCtx* ctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize, int compressionLevel)
+{
+    return ZSTD_compress_usingDict(ctx, dst, dstCapacity, src, srcSize, NULL, 0, compressionLevel);
+}
+
+size_t ZSTD_compress(void* dst, size_t dstCapacity, const void* src, size_t srcSize, int compressionLevel)
+{
+    size_t result;
+    ZSTD_CCtx ctxBody;
+    memset(&ctxBody, 0, sizeof(ctxBody));
+    memcpy(&ctxBody.customMem, &defaultCustomMem, sizeof(ZSTD_customMem));
+    result = ZSTD_compressCCtx(&ctxBody, dst, dstCapacity, src, srcSize, compressionLevel);
+    ZSTD_free(ctxBody.workSpace, defaultCustomMem);  /* can't free ctxBody itself, as it's on stack; free only heap content */
+    return result;
+}
+
+
+/* =====  Dictionary API  ===== */
+
+struct ZSTD_CDict_s {
+    void* dictContent;
+    size_t dictContentSize;
+    ZSTD_CCtx* refContext;
+};  /* typedef'd tp ZSTD_CDict within "zstd.h" */
+
+size_t ZSTD_sizeof_CDict(const ZSTD_CDict* cdict)
+{
+    if (cdict==NULL) return 0;   /* support sizeof on NULL */
+    return ZSTD_sizeof_CCtx(cdict->refContext) + cdict->dictContentSize;
+}
+
+ZSTD_CDict* ZSTD_createCDict_advanced(const void* dict, size_t dictSize, ZSTD_parameters params, ZSTD_customMem customMem)
+{
+    if (!customMem.customAlloc && !customMem.customFree) customMem = defaultCustomMem;
+    if (!customMem.customAlloc || !customMem.customFree) return NULL;
+
+    {   ZSTD_CDict* const cdict = (ZSTD_CDict*) ZSTD_malloc(sizeof(ZSTD_CDict), customMem);
+        void* const dictContent = ZSTD_malloc(dictSize, customMem);
+        ZSTD_CCtx* const cctx = ZSTD_createCCtx_advanced(customMem);
+
+        if (!dictContent || !cdict || !cctx) {
+            ZSTD_free(dictContent, customMem);
+            ZSTD_free(cdict, customMem);
+            ZSTD_free(cctx, customMem);
+            return NULL;
+        }
+
+        if (dictSize) {
+            memcpy(dictContent, dict, dictSize);
+        }
+        {   size_t const errorCode = ZSTD_compressBegin_advanced(cctx, dictContent, dictSize, params, 0);
+            if (ZSTD_isError(errorCode)) {
+                ZSTD_free(dictContent, customMem);
+                ZSTD_free(cdict, customMem);
+                ZSTD_free(cctx, customMem);
+                return NULL;
+        }   }
+
+        cdict->dictContent = dictContent;
+        cdict->dictContentSize = dictSize;
+        cdict->refContext = cctx;
+        return cdict;
+    }
+}
+
+ZSTD_CDict* ZSTD_createCDict(const void* dict, size_t dictSize, int compressionLevel)
+{
+    ZSTD_customMem const allocator = { NULL, NULL, NULL };
+    ZSTD_parameters params = ZSTD_getParams(compressionLevel, 0, dictSize);
+    params.fParams.contentSizeFlag = 1;
+    return ZSTD_createCDict_advanced(dict, dictSize, params, allocator);
+}
+
+size_t ZSTD_freeCDict(ZSTD_CDict* cdict)
+{
+    if (cdict==NULL) return 0;   /* support free on NULL */
+    {   ZSTD_customMem const cMem = cdict->refContext->customMem;
+        ZSTD_freeCCtx(cdict->refContext);
+        ZSTD_free(cdict->dictContent, cMem);
+        ZSTD_free(cdict, cMem);
+        return 0;
+    }
+}
+
+static ZSTD_parameters ZSTD_getParamsFromCDict(const ZSTD_CDict* cdict) {
+    return ZSTD_getParamsFromCCtx(cdict->refContext);
+}
+
+size_t ZSTD_compressBegin_usingCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict, U64 pledgedSrcSize)
+{
+    if (cdict->dictContentSize) CHECK_F(ZSTD_copyCCtx(cctx, cdict->refContext, pledgedSrcSize))
+    else CHECK_F(ZSTD_compressBegin_advanced(cctx, NULL, 0, cdict->refContext->params, pledgedSrcSize));
+    return 0;
+}
+
+/*! ZSTD_compress_usingCDict() :
+*   Compression using a digested Dictionary.
+*   Faster startup than ZSTD_compress_usingDict(), recommended when same dictionary is used multiple times.
+*   Note that compression level is decided during dictionary creation */
+size_t ZSTD_compress_usingCDict(ZSTD_CCtx* cctx,
+                                void* dst, size_t dstCapacity,
+                                const void* src, size_t srcSize,
+                                const ZSTD_CDict* cdict)
+{
+    CHECK_F(ZSTD_compressBegin_usingCDict(cctx, cdict, srcSize));
+
+    if (cdict->refContext->params.fParams.contentSizeFlag==1) {
+        cctx->params.fParams.contentSizeFlag = 1;
+        cctx->frameContentSize = srcSize;
+    }
+
+    return ZSTD_compressEnd(cctx, dst, dstCapacity, src, srcSize);
+}
+
+
+
+/* ******************************************************************
+*  Streaming
+********************************************************************/
+
+typedef enum { zcss_init, zcss_load, zcss_flush, zcss_final } ZSTD_cStreamStage;
+
+struct ZSTD_CStream_s {
+    ZSTD_CCtx* cctx;
+    ZSTD_CDict* cdictLocal;
+    const ZSTD_CDict* cdict;
+    char*  inBuff;
+    size_t inBuffSize;
+    size_t inToCompress;
+    size_t inBuffPos;
+    size_t inBuffTarget;
+    size_t blockSize;
+    char*  outBuff;
+    size_t outBuffSize;
+    size_t outBuffContentSize;
+    size_t outBuffFlushedSize;
+    ZSTD_cStreamStage stage;
+    U32    checksum;
+    U32    frameEnded;
+    U64    pledgedSrcSize;
+    U64    inputProcessed;
+    ZSTD_parameters params;
+    ZSTD_customMem customMem;
+};   /* typedef'd to ZSTD_CStream within "zstd.h" */
+
+ZSTD_CStream* ZSTD_createCStream(void)
+{
+    return ZSTD_createCStream_advanced(defaultCustomMem);
+}
+
+ZSTD_CStream* ZSTD_createCStream_advanced(ZSTD_customMem customMem)
+{
+    ZSTD_CStream* zcs;
+
+    if (!customMem.customAlloc && !customMem.customFree) customMem = defaultCustomMem;
+    if (!customMem.customAlloc || !customMem.customFree) return NULL;
+
+    zcs = (ZSTD_CStream*)ZSTD_malloc(sizeof(ZSTD_CStream), customMem);
+    if (zcs==NULL) return NULL;
+    memset(zcs, 0, sizeof(ZSTD_CStream));
+    memcpy(&zcs->customMem, &customMem, sizeof(ZSTD_customMem));
+    zcs->cctx = ZSTD_createCCtx_advanced(customMem);
+    if (zcs->cctx == NULL) { ZSTD_freeCStream(zcs); return NULL; }
+    return zcs;
+}
+
+size_t ZSTD_freeCStream(ZSTD_CStream* zcs)
+{
+    if (zcs==NULL) return 0;   /* support free on NULL */
+    {   ZSTD_customMem const cMem = zcs->customMem;
+        ZSTD_freeCCtx(zcs->cctx);
+        ZSTD_freeCDict(zcs->cdictLocal);
+        ZSTD_free(zcs->inBuff, cMem);
+        ZSTD_free(zcs->outBuff, cMem);
+        ZSTD_free(zcs, cMem);
+        return 0;
+    }
+}
+
+
+/*======   Initialization   ======*/
+
+size_t ZSTD_CStreamInSize(void)  { return ZSTD_BLOCKSIZE_ABSOLUTEMAX; }
+size_t ZSTD_CStreamOutSize(void) { return ZSTD_compressBound(ZSTD_BLOCKSIZE_ABSOLUTEMAX) + ZSTD_blockHeaderSize + 4 /* 32-bits hash */ ; }
+
+size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pledgedSrcSize)
+{
+    if (zcs->inBuffSize==0) return ERROR(stage_wrong);   /* zcs has not been init at least once */
+
+    if (zcs->cdict) CHECK_F(ZSTD_compressBegin_usingCDict(zcs->cctx, zcs->cdict, pledgedSrcSize))
+    else CHECK_F(ZSTD_compressBegin_advanced(zcs->cctx, NULL, 0, zcs->params, pledgedSrcSize));
+
+    zcs->inToCompress = 0;
+    zcs->inBuffPos = 0;
+    zcs->inBuffTarget = zcs->blockSize;
+    zcs->outBuffContentSize = zcs->outBuffFlushedSize = 0;
+    zcs->stage = zcss_load;
+    zcs->frameEnded = 0;
+    zcs->pledgedSrcSize = pledgedSrcSize;
+    zcs->inputProcessed = 0;
+    return 0;   /* ready to go */
+}
+
+size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs,
+                                 const void* dict, size_t dictSize,
+                                 ZSTD_parameters params, unsigned long long pledgedSrcSize)
+{
+    /* allocate buffers */
+    {   size_t const neededInBuffSize = (size_t)1 << params.cParams.windowLog;
+        if (zcs->inBuffSize < neededInBuffSize) {
+            zcs->inBuffSize = neededInBuffSize;
+            ZSTD_free(zcs->inBuff, zcs->customMem);
+            zcs->inBuff = (char*) ZSTD_malloc(neededInBuffSize, zcs->customMem);
+            if (zcs->inBuff == NULL) return ERROR(memory_allocation);
+        }
+        zcs->blockSize = MIN(ZSTD_BLOCKSIZE_ABSOLUTEMAX, neededInBuffSize);
+    }
+    if (zcs->outBuffSize < ZSTD_compressBound(zcs->blockSize)+1) {
+        zcs->outBuffSize = ZSTD_compressBound(zcs->blockSize)+1;
+        ZSTD_free(zcs->outBuff, zcs->customMem);
+        zcs->outBuff = (char*) ZSTD_malloc(zcs->outBuffSize, zcs->customMem);
+        if (zcs->outBuff == NULL) return ERROR(memory_allocation);
+    }
+
+    if (dict) {
+        ZSTD_freeCDict(zcs->cdictLocal);
+        zcs->cdictLocal = ZSTD_createCDict_advanced(dict, dictSize, params, zcs->customMem);
+        if (zcs->cdictLocal == NULL) return ERROR(memory_allocation);
+        zcs->cdict = zcs->cdictLocal;
+    } else zcs->cdict = NULL;
+
+    zcs->checksum = params.fParams.checksumFlag > 0;
+    zcs->params = params;
+
+    return ZSTD_resetCStream(zcs, pledgedSrcSize);
+}
+
+/* note : cdict must outlive compression session */
+size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDict* cdict)
+{
+    ZSTD_parameters const params = ZSTD_getParamsFromCDict(cdict);
+    size_t const initError =  ZSTD_initCStream_advanced(zcs, NULL, 0, params, 0);
+    zcs->cdict = cdict;
+    return initError;
+}
+
+size_t ZSTD_initCStream_usingDict(ZSTD_CStream* zcs, const void* dict, size_t dictSize, int compressionLevel)
+{
+    ZSTD_parameters const params = ZSTD_getParams(compressionLevel, 0, dictSize);
+    return ZSTD_initCStream_advanced(zcs, dict, dictSize, params, 0);
+}
+
+size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLevel, unsigned long long pledgedSrcSize)
+{
+    ZSTD_parameters const params = ZSTD_getParams(compressionLevel, pledgedSrcSize, 0);
+    return ZSTD_initCStream_advanced(zcs, NULL, 0, params, pledgedSrcSize);
+}
+
+size_t ZSTD_initCStream(ZSTD_CStream* zcs, int compressionLevel)
+{
+    return ZSTD_initCStream_usingDict(zcs, NULL, 0, compressionLevel);
+}
+
+size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs)
+{
+    if (zcs==NULL) return 0;   /* support sizeof on NULL */
+    return sizeof(zcs) + ZSTD_sizeof_CCtx(zcs->cctx) + ZSTD_sizeof_CDict(zcs->cdictLocal) + zcs->outBuffSize + zcs->inBuffSize;
+}
+
+/*======   Compression   ======*/
+
+typedef enum { zsf_gather, zsf_flush, zsf_end } ZSTD_flush_e;
+
+MEM_STATIC size_t ZSTD_limitCopy(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+    size_t const length = MIN(dstCapacity, srcSize);
+    memcpy(dst, src, length);
+    return length;
+}
+
+static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
+                              void* dst, size_t* dstCapacityPtr,
+                        const void* src, size_t* srcSizePtr,
+                              ZSTD_flush_e const flush)
+{
+    U32 someMoreWork = 1;
+    const char* const istart = (const char*)src;
+    const char* const iend = istart + *srcSizePtr;
+    const char* ip = istart;
+    char* const ostart = (char*)dst;
+    char* const oend = ostart + *dstCapacityPtr;
+    char* op = ostart;
+
+    while (someMoreWork) {
+        switch(zcs->stage)
+        {
+        case zcss_init: return ERROR(init_missing);   /* call ZBUFF_compressInit() first ! */
+
+        case zcss_load:
+            /* complete inBuffer */
+            {   size_t const toLoad = zcs->inBuffTarget - zcs->inBuffPos;
+                size_t const loaded = ZSTD_limitCopy(zcs->inBuff + zcs->inBuffPos, toLoad, ip, iend-ip);
+                zcs->inBuffPos += loaded;
+                ip += loaded;
+                if ( (zcs->inBuffPos==zcs->inToCompress) || (!flush && (toLoad != loaded)) ) {
+                    someMoreWork = 0; break;  /* not enough input to get a full block : stop there, wait for more */
+            }   }
+            /* compress current block (note : this stage cannot be stopped in the middle) */
+            {   void* cDst;
+                size_t cSize;
+                size_t const iSize = zcs->inBuffPos - zcs->inToCompress;
+                size_t oSize = oend-op;
+                if (oSize >= ZSTD_compressBound(iSize))
+                    cDst = op;   /* compress directly into output buffer (avoid flush stage) */
+                else
+                    cDst = zcs->outBuff, oSize = zcs->outBuffSize;
+                cSize = (flush == zsf_end) ?
+                        ZSTD_compressEnd(zcs->cctx, cDst, oSize, zcs->inBuff + zcs->inToCompress, iSize) :
+                        ZSTD_compressContinue(zcs->cctx, cDst, oSize, zcs->inBuff + zcs->inToCompress, iSize);
+                if (ZSTD_isError(cSize)) return cSize;
+                if (flush == zsf_end) zcs->frameEnded = 1;
+                /* prepare next block */
+                zcs->inBuffTarget = zcs->inBuffPos + zcs->blockSize;
+                if (zcs->inBuffTarget > zcs->inBuffSize)
+                    zcs->inBuffPos = 0, zcs->inBuffTarget = zcs->blockSize;   /* note : inBuffSize >= blockSize */
+                zcs->inToCompress = zcs->inBuffPos;
+                if (cDst == op) { op += cSize; break; }   /* no need to flush */
+                zcs->outBuffContentSize = cSize;
+                zcs->outBuffFlushedSize = 0;
+                zcs->stage = zcss_flush;   /* pass-through to flush stage */
+            }
+
+        case zcss_flush:
+            {   size_t const toFlush = zcs->outBuffContentSize - zcs->outBuffFlushedSize;
+                size_t const flushed = ZSTD_limitCopy(op, oend-op, zcs->outBuff + zcs->outBuffFlushedSize, toFlush);
+                op += flushed;
+                zcs->outBuffFlushedSize += flushed;
+                if (toFlush!=flushed) { someMoreWork = 0; break; }  /* dst too small to store flushed data : stop there */
+                zcs->outBuffContentSize = zcs->outBuffFlushedSize = 0;
+                zcs->stage = zcss_load;
+                break;
+            }
+
+        case zcss_final:
+            someMoreWork = 0;   /* do nothing */
+            break;
+
+        default:
+            return ERROR(GENERIC);   /* impossible */
+        }
+    }
+
+    *srcSizePtr = ip - istart;
+    *dstCapacityPtr = op - ostart;
+    zcs->inputProcessed += *srcSizePtr;
+    if (zcs->frameEnded) return 0;
+    {   size_t hintInSize = zcs->inBuffTarget - zcs->inBuffPos;
+        if (hintInSize==0) hintInSize = zcs->blockSize;
+        return hintInSize;
+    }
+}
+
+size_t ZSTD_compressStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
+{
+    size_t sizeRead = input->size - input->pos;
+    size_t sizeWritten = output->size - output->pos;
+    size_t const result = ZSTD_compressStream_generic(zcs,
+                                                      (char*)(output->dst) + output->pos, &sizeWritten,
+                                                      (const char*)(input->src) + input->pos, &sizeRead, zsf_gather);
+    input->pos += sizeRead;
+    output->pos += sizeWritten;
+    return result;
+}
+
+
+/*======   Finalize   ======*/
+
+/*! ZSTD_flushStream() :
+*   @return : amount of data remaining to flush */
+size_t ZSTD_flushStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)
+{
+    size_t srcSize = 0;
+    size_t sizeWritten = output->size - output->pos;
+    size_t const result = ZSTD_compressStream_generic(zcs,
+                                                     (char*)(output->dst) + output->pos, &sizeWritten,
+                                                     &srcSize, &srcSize, /* use a valid src address instead of NULL */
+                                                      zsf_flush);
+    output->pos += sizeWritten;
+    if (ZSTD_isError(result)) return result;
+    return zcs->outBuffContentSize - zcs->outBuffFlushedSize;   /* remaining to flush */
+}
+
+
+size_t ZSTD_endStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)
+{
+    BYTE* const ostart = (BYTE*)(output->dst) + output->pos;
+    BYTE* const oend = (BYTE*)(output->dst) + output->size;
+    BYTE* op = ostart;
+
+    if ((zcs->pledgedSrcSize) && (zcs->inputProcessed != zcs->pledgedSrcSize))
+        return ERROR(srcSize_wrong);   /* pledgedSrcSize not respected */
+
+    if (zcs->stage != zcss_final) {
+        /* flush whatever remains */
+        size_t srcSize = 0;
+        size_t sizeWritten = output->size - output->pos;
+        size_t const notEnded = ZSTD_compressStream_generic(zcs, ostart, &sizeWritten, &srcSize, &srcSize, zsf_end);  /* use a valid src address instead of NULL */
+        size_t const remainingToFlush = zcs->outBuffContentSize - zcs->outBuffFlushedSize;
+        op += sizeWritten;
+        if (remainingToFlush) {
+            output->pos += sizeWritten;
+            return remainingToFlush + ZSTD_BLOCKHEADERSIZE /* final empty block */ + (zcs->checksum * 4);
+        }
+        /* create epilogue */
+        zcs->stage = zcss_final;
+        zcs->outBuffContentSize = !notEnded ? 0 :
+            ZSTD_compressEnd(zcs->cctx, zcs->outBuff, zcs->outBuffSize, NULL, 0);  /* write epilogue, including final empty block, into outBuff */
+    }
+
+    /* flush epilogue */
+    {   size_t const toFlush = zcs->outBuffContentSize - zcs->outBuffFlushedSize;
+        size_t const flushed = ZSTD_limitCopy(op, oend-op, zcs->outBuff + zcs->outBuffFlushedSize, toFlush);
+        op += flushed;
+        zcs->outBuffFlushedSize += flushed;
+        output->pos += op-ostart;
+        if (toFlush==flushed) zcs->stage = zcss_init;  /* end reached */
+        return toFlush - flushed;
+    }
+}
+
+
+
+/*-=====  Pre-defined compression levels  =====-*/
+
+#define ZSTD_DEFAULT_CLEVEL 1
+#define ZSTD_MAX_CLEVEL     22
+int ZSTD_maxCLevel(void) { return ZSTD_MAX_CLEVEL; }
+
+static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEVEL+1] = {
+{   /* "default" */
+    /* W,  C,  H,  S,  L, TL, strat */
+    { 18, 12, 12,  1,  7, 16, ZSTD_fast    },  /* level  0 - never used */
+    { 19, 13, 14,  1,  7, 16, ZSTD_fast    },  /* level  1 */
+    { 19, 15, 16,  1,  6, 16, ZSTD_fast    },  /* level  2 */
+    { 20, 16, 17,  1,  5, 16, ZSTD_dfast   },  /* level  3.*/
+    { 20, 18, 18,  1,  5, 16, ZSTD_dfast   },  /* level  4.*/
+    { 20, 15, 18,  3,  5, 16, ZSTD_greedy  },  /* level  5 */
+    { 21, 16, 19,  2,  5, 16, ZSTD_lazy    },  /* level  6 */
+    { 21, 17, 20,  3,  5, 16, ZSTD_lazy    },  /* level  7 */
+    { 21, 18, 20,  3,  5, 16, ZSTD_lazy2   },  /* level  8 */
+    { 21, 20, 20,  3,  5, 16, ZSTD_lazy2   },  /* level  9 */
+    { 21, 19, 21,  4,  5, 16, ZSTD_lazy2   },  /* level 10 */
+    { 22, 20, 22,  4,  5, 16, ZSTD_lazy2   },  /* level 11 */
+    { 22, 20, 22,  5,  5, 16, ZSTD_lazy2   },  /* level 12 */
+    { 22, 21, 22,  5,  5, 16, ZSTD_lazy2   },  /* level 13 */
+    { 22, 21, 22,  6,  5, 16, ZSTD_lazy2   },  /* level 14 */
+    { 22, 21, 21,  5,  5, 16, ZSTD_btlazy2 },  /* level 15 */
+    { 23, 22, 22,  5,  5, 16, ZSTD_btlazy2 },  /* level 16 */
+    { 23, 21, 22,  4,  5, 24, ZSTD_btopt   },  /* level 17 */
+    { 23, 23, 22,  6,  5, 32, ZSTD_btopt   },  /* level 18 */
+    { 23, 23, 22,  6,  3, 48, ZSTD_btopt   },  /* level 19 */
+    { 25, 25, 23,  7,  3, 64, ZSTD_btopt2  },  /* level 20 */
+    { 26, 26, 23,  7,  3,256, ZSTD_btopt2  },  /* level 21 */
+    { 27, 27, 25,  9,  3,512, ZSTD_btopt2  },  /* level 22 */
+},
+{   /* for srcSize <= 256 KB */
+    /* W,  C,  H,  S,  L,  T, strat */
+    {  0,  0,  0,  0,  0,  0, ZSTD_fast    },  /* level  0 - not used */
+    { 18, 13, 14,  1,  6,  8, ZSTD_fast    },  /* level  1 */
+    { 18, 14, 13,  1,  5,  8, ZSTD_dfast   },  /* level  2 */
+    { 18, 16, 15,  1,  5,  8, ZSTD_dfast   },  /* level  3 */
+    { 18, 15, 17,  1,  5,  8, ZSTD_greedy  },  /* level  4.*/
+    { 18, 16, 17,  4,  5,  8, ZSTD_greedy  },  /* level  5.*/
+    { 18, 16, 17,  3,  5,  8, ZSTD_lazy    },  /* level  6.*/
+    { 18, 17, 17,  4,  4,  8, ZSTD_lazy    },  /* level  7 */
+    { 18, 17, 17,  4,  4,  8, ZSTD_lazy2   },  /* level  8 */
+    { 18, 17, 17,  5,  4,  8, ZSTD_lazy2   },  /* level  9 */
+    { 18, 17, 17,  6,  4,  8, ZSTD_lazy2   },  /* level 10 */
+    { 18, 18, 17,  6,  4,  8, ZSTD_lazy2   },  /* level 11.*/
+    { 18, 18, 17,  7,  4,  8, ZSTD_lazy2   },  /* level 12.*/
+    { 18, 19, 17,  6,  4,  8, ZSTD_btlazy2 },  /* level 13 */
+    { 18, 18, 18,  4,  4, 16, ZSTD_btopt   },  /* level 14.*/
+    { 18, 18, 18,  4,  3, 16, ZSTD_btopt   },  /* level 15.*/
+    { 18, 19, 18,  6,  3, 32, ZSTD_btopt   },  /* level 16.*/
+    { 18, 19, 18,  8,  3, 64, ZSTD_btopt   },  /* level 17.*/
+    { 18, 19, 18,  9,  3,128, ZSTD_btopt   },  /* level 18.*/
+    { 18, 19, 18, 10,  3,256, ZSTD_btopt   },  /* level 19.*/
+    { 18, 19, 18, 11,  3,512, ZSTD_btopt2  },  /* level 20.*/
+    { 18, 19, 18, 12,  3,512, ZSTD_btopt2  },  /* level 21.*/
+    { 18, 19, 18, 13,  3,512, ZSTD_btopt2  },  /* level 22.*/
+},
+{   /* for srcSize <= 128 KB */
+    /* W,  C,  H,  S,  L,  T, strat */
+    { 17, 12, 12,  1,  7,  8, ZSTD_fast    },  /* level  0 - not used */
+    { 17, 12, 13,  1,  6,  8, ZSTD_fast    },  /* level  1 */
+    { 17, 13, 16,  1,  5,  8, ZSTD_fast    },  /* level  2 */
+    { 17, 16, 16,  2,  5,  8, ZSTD_dfast   },  /* level  3 */
+    { 17, 13, 15,  3,  4,  8, ZSTD_greedy  },  /* level  4 */
+    { 17, 15, 17,  4,  4,  8, ZSTD_greedy  },  /* level  5 */
+    { 17, 16, 17,  3,  4,  8, ZSTD_lazy    },  /* level  6 */
+    { 17, 15, 17,  4,  4,  8, ZSTD_lazy2   },  /* level  7 */
+    { 17, 17, 17,  4,  4,  8, ZSTD_lazy2   },  /* level  8 */
+    { 17, 17, 17,  5,  4,  8, ZSTD_lazy2   },  /* level  9 */
+    { 17, 17, 17,  6,  4,  8, ZSTD_lazy2   },  /* level 10 */
+    { 17, 17, 17,  7,  4,  8, ZSTD_lazy2   },  /* level 11 */
+    { 17, 17, 17,  8,  4,  8, ZSTD_lazy2   },  /* level 12 */
+    { 17, 18, 17,  6,  4,  8, ZSTD_btlazy2 },  /* level 13.*/
+    { 17, 17, 17,  7,  3,  8, ZSTD_btopt   },  /* level 14.*/
+    { 17, 17, 17,  7,  3, 16, ZSTD_btopt   },  /* level 15.*/
+    { 17, 18, 17,  7,  3, 32, ZSTD_btopt   },  /* level 16.*/
+    { 17, 18, 17,  7,  3, 64, ZSTD_btopt   },  /* level 17.*/
+    { 17, 18, 17,  7,  3,256, ZSTD_btopt   },  /* level 18.*/
+    { 17, 18, 17,  8,  3,256, ZSTD_btopt   },  /* level 19.*/
+    { 17, 18, 17,  9,  3,256, ZSTD_btopt2  },  /* level 20.*/
+    { 17, 18, 17, 10,  3,256, ZSTD_btopt2  },  /* level 21.*/
+    { 17, 18, 17, 11,  3,512, ZSTD_btopt2  },  /* level 22.*/
+},
+{   /* for srcSize <= 16 KB */
+    /* W,  C,  H,  S,  L,  T, strat */
+    { 14, 12, 12,  1,  7,  6, ZSTD_fast    },  /* level  0 - not used */
+    { 14, 14, 14,  1,  6,  6, ZSTD_fast    },  /* level  1 */
+    { 14, 14, 14,  1,  4,  6, ZSTD_fast    },  /* level  2 */
+    { 14, 14, 14,  1,  4,  6, ZSTD_dfast   },  /* level  3.*/
+    { 14, 14, 14,  4,  4,  6, ZSTD_greedy  },  /* level  4.*/
+    { 14, 14, 14,  3,  4,  6, ZSTD_lazy    },  /* level  5.*/
+    { 14, 14, 14,  4,  4,  6, ZSTD_lazy2   },  /* level  6 */
+    { 14, 14, 14,  5,  4,  6, ZSTD_lazy2   },  /* level  7 */
+    { 14, 14, 14,  6,  4,  6, ZSTD_lazy2   },  /* level  8.*/
+    { 14, 15, 14,  6,  4,  6, ZSTD_btlazy2 },  /* level  9.*/
+    { 14, 15, 14,  3,  3,  6, ZSTD_btopt   },  /* level 10.*/
+    { 14, 15, 14,  6,  3,  8, ZSTD_btopt   },  /* level 11.*/
+    { 14, 15, 14,  6,  3, 16, ZSTD_btopt   },  /* level 12.*/
+    { 14, 15, 14,  6,  3, 24, ZSTD_btopt   },  /* level 13.*/
+    { 14, 15, 15,  6,  3, 48, ZSTD_btopt   },  /* level 14.*/
+    { 14, 15, 15,  6,  3, 64, ZSTD_btopt   },  /* level 15.*/
+    { 14, 15, 15,  6,  3, 96, ZSTD_btopt   },  /* level 16.*/
+    { 14, 15, 15,  6,  3,128, ZSTD_btopt   },  /* level 17.*/
+    { 14, 15, 15,  6,  3,256, ZSTD_btopt   },  /* level 18.*/
+    { 14, 15, 15,  7,  3,256, ZSTD_btopt   },  /* level 19.*/
+    { 14, 15, 15,  8,  3,256, ZSTD_btopt2  },  /* level 20.*/
+    { 14, 15, 15,  9,  3,256, ZSTD_btopt2  },  /* level 21.*/
+    { 14, 15, 15, 10,  3,256, ZSTD_btopt2  },  /* level 22.*/
+},
+};
+
+/*! ZSTD_getCParams() :
+*   @return ZSTD_compressionParameters structure for a selected compression level, `srcSize` and `dictSize`.
+*   Size values are optional, provide 0 if not known or unused */
+ZSTD_compressionParameters ZSTD_getCParams(int compressionLevel, unsigned long long srcSize, size_t dictSize)
+{
+    ZSTD_compressionParameters cp;
+    size_t const addedSize = srcSize ? 0 : 500;
+    U64 const rSize = srcSize+dictSize ? srcSize+dictSize+addedSize : (U64)-1;
+    U32 const tableID = (rSize <= 256 KB) + (rSize <= 128 KB) + (rSize <= 16 KB);   /* intentional underflow for srcSizeHint == 0 */
+    if (compressionLevel <= 0) compressionLevel = ZSTD_DEFAULT_CLEVEL;   /* 0 == default; no negative compressionLevel yet */
+    if (compressionLevel > ZSTD_MAX_CLEVEL) compressionLevel = ZSTD_MAX_CLEVEL;
+    cp = ZSTD_defaultCParameters[tableID][compressionLevel];
+    if (MEM_32bits()) {   /* auto-correction, for 32-bits mode */
+        if (cp.windowLog > ZSTD_WINDOWLOG_MAX) cp.windowLog = ZSTD_WINDOWLOG_MAX;
+        if (cp.chainLog > ZSTD_CHAINLOG_MAX) cp.chainLog = ZSTD_CHAINLOG_MAX;
+        if (cp.hashLog > ZSTD_HASHLOG_MAX) cp.hashLog = ZSTD_HASHLOG_MAX;
+    }
+    cp = ZSTD_adjustCParams(cp, srcSize, dictSize);
+    return cp;
+}
+
+/*! ZSTD_getParams() :
+*   same as ZSTD_getCParams(), but @return a `ZSTD_parameters` object (instead of `ZSTD_compressionParameters`).
+*   All fields of `ZSTD_frameParameters` are set to default (0) */
+ZSTD_parameters ZSTD_getParams(int compressionLevel, unsigned long long srcSize, size_t dictSize) {
+    ZSTD_parameters params;
+    ZSTD_compressionParameters const cParams = ZSTD_getCParams(compressionLevel, srcSize, dictSize);
+    memset(¶ms, 0, sizeof(params));
+    params.cParams = cParams;
+    return params;
+}
diff --git a/lib/compress/zstd_opt.h b/lib/compress/zstd_opt.h
new file mode 100644
index 0000000..f071c4f
--- /dev/null
+++ b/lib/compress/zstd_opt.h
@@ -0,0 +1,919 @@
+/**
+ * Copyright (c) 2016-present, Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+
+/* Note : this file is intended to be included within zstd_compress.c */
+
+
+#ifndef ZSTD_OPT_H_91842398743
+#define ZSTD_OPT_H_91842398743
+
+
+#define ZSTD_LITFREQ_ADD    2
+#define ZSTD_FREQ_DIV       4
+#define ZSTD_MAX_PRICE      (1<<30)
+
+/*-*************************************
+*  Price functions for optimal parser
+***************************************/
+FORCE_INLINE void ZSTD_setLog2Prices(seqStore_t* ssPtr)
+{
+    ssPtr->log2matchLengthSum = ZSTD_highbit32(ssPtr->matchLengthSum+1);
+    ssPtr->log2litLengthSum = ZSTD_highbit32(ssPtr->litLengthSum+1);
+    ssPtr->log2litSum = ZSTD_highbit32(ssPtr->litSum+1);
+    ssPtr->log2offCodeSum = ZSTD_highbit32(ssPtr->offCodeSum+1);
+    ssPtr->factor = 1 + ((ssPtr->litSum>>5) / ssPtr->litLengthSum) + ((ssPtr->litSum<<1) / (ssPtr->litSum + ssPtr->matchSum));
+}
+
+
+MEM_STATIC void ZSTD_rescaleFreqs(seqStore_t* ssPtr, const BYTE* src, size_t srcSize)
+{
+    unsigned u;
+
+    ssPtr->cachedLiterals = NULL;
+    ssPtr->cachedPrice = ssPtr->cachedLitLength = 0;
+    ssPtr->staticPrices = 0; 
+
+    if (ssPtr->litLengthSum == 0) {
+        if (srcSize <= 1024) ssPtr->staticPrices = 1;
+
+        for (u=0; u<=MaxLit; u++)
+            ssPtr->litFreq[u] = 0;
+        for (u=0; ulitFreq[src[u]]++;
+
+        ssPtr->litSum = 0;
+        ssPtr->litLengthSum = MaxLL+1;
+        ssPtr->matchLengthSum = MaxML+1;
+        ssPtr->offCodeSum = (MaxOff+1);
+        ssPtr->matchSum = (ZSTD_LITFREQ_ADD<litFreq[u] = 1 + (ssPtr->litFreq[u]>>ZSTD_FREQ_DIV);
+            ssPtr->litSum += ssPtr->litFreq[u]; 
+        }
+        for (u=0; u<=MaxLL; u++)
+            ssPtr->litLengthFreq[u] = 1;
+        for (u=0; u<=MaxML; u++)
+            ssPtr->matchLengthFreq[u] = 1;
+        for (u=0; u<=MaxOff; u++)
+            ssPtr->offCodeFreq[u] = 1;
+    } else {
+        ssPtr->matchLengthSum = 0;
+        ssPtr->litLengthSum = 0;
+        ssPtr->offCodeSum = 0;
+        ssPtr->matchSum = 0;
+        ssPtr->litSum = 0;
+
+        for (u=0; u<=MaxLit; u++) {
+            ssPtr->litFreq[u] = 1 + (ssPtr->litFreq[u]>>(ZSTD_FREQ_DIV+1));
+            ssPtr->litSum += ssPtr->litFreq[u];
+        }
+        for (u=0; u<=MaxLL; u++) {
+            ssPtr->litLengthFreq[u] = 1 + (ssPtr->litLengthFreq[u]>>(ZSTD_FREQ_DIV+1));
+            ssPtr->litLengthSum += ssPtr->litLengthFreq[u];
+        }
+        for (u=0; u<=MaxML; u++) {
+            ssPtr->matchLengthFreq[u] = 1 + (ssPtr->matchLengthFreq[u]>>ZSTD_FREQ_DIV);
+            ssPtr->matchLengthSum += ssPtr->matchLengthFreq[u];
+            ssPtr->matchSum += ssPtr->matchLengthFreq[u] * (u + 3);
+        }
+        ssPtr->matchSum *= ZSTD_LITFREQ_ADD;
+        for (u=0; u<=MaxOff; u++) {
+            ssPtr->offCodeFreq[u] = 1 + (ssPtr->offCodeFreq[u]>>ZSTD_FREQ_DIV);
+            ssPtr->offCodeSum += ssPtr->offCodeFreq[u];
+        }
+    }
+
+    ZSTD_setLog2Prices(ssPtr);
+}
+
+
+FORCE_INLINE U32 ZSTD_getLiteralPrice(seqStore_t* ssPtr, U32 litLength, const BYTE* literals)
+{
+    U32 price, u;
+
+    if (ssPtr->staticPrices)
+        return ZSTD_highbit32((U32)litLength+1) + (litLength*6);
+
+    if (litLength == 0)
+        return ssPtr->log2litLengthSum - ZSTD_highbit32(ssPtr->litLengthFreq[0]+1);
+
+    /* literals */
+    if (ssPtr->cachedLiterals == literals) {
+        U32 const additional = litLength - ssPtr->cachedLitLength;
+        const BYTE* literals2 = ssPtr->cachedLiterals + ssPtr->cachedLitLength;
+        price = ssPtr->cachedPrice + additional * ssPtr->log2litSum;
+        for (u=0; u < additional; u++)
+            price -= ZSTD_highbit32(ssPtr->litFreq[literals2[u]]+1);
+        ssPtr->cachedPrice = price;
+        ssPtr->cachedLitLength = litLength;
+    } else {
+        price = litLength * ssPtr->log2litSum;
+        for (u=0; u < litLength; u++)
+            price -= ZSTD_highbit32(ssPtr->litFreq[literals[u]]+1);
+
+        if (litLength >= 12) {
+            ssPtr->cachedLiterals = literals;
+            ssPtr->cachedPrice = price;
+            ssPtr->cachedLitLength = litLength;
+        }
+    }
+
+    /* literal Length */
+    {   const BYTE LL_deltaCode = 19;
+        const BYTE llCode = (litLength>63) ? (BYTE)ZSTD_highbit32(litLength) + LL_deltaCode : LL_Code[litLength];
+        price += LL_bits[llCode] + ssPtr->log2litLengthSum - ZSTD_highbit32(ssPtr->litLengthFreq[llCode]+1);
+    }
+
+    return price;
+}
+
+
+FORCE_INLINE U32 ZSTD_getPrice(seqStore_t* seqStorePtr, U32 litLength, const BYTE* literals, U32 offset, U32 matchLength, const int ultra)
+{
+    /* offset */
+    U32 price;
+    BYTE const offCode = (BYTE)ZSTD_highbit32(offset+1);
+
+    if (seqStorePtr->staticPrices)
+        return ZSTD_getLiteralPrice(seqStorePtr, litLength, literals) + ZSTD_highbit32((U32)matchLength+1) + 16 + offCode;
+
+    price = offCode + seqStorePtr->log2offCodeSum - ZSTD_highbit32(seqStorePtr->offCodeFreq[offCode]+1);
+    if (!ultra && offCode >= 20) price += (offCode-19)*2;
+
+    /* match Length */
+    {   const BYTE ML_deltaCode = 36;
+        const BYTE mlCode = (matchLength>127) ? (BYTE)ZSTD_highbit32(matchLength) + ML_deltaCode : ML_Code[matchLength];
+        price += ML_bits[mlCode] + seqStorePtr->log2matchLengthSum - ZSTD_highbit32(seqStorePtr->matchLengthFreq[mlCode]+1);
+    }
+
+    return price + ZSTD_getLiteralPrice(seqStorePtr, litLength, literals) + seqStorePtr->factor;
+}
+
+
+MEM_STATIC void ZSTD_updatePrice(seqStore_t* seqStorePtr, U32 litLength, const BYTE* literals, U32 offset, U32 matchLength)
+{
+    U32 u;
+
+    /* literals */
+    seqStorePtr->litSum += litLength*ZSTD_LITFREQ_ADD;
+    for (u=0; u < litLength; u++)
+        seqStorePtr->litFreq[literals[u]] += ZSTD_LITFREQ_ADD;
+
+    /* literal Length */
+    {   const BYTE LL_deltaCode = 19;
+        const BYTE llCode = (litLength>63) ? (BYTE)ZSTD_highbit32(litLength) + LL_deltaCode : LL_Code[litLength];
+        seqStorePtr->litLengthFreq[llCode]++;
+        seqStorePtr->litLengthSum++;
+    }
+
+    /* match offset */
+	{   BYTE const offCode = (BYTE)ZSTD_highbit32(offset+1);
+		seqStorePtr->offCodeSum++;
+		seqStorePtr->offCodeFreq[offCode]++;
+	}
+
+    /* match Length */
+    {   const BYTE ML_deltaCode = 36;
+        const BYTE mlCode = (matchLength>127) ? (BYTE)ZSTD_highbit32(matchLength) + ML_deltaCode : ML_Code[matchLength];
+        seqStorePtr->matchLengthFreq[mlCode]++;
+        seqStorePtr->matchLengthSum++;
+    }
+
+    ZSTD_setLog2Prices(seqStorePtr);
+}
+
+
+#define SET_PRICE(pos, mlen_, offset_, litlen_, price_)   \
+    {                                                 \
+        while (last_pos < pos)  { opt[last_pos+1].price = ZSTD_MAX_PRICE; last_pos++; } \
+        opt[pos].mlen = mlen_;                         \
+        opt[pos].off = offset_;                        \
+        opt[pos].litlen = litlen_;                     \
+        opt[pos].price = price_;                       \
+    }
+
+
+
+/* Update hashTable3 up to ip (excluded)
+   Assumption : always within prefix (ie. not within extDict) */
+FORCE_INLINE
+U32 ZSTD_insertAndFindFirstIndexHash3 (ZSTD_CCtx* zc, const BYTE* ip)
+{
+    U32* const hashTable3  = zc->hashTable3;
+    U32 const hashLog3  = zc->hashLog3;
+    const BYTE* const base = zc->base;
+    U32 idx = zc->nextToUpdate3;
+    const U32 target = zc->nextToUpdate3 = (U32)(ip - base);
+    const size_t hash3 = ZSTD_hash3Ptr(ip, hashLog3);
+
+    while(idx < target) {
+        hashTable3[ZSTD_hash3Ptr(base+idx, hashLog3)] = idx;
+        idx++;
+    }
+
+    return hashTable3[hash3];
+}
+
+
+/*-*************************************
+*  Binary Tree search
+***************************************/
+static U32 ZSTD_insertBtAndGetAllMatches (
+                        ZSTD_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iLimit,
+                        U32 nbCompares, const U32 mls,
+                        U32 extDict, ZSTD_match_t* matches, const U32 minMatchLen)
+{
+    const BYTE* const base = zc->base;
+    const U32 current = (U32)(ip-base);
+    const U32 hashLog = zc->params.cParams.hashLog;
+    const size_t h  = ZSTD_hashPtr(ip, hashLog, mls);
+    U32* const hashTable = zc->hashTable;
+    U32 matchIndex  = hashTable[h];
+    U32* const bt   = zc->chainTable;
+    const U32 btLog = zc->params.cParams.chainLog - 1;
+    const U32 btMask= (1U << btLog) - 1;
+    size_t commonLengthSmaller=0, commonLengthLarger=0;
+    const BYTE* const dictBase = zc->dictBase;
+    const U32 dictLimit = zc->dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const BYTE* const prefixStart = base + dictLimit;
+    const U32 btLow = btMask >= current ? 0 : current - btMask;
+    const U32 windowLow = zc->lowLimit;
+    U32* smallerPtr = bt + 2*(current&btMask);
+    U32* largerPtr  = bt + 2*(current&btMask) + 1;
+    U32 matchEndIdx = current+8;
+    U32 dummy32;   /* to be nullified at the end */
+    U32 mnum = 0;
+
+    const U32 minMatch = (mls == 3) ? 3 : 4;
+    size_t bestLength = minMatchLen-1;
+
+    if (minMatch == 3) { /* HC3 match finder */
+        U32 const matchIndex3 = ZSTD_insertAndFindFirstIndexHash3 (zc, ip);
+        if (matchIndex3>windowLow && (current - matchIndex3 < (1<<18))) {
+            const BYTE* match;
+            size_t currentMl=0;
+            if ((!extDict) || matchIndex3 >= dictLimit) {
+                match = base + matchIndex3;
+                if (match[bestLength] == ip[bestLength]) currentMl = ZSTD_count(ip, match, iLimit);
+            } else {
+                match = dictBase + matchIndex3;
+                if (MEM_readMINMATCH(match, MINMATCH) == MEM_readMINMATCH(ip, MINMATCH))    /* assumption : matchIndex3 <= dictLimit-4 (by table construction) */
+                    currentMl = ZSTD_count_2segments(ip+MINMATCH, match+MINMATCH, iLimit, dictEnd, prefixStart) + MINMATCH;
+            }
+
+            /* save best solution */
+            if (currentMl > bestLength) {
+                bestLength = currentMl;
+                matches[mnum].off = ZSTD_REP_MOVE_OPT + current - matchIndex3;
+                matches[mnum].len = (U32)currentMl;
+                mnum++;
+                if (currentMl > ZSTD_OPT_NUM) goto update;
+                if (ip+currentMl == iLimit) goto update; /* best possible, and avoid read overflow*/
+            }
+        }
+    }
+
+    hashTable[h] = current;   /* Update Hash Table */
+
+    while (nbCompares-- && (matchIndex > windowLow)) {
+        U32* nextPtr = bt + 2*(matchIndex & btMask);
+        size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
+        const BYTE* match;
+
+        if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
+            match = base + matchIndex;
+            if (match[matchLength] == ip[matchLength]) {
+                matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iLimit) +1;
+            }
+        } else {
+            match = dictBase + matchIndex;
+            matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iLimit, dictEnd, prefixStart);
+            if (matchIndex+matchLength >= dictLimit)
+                match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
+        }
+
+        if (matchLength > bestLength) {
+            if (matchLength > matchEndIdx - matchIndex) matchEndIdx = matchIndex + (U32)matchLength;
+            bestLength = matchLength;
+            matches[mnum].off = ZSTD_REP_MOVE_OPT + current - matchIndex;
+            matches[mnum].len = (U32)matchLength;
+            mnum++;
+            if (matchLength > ZSTD_OPT_NUM) break;
+            if (ip+matchLength == iLimit)   /* equal : no way to know if inf or sup */
+                break;   /* drop, to guarantee consistency (miss a little bit of compression) */
+        }
+
+        if (match[matchLength] < ip[matchLength]) {
+            /* match is smaller than current */
+            *smallerPtr = matchIndex;             /* update smaller idx */
+            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
+            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
+            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+        } else {
+            /* match is larger than current */
+            *largerPtr = matchIndex;
+            commonLengthLarger = matchLength;
+            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            largerPtr = nextPtr;
+            matchIndex = nextPtr[0];
+    }   }
+
+    *smallerPtr = *largerPtr = 0;
+
+update:
+    zc->nextToUpdate = (matchEndIdx > current + 8) ? matchEndIdx - 8 : current+1;
+    return mnum;
+}
+
+
+/** Tree updater, providing best match */
+static U32 ZSTD_BtGetAllMatches (
+                        ZSTD_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iLimit,
+                        const U32 maxNbAttempts, const U32 mls, ZSTD_match_t* matches, const U32 minMatchLen)
+{
+    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
+    ZSTD_updateTree(zc, ip, iLimit, maxNbAttempts, mls);
+    return ZSTD_insertBtAndGetAllMatches(zc, ip, iLimit, maxNbAttempts, mls, 0, matches, minMatchLen);
+}
+
+
+static U32 ZSTD_BtGetAllMatches_selectMLS (
+                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        const BYTE* ip, const BYTE* const iHighLimit,
+                        const U32 maxNbAttempts, const U32 matchLengthSearch, ZSTD_match_t* matches, const U32 minMatchLen)
+{
+    switch(matchLengthSearch)
+    {
+    case 3 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 3, matches, minMatchLen);
+    default :
+    case 4 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 4, matches, minMatchLen);
+    case 5 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 5, matches, minMatchLen);
+    case 6 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 6, matches, minMatchLen);
+    }
+}
+
+/** Tree updater, providing best match */
+static U32 ZSTD_BtGetAllMatches_extDict (
+                        ZSTD_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iLimit,
+                        const U32 maxNbAttempts, const U32 mls, ZSTD_match_t* matches, const U32 minMatchLen)
+{
+    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
+    ZSTD_updateTree_extDict(zc, ip, iLimit, maxNbAttempts, mls);
+    return ZSTD_insertBtAndGetAllMatches(zc, ip, iLimit, maxNbAttempts, mls, 1, matches, minMatchLen);
+}
+
+
+static U32 ZSTD_BtGetAllMatches_selectMLS_extDict (
+                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        const BYTE* ip, const BYTE* const iHighLimit,
+                        const U32 maxNbAttempts, const U32 matchLengthSearch, ZSTD_match_t* matches, const U32 minMatchLen)
+{
+    switch(matchLengthSearch)
+    {
+    case 3 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 3, matches, minMatchLen);
+    default :
+    case 4 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 4, matches, minMatchLen);
+    case 5 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 5, matches, minMatchLen);
+    case 6 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 6, matches, minMatchLen);
+    }
+}
+
+
+/*-*******************************
+*  Optimal parser
+*********************************/
+FORCE_INLINE
+void ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
+                                    const void* src, size_t srcSize, const int ultra)
+{
+    seqStore_t* seqStorePtr = &(ctx->seqStore);
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - 8;
+    const BYTE* const base = ctx->base;
+    const BYTE* const prefixStart = base + ctx->dictLimit;
+
+    const U32 maxSearches = 1U << ctx->params.cParams.searchLog;
+    const U32 sufficient_len = ctx->params.cParams.targetLength;
+    const U32 mls = ctx->params.cParams.searchLength;
+    const U32 minMatch = (ctx->params.cParams.searchLength == 3) ? 3 : 4;
+
+    ZSTD_optimal_t* opt = seqStorePtr->priceTable;
+    ZSTD_match_t* matches = seqStorePtr->matchTable;
+    const BYTE* inr;
+    U32 offset, rep[ZSTD_REP_NUM];
+
+    /* init */
+    ctx->nextToUpdate3 = ctx->nextToUpdate;
+    ZSTD_rescaleFreqs(seqStorePtr, (const BYTE*)src, srcSize);
+    ip += (ip==prefixStart);
+    { U32 i; for (i=0; irep[i]; }
+
+    /* Match Loop */
+    while (ip < ilimit) {
+        U32 cur, match_num, last_pos, litlen, price;
+        U32 u, mlen, best_mlen, best_off, litLength;
+        memset(opt, 0, sizeof(ZSTD_optimal_t));
+        last_pos = 0;
+        litlen = (U32)(ip - anchor);
+
+        /* check repCode */
+        {   U32 i, last_i = ZSTD_REP_CHECK + (ip==anchor);
+            for (i=(ip == anchor); i 0) && (repCur < (S32)(ip-prefixStart))
+                    && (MEM_readMINMATCH(ip, minMatch) == MEM_readMINMATCH(ip - repCur, minMatch))) {
+                    mlen = (U32)ZSTD_count(ip+minMatch, ip+minMatch-repCur, iend) + minMatch;
+                    if (mlen > sufficient_len || mlen >= ZSTD_OPT_NUM) {
+                        best_mlen = mlen; best_off = i; cur = 0; last_pos = 1;
+                        goto _storeSequence;
+                    }
+                    best_off = i - (ip == anchor);
+                    do {
+                        price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
+                        if (mlen > last_pos || price < opt[mlen].price)
+                            SET_PRICE(mlen, mlen, i, litlen, price);   /* note : macro modifies last_pos */
+                        mlen--;
+                    } while (mlen >= minMatch);
+        }   }   }
+
+        match_num = ZSTD_BtGetAllMatches_selectMLS(ctx, ip, iend, maxSearches, mls, matches, minMatch);
+
+        if (!last_pos && !match_num) { ip++; continue; }
+
+        if (match_num && (matches[match_num-1].len > sufficient_len || matches[match_num-1].len >= ZSTD_OPT_NUM)) {
+            best_mlen = matches[match_num-1].len;
+            best_off = matches[match_num-1].off;
+            cur = 0;
+            last_pos = 1;
+            goto _storeSequence;
+        }
+
+        /* set prices using matches at position = 0 */
+        best_mlen = (last_pos) ? last_pos : minMatch;
+        for (u = 0; u < match_num; u++) {
+            mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
+            best_mlen = matches[u].len;
+            while (mlen <= best_mlen) {
+                price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
+                if (mlen > last_pos || price < opt[mlen].price)
+                    SET_PRICE(mlen, mlen, matches[u].off, litlen, price);   /* note : macro modifies last_pos */
+                mlen++;
+        }   }
+
+        if (last_pos < minMatch) { ip++; continue; }
+
+        /* initialize opt[0] */
+        { U32 i ; for (i=0; i litlen) {
+                    price = opt[cur - litlen].price + ZSTD_getLiteralPrice(seqStorePtr, litlen, inr-litlen);
+                } else
+                    price = ZSTD_getLiteralPrice(seqStorePtr, litlen, anchor);
+           } else {
+                litlen = 1;
+                price = opt[cur - 1].price + ZSTD_getLiteralPrice(seqStorePtr, litlen, inr-1);
+           }
+
+           if (cur > last_pos || price <= opt[cur].price)
+                SET_PRICE(cur, 1, 0, litlen, price);
+
+           if (cur == last_pos) break;
+
+           if (inr > ilimit)  /* last match must start at a minimum distance of 8 from oend */
+               continue;
+
+           mlen = opt[cur].mlen;
+           if (opt[cur].off > ZSTD_REP_MOVE_OPT) {
+                opt[cur].rep[2] = opt[cur-mlen].rep[1];
+                opt[cur].rep[1] = opt[cur-mlen].rep[0];
+                opt[cur].rep[0] = opt[cur].off - ZSTD_REP_MOVE_OPT;
+           } else {
+                opt[cur].rep[2] = (opt[cur].off > 1) ? opt[cur-mlen].rep[1] : opt[cur-mlen].rep[2];
+                opt[cur].rep[1] = (opt[cur].off > 0) ? opt[cur-mlen].rep[0] : opt[cur-mlen].rep[1];
+                opt[cur].rep[0] = ((opt[cur].off==ZSTD_REP_MOVE_OPT) && (mlen != 1)) ? (opt[cur-mlen].rep[0] - 1) : (opt[cur-mlen].rep[opt[cur].off]);
+           }
+
+            best_mlen = minMatch;
+            {   U32 i, last_i = ZSTD_REP_CHECK + (mlen != 1);
+                for (i=(opt[cur].mlen != 1); i 0) && (repCur < (S32)(inr-prefixStart))
+                       && (MEM_readMINMATCH(inr, minMatch) == MEM_readMINMATCH(inr - repCur, minMatch))) {
+                       mlen = (U32)ZSTD_count(inr+minMatch, inr+minMatch - repCur, iend) + minMatch;
+
+                       if (mlen > sufficient_len || cur + mlen >= ZSTD_OPT_NUM) {
+                            best_mlen = mlen; best_off = i; last_pos = cur + 1;
+                            goto _storeSequence;
+                       }
+
+                       best_off = i - (opt[cur].mlen != 1);
+                       if (mlen > best_mlen) best_mlen = mlen;
+
+                       do {
+                           if (opt[cur].mlen == 1) {
+                                litlen = opt[cur].litlen;
+                                if (cur > litlen) {
+                                    price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, inr-litlen, best_off, mlen - MINMATCH, ultra);
+                                } else
+                                    price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
+                            } else {
+                                litlen = 0;
+                                price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, best_off, mlen - MINMATCH, ultra);
+                            }
+
+                            if (cur + mlen > last_pos || price <= opt[cur + mlen].price)
+                                SET_PRICE(cur + mlen, mlen, i, litlen, price);
+                            mlen--;
+                        } while (mlen >= minMatch);
+            }   }   }
+
+            match_num = ZSTD_BtGetAllMatches_selectMLS(ctx, inr, iend, maxSearches, mls, matches, best_mlen);
+
+            if (match_num > 0 && (matches[match_num-1].len > sufficient_len || cur + matches[match_num-1].len >= ZSTD_OPT_NUM)) {
+                best_mlen = matches[match_num-1].len;
+                best_off = matches[match_num-1].off;
+                last_pos = cur + 1;
+                goto _storeSequence;
+            }
+
+            /* set prices using matches at position = cur */
+            for (u = 0; u < match_num; u++) {
+                mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
+                best_mlen = matches[u].len;
+
+                while (mlen <= best_mlen) {
+                    if (opt[cur].mlen == 1) {
+                        litlen = opt[cur].litlen;
+                        if (cur > litlen)
+                            price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, ip+cur-litlen, matches[u].off-1, mlen - MINMATCH, ultra);
+                        else
+                            price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
+                    } else {
+                        litlen = 0;
+                        price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, matches[u].off-1, mlen - MINMATCH, ultra);
+                    }
+
+                    if (cur + mlen > last_pos || (price < opt[cur + mlen].price))
+                        SET_PRICE(cur + mlen, mlen, matches[u].off, litlen, price);
+
+                    mlen++;
+        }   }   }
+
+        best_mlen = opt[last_pos].mlen;
+        best_off = opt[last_pos].off;
+        cur = last_pos - best_mlen;
+
+        /* store sequence */
+_storeSequence:   /* cur, last_pos, best_mlen, best_off have to be set */
+        opt[0].mlen = 1;
+
+        while (1) {
+            mlen = opt[cur].mlen;
+            offset = opt[cur].off;
+            opt[cur].mlen = best_mlen;
+            opt[cur].off = best_off;
+            best_mlen = mlen;
+            best_off = offset;
+            if (mlen > cur) break;
+            cur -= mlen;
+        }
+
+        for (u = 0; u <= last_pos;) {
+            u += opt[u].mlen;
+        }
+
+        for (cur=0; cur < last_pos; ) {
+            mlen = opt[cur].mlen;
+            if (mlen == 1) { ip++; cur++; continue; }
+            offset = opt[cur].off;
+            cur += mlen;
+            litLength = (U32)(ip - anchor);
+
+            if (offset > ZSTD_REP_MOVE_OPT) {
+                rep[2] = rep[1];
+                rep[1] = rep[0];
+                rep[0] = offset - ZSTD_REP_MOVE_OPT;
+                offset--;
+            } else {
+                if (offset != 0) {
+                    best_off = (offset==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : (rep[offset]);
+                    if (offset != 1) rep[2] = rep[1];
+                    rep[1] = rep[0];
+                    rep[0] = best_off;
+                }
+                if (litLength==0) offset--;
+            }
+
+            ZSTD_updatePrice(seqStorePtr, litLength, anchor, offset, mlen-MINMATCH);
+            ZSTD_storeSeq(seqStorePtr, litLength, anchor, offset, mlen-MINMATCH);
+            anchor = ip = ip + mlen;
+    }    }   /* for (cur=0; cur < last_pos; ) */
+
+    /* Save reps for next block */
+    { int i; for (i=0; isavedRep[i] = rep[i]; }
+
+    /* Last Literals */
+    {   size_t const lastLLSize = iend - anchor;
+        memcpy(seqStorePtr->lit, anchor, lastLLSize);
+        seqStorePtr->lit += lastLLSize;
+    }
+}
+
+
+FORCE_INLINE
+void ZSTD_compressBlock_opt_extDict_generic(ZSTD_CCtx* ctx,
+                                     const void* src, size_t srcSize, const int ultra)
+{
+    seqStore_t* seqStorePtr = &(ctx->seqStore);
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - 8;
+    const BYTE* const base = ctx->base;
+    const U32 lowestIndex = ctx->lowLimit;
+    const U32 dictLimit = ctx->dictLimit;
+    const BYTE* const prefixStart = base + dictLimit;
+    const BYTE* const dictBase = ctx->dictBase;
+    const BYTE* const dictEnd  = dictBase + dictLimit;
+
+    const U32 maxSearches = 1U << ctx->params.cParams.searchLog;
+    const U32 sufficient_len = ctx->params.cParams.targetLength;
+    const U32 mls = ctx->params.cParams.searchLength;
+    const U32 minMatch = (ctx->params.cParams.searchLength == 3) ? 3 : 4;
+
+    ZSTD_optimal_t* opt = seqStorePtr->priceTable;
+    ZSTD_match_t* matches = seqStorePtr->matchTable;
+    const BYTE* inr;
+
+    /* init */
+    U32 offset, rep[ZSTD_REP_NUM];
+    { U32 i; for (i=0; irep[i]; }
+
+    ctx->nextToUpdate3 = ctx->nextToUpdate;
+    ZSTD_rescaleFreqs(seqStorePtr, (const BYTE*)src, srcSize);
+    ip += (ip==prefixStart);
+
+    /* Match Loop */
+    while (ip < ilimit) {
+        U32 cur, match_num, last_pos, litlen, price;
+        U32 u, mlen, best_mlen, best_off, litLength;
+        U32 current = (U32)(ip-base);
+        memset(opt, 0, sizeof(ZSTD_optimal_t));
+        last_pos = 0;
+        opt[0].litlen = (U32)(ip - anchor);
+
+        /* check repCode */
+        {   U32 i, last_i = ZSTD_REP_CHECK + (ip==anchor);
+            for (i = (ip==anchor); i 0 && repCur <= (S32)current)
+                   && (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex>lowestIndex))  /* intentional overflow */
+                   && (MEM_readMINMATCH(ip, minMatch) == MEM_readMINMATCH(repMatch, minMatch)) ) {
+                    /* repcode detected we should take it */
+                    const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+                    mlen = (U32)ZSTD_count_2segments(ip+minMatch, repMatch+minMatch, iend, repEnd, prefixStart) + minMatch;
+
+                    if (mlen > sufficient_len || mlen >= ZSTD_OPT_NUM) {
+                        best_mlen = mlen; best_off = i; cur = 0; last_pos = 1;
+                        goto _storeSequence;
+                    }
+
+                    best_off = i - (ip==anchor);
+                    litlen = opt[0].litlen;
+                    do {
+                        price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
+                        if (mlen > last_pos || price < opt[mlen].price)
+                            SET_PRICE(mlen, mlen, i, litlen, price);   /* note : macro modifies last_pos */
+                        mlen--;
+                    } while (mlen >= minMatch);
+        }   }   }
+
+        match_num = ZSTD_BtGetAllMatches_selectMLS_extDict(ctx, ip, iend, maxSearches, mls, matches, minMatch);  /* first search (depth 0) */
+
+        if (!last_pos && !match_num) { ip++; continue; }
+
+        { U32 i; for (i=0; i sufficient_len || matches[match_num-1].len >= ZSTD_OPT_NUM)) {
+            best_mlen = matches[match_num-1].len;
+            best_off = matches[match_num-1].off;
+            cur = 0;
+            last_pos = 1;
+            goto _storeSequence;
+        }
+
+        best_mlen = (last_pos) ? last_pos : minMatch;
+
+        /* set prices using matches at position = 0 */
+        for (u = 0; u < match_num; u++) {
+            mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
+            best_mlen = matches[u].len;
+            litlen = opt[0].litlen;
+            while (mlen <= best_mlen) {
+                price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
+                if (mlen > last_pos || price < opt[mlen].price)
+                    SET_PRICE(mlen, mlen, matches[u].off, litlen, price);
+                mlen++;
+        }   }
+
+        if (last_pos < minMatch) {
+            ip++; continue;
+        }
+
+        /* check further positions */
+        for (cur = 1; cur <= last_pos; cur++) {
+            inr = ip + cur;
+
+            if (opt[cur-1].mlen == 1) {
+                litlen = opt[cur-1].litlen + 1;
+                if (cur > litlen) {
+                    price = opt[cur - litlen].price + ZSTD_getLiteralPrice(seqStorePtr, litlen, inr-litlen);
+                } else
+                    price = ZSTD_getLiteralPrice(seqStorePtr, litlen, anchor);
+            } else {
+                litlen = 1;
+                price = opt[cur - 1].price + ZSTD_getLiteralPrice(seqStorePtr, litlen, inr-1);
+            }
+
+            if (cur > last_pos || price <= opt[cur].price)
+                SET_PRICE(cur, 1, 0, litlen, price);
+
+            if (cur == last_pos) break;
+
+            if (inr > ilimit)  /* last match must start at a minimum distance of 8 from oend */
+                continue;
+
+            mlen = opt[cur].mlen;
+            if (opt[cur].off > ZSTD_REP_MOVE_OPT) {
+                opt[cur].rep[2] = opt[cur-mlen].rep[1];
+                opt[cur].rep[1] = opt[cur-mlen].rep[0];
+                opt[cur].rep[0] = opt[cur].off - ZSTD_REP_MOVE_OPT;
+            } else {
+                opt[cur].rep[2] = (opt[cur].off > 1) ? opt[cur-mlen].rep[1] : opt[cur-mlen].rep[2];
+                opt[cur].rep[1] = (opt[cur].off > 0) ? opt[cur-mlen].rep[0] : opt[cur-mlen].rep[1];
+                opt[cur].rep[0] = ((opt[cur].off==ZSTD_REP_MOVE_OPT) && (mlen != 1)) ? (opt[cur-mlen].rep[0] - 1) : (opt[cur-mlen].rep[opt[cur].off]);
+            }
+
+            best_mlen = minMatch;
+            {   U32 i, last_i = ZSTD_REP_CHECK + (mlen != 1);
+                for (i = (mlen != 1); i 0 && repCur <= (S32)(current+cur))
+                      && (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex>lowestIndex))  /* intentional overflow */
+                      && (MEM_readMINMATCH(inr, minMatch) == MEM_readMINMATCH(repMatch, minMatch)) ) {
+                        /* repcode detected */
+                        const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+                        mlen = (U32)ZSTD_count_2segments(inr+minMatch, repMatch+minMatch, iend, repEnd, prefixStart) + minMatch;
+
+                        if (mlen > sufficient_len || cur + mlen >= ZSTD_OPT_NUM) {
+                            best_mlen = mlen; best_off = i; last_pos = cur + 1;
+                            goto _storeSequence;
+                        }
+
+                        best_off = i - (opt[cur].mlen != 1);
+                        if (mlen > best_mlen) best_mlen = mlen;
+
+                        do {
+                            if (opt[cur].mlen == 1) {
+                                litlen = opt[cur].litlen;
+                                if (cur > litlen) {
+                                    price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, inr-litlen, best_off, mlen - MINMATCH, ultra);
+                                } else
+                                    price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
+                            } else {
+                                litlen = 0;
+                                price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, best_off, mlen - MINMATCH, ultra);
+                            }
+
+                            if (cur + mlen > last_pos || price <= opt[cur + mlen].price)
+                                SET_PRICE(cur + mlen, mlen, i, litlen, price);
+                            mlen--;
+                        } while (mlen >= minMatch);
+            }   }   }
+
+            match_num = ZSTD_BtGetAllMatches_selectMLS_extDict(ctx, inr, iend, maxSearches, mls, matches, minMatch);
+
+            if (match_num > 0 && matches[match_num-1].len > sufficient_len) {
+                best_mlen = matches[match_num-1].len;
+                best_off = matches[match_num-1].off;
+                last_pos = cur + 1;
+                goto _storeSequence;
+            }
+
+            /* set prices using matches at position = cur */
+            for (u = 0; u < match_num; u++) {
+                mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
+                best_mlen = (cur + matches[u].len < ZSTD_OPT_NUM) ? matches[u].len : ZSTD_OPT_NUM - cur;
+
+                while (mlen <= best_mlen) {
+                    if (opt[cur].mlen == 1) {
+                        litlen = opt[cur].litlen;
+                        if (cur > litlen)
+                            price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, ip+cur-litlen, matches[u].off-1, mlen - MINMATCH, ultra);
+                        else
+                            price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
+                    } else {
+                        litlen = 0;
+                        price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, matches[u].off-1, mlen - MINMATCH, ultra);
+                    }
+
+                    if (cur + mlen > last_pos || (price < opt[cur + mlen].price))
+                        SET_PRICE(cur + mlen, mlen, matches[u].off, litlen, price);
+
+                    mlen++;
+        }   }   }   /* for (cur = 1; cur <= last_pos; cur++) */
+
+        best_mlen = opt[last_pos].mlen;
+        best_off = opt[last_pos].off;
+        cur = last_pos - best_mlen;
+
+        /* store sequence */
+_storeSequence:   /* cur, last_pos, best_mlen, best_off have to be set */
+        opt[0].mlen = 1;
+
+        while (1) {
+            mlen = opt[cur].mlen;
+            offset = opt[cur].off;
+            opt[cur].mlen = best_mlen;
+            opt[cur].off = best_off;
+            best_mlen = mlen;
+            best_off = offset;
+            if (mlen > cur) break;
+            cur -= mlen;
+        }
+
+        for (u = 0; u <= last_pos; ) {
+            u += opt[u].mlen;
+        }
+
+        for (cur=0; cur < last_pos; ) {
+            mlen = opt[cur].mlen;
+            if (mlen == 1) { ip++; cur++; continue; }
+            offset = opt[cur].off;
+            cur += mlen;
+            litLength = (U32)(ip - anchor);
+
+            if (offset > ZSTD_REP_MOVE_OPT) {
+                rep[2] = rep[1];
+                rep[1] = rep[0];
+                rep[0] = offset - ZSTD_REP_MOVE_OPT;
+                offset--;
+            } else {
+                if (offset != 0) {
+                    best_off = (offset==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : (rep[offset]);
+                    if (offset != 1) rep[2] = rep[1];
+                    rep[1] = rep[0];
+                    rep[0] = best_off;
+                }
+
+                if (litLength==0) offset--;
+            }
+
+            ZSTD_updatePrice(seqStorePtr, litLength, anchor, offset, mlen-MINMATCH);
+            ZSTD_storeSeq(seqStorePtr, litLength, anchor, offset, mlen-MINMATCH);
+            anchor = ip = ip + mlen;
+    }    }   /* for (cur=0; cur < last_pos; ) */
+
+    /* Save reps for next block */
+    { int i; for (i=0; isavedRep[i] = rep[i]; }
+
+    /* Last Literals */
+    {   size_t lastLLSize = iend - anchor;
+        memcpy(seqStorePtr->lit, anchor, lastLLSize);
+        seqStorePtr->lit += lastLLSize;
+    }
+}
+
+#endif  /* ZSTD_OPT_H_91842398743 */
diff --git a/lib/decompress/huf_decompress.c b/lib/decompress/huf_decompress.c
new file mode 100644
index 0000000..a342dfb
--- /dev/null
+++ b/lib/decompress/huf_decompress.c
@@ -0,0 +1,885 @@
+/* ******************************************************************
+   Huffman decoder, part of New Generation Entropy library
+   Copyright (C) 2013-2016, Yann Collet.
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+    You can contact the author at :
+    - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
+    - Public forum : https://groups.google.com/forum/#!forum/lz4c
+****************************************************************** */
+
+/* **************************************************************
+*  Compiler specifics
+****************************************************************/
+#if defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
+/* inline is defined */
+#elif defined(_MSC_VER) || defined(__GNUC__)
+#  define inline __inline
+#else
+#  define inline /* disable inline */
+#endif
+
+#ifdef _MSC_VER    /* Visual Studio */
+#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
+#endif
+
+
+/* **************************************************************
+*  Dependencies
+****************************************************************/
+#include      /* memcpy, memset */
+#include "bitstream.h"  /* BIT_* */
+#include "fse.h"        /* header compression */
+#define HUF_STATIC_LINKING_ONLY
+#include "huf.h"
+
+
+/* **************************************************************
+*  Error Management
+****************************************************************/
+#define HUF_STATIC_ASSERT(c) { enum { HUF_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
+
+
+/*-***************************/
+/*  generic DTableDesc       */
+/*-***************************/
+
+typedef struct { BYTE maxTableLog; BYTE tableType; BYTE tableLog; BYTE reserved; } DTableDesc;
+
+static DTableDesc HUF_getDTableDesc(const HUF_DTable* table)
+{
+    DTableDesc dtd;
+    memcpy(&dtd, table, sizeof(dtd));
+    return dtd;
+}
+
+
+/*-***************************/
+/*  single-symbol decoding   */
+/*-***************************/
+
+typedef struct { BYTE byte; BYTE nbBits; } HUF_DEltX2;   /* single-symbol decoding */
+
+size_t HUF_readDTableX2 (HUF_DTable* DTable, const void* src, size_t srcSize)
+{
+    BYTE huffWeight[HUF_SYMBOLVALUE_MAX + 1];
+    U32 rankVal[HUF_TABLELOG_ABSOLUTEMAX + 1];   /* large enough for values from 0 to 16 */
+    U32 tableLog = 0;
+    U32 nbSymbols = 0;
+    size_t iSize;
+    void* const dtPtr = DTable + 1;
+    HUF_DEltX2* const dt = (HUF_DEltX2*)dtPtr;
+
+    HUF_STATIC_ASSERT(sizeof(DTableDesc) == sizeof(HUF_DTable));
+    /* memset(huffWeight, 0, sizeof(huffWeight)); */   /* is not necessary, even though some analyzer complain ... */
+
+    iSize = HUF_readStats(huffWeight, HUF_SYMBOLVALUE_MAX + 1, rankVal, &nbSymbols, &tableLog, src, srcSize);
+    if (HUF_isError(iSize)) return iSize;
+
+    /* Table header */
+    {   DTableDesc dtd = HUF_getDTableDesc(DTable);
+        if (tableLog > (U32)(dtd.maxTableLog+1)) return ERROR(tableLog_tooLarge);   /* DTable too small, huffman tree cannot fit in */
+        dtd.tableType = 0;
+        dtd.tableLog = (BYTE)tableLog;
+        memcpy(DTable, &dtd, sizeof(dtd));
+    }
+
+    /* Prepare ranks */
+    {   U32 n, nextRankStart = 0;
+        for (n=1; n> 1;
+            U32 i;
+            HUF_DEltX2 D;
+            D.byte = (BYTE)n; D.nbBits = (BYTE)(tableLog + 1 - w);
+            for (i = rankVal[w]; i < rankVal[w] + length; i++)
+                dt[i] = D;
+            rankVal[w] += length;
+    }   }
+
+    return iSize;
+}
+
+
+static BYTE HUF_decodeSymbolX2(BIT_DStream_t* Dstream, const HUF_DEltX2* dt, const U32 dtLog)
+{
+    size_t const val = BIT_lookBitsFast(Dstream, dtLog); /* note : dtLog >= 1 */
+    BYTE const c = dt[val].byte;
+    BIT_skipBits(Dstream, dt[val].nbBits);
+    return c;
+}
+
+#define HUF_DECODE_SYMBOLX2_0(ptr, DStreamPtr) \
+    *ptr++ = HUF_decodeSymbolX2(DStreamPtr, dt, dtLog)
+
+#define HUF_DECODE_SYMBOLX2_1(ptr, DStreamPtr) \
+    if (MEM_64bits() || (HUF_TABLELOG_MAX<=12)) \
+        HUF_DECODE_SYMBOLX2_0(ptr, DStreamPtr)
+
+#define HUF_DECODE_SYMBOLX2_2(ptr, DStreamPtr) \
+    if (MEM_64bits()) \
+        HUF_DECODE_SYMBOLX2_0(ptr, DStreamPtr)
+
+static inline size_t HUF_decodeStreamX2(BYTE* p, BIT_DStream_t* const bitDPtr, BYTE* const pEnd, const HUF_DEltX2* const dt, const U32 dtLog)
+{
+    BYTE* const pStart = p;
+
+    /* up to 4 symbols at a time */
+    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) && (p <= pEnd-4)) {
+        HUF_DECODE_SYMBOLX2_2(p, bitDPtr);
+        HUF_DECODE_SYMBOLX2_1(p, bitDPtr);
+        HUF_DECODE_SYMBOLX2_2(p, bitDPtr);
+        HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+    }
+
+    /* closer to the end */
+    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) && (p < pEnd))
+        HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+
+    /* no more data to retrieve from bitstream, hence no need to reload */
+    while (p < pEnd)
+        HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+
+    return pEnd-pStart;
+}
+
+static size_t HUF_decompress1X2_usingDTable_internal(
+          void* dst,  size_t dstSize,
+    const void* cSrc, size_t cSrcSize,
+    const HUF_DTable* DTable)
+{
+    BYTE* op = (BYTE*)dst;
+    BYTE* const oend = op + dstSize;
+    const void* dtPtr = DTable + 1;
+    const HUF_DEltX2* const dt = (const HUF_DEltX2*)dtPtr;
+    BIT_DStream_t bitD;
+    DTableDesc const dtd = HUF_getDTableDesc(DTable);
+    U32 const dtLog = dtd.tableLog;
+
+    { size_t const errorCode = BIT_initDStream(&bitD, cSrc, cSrcSize);
+      if (HUF_isError(errorCode)) return errorCode; }
+
+    HUF_decodeStreamX2(op, &bitD, oend, dt, dtLog);
+
+    /* check */
+    if (!BIT_endOfDStream(&bitD)) return ERROR(corruption_detected);
+
+    return dstSize;
+}
+
+size_t HUF_decompress1X2_usingDTable(
+          void* dst,  size_t dstSize,
+    const void* cSrc, size_t cSrcSize,
+    const HUF_DTable* DTable)
+{
+    DTableDesc dtd = HUF_getDTableDesc(DTable);
+    if (dtd.tableType != 0) return ERROR(GENERIC);
+    return HUF_decompress1X2_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable);
+}
+
+size_t HUF_decompress1X2_DCtx (HUF_DTable* DCtx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    const BYTE* ip = (const BYTE*) cSrc;
+
+    size_t const hSize = HUF_readDTableX2 (DCtx, cSrc, cSrcSize);
+    if (HUF_isError(hSize)) return hSize;
+    if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
+    ip += hSize; cSrcSize -= hSize;
+
+    return HUF_decompress1X2_usingDTable_internal (dst, dstSize, ip, cSrcSize, DCtx);
+}
+
+size_t HUF_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_TABLELOG_MAX);
+    return HUF_decompress1X2_DCtx (DTable, dst, dstSize, cSrc, cSrcSize);
+}
+
+
+static size_t HUF_decompress4X2_usingDTable_internal(
+          void* dst,  size_t dstSize,
+    const void* cSrc, size_t cSrcSize,
+    const HUF_DTable* DTable)
+{
+    /* Check */
+    if (cSrcSize < 10) return ERROR(corruption_detected);  /* strict minimum : jump table + 1 byte per stream */
+
+    {   const BYTE* const istart = (const BYTE*) cSrc;
+        BYTE* const ostart = (BYTE*) dst;
+        BYTE* const oend = ostart + dstSize;
+        const void* const dtPtr = DTable + 1;
+        const HUF_DEltX2* const dt = (const HUF_DEltX2*)dtPtr;
+
+        /* Init */
+        BIT_DStream_t bitD1;
+        BIT_DStream_t bitD2;
+        BIT_DStream_t bitD3;
+        BIT_DStream_t bitD4;
+        size_t const length1 = MEM_readLE16(istart);
+        size_t const length2 = MEM_readLE16(istart+2);
+        size_t const length3 = MEM_readLE16(istart+4);
+        size_t const length4 = cSrcSize - (length1 + length2 + length3 + 6);
+        const BYTE* const istart1 = istart + 6;  /* jumpTable */
+        const BYTE* const istart2 = istart1 + length1;
+        const BYTE* const istart3 = istart2 + length2;
+        const BYTE* const istart4 = istart3 + length3;
+        const size_t segmentSize = (dstSize+3) / 4;
+        BYTE* const opStart2 = ostart + segmentSize;
+        BYTE* const opStart3 = opStart2 + segmentSize;
+        BYTE* const opStart4 = opStart3 + segmentSize;
+        BYTE* op1 = ostart;
+        BYTE* op2 = opStart2;
+        BYTE* op3 = opStart3;
+        BYTE* op4 = opStart4;
+        U32 endSignal;
+        DTableDesc const dtd = HUF_getDTableDesc(DTable);
+        U32 const dtLog = dtd.tableLog;
+
+        if (length4 > cSrcSize) return ERROR(corruption_detected);   /* overflow */
+        { size_t const errorCode = BIT_initDStream(&bitD1, istart1, length1);
+          if (HUF_isError(errorCode)) return errorCode; }
+        { size_t const errorCode = BIT_initDStream(&bitD2, istart2, length2);
+          if (HUF_isError(errorCode)) return errorCode; }
+        { size_t const errorCode = BIT_initDStream(&bitD3, istart3, length3);
+          if (HUF_isError(errorCode)) return errorCode; }
+        { size_t const errorCode = BIT_initDStream(&bitD4, istart4, length4);
+          if (HUF_isError(errorCode)) return errorCode; }
+
+        /* 16-32 symbols per loop (4-8 symbols per stream) */
+        endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
+        for ( ; (endSignal==BIT_DStream_unfinished) && (op4<(oend-7)) ; ) {
+            HUF_DECODE_SYMBOLX2_2(op1, &bitD1);
+            HUF_DECODE_SYMBOLX2_2(op2, &bitD2);
+            HUF_DECODE_SYMBOLX2_2(op3, &bitD3);
+            HUF_DECODE_SYMBOLX2_2(op4, &bitD4);
+            HUF_DECODE_SYMBOLX2_1(op1, &bitD1);
+            HUF_DECODE_SYMBOLX2_1(op2, &bitD2);
+            HUF_DECODE_SYMBOLX2_1(op3, &bitD3);
+            HUF_DECODE_SYMBOLX2_1(op4, &bitD4);
+            HUF_DECODE_SYMBOLX2_2(op1, &bitD1);
+            HUF_DECODE_SYMBOLX2_2(op2, &bitD2);
+            HUF_DECODE_SYMBOLX2_2(op3, &bitD3);
+            HUF_DECODE_SYMBOLX2_2(op4, &bitD4);
+            HUF_DECODE_SYMBOLX2_0(op1, &bitD1);
+            HUF_DECODE_SYMBOLX2_0(op2, &bitD2);
+            HUF_DECODE_SYMBOLX2_0(op3, &bitD3);
+            HUF_DECODE_SYMBOLX2_0(op4, &bitD4);
+            endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
+        }
+
+        /* check corruption */
+        if (op1 > opStart2) return ERROR(corruption_detected);
+        if (op2 > opStart3) return ERROR(corruption_detected);
+        if (op3 > opStart4) return ERROR(corruption_detected);
+        /* note : op4 supposed already verified within main loop */
+
+        /* finish bitStreams one by one */
+        HUF_decodeStreamX2(op1, &bitD1, opStart2, dt, dtLog);
+        HUF_decodeStreamX2(op2, &bitD2, opStart3, dt, dtLog);
+        HUF_decodeStreamX2(op3, &bitD3, opStart4, dt, dtLog);
+        HUF_decodeStreamX2(op4, &bitD4, oend,     dt, dtLog);
+
+        /* check */
+        endSignal = BIT_endOfDStream(&bitD1) & BIT_endOfDStream(&bitD2) & BIT_endOfDStream(&bitD3) & BIT_endOfDStream(&bitD4);
+        if (!endSignal) return ERROR(corruption_detected);
+
+        /* decoded size */
+        return dstSize;
+    }
+}
+
+
+size_t HUF_decompress4X2_usingDTable(
+          void* dst,  size_t dstSize,
+    const void* cSrc, size_t cSrcSize,
+    const HUF_DTable* DTable)
+{
+    DTableDesc dtd = HUF_getDTableDesc(DTable);
+    if (dtd.tableType != 0) return ERROR(GENERIC);
+    return HUF_decompress4X2_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable);
+}
+
+
+size_t HUF_decompress4X2_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    const BYTE* ip = (const BYTE*) cSrc;
+
+    size_t const hSize = HUF_readDTableX2 (dctx, cSrc, cSrcSize);
+    if (HUF_isError(hSize)) return hSize;
+    if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
+    ip += hSize; cSrcSize -= hSize;
+
+    return HUF_decompress4X2_usingDTable_internal (dst, dstSize, ip, cSrcSize, dctx);
+}
+
+size_t HUF_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_TABLELOG_MAX);
+    return HUF_decompress4X2_DCtx(DTable, dst, dstSize, cSrc, cSrcSize);
+}
+
+
+/* *************************/
+/* double-symbols decoding */
+/* *************************/
+typedef struct { U16 sequence; BYTE nbBits; BYTE length; } HUF_DEltX4;  /* double-symbols decoding */
+
+typedef struct { BYTE symbol; BYTE weight; } sortedSymbol_t;
+
+/* HUF_fillDTableX4Level2() :
+ * `rankValOrigin` must be a table of at least (HUF_TABLELOG_MAX + 1) U32 */
+static void HUF_fillDTableX4Level2(HUF_DEltX4* DTable, U32 sizeLog, const U32 consumed,
+                           const U32* rankValOrigin, const int minWeight,
+                           const sortedSymbol_t* sortedSymbols, const U32 sortedListSize,
+                           U32 nbBitsBaseline, U16 baseSeq)
+{
+    HUF_DEltX4 DElt;
+    U32 rankVal[HUF_TABLELOG_MAX + 1];
+
+    /* get pre-calculated rankVal */
+    memcpy(rankVal, rankValOrigin, sizeof(rankVal));
+
+    /* fill skipped values */
+    if (minWeight>1) {
+        U32 i, skipSize = rankVal[minWeight];
+        MEM_writeLE16(&(DElt.sequence), baseSeq);
+        DElt.nbBits   = (BYTE)(consumed);
+        DElt.length   = 1;
+        for (i = 0; i < skipSize; i++)
+            DTable[i] = DElt;
+    }
+
+    /* fill DTable */
+    {   U32 s; for (s=0; s= 1 */
+
+            rankVal[weight] += length;
+    }   }
+}
+
+typedef U32 rankVal_t[HUF_TABLELOG_MAX][HUF_TABLELOG_MAX + 1];
+
+static void HUF_fillDTableX4(HUF_DEltX4* DTable, const U32 targetLog,
+                           const sortedSymbol_t* sortedList, const U32 sortedListSize,
+                           const U32* rankStart, rankVal_t rankValOrigin, const U32 maxWeight,
+                           const U32 nbBitsBaseline)
+{
+    U32 rankVal[HUF_TABLELOG_MAX + 1];
+    const int scaleLog = nbBitsBaseline - targetLog;   /* note : targetLog >= srcLog, hence scaleLog <= 1 */
+    const U32 minBits  = nbBitsBaseline - maxWeight;
+    U32 s;
+
+    memcpy(rankVal, rankValOrigin, sizeof(rankVal));
+
+    /* fill DTable */
+    for (s=0; s= minBits) {   /* enough room for a second symbol */
+            U32 sortedRank;
+            int minWeight = nbBits + scaleLog;
+            if (minWeight < 1) minWeight = 1;
+            sortedRank = rankStart[minWeight];
+            HUF_fillDTableX4Level2(DTable+start, targetLog-nbBits, nbBits,
+                           rankValOrigin[nbBits], minWeight,
+                           sortedList+sortedRank, sortedListSize-sortedRank,
+                           nbBitsBaseline, symbol);
+        } else {
+            HUF_DEltX4 DElt;
+            MEM_writeLE16(&(DElt.sequence), symbol);
+            DElt.nbBits = (BYTE)(nbBits);
+            DElt.length = 1;
+            {   U32 const end = start + length;
+                U32 u;
+                for (u = start; u < end; u++) DTable[u] = DElt;
+        }   }
+        rankVal[weight] += length;
+    }
+}
+
+size_t HUF_readDTableX4 (HUF_DTable* DTable, const void* src, size_t srcSize)
+{
+    BYTE weightList[HUF_SYMBOLVALUE_MAX + 1];
+    sortedSymbol_t sortedSymbol[HUF_SYMBOLVALUE_MAX + 1];
+    U32 rankStats[HUF_TABLELOG_MAX + 1] = { 0 };
+    U32 rankStart0[HUF_TABLELOG_MAX + 2] = { 0 };
+    U32* const rankStart = rankStart0+1;
+    rankVal_t rankVal;
+    U32 tableLog, maxW, sizeOfSort, nbSymbols;
+    DTableDesc dtd = HUF_getDTableDesc(DTable);
+    U32 const maxTableLog = dtd.maxTableLog;
+    size_t iSize;
+    void* dtPtr = DTable+1;   /* force compiler to avoid strict-aliasing */
+    HUF_DEltX4* const dt = (HUF_DEltX4*)dtPtr;
+
+    HUF_STATIC_ASSERT(sizeof(HUF_DEltX4) == sizeof(HUF_DTable));   /* if compilation fails here, assertion is false */
+    if (maxTableLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
+    /* memset(weightList, 0, sizeof(weightList)); */  /* is not necessary, even though some analyzer complain ... */
+
+    iSize = HUF_readStats(weightList, HUF_SYMBOLVALUE_MAX + 1, rankStats, &nbSymbols, &tableLog, src, srcSize);
+    if (HUF_isError(iSize)) return iSize;
+
+    /* check result */
+    if (tableLog > maxTableLog) return ERROR(tableLog_tooLarge);   /* DTable can't fit code depth */
+
+    /* find maxWeight */
+    for (maxW = tableLog; rankStats[maxW]==0; maxW--) {}  /* necessarily finds a solution before 0 */
+
+    /* Get start index of each weight */
+    {   U32 w, nextRankStart = 0;
+        for (w=1; w> consumed;
+    }   }   }   }
+
+    HUF_fillDTableX4(dt, maxTableLog,
+                   sortedSymbol, sizeOfSort,
+                   rankStart0, rankVal, maxW,
+                   tableLog+1);
+
+    dtd.tableLog = (BYTE)maxTableLog;
+    dtd.tableType = 1;
+    memcpy(DTable, &dtd, sizeof(dtd));
+    return iSize;
+}
+
+
+static U32 HUF_decodeSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DEltX4* dt, const U32 dtLog)
+{
+    size_t const val = BIT_lookBitsFast(DStream, dtLog);   /* note : dtLog >= 1 */
+    memcpy(op, dt+val, 2);
+    BIT_skipBits(DStream, dt[val].nbBits);
+    return dt[val].length;
+}
+
+static U32 HUF_decodeLastSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DEltX4* dt, const U32 dtLog)
+{
+    size_t const val = BIT_lookBitsFast(DStream, dtLog);   /* note : dtLog >= 1 */
+    memcpy(op, dt+val, 1);
+    if (dt[val].length==1) BIT_skipBits(DStream, dt[val].nbBits);
+    else {
+        if (DStream->bitsConsumed < (sizeof(DStream->bitContainer)*8)) {
+            BIT_skipBits(DStream, dt[val].nbBits);
+            if (DStream->bitsConsumed > (sizeof(DStream->bitContainer)*8))
+                DStream->bitsConsumed = (sizeof(DStream->bitContainer)*8);   /* ugly hack; works only because it's the last symbol. Note : can't easily extract nbBits from just this symbol */
+    }   }
+    return 1;
+}
+
+
+#define HUF_DECODE_SYMBOLX4_0(ptr, DStreamPtr) \
+    ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
+
+#define HUF_DECODE_SYMBOLX4_1(ptr, DStreamPtr) \
+    if (MEM_64bits() || (HUF_TABLELOG_MAX<=12)) \
+        ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
+
+#define HUF_DECODE_SYMBOLX4_2(ptr, DStreamPtr) \
+    if (MEM_64bits()) \
+        ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
+
+static inline size_t HUF_decodeStreamX4(BYTE* p, BIT_DStream_t* bitDPtr, BYTE* const pEnd, const HUF_DEltX4* const dt, const U32 dtLog)
+{
+    BYTE* const pStart = p;
+
+    /* up to 8 symbols at a time */
+    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) & (p < pEnd-(sizeof(bitDPtr->bitContainer)-1))) {
+        HUF_DECODE_SYMBOLX4_2(p, bitDPtr);
+        HUF_DECODE_SYMBOLX4_1(p, bitDPtr);
+        HUF_DECODE_SYMBOLX4_2(p, bitDPtr);
+        HUF_DECODE_SYMBOLX4_0(p, bitDPtr);
+    }
+
+    /* closer to end : up to 2 symbols at a time */
+    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) & (p <= pEnd-2))
+        HUF_DECODE_SYMBOLX4_0(p, bitDPtr);
+
+    while (p <= pEnd-2)
+        HUF_DECODE_SYMBOLX4_0(p, bitDPtr);   /* no need to reload : reached the end of DStream */
+
+    if (p < pEnd)
+        p += HUF_decodeLastSymbolX4(p, bitDPtr, dt, dtLog);
+
+    return p-pStart;
+}
+
+
+static size_t HUF_decompress1X4_usingDTable_internal(
+          void* dst,  size_t dstSize,
+    const void* cSrc, size_t cSrcSize,
+    const HUF_DTable* DTable)
+{
+    BIT_DStream_t bitD;
+
+    /* Init */
+    {   size_t const errorCode = BIT_initDStream(&bitD, cSrc, cSrcSize);
+        if (HUF_isError(errorCode)) return errorCode;
+    }
+
+    /* decode */
+    {   BYTE* const ostart = (BYTE*) dst;
+        BYTE* const oend = ostart + dstSize;
+        const void* const dtPtr = DTable+1;   /* force compiler to not use strict-aliasing */
+        const HUF_DEltX4* const dt = (const HUF_DEltX4*)dtPtr;
+        DTableDesc const dtd = HUF_getDTableDesc(DTable);
+        HUF_decodeStreamX4(ostart, &bitD, oend, dt, dtd.tableLog);
+    }
+
+    /* check */
+    if (!BIT_endOfDStream(&bitD)) return ERROR(corruption_detected);
+
+    /* decoded size */
+    return dstSize;
+}
+
+size_t HUF_decompress1X4_usingDTable(
+          void* dst,  size_t dstSize,
+    const void* cSrc, size_t cSrcSize,
+    const HUF_DTable* DTable)
+{
+    DTableDesc dtd = HUF_getDTableDesc(DTable);
+    if (dtd.tableType != 1) return ERROR(GENERIC);
+    return HUF_decompress1X4_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable);
+}
+
+size_t HUF_decompress1X4_DCtx (HUF_DTable* DCtx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    const BYTE* ip = (const BYTE*) cSrc;
+
+    size_t const hSize = HUF_readDTableX4 (DCtx, cSrc, cSrcSize);
+    if (HUF_isError(hSize)) return hSize;
+    if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
+    ip += hSize; cSrcSize -= hSize;
+
+    return HUF_decompress1X4_usingDTable_internal (dst, dstSize, ip, cSrcSize, DCtx);
+}
+
+size_t HUF_decompress1X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    HUF_CREATE_STATIC_DTABLEX4(DTable, HUF_TABLELOG_MAX);
+    return HUF_decompress1X4_DCtx(DTable, dst, dstSize, cSrc, cSrcSize);
+}
+
+static size_t HUF_decompress4X4_usingDTable_internal(
+          void* dst,  size_t dstSize,
+    const void* cSrc, size_t cSrcSize,
+    const HUF_DTable* DTable)
+{
+    if (cSrcSize < 10) return ERROR(corruption_detected);   /* strict minimum : jump table + 1 byte per stream */
+
+    {   const BYTE* const istart = (const BYTE*) cSrc;
+        BYTE* const ostart = (BYTE*) dst;
+        BYTE* const oend = ostart + dstSize;
+        const void* const dtPtr = DTable+1;
+        const HUF_DEltX4* const dt = (const HUF_DEltX4*)dtPtr;
+
+        /* Init */
+        BIT_DStream_t bitD1;
+        BIT_DStream_t bitD2;
+        BIT_DStream_t bitD3;
+        BIT_DStream_t bitD4;
+        size_t const length1 = MEM_readLE16(istart);
+        size_t const length2 = MEM_readLE16(istart+2);
+        size_t const length3 = MEM_readLE16(istart+4);
+        size_t const length4 = cSrcSize - (length1 + length2 + length3 + 6);
+        const BYTE* const istart1 = istart + 6;  /* jumpTable */
+        const BYTE* const istart2 = istart1 + length1;
+        const BYTE* const istart3 = istart2 + length2;
+        const BYTE* const istart4 = istart3 + length3;
+        size_t const segmentSize = (dstSize+3) / 4;
+        BYTE* const opStart2 = ostart + segmentSize;
+        BYTE* const opStart3 = opStart2 + segmentSize;
+        BYTE* const opStart4 = opStart3 + segmentSize;
+        BYTE* op1 = ostart;
+        BYTE* op2 = opStart2;
+        BYTE* op3 = opStart3;
+        BYTE* op4 = opStart4;
+        U32 endSignal;
+        DTableDesc const dtd = HUF_getDTableDesc(DTable);
+        U32 const dtLog = dtd.tableLog;
+
+        if (length4 > cSrcSize) return ERROR(corruption_detected);   /* overflow */
+        { size_t const errorCode = BIT_initDStream(&bitD1, istart1, length1);
+          if (HUF_isError(errorCode)) return errorCode; }
+        { size_t const errorCode = BIT_initDStream(&bitD2, istart2, length2);
+          if (HUF_isError(errorCode)) return errorCode; }
+        { size_t const errorCode = BIT_initDStream(&bitD3, istart3, length3);
+          if (HUF_isError(errorCode)) return errorCode; }
+        { size_t const errorCode = BIT_initDStream(&bitD4, istart4, length4);
+          if (HUF_isError(errorCode)) return errorCode; }
+
+        /* 16-32 symbols per loop (4-8 symbols per stream) */
+        endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
+        for ( ; (endSignal==BIT_DStream_unfinished) & (op4<(oend-(sizeof(bitD4.bitContainer)-1))) ; ) {
+            HUF_DECODE_SYMBOLX4_2(op1, &bitD1);
+            HUF_DECODE_SYMBOLX4_2(op2, &bitD2);
+            HUF_DECODE_SYMBOLX4_2(op3, &bitD3);
+            HUF_DECODE_SYMBOLX4_2(op4, &bitD4);
+            HUF_DECODE_SYMBOLX4_1(op1, &bitD1);
+            HUF_DECODE_SYMBOLX4_1(op2, &bitD2);
+            HUF_DECODE_SYMBOLX4_1(op3, &bitD3);
+            HUF_DECODE_SYMBOLX4_1(op4, &bitD4);
+            HUF_DECODE_SYMBOLX4_2(op1, &bitD1);
+            HUF_DECODE_SYMBOLX4_2(op2, &bitD2);
+            HUF_DECODE_SYMBOLX4_2(op3, &bitD3);
+            HUF_DECODE_SYMBOLX4_2(op4, &bitD4);
+            HUF_DECODE_SYMBOLX4_0(op1, &bitD1);
+            HUF_DECODE_SYMBOLX4_0(op2, &bitD2);
+            HUF_DECODE_SYMBOLX4_0(op3, &bitD3);
+            HUF_DECODE_SYMBOLX4_0(op4, &bitD4);
+
+            endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
+        }
+
+        /* check corruption */
+        if (op1 > opStart2) return ERROR(corruption_detected);
+        if (op2 > opStart3) return ERROR(corruption_detected);
+        if (op3 > opStart4) return ERROR(corruption_detected);
+        /* note : op4 already verified within main loop */
+
+        /* finish bitStreams one by one */
+        HUF_decodeStreamX4(op1, &bitD1, opStart2, dt, dtLog);
+        HUF_decodeStreamX4(op2, &bitD2, opStart3, dt, dtLog);
+        HUF_decodeStreamX4(op3, &bitD3, opStart4, dt, dtLog);
+        HUF_decodeStreamX4(op4, &bitD4, oend,     dt, dtLog);
+
+        /* check */
+        { U32 const endCheck = BIT_endOfDStream(&bitD1) & BIT_endOfDStream(&bitD2) & BIT_endOfDStream(&bitD3) & BIT_endOfDStream(&bitD4);
+          if (!endCheck) return ERROR(corruption_detected); }
+
+        /* decoded size */
+        return dstSize;
+    }
+}
+
+
+size_t HUF_decompress4X4_usingDTable(
+          void* dst,  size_t dstSize,
+    const void* cSrc, size_t cSrcSize,
+    const HUF_DTable* DTable)
+{
+    DTableDesc dtd = HUF_getDTableDesc(DTable);
+    if (dtd.tableType != 1) return ERROR(GENERIC);
+    return HUF_decompress4X4_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable);
+}
+
+
+size_t HUF_decompress4X4_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    const BYTE* ip = (const BYTE*) cSrc;
+
+    size_t hSize = HUF_readDTableX4 (dctx, cSrc, cSrcSize);
+    if (HUF_isError(hSize)) return hSize;
+    if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
+    ip += hSize; cSrcSize -= hSize;
+
+    return HUF_decompress4X4_usingDTable_internal(dst, dstSize, ip, cSrcSize, dctx);
+}
+
+size_t HUF_decompress4X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    HUF_CREATE_STATIC_DTABLEX4(DTable, HUF_TABLELOG_MAX);
+    return HUF_decompress4X4_DCtx(DTable, dst, dstSize, cSrc, cSrcSize);
+}
+
+
+/* ********************************/
+/* Generic decompression selector */
+/* ********************************/
+
+size_t HUF_decompress1X_usingDTable(void* dst, size_t maxDstSize,
+                                    const void* cSrc, size_t cSrcSize,
+                                    const HUF_DTable* DTable)
+{
+    DTableDesc const dtd = HUF_getDTableDesc(DTable);
+    return dtd.tableType ? HUF_decompress1X4_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable) :
+                           HUF_decompress1X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable);
+}
+
+size_t HUF_decompress4X_usingDTable(void* dst, size_t maxDstSize,
+                                    const void* cSrc, size_t cSrcSize,
+                                    const HUF_DTable* DTable)
+{
+    DTableDesc const dtd = HUF_getDTableDesc(DTable);
+    return dtd.tableType ? HUF_decompress4X4_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable) :
+                           HUF_decompress4X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable);
+}
+
+
+typedef struct { U32 tableTime; U32 decode256Time; } algo_time_t;
+static const algo_time_t algoTime[16 /* Quantization */][3 /* single, double, quad */] =
+{
+    /* single, double, quad */
+    {{0,0}, {1,1}, {2,2}},  /* Q==0 : impossible */
+    {{0,0}, {1,1}, {2,2}},  /* Q==1 : impossible */
+    {{  38,130}, {1313, 74}, {2151, 38}},   /* Q == 2 : 12-18% */
+    {{ 448,128}, {1353, 74}, {2238, 41}},   /* Q == 3 : 18-25% */
+    {{ 556,128}, {1353, 74}, {2238, 47}},   /* Q == 4 : 25-32% */
+    {{ 714,128}, {1418, 74}, {2436, 53}},   /* Q == 5 : 32-38% */
+    {{ 883,128}, {1437, 74}, {2464, 61}},   /* Q == 6 : 38-44% */
+    {{ 897,128}, {1515, 75}, {2622, 68}},   /* Q == 7 : 44-50% */
+    {{ 926,128}, {1613, 75}, {2730, 75}},   /* Q == 8 : 50-56% */
+    {{ 947,128}, {1729, 77}, {3359, 77}},   /* Q == 9 : 56-62% */
+    {{1107,128}, {2083, 81}, {4006, 84}},   /* Q ==10 : 62-69% */
+    {{1177,128}, {2379, 87}, {4785, 88}},   /* Q ==11 : 69-75% */
+    {{1242,128}, {2415, 93}, {5155, 84}},   /* Q ==12 : 75-81% */
+    {{1349,128}, {2644,106}, {5260,106}},   /* Q ==13 : 81-87% */
+    {{1455,128}, {2422,124}, {4174,124}},   /* Q ==14 : 87-93% */
+    {{ 722,128}, {1891,145}, {1936,146}},   /* Q ==15 : 93-99% */
+};
+
+/** HUF_selectDecoder() :
+*   Tells which decoder is likely to decode faster,
+*   based on a set of pre-determined metrics.
+*   @return : 0==HUF_decompress4X2, 1==HUF_decompress4X4 .
+*   Assumption : 0 < cSrcSize < dstSize <= 128 KB */
+U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize)
+{
+    /* decoder timing evaluation */
+    U32 const Q = (U32)(cSrcSize * 16 / dstSize);   /* Q < 16 since dstSize > cSrcSize */
+    U32 const D256 = (U32)(dstSize >> 8);
+    U32 const DTime0 = algoTime[Q][0].tableTime + (algoTime[Q][0].decode256Time * D256);
+    U32 DTime1 = algoTime[Q][1].tableTime + (algoTime[Q][1].decode256Time * D256);
+    DTime1 += DTime1 >> 3;  /* advantage to algorithm using less memory, for cache eviction */
+
+    return DTime1 < DTime0;
+}
+
+
+typedef size_t (*decompressionAlgo)(void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);
+
+size_t HUF_decompress (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    static const decompressionAlgo decompress[2] = { HUF_decompress4X2, HUF_decompress4X4 };
+
+    /* validation checks */
+    if (dstSize == 0) return ERROR(dstSize_tooSmall);
+    if (cSrcSize > dstSize) return ERROR(corruption_detected);   /* invalid */
+    if (cSrcSize == dstSize) { memcpy(dst, cSrc, dstSize); return dstSize; }   /* not compressed */
+    if (cSrcSize == 1) { memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; }   /* RLE */
+
+    {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
+        return decompress[algoNb](dst, dstSize, cSrc, cSrcSize);
+    }
+}
+
+size_t HUF_decompress4X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    /* validation checks */
+    if (dstSize == 0) return ERROR(dstSize_tooSmall);
+    if (cSrcSize > dstSize) return ERROR(corruption_detected);   /* invalid */
+    if (cSrcSize == dstSize) { memcpy(dst, cSrc, dstSize); return dstSize; }   /* not compressed */
+    if (cSrcSize == 1) { memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; }   /* RLE */
+
+    {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
+        return algoNb ? HUF_decompress4X4_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) :
+                        HUF_decompress4X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) ;
+    }
+}
+
+size_t HUF_decompress4X_hufOnly (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    /* validation checks */
+    if (dstSize == 0) return ERROR(dstSize_tooSmall);
+    if ((cSrcSize >= dstSize) || (cSrcSize <= 1)) return ERROR(corruption_detected);   /* invalid */
+
+    {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
+        return algoNb ? HUF_decompress4X4_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) :
+                        HUF_decompress4X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) ;
+    }
+}
+
+size_t HUF_decompress1X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    /* validation checks */
+    if (dstSize == 0) return ERROR(dstSize_tooSmall);
+    if (cSrcSize > dstSize) return ERROR(corruption_detected);   /* invalid */
+    if (cSrcSize == dstSize) { memcpy(dst, cSrc, dstSize); return dstSize; }   /* not compressed */
+    if (cSrcSize == 1) { memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; }   /* RLE */
+
+    {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
+        return algoNb ? HUF_decompress1X4_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) :
+                        HUF_decompress1X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) ;
+    }
+}
diff --git a/lib/decompress/zstd_decompress.c b/lib/decompress/zstd_decompress.c
new file mode 100644
index 0000000..085477b
--- /dev/null
+++ b/lib/decompress/zstd_decompress.c
@@ -0,0 +1,2154 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+
+/* ***************************************************************
+*  Tuning parameters
+*****************************************************************/
+/*!
+ * HEAPMODE :
+ * Select how default decompression function ZSTD_decompress() will allocate memory,
+ * in memory stack (0), or in memory heap (1, requires malloc())
+ */
+#ifndef ZSTD_HEAPMODE
+#  define ZSTD_HEAPMODE 1
+#endif
+
+/*!
+*  LEGACY_SUPPORT :
+*  if set to 1, ZSTD_decompress() can decode older formats (v0.1+)
+*/
+#ifndef ZSTD_LEGACY_SUPPORT
+#  define ZSTD_LEGACY_SUPPORT 0
+#endif
+
+/*!
+*  MAXWINDOWSIZE_DEFAULT :
+*  maximum window size accepted by DStream, by default.
+*  Frames requiring more memory will be rejected.
+*/
+#ifndef ZSTD_MAXWINDOWSIZE_DEFAULT
+#  define ZSTD_MAXWINDOWSIZE_DEFAULT ((1 << ZSTD_WINDOWLOG_MAX) + 1)   /* defined within zstd.h */
+#endif
+
+
+/*-*******************************************************
+*  Dependencies
+*********************************************************/
+#include       /* memcpy, memmove, memset */
+#include "mem.h"         /* low level memory routines */
+#define XXH_STATIC_LINKING_ONLY   /* XXH64_state_t */
+#include "xxhash.h"      /* XXH64_* */
+#define FSE_STATIC_LINKING_ONLY
+#include "fse.h"
+#define HUF_STATIC_LINKING_ONLY
+#include "huf.h"
+#include "zstd_internal.h"
+
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
+#  include "zstd_legacy.h"
+#endif
+
+
+#if defined(_MSC_VER)
+#  include    /* https://msdn.microsoft.com/fr-fr/library/84szxsww(v=vs.90).aspx */
+#  define ZSTD_PREFETCH(ptr)   _mm_prefetch((const char*)ptr, _MM_HINT_T0)
+#elif defined(__GNUC__)
+#  define ZSTD_PREFETCH(ptr)   __builtin_prefetch(ptr, 0, 0)
+#else
+#  define ZSTD_PREFETCH(ptr)   /* disabled */
+#endif
+
+/*-*************************************
+*  Macros
+***************************************/
+#define ZSTD_isError ERR_isError   /* for inlining */
+#define FSE_isError  ERR_isError
+#define HUF_isError  ERR_isError
+
+
+/*_*******************************************************
+*  Memory operations
+**********************************************************/
+static void ZSTD_copy4(void* dst, const void* src) { memcpy(dst, src, 4); }
+
+
+/*-*************************************************************
+*   Context management
+***************************************************************/
+typedef enum { ZSTDds_getFrameHeaderSize, ZSTDds_decodeFrameHeader,
+               ZSTDds_decodeBlockHeader, ZSTDds_decompressBlock,
+               ZSTDds_decompressLastBlock, ZSTDds_checkChecksum,
+               ZSTDds_decodeSkippableHeader, ZSTDds_skipFrame } ZSTD_dStage;
+
+struct ZSTD_DCtx_s
+{
+    const FSE_DTable* LLTptr;
+    const FSE_DTable* MLTptr;
+    const FSE_DTable* OFTptr;
+    const HUF_DTable* HUFptr;
+    FSE_DTable LLTable[FSE_DTABLE_SIZE_U32(LLFSELog)];
+    FSE_DTable OFTable[FSE_DTABLE_SIZE_U32(OffFSELog)];
+    FSE_DTable MLTable[FSE_DTABLE_SIZE_U32(MLFSELog)];
+    HUF_DTable hufTable[HUF_DTABLE_SIZE(HufLog)];  /* can accommodate HUF_decompress4X */
+    const void* previousDstEnd;
+    const void* base;
+    const void* vBase;
+    const void* dictEnd;
+    size_t expected;
+    U32 rep[ZSTD_REP_NUM];
+    ZSTD_frameParams fParams;
+    blockType_e bType;   /* used in ZSTD_decompressContinue(), to transfer blockType between header decoding and block decoding stages */
+    ZSTD_dStage stage;
+    U32 litEntropy;
+    U32 fseEntropy;
+    XXH64_state_t xxhState;
+    size_t headerSize;
+    U32 dictID;
+    const BYTE* litPtr;
+    ZSTD_customMem customMem;
+    size_t litSize;
+    size_t rleSize;
+    BYTE litBuffer[ZSTD_BLOCKSIZE_ABSOLUTEMAX + WILDCOPY_OVERLENGTH];
+    BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX];
+};  /* typedef'd to ZSTD_DCtx within "zstd.h" */
+
+size_t ZSTD_sizeof_DCtx (const ZSTD_DCtx* dctx) { return (dctx==NULL) ? 0 : sizeof(ZSTD_DCtx); }
+
+size_t ZSTD_estimateDCtxSize(void) { return sizeof(ZSTD_DCtx); }
+
+size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx)
+{
+    dctx->expected = ZSTD_frameHeaderSize_prefix;
+    dctx->stage = ZSTDds_getFrameHeaderSize;
+    dctx->previousDstEnd = NULL;
+    dctx->base = NULL;
+    dctx->vBase = NULL;
+    dctx->dictEnd = NULL;
+    dctx->hufTable[0] = (HUF_DTable)((HufLog)*0x1000001);  /* cover both little and big endian */
+    dctx->litEntropy = dctx->fseEntropy = 0;
+    dctx->dictID = 0;
+    MEM_STATIC_ASSERT(sizeof(dctx->rep) == sizeof(repStartValue));
+    memcpy(dctx->rep, repStartValue, sizeof(repStartValue));  /* initial repcodes */
+    dctx->LLTptr = dctx->LLTable;
+    dctx->MLTptr = dctx->MLTable;
+    dctx->OFTptr = dctx->OFTable;
+    dctx->HUFptr = dctx->hufTable;
+    return 0;
+}
+
+ZSTD_DCtx* ZSTD_createDCtx_advanced(ZSTD_customMem customMem)
+{
+    ZSTD_DCtx* dctx;
+
+    if (!customMem.customAlloc && !customMem.customFree) customMem = defaultCustomMem;
+    if (!customMem.customAlloc || !customMem.customFree) return NULL;
+
+    dctx = (ZSTD_DCtx*)ZSTD_malloc(sizeof(ZSTD_DCtx), customMem);
+    if (!dctx) return NULL;
+    memcpy(&dctx->customMem, &customMem, sizeof(customMem));
+    ZSTD_decompressBegin(dctx);
+    return dctx;
+}
+
+ZSTD_DCtx* ZSTD_createDCtx(void)
+{
+    return ZSTD_createDCtx_advanced(defaultCustomMem);
+}
+
+size_t ZSTD_freeDCtx(ZSTD_DCtx* dctx)
+{
+    if (dctx==NULL) return 0;   /* support free on NULL */
+    ZSTD_free(dctx, dctx->customMem);
+    return 0;   /* reserved as a potential error code in the future */
+}
+
+void ZSTD_copyDCtx(ZSTD_DCtx* dstDCtx, const ZSTD_DCtx* srcDCtx)
+{
+    size_t const workSpaceSize = (ZSTD_BLOCKSIZE_ABSOLUTEMAX+WILDCOPY_OVERLENGTH) + ZSTD_frameHeaderSize_max;
+    memcpy(dstDCtx, srcDCtx, sizeof(ZSTD_DCtx) - workSpaceSize);  /* no need to copy workspace */
+}
+
+static void ZSTD_refDCtx(ZSTD_DCtx* dstDCtx, const ZSTD_DCtx* srcDCtx)
+{
+    ZSTD_decompressBegin(dstDCtx);  /* init */
+    if (srcDCtx) {   /* support refDCtx on NULL */
+        dstDCtx->dictEnd = srcDCtx->dictEnd;
+        dstDCtx->vBase = srcDCtx->vBase;
+        dstDCtx->base = srcDCtx->base;
+        dstDCtx->previousDstEnd = srcDCtx->previousDstEnd;
+        dstDCtx->dictID = srcDCtx->dictID;
+        dstDCtx->litEntropy = srcDCtx->litEntropy;
+        dstDCtx->fseEntropy = srcDCtx->fseEntropy;
+        dstDCtx->LLTptr = srcDCtx->LLTable;
+        dstDCtx->MLTptr = srcDCtx->MLTable;
+        dstDCtx->OFTptr = srcDCtx->OFTable;
+        dstDCtx->HUFptr = srcDCtx->hufTable;
+        dstDCtx->rep[0] = srcDCtx->rep[0];
+        dstDCtx->rep[1] = srcDCtx->rep[1];
+        dstDCtx->rep[2] = srcDCtx->rep[2];
+    }
+}
+
+
+/*-*************************************************************
+*   Decompression section
+***************************************************************/
+
+/*! ZSTD_isFrame() :
+ *  Tells if the content of `buffer` starts with a valid Frame Identifier.
+ *  Note : Frame Identifier is 4 bytes. If `size < 4`, @return will always be 0.
+ *  Note 2 : Legacy Frame Identifiers are considered valid only if Legacy Support is enabled.
+ *  Note 3 : Skippable Frame Identifiers are considered valid. */
+unsigned ZSTD_isFrame(const void* buffer, size_t size)
+{
+    if (size < 4) return 0;
+    {   U32 const magic = MEM_readLE32(buffer);
+        if (magic == ZSTD_MAGICNUMBER) return 1;
+        if ((magic & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) return 1;
+    }
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
+    if (ZSTD_isLegacy(buffer, size)) return 1;
+#endif
+    return 0;
+}
+
+
+/** ZSTD_frameHeaderSize() :
+*   srcSize must be >= ZSTD_frameHeaderSize_prefix.
+*   @return : size of the Frame Header */
+static size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize)
+{
+    if (srcSize < ZSTD_frameHeaderSize_prefix) return ERROR(srcSize_wrong);
+    {   BYTE const fhd = ((const BYTE*)src)[4];
+        U32 const dictID= fhd & 3;
+        U32 const singleSegment = (fhd >> 5) & 1;
+        U32 const fcsId = fhd >> 6;
+        return ZSTD_frameHeaderSize_prefix + !singleSegment + ZSTD_did_fieldSize[dictID] + ZSTD_fcs_fieldSize[fcsId]
+                + (singleSegment && !fcsId);
+    }
+}
+
+
+/** ZSTD_getFrameParams() :
+*   decode Frame Header, or require larger `srcSize`.
+*   @return : 0, `fparamsPtr` is correctly filled,
+*            >0, `srcSize` is too small, result is expected `srcSize`,
+*             or an error code, which can be tested using ZSTD_isError() */
+size_t ZSTD_getFrameParams(ZSTD_frameParams* fparamsPtr, const void* src, size_t srcSize)
+{
+    const BYTE* ip = (const BYTE*)src;
+
+    if (srcSize < ZSTD_frameHeaderSize_prefix) return ZSTD_frameHeaderSize_prefix;
+    if (MEM_readLE32(src) != ZSTD_MAGICNUMBER) {
+        if ((MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {
+            if (srcSize < ZSTD_skippableHeaderSize) return ZSTD_skippableHeaderSize; /* magic number + skippable frame length */
+            memset(fparamsPtr, 0, sizeof(*fparamsPtr));
+            fparamsPtr->frameContentSize = MEM_readLE32((const char *)src + 4);
+            fparamsPtr->windowSize = 0; /* windowSize==0 means a frame is skippable */
+            return 0;
+        }
+        return ERROR(prefix_unknown);
+    }
+
+    /* ensure there is enough `srcSize` to fully read/decode frame header */
+    { size_t const fhsize = ZSTD_frameHeaderSize(src, srcSize);
+      if (srcSize < fhsize) return fhsize; }
+
+    {   BYTE const fhdByte = ip[4];
+        size_t pos = 5;
+        U32 const dictIDSizeCode = fhdByte&3;
+        U32 const checksumFlag = (fhdByte>>2)&1;
+        U32 const singleSegment = (fhdByte>>5)&1;
+        U32 const fcsID = fhdByte>>6;
+        U32 const windowSizeMax = 1U << ZSTD_WINDOWLOG_MAX;
+        U32 windowSize = 0;
+        U32 dictID = 0;
+        U64 frameContentSize = 0;
+        if ((fhdByte & 0x08) != 0) return ERROR(frameParameter_unsupported);   /* reserved bits, which must be zero */
+        if (!singleSegment) {
+            BYTE const wlByte = ip[pos++];
+            U32 const windowLog = (wlByte >> 3) + ZSTD_WINDOWLOG_ABSOLUTEMIN;
+            if (windowLog > ZSTD_WINDOWLOG_MAX) return ERROR(frameParameter_windowTooLarge);  /* avoids issue with 1 << windowLog */
+            windowSize = (1U << windowLog);
+            windowSize += (windowSize >> 3) * (wlByte&7);
+        }
+
+        switch(dictIDSizeCode)
+        {
+            default:   /* impossible */
+            case 0 : break;
+            case 1 : dictID = ip[pos]; pos++; break;
+            case 2 : dictID = MEM_readLE16(ip+pos); pos+=2; break;
+            case 3 : dictID = MEM_readLE32(ip+pos); pos+=4; break;
+        }
+        switch(fcsID)
+        {
+            default:   /* impossible */
+            case 0 : if (singleSegment) frameContentSize = ip[pos]; break;
+            case 1 : frameContentSize = MEM_readLE16(ip+pos)+256; break;
+            case 2 : frameContentSize = MEM_readLE32(ip+pos); break;
+            case 3 : frameContentSize = MEM_readLE64(ip+pos); break;
+        }
+        if (!windowSize) windowSize = (U32)frameContentSize;
+        if (windowSize > windowSizeMax) return ERROR(frameParameter_windowTooLarge);
+        fparamsPtr->frameContentSize = frameContentSize;
+        fparamsPtr->windowSize = windowSize;
+        fparamsPtr->dictID = dictID;
+        fparamsPtr->checksumFlag = checksumFlag;
+    }
+    return 0;
+}
+
+
+/** ZSTD_getDecompressedSize() :
+*   compatible with legacy mode
+*   @return : decompressed size if known, 0 otherwise
+              note : 0 can mean any of the following :
+                   - decompressed size is not present within frame header
+                   - frame header unknown / not supported
+                   - frame header not complete (`srcSize` too small) */
+unsigned long long ZSTD_getDecompressedSize(const void* src, size_t srcSize)
+{
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT==1)
+    if (ZSTD_isLegacy(src, srcSize)) return ZSTD_getDecompressedSize_legacy(src, srcSize);
+#endif
+    {   ZSTD_frameParams fparams;
+        size_t const frResult = ZSTD_getFrameParams(&fparams, src, srcSize);
+        if (frResult!=0) return 0;
+        return fparams.frameContentSize;
+    }
+}
+
+
+/** ZSTD_decodeFrameHeader() :
+*   `headerSize` must be the size provided by ZSTD_frameHeaderSize().
+*   @return : 0 if success, or an error code, which can be tested using ZSTD_isError() */
+static size_t ZSTD_decodeFrameHeader(ZSTD_DCtx* dctx, const void* src, size_t headerSize)
+{
+    size_t const result = ZSTD_getFrameParams(&(dctx->fParams), src, headerSize);
+    if (ZSTD_isError(result)) return result;  /* invalid header */
+    if (result>0) return ERROR(srcSize_wrong);   /* headerSize too small */
+    if (dctx->fParams.dictID && (dctx->dictID != dctx->fParams.dictID)) return ERROR(dictionary_wrong);
+    if (dctx->fParams.checksumFlag) XXH64_reset(&dctx->xxhState, 0);
+    return 0;
+}
+
+
+typedef struct
+{
+    blockType_e blockType;
+    U32 lastBlock;
+    U32 origSize;
+} blockProperties_t;
+
+/*! ZSTD_getcBlockSize() :
+*   Provides the size of compressed block from block header `src` */
+size_t ZSTD_getcBlockSize(const void* src, size_t srcSize, blockProperties_t* bpPtr)
+{
+    if (srcSize < ZSTD_blockHeaderSize) return ERROR(srcSize_wrong);
+    {   U32 const cBlockHeader = MEM_readLE24(src);
+        U32 const cSize = cBlockHeader >> 3;
+        bpPtr->lastBlock = cBlockHeader & 1;
+        bpPtr->blockType = (blockType_e)((cBlockHeader >> 1) & 3);
+        bpPtr->origSize = cSize;   /* only useful for RLE */
+        if (bpPtr->blockType == bt_rle) return 1;
+        if (bpPtr->blockType == bt_reserved) return ERROR(corruption_detected);
+        return cSize;
+    }
+}
+
+
+static size_t ZSTD_copyRawBlock(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+    if (srcSize > dstCapacity) return ERROR(dstSize_tooSmall);
+    memcpy(dst, src, srcSize);
+    return srcSize;
+}
+
+
+static size_t ZSTD_setRleBlock(void* dst, size_t dstCapacity, const void* src, size_t srcSize, size_t regenSize)
+{
+    if (srcSize != 1) return ERROR(srcSize_wrong);
+    if (regenSize > dstCapacity) return ERROR(dstSize_tooSmall);
+    memset(dst, *(const BYTE*)src, regenSize);
+    return regenSize;
+}
+
+/*! ZSTD_decodeLiteralsBlock() :
+    @return : nb of bytes read from src (< srcSize ) */
+size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
+                          const void* src, size_t srcSize)   /* note : srcSize < BLOCKSIZE */
+{
+    if (srcSize < MIN_CBLOCK_SIZE) return ERROR(corruption_detected);
+
+    {   const BYTE* const istart = (const BYTE*) src;
+        symbolEncodingType_e const litEncType = (symbolEncodingType_e)(istart[0] & 3);
+
+        switch(litEncType)
+        {
+        case set_repeat:
+            if (dctx->litEntropy==0) return ERROR(dictionary_corrupted);
+            /* fall-through */
+        case set_compressed:
+            if (srcSize < 5) return ERROR(corruption_detected);   /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need up to 5 for case 3 */
+            {   size_t lhSize, litSize, litCSize;
+                U32 singleStream=0;
+                U32 const lhlCode = (istart[0] >> 2) & 3;
+                U32 const lhc = MEM_readLE32(istart);
+                switch(lhlCode)
+                {
+                case 0: case 1: default:   /* note : default is impossible, since lhlCode into [0..3] */
+                    /* 2 - 2 - 10 - 10 */
+                    singleStream = !lhlCode;
+                    lhSize = 3;
+                    litSize  = (lhc >> 4) & 0x3FF;
+                    litCSize = (lhc >> 14) & 0x3FF;
+                    break;
+                case 2:
+                    /* 2 - 2 - 14 - 14 */
+                    lhSize = 4;
+                    litSize  = (lhc >> 4) & 0x3FFF;
+                    litCSize = lhc >> 18;
+                    break;
+                case 3:
+                    /* 2 - 2 - 18 - 18 */
+                    lhSize = 5;
+                    litSize  = (lhc >> 4) & 0x3FFFF;
+                    litCSize = (lhc >> 22) + (istart[4] << 10);
+                    break;
+                }
+                if (litSize > ZSTD_BLOCKSIZE_ABSOLUTEMAX) return ERROR(corruption_detected);
+                if (litCSize + lhSize > srcSize) return ERROR(corruption_detected);
+
+                if (HUF_isError((litEncType==set_repeat) ?
+                                    ( singleStream ?
+                                        HUF_decompress1X_usingDTable(dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->HUFptr) :
+                                        HUF_decompress4X_usingDTable(dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->HUFptr) ) :
+                                    ( singleStream ?
+                                        HUF_decompress1X2_DCtx(dctx->hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize) :
+                                        HUF_decompress4X_hufOnly (dctx->hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize)) ))
+                    return ERROR(corruption_detected);
+
+                dctx->litPtr = dctx->litBuffer;
+                dctx->litSize = litSize;
+                dctx->litEntropy = 1;
+                if (litEncType==set_compressed) dctx->HUFptr = dctx->hufTable;
+                memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH);
+                return litCSize + lhSize;
+            }
+
+        case set_basic:
+            {   size_t litSize, lhSize;
+                U32 const lhlCode = ((istart[0]) >> 2) & 3;
+                switch(lhlCode)
+                {
+                case 0: case 2: default:   /* note : default is impossible, since lhlCode into [0..3] */
+                    lhSize = 1;
+                    litSize = istart[0] >> 3;
+                    break;
+                case 1:
+                    lhSize = 2;
+                    litSize = MEM_readLE16(istart) >> 4;
+                    break;
+                case 3:
+                    lhSize = 3;
+                    litSize = MEM_readLE24(istart) >> 4;
+                    break;
+                }
+
+                if (lhSize+litSize+WILDCOPY_OVERLENGTH > srcSize) {  /* risk reading beyond src buffer with wildcopy */
+                    if (litSize+lhSize > srcSize) return ERROR(corruption_detected);
+                    memcpy(dctx->litBuffer, istart+lhSize, litSize);
+                    dctx->litPtr = dctx->litBuffer;
+                    dctx->litSize = litSize;
+                    memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH);
+                    return lhSize+litSize;
+                }
+                /* direct reference into compressed stream */
+                dctx->litPtr = istart+lhSize;
+                dctx->litSize = litSize;
+                return lhSize+litSize;
+            }
+
+        case set_rle:
+            {   U32 const lhlCode = ((istart[0]) >> 2) & 3;
+                size_t litSize, lhSize;
+                switch(lhlCode)
+                {
+                case 0: case 2: default:   /* note : default is impossible, since lhlCode into [0..3] */
+                    lhSize = 1;
+                    litSize = istart[0] >> 3;
+                    break;
+                case 1:
+                    lhSize = 2;
+                    litSize = MEM_readLE16(istart) >> 4;
+                    break;
+                case 3:
+                    lhSize = 3;
+                    litSize = MEM_readLE24(istart) >> 4;
+                    if (srcSize<4) return ERROR(corruption_detected);   /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need lhSize+1 = 4 */
+                    break;
+                }
+                if (litSize > ZSTD_BLOCKSIZE_ABSOLUTEMAX) return ERROR(corruption_detected);
+                memset(dctx->litBuffer, istart[lhSize], litSize + WILDCOPY_OVERLENGTH);
+                dctx->litPtr = dctx->litBuffer;
+                dctx->litSize = litSize;
+                return lhSize+1;
+            }
+        default:
+            return ERROR(corruption_detected);   /* impossible */
+        }
+    }
+}
+
+
+typedef union {
+    FSE_decode_t realData;
+    U32 alignedBy4;
+} FSE_decode_t4;
+
+static const FSE_decode_t4 LL_defaultDTable[(1< max) return ERROR(corruption_detected);
+        FSE_buildDTable_rle(DTableSpace, *(const BYTE*)src);
+        *DTablePtr = DTableSpace;
+        return 1;
+    case set_basic :
+        *DTablePtr = (const FSE_DTable*)tmpPtr;
+        return 0;
+    case set_repeat:
+        if (!flagRepeatTable) return ERROR(corruption_detected);
+        return 0;
+    default :   /* impossible */
+    case set_compressed :
+        {   U32 tableLog;
+            S16 norm[MaxSeq+1];
+            size_t const headerSize = FSE_readNCount(norm, &max, &tableLog, src, srcSize);
+            if (FSE_isError(headerSize)) return ERROR(corruption_detected);
+            if (tableLog > maxLog) return ERROR(corruption_detected);
+            FSE_buildDTable(DTableSpace, norm, max, tableLog);
+            *DTablePtr = DTableSpace;
+            return headerSize;
+    }   }
+}
+
+size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
+                             const void* src, size_t srcSize)
+{
+    const BYTE* const istart = (const BYTE* const)src;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* ip = istart;
+
+    /* check */
+    if (srcSize < MIN_SEQUENCES_SIZE) return ERROR(srcSize_wrong);
+
+    /* SeqHead */
+    {   int nbSeq = *ip++;
+        if (!nbSeq) { *nbSeqPtr=0; return 1; }
+        if (nbSeq > 0x7F) {
+            if (nbSeq == 0xFF) {
+                if (ip+2 > iend) return ERROR(srcSize_wrong);
+                nbSeq = MEM_readLE16(ip) + LONGNBSEQ, ip+=2;
+            } else {
+                if (ip >= iend) return ERROR(srcSize_wrong);
+                nbSeq = ((nbSeq-0x80)<<8) + *ip++;
+            }
+        }
+        *nbSeqPtr = nbSeq;
+    }
+
+    /* FSE table descriptors */
+    if (ip+4 > iend) return ERROR(srcSize_wrong); /* minimum possible size */
+    {   symbolEncodingType_e const LLtype = (symbolEncodingType_e)(*ip >> 6);
+        symbolEncodingType_e const OFtype = (symbolEncodingType_e)((*ip >> 4) & 3);
+        symbolEncodingType_e const MLtype = (symbolEncodingType_e)((*ip >> 2) & 3);
+        ip++;
+
+        /* Build DTables */
+        {   size_t const llhSize = ZSTD_buildSeqTable(dctx->LLTable, &dctx->LLTptr,
+                                                      LLtype, MaxLL, LLFSELog,
+                                                      ip, iend-ip, LL_defaultDTable, dctx->fseEntropy);
+            if (ZSTD_isError(llhSize)) return ERROR(corruption_detected);
+            ip += llhSize;
+        }
+        {   size_t const ofhSize = ZSTD_buildSeqTable(dctx->OFTable, &dctx->OFTptr,
+                                                      OFtype, MaxOff, OffFSELog,
+                                                      ip, iend-ip, OF_defaultDTable, dctx->fseEntropy);
+            if (ZSTD_isError(ofhSize)) return ERROR(corruption_detected);
+            ip += ofhSize;
+        }
+        {   size_t const mlhSize = ZSTD_buildSeqTable(dctx->MLTable, &dctx->MLTptr,
+                                                      MLtype, MaxML, MLFSELog,
+                                                      ip, iend-ip, ML_defaultDTable, dctx->fseEntropy);
+            if (ZSTD_isError(mlhSize)) return ERROR(corruption_detected);
+            ip += mlhSize;
+        }
+    }
+
+    return ip-istart;
+}
+
+
+typedef struct {
+    size_t litLength;
+    size_t matchLength;
+    size_t offset;
+    const BYTE* match;
+} seq_t;
+
+typedef struct {
+    BIT_DStream_t DStream;
+    FSE_DState_t stateLL;
+    FSE_DState_t stateOffb;
+    FSE_DState_t stateML;
+    size_t prevOffset[ZSTD_REP_NUM];
+    const BYTE* base;
+    size_t pos;
+    iPtrDiff gotoDict;
+} seqState_t;
+
+
+FORCE_NOINLINE
+size_t ZSTD_execSequenceLast7(BYTE* op,
+                              BYTE* const oend, seq_t sequence,
+                              const BYTE** litPtr, const BYTE* const litLimit,
+                              const BYTE* const base, const BYTE* const vBase, const BYTE* const dictEnd)
+{
+    BYTE* const oLitEnd = op + sequence.litLength;
+    size_t const sequenceLength = sequence.litLength + sequence.matchLength;
+    BYTE* const oMatchEnd = op + sequenceLength;   /* risk : address space overflow (32-bits) */
+    BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;
+    const BYTE* const iLitEnd = *litPtr + sequence.litLength;
+    const BYTE* match = oLitEnd - sequence.offset;
+
+    /* check */
+    if (oMatchEnd>oend) return ERROR(dstSize_tooSmall); /* last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend */
+    if (iLitEnd > litLimit) return ERROR(corruption_detected);   /* over-read beyond lit buffer */
+    if (oLitEnd <= oend_w) return ERROR(GENERIC);   /* Precondition */
+
+    /* copy literals */
+    if (op < oend_w) {
+        ZSTD_wildcopy(op, *litPtr, oend_w - op);
+        *litPtr += oend_w - op;
+        op = oend_w;
+    }
+    while (op < oLitEnd) *op++ = *(*litPtr)++;
+
+    /* copy Match */
+    if (sequence.offset > (size_t)(oLitEnd - base)) {
+        /* offset beyond prefix */
+        if (sequence.offset > (size_t)(oLitEnd - vBase)) return ERROR(corruption_detected);
+        match = dictEnd - (base-match);
+        if (match + sequence.matchLength <= dictEnd) {
+            memmove(oLitEnd, match, sequence.matchLength);
+            return sequenceLength;
+        }
+        /* span extDict & currentPrefixSegment */
+        {   size_t const length1 = dictEnd - match;
+            memmove(oLitEnd, match, length1);
+            op = oLitEnd + length1;
+            sequence.matchLength -= length1;
+            match = base;
+    }   }
+    while (op < oMatchEnd) *op++ = *match++;
+    return sequenceLength;
+}
+
+
+
+
+static seq_t ZSTD_decodeSequence(seqState_t* seqState)
+{
+    seq_t seq;
+
+    U32 const llCode = FSE_peekSymbol(&seqState->stateLL);
+    U32 const mlCode = FSE_peekSymbol(&seqState->stateML);
+    U32 const ofCode = FSE_peekSymbol(&seqState->stateOffb);   /* <= maxOff, by table construction */
+
+    U32 const llBits = LL_bits[llCode];
+    U32 const mlBits = ML_bits[mlCode];
+    U32 const ofBits = ofCode;
+    U32 const totalBits = llBits+mlBits+ofBits;
+
+    static const U32 LL_base[MaxLL+1] = {
+                             0,  1,  2,  3,  4,  5,  6,  7,  8,  9,   10,    11,    12,    13,    14,     15,
+                            16, 18, 20, 22, 24, 28, 32, 40, 48, 64, 0x80, 0x100, 0x200, 0x400, 0x800, 0x1000,
+                            0x2000, 0x4000, 0x8000, 0x10000 };
+
+    static const U32 ML_base[MaxML+1] = {
+                             3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13,   14,    15,    16,    17,    18,
+                            19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29,   30,    31,    32,    33,    34,
+                            35, 37, 39, 41, 43, 47, 51, 59, 67, 83, 99, 0x83, 0x103, 0x203, 0x403, 0x803,
+                            0x1003, 0x2003, 0x4003, 0x8003, 0x10003 };
+
+    static const U32 OF_base[MaxOff+1] = {
+                             0,        1,       1,       5,     0xD,     0x1D,     0x3D,     0x7D,
+                             0xFD,   0x1FD,   0x3FD,   0x7FD,   0xFFD,   0x1FFD,   0x3FFD,   0x7FFD,
+                             0xFFFD, 0x1FFFD, 0x3FFFD, 0x7FFFD, 0xFFFFD, 0x1FFFFD, 0x3FFFFD, 0x7FFFFD,
+                             0xFFFFFD, 0x1FFFFFD, 0x3FFFFFD, 0x7FFFFFD, 0xFFFFFFD };
+
+    /* sequence */
+    {   size_t offset;
+        if (!ofCode)
+            offset = 0;
+        else {
+            offset = OF_base[ofCode] + BIT_readBitsFast(&seqState->DStream, ofBits);   /* <=  (ZSTD_WINDOWLOG_MAX-1) bits */
+            if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);
+        }
+
+        if (ofCode <= 1) {
+            offset += (llCode==0);
+            if (offset) {
+                size_t temp = (offset==3) ? seqState->prevOffset[0] - 1 : seqState->prevOffset[offset];
+                temp += !temp;   /* 0 is not valid; input is corrupted; force offset to 1 */
+                if (offset != 1) seqState->prevOffset[2] = seqState->prevOffset[1];
+                seqState->prevOffset[1] = seqState->prevOffset[0];
+                seqState->prevOffset[0] = offset = temp;
+            } else {
+                offset = seqState->prevOffset[0];
+            }
+        } else {
+            seqState->prevOffset[2] = seqState->prevOffset[1];
+            seqState->prevOffset[1] = seqState->prevOffset[0];
+            seqState->prevOffset[0] = offset;
+        }
+        seq.offset = offset;
+    }
+
+    seq.matchLength = ML_base[mlCode] + ((mlCode>31) ? BIT_readBitsFast(&seqState->DStream, mlBits) : 0);  /* <=  16 bits */
+    if (MEM_32bits() && (mlBits+llBits>24)) BIT_reloadDStream(&seqState->DStream);
+
+    seq.litLength = LL_base[llCode] + ((llCode>15) ? BIT_readBitsFast(&seqState->DStream, llBits) : 0);    /* <=  16 bits */
+    if (MEM_32bits() ||
+       (totalBits > 64 - 7 - (LLFSELog+MLFSELog+OffFSELog)) ) BIT_reloadDStream(&seqState->DStream);
+
+    /* ANS state update */
+    FSE_updateState(&seqState->stateLL, &seqState->DStream);    /* <=  9 bits */
+    FSE_updateState(&seqState->stateML, &seqState->DStream);    /* <=  9 bits */
+    if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);    /* <= 18 bits */
+    FSE_updateState(&seqState->stateOffb, &seqState->DStream);  /* <=  8 bits */
+
+    return seq;
+}
+
+
+FORCE_INLINE
+size_t ZSTD_execSequence(BYTE* op,
+                                BYTE* const oend, seq_t sequence,
+                                const BYTE** litPtr, const BYTE* const litLimit,
+                                const BYTE* const base, const BYTE* const vBase, const BYTE* const dictEnd)
+{
+    BYTE* const oLitEnd = op + sequence.litLength;
+    size_t const sequenceLength = sequence.litLength + sequence.matchLength;
+    BYTE* const oMatchEnd = op + sequenceLength;   /* risk : address space overflow (32-bits) */
+    BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;
+    const BYTE* const iLitEnd = *litPtr + sequence.litLength;
+    const BYTE* match = oLitEnd - sequence.offset;
+
+    /* check */
+    if (oMatchEnd>oend) return ERROR(dstSize_tooSmall); /* last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend */
+    if (iLitEnd > litLimit) return ERROR(corruption_detected);   /* over-read beyond lit buffer */
+    if (oLitEnd>oend_w) return ZSTD_execSequenceLast7(op, oend, sequence, litPtr, litLimit, base, vBase, dictEnd);
+
+    /* copy Literals */
+    ZSTD_copy8(op, *litPtr);
+    if (sequence.litLength > 8)
+        ZSTD_wildcopy(op+8, (*litPtr)+8, sequence.litLength - 8);   /* note : since oLitEnd <= oend-WILDCOPY_OVERLENGTH, no risk of overwrite beyond oend */
+    op = oLitEnd;
+    *litPtr = iLitEnd;   /* update for next sequence */
+
+    /* copy Match */
+    if (sequence.offset > (size_t)(oLitEnd - base)) {
+        /* offset beyond prefix */
+        if (sequence.offset > (size_t)(oLitEnd - vBase)) return ERROR(corruption_detected);
+        match += (dictEnd-base);
+        if (match + sequence.matchLength <= dictEnd) {
+            memmove(oLitEnd, match, sequence.matchLength);
+            return sequenceLength;
+        }
+        /* span extDict & currentPrefixSegment */
+        {   size_t const length1 = dictEnd - match;
+            memmove(oLitEnd, match, length1);
+            op = oLitEnd + length1;
+            sequence.matchLength -= length1;
+            match = base;
+            if (op > oend_w || sequence.matchLength < MINMATCH) {
+              U32 i;
+              for (i = 0; i < sequence.matchLength; ++i) op[i] = match[i];
+              return sequenceLength;
+            }
+    }   }
+    /* Requirement: op <= oend_w && sequence.matchLength >= MINMATCH */
+
+    /* match within prefix */
+    if (sequence.offset < 8) {
+        /* close range match, overlap */
+        static const U32 dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 };   /* added */
+        static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 };   /* substracted */
+        int const sub2 = dec64table[sequence.offset];
+        op[0] = match[0];
+        op[1] = match[1];
+        op[2] = match[2];
+        op[3] = match[3];
+        match += dec32table[sequence.offset];
+        ZSTD_copy4(op+4, match);
+        match -= sub2;
+    } else {
+        ZSTD_copy8(op, match);
+    }
+    op += 8; match += 8;
+
+    if (oMatchEnd > oend-(16-MINMATCH)) {
+        if (op < oend_w) {
+            ZSTD_wildcopy(op, match, oend_w - op);
+            match += oend_w - op;
+            op = oend_w;
+        }
+        while (op < oMatchEnd) *op++ = *match++;
+    } else {
+        ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8);   /* works even if matchLength < 8 */
+    }
+    return sequenceLength;
+}
+
+
+static size_t ZSTD_decompressSequences(
+                               ZSTD_DCtx* dctx,
+                               void* dst, size_t maxDstSize,
+                         const void* seqStart, size_t seqSize)
+{
+    const BYTE* ip = (const BYTE*)seqStart;
+    const BYTE* const iend = ip + seqSize;
+    BYTE* const ostart = (BYTE* const)dst;
+    BYTE* const oend = ostart + maxDstSize;
+    BYTE* op = ostart;
+    const BYTE* litPtr = dctx->litPtr;
+    const BYTE* const litEnd = litPtr + dctx->litSize;
+    const BYTE* const base = (const BYTE*) (dctx->base);
+    const BYTE* const vBase = (const BYTE*) (dctx->vBase);
+    const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
+    int nbSeq;
+
+    /* Build Decoding Tables */
+    {   size_t const seqHSize = ZSTD_decodeSeqHeaders(dctx, &nbSeq, ip, seqSize);
+        if (ZSTD_isError(seqHSize)) return seqHSize;
+        ip += seqHSize;
+    }
+
+    /* Regen sequences */
+    if (nbSeq) {
+        seqState_t seqState;
+        dctx->fseEntropy = 1;
+        { U32 i; for (i=0; irep[i]; }
+        CHECK_E(BIT_initDStream(&seqState.DStream, ip, iend-ip), corruption_detected);
+        FSE_initDState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr);
+        FSE_initDState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr);
+        FSE_initDState(&seqState.stateML, &seqState.DStream, dctx->MLTptr);
+
+        for ( ; (BIT_reloadDStream(&(seqState.DStream)) <= BIT_DStream_completed) && nbSeq ; ) {
+            nbSeq--;
+            {   seq_t const sequence = ZSTD_decodeSequence(&seqState);
+                size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litEnd, base, vBase, dictEnd);
+                if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
+                op += oneSeqSize;
+        }   }
+
+        /* check if reached exact end */
+        if (nbSeq) return ERROR(corruption_detected);
+        /* save reps for next block */
+        { U32 i; for (i=0; irep[i] = (U32)(seqState.prevOffset[i]); }
+    }
+
+    /* last literal segment */
+    {   size_t const lastLLSize = litEnd - litPtr;
+        if (lastLLSize > (size_t)(oend-op)) return ERROR(dstSize_tooSmall);
+        memcpy(op, litPtr, lastLLSize);
+        op += lastLLSize;
+    }
+
+    return op-ostart;
+}
+
+
+static seq_t ZSTD_decodeSequenceLong(seqState_t* seqState)
+{
+    seq_t seq;
+
+    U32 const llCode = FSE_peekSymbol(&seqState->stateLL);
+    U32 const mlCode = FSE_peekSymbol(&seqState->stateML);
+    U32 const ofCode = FSE_peekSymbol(&seqState->stateOffb);   /* <= maxOff, by table construction */
+
+    U32 const llBits = LL_bits[llCode];
+    U32 const mlBits = ML_bits[mlCode];
+    U32 const ofBits = ofCode;
+    U32 const totalBits = llBits+mlBits+ofBits;
+
+    static const U32 LL_base[MaxLL+1] = {
+                             0,  1,  2,  3,  4,  5,  6,  7,  8,  9,   10,    11,    12,    13,    14,     15,
+                            16, 18, 20, 22, 24, 28, 32, 40, 48, 64, 0x80, 0x100, 0x200, 0x400, 0x800, 0x1000,
+                            0x2000, 0x4000, 0x8000, 0x10000 };
+
+    static const U32 ML_base[MaxML+1] = {
+                             3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13,   14,    15,    16,    17,    18,
+                            19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29,   30,    31,    32,    33,    34,
+                            35, 37, 39, 41, 43, 47, 51, 59, 67, 83, 99, 0x83, 0x103, 0x203, 0x403, 0x803,
+                            0x1003, 0x2003, 0x4003, 0x8003, 0x10003 };
+
+    static const U32 OF_base[MaxOff+1] = {
+                             0,        1,       1,       5,     0xD,     0x1D,     0x3D,     0x7D,
+                             0xFD,   0x1FD,   0x3FD,   0x7FD,   0xFFD,   0x1FFD,   0x3FFD,   0x7FFD,
+                             0xFFFD, 0x1FFFD, 0x3FFFD, 0x7FFFD, 0xFFFFD, 0x1FFFFD, 0x3FFFFD, 0x7FFFFD,
+                             0xFFFFFD, 0x1FFFFFD, 0x3FFFFFD, 0x7FFFFFD, 0xFFFFFFD };
+
+    /* sequence */
+    {   size_t offset;
+        if (!ofCode)
+            offset = 0;
+        else {
+            offset = OF_base[ofCode] + BIT_readBitsFast(&seqState->DStream, ofBits);   /* <=  (ZSTD_WINDOWLOG_MAX-1) bits */
+            if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);
+        }
+
+        if (ofCode <= 1) {
+            offset += (llCode==0);
+            if (offset) {
+                size_t temp = (offset==3) ? seqState->prevOffset[0] - 1 : seqState->prevOffset[offset];
+                temp += !temp;   /* 0 is not valid; input is corrupted; force offset to 1 */
+                if (offset != 1) seqState->prevOffset[2] = seqState->prevOffset[1];
+                seqState->prevOffset[1] = seqState->prevOffset[0];
+                seqState->prevOffset[0] = offset = temp;
+            } else {
+                offset = seqState->prevOffset[0];
+            }
+        } else {
+            seqState->prevOffset[2] = seqState->prevOffset[1];
+            seqState->prevOffset[1] = seqState->prevOffset[0];
+            seqState->prevOffset[0] = offset;
+        }
+        seq.offset = offset;
+    }
+
+    seq.matchLength = ML_base[mlCode] + ((mlCode>31) ? BIT_readBitsFast(&seqState->DStream, mlBits) : 0);  /* <=  16 bits */
+    if (MEM_32bits() && (mlBits+llBits>24)) BIT_reloadDStream(&seqState->DStream);
+
+    seq.litLength = LL_base[llCode] + ((llCode>15) ? BIT_readBitsFast(&seqState->DStream, llBits) : 0);    /* <=  16 bits */
+    if (MEM_32bits() ||
+       (totalBits > 64 - 7 - (LLFSELog+MLFSELog+OffFSELog)) ) BIT_reloadDStream(&seqState->DStream);
+
+    {   size_t const pos = seqState->pos + seq.litLength;
+        seq.match = seqState->base + pos - seq.offset;    /* single memory segment */
+        if (seq.offset > pos) seq.match += seqState->gotoDict;   /* separate memory segment */
+        seqState->pos = pos + seq.matchLength;
+    }
+
+    /* ANS state update */
+    FSE_updateState(&seqState->stateLL, &seqState->DStream);    /* <=  9 bits */
+    FSE_updateState(&seqState->stateML, &seqState->DStream);    /* <=  9 bits */
+    if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);    /* <= 18 bits */
+    FSE_updateState(&seqState->stateOffb, &seqState->DStream);  /* <=  8 bits */
+
+    return seq;
+}
+
+FORCE_INLINE
+size_t ZSTD_execSequenceLong(BYTE* op,
+                                BYTE* const oend, seq_t sequence,
+                                const BYTE** litPtr, const BYTE* const litLimit,
+                                const BYTE* const base, const BYTE* const vBase, const BYTE* const dictEnd)
+{
+    BYTE* const oLitEnd = op + sequence.litLength;
+    size_t const sequenceLength = sequence.litLength + sequence.matchLength;
+    BYTE* const oMatchEnd = op + sequenceLength;   /* risk : address space overflow (32-bits) */
+    BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;
+    const BYTE* const iLitEnd = *litPtr + sequence.litLength;
+    const BYTE* match = sequence.match;
+
+    /* check */
+#if 1
+    if (oMatchEnd>oend) return ERROR(dstSize_tooSmall); /* last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend */
+    if (iLitEnd > litLimit) return ERROR(corruption_detected);   /* over-read beyond lit buffer */
+    if (oLitEnd>oend_w) return ZSTD_execSequenceLast7(op, oend, sequence, litPtr, litLimit, base, vBase, dictEnd);
+#endif
+
+    /* copy Literals */
+    ZSTD_copy8(op, *litPtr);
+    if (sequence.litLength > 8)
+        ZSTD_wildcopy(op+8, (*litPtr)+8, sequence.litLength - 8);   /* note : since oLitEnd <= oend-WILDCOPY_OVERLENGTH, no risk of overwrite beyond oend */
+    op = oLitEnd;
+    *litPtr = iLitEnd;   /* update for next sequence */
+
+    /* copy Match */
+#if 1
+    if (sequence.offset > (size_t)(oLitEnd - base)) {
+        /* offset beyond prefix */
+        if (sequence.offset > (size_t)(oLitEnd - vBase)) return ERROR(corruption_detected);
+        if (match + sequence.matchLength <= dictEnd) {
+            memmove(oLitEnd, match, sequence.matchLength);
+            return sequenceLength;
+        }
+        /* span extDict & currentPrefixSegment */
+        {   size_t const length1 = dictEnd - match;
+            memmove(oLitEnd, match, length1);
+            op = oLitEnd + length1;
+            sequence.matchLength -= length1;
+            match = base;
+            if (op > oend_w || sequence.matchLength < MINMATCH) {
+              U32 i;
+              for (i = 0; i < sequence.matchLength; ++i) op[i] = match[i];
+              return sequenceLength;
+            }
+    }   }
+    /* Requirement: op <= oend_w && sequence.matchLength >= MINMATCH */
+#endif
+
+    /* match within prefix */
+    if (sequence.offset < 8) {
+        /* close range match, overlap */
+        static const U32 dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 };   /* added */
+        static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 };   /* substracted */
+        int const sub2 = dec64table[sequence.offset];
+        op[0] = match[0];
+        op[1] = match[1];
+        op[2] = match[2];
+        op[3] = match[3];
+        match += dec32table[sequence.offset];
+        ZSTD_copy4(op+4, match);
+        match -= sub2;
+    } else {
+        ZSTD_copy8(op, match);
+    }
+    op += 8; match += 8;
+
+    if (oMatchEnd > oend-(16-MINMATCH)) {
+        if (op < oend_w) {
+            ZSTD_wildcopy(op, match, oend_w - op);
+            match += oend_w - op;
+            op = oend_w;
+        }
+        while (op < oMatchEnd) *op++ = *match++;
+    } else {
+        ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8);   /* works even if matchLength < 8 */
+    }
+    return sequenceLength;
+}
+
+static size_t ZSTD_decompressSequencesLong(
+                               ZSTD_DCtx* dctx,
+                               void* dst, size_t maxDstSize,
+                         const void* seqStart, size_t seqSize)
+{
+    const BYTE* ip = (const BYTE*)seqStart;
+    const BYTE* const iend = ip + seqSize;
+    BYTE* const ostart = (BYTE* const)dst;
+    BYTE* const oend = ostart + maxDstSize;
+    BYTE* op = ostart;
+    const BYTE* litPtr = dctx->litPtr;
+    const BYTE* const litEnd = litPtr + dctx->litSize;
+    const BYTE* const base = (const BYTE*) (dctx->base);
+    const BYTE* const vBase = (const BYTE*) (dctx->vBase);
+    const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
+    int nbSeq;
+
+    /* Build Decoding Tables */
+    {   size_t const seqHSize = ZSTD_decodeSeqHeaders(dctx, &nbSeq, ip, seqSize);
+        if (ZSTD_isError(seqHSize)) return seqHSize;
+        ip += seqHSize;
+    }
+
+    /* Regen sequences */
+    if (nbSeq) {
+#define STORED_SEQS 4
+#define STOSEQ_MASK (STORED_SEQS-1)
+#define ADVANCED_SEQS 4
+        seq_t sequences[STORED_SEQS];
+        int const seqAdvance = MIN(nbSeq, ADVANCED_SEQS);
+        seqState_t seqState;
+        int seqNb;
+        dctx->fseEntropy = 1;
+        { U32 i; for (i=0; irep[i]; }
+        seqState.base = base;
+        seqState.pos = (size_t)(op-base);
+        seqState.gotoDict = (iPtrDiff)(dictEnd - base);
+        CHECK_E(BIT_initDStream(&seqState.DStream, ip, iend-ip), corruption_detected);
+        FSE_initDState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr);
+        FSE_initDState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr);
+        FSE_initDState(&seqState.stateML, &seqState.DStream, dctx->MLTptr);
+
+        /* prepare in advance */
+        for (seqNb=0; (BIT_reloadDStream(&seqState.DStream) <= BIT_DStream_completed) && seqNbrep[i] = (U32)(seqState.prevOffset[i]); }
+    }
+
+    /* last literal segment */
+    {   size_t const lastLLSize = litEnd - litPtr;
+        if (lastLLSize > (size_t)(oend-op)) return ERROR(dstSize_tooSmall);
+        memcpy(op, litPtr, lastLLSize);
+        op += lastLLSize;
+    }
+
+    return op-ostart;
+}
+
+
+static size_t ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
+                            void* dst, size_t dstCapacity,
+                      const void* src, size_t srcSize)
+{   /* blockType == blockCompressed */
+    const BYTE* ip = (const BYTE*)src;
+
+    if (srcSize >= ZSTD_BLOCKSIZE_ABSOLUTEMAX) return ERROR(srcSize_wrong);
+
+    /* Decode literals sub-block */
+    {   size_t const litCSize = ZSTD_decodeLiteralsBlock(dctx, src, srcSize);
+        if (ZSTD_isError(litCSize)) return litCSize;
+        ip += litCSize;
+        srcSize -= litCSize;
+    }
+    if (dctx->fParams.windowSize > (1<<23)) return ZSTD_decompressSequencesLong(dctx, dst, dstCapacity, ip, srcSize);
+    return ZSTD_decompressSequences(dctx, dst, dstCapacity, ip, srcSize);
+}
+
+
+static void ZSTD_checkContinuity(ZSTD_DCtx* dctx, const void* dst)
+{
+    if (dst != dctx->previousDstEnd) {   /* not contiguous */
+        dctx->dictEnd = dctx->previousDstEnd;
+        dctx->vBase = (const char*)dst - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->base));
+        dctx->base = dst;
+        dctx->previousDstEnd = dst;
+    }
+}
+
+size_t ZSTD_decompressBlock(ZSTD_DCtx* dctx,
+                            void* dst, size_t dstCapacity,
+                      const void* src, size_t srcSize)
+{
+    size_t dSize;
+    ZSTD_checkContinuity(dctx, dst);
+    dSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize);
+    dctx->previousDstEnd = (char*)dst + dSize;
+    return dSize;
+}
+
+
+/** ZSTD_insertBlock() :
+    insert `src` block into `dctx` history. Useful to track uncompressed blocks. */
+ZSTDLIB_API size_t ZSTD_insertBlock(ZSTD_DCtx* dctx, const void* blockStart, size_t blockSize)
+{
+    ZSTD_checkContinuity(dctx, blockStart);
+    dctx->previousDstEnd = (const char*)blockStart + blockSize;
+    return blockSize;
+}
+
+
+size_t ZSTD_generateNxBytes(void* dst, size_t dstCapacity, BYTE byte, size_t length)
+{
+    if (length > dstCapacity) return ERROR(dstSize_tooSmall);
+    memset(dst, byte, length);
+    return length;
+}
+
+
+/*! ZSTD_decompressFrame() :
+*   `dctx` must be properly initialized */
+static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
+                                 void* dst, size_t dstCapacity,
+                                 const void* src, size_t srcSize)
+{
+    const BYTE* ip = (const BYTE*)src;
+    BYTE* const ostart = (BYTE* const)dst;
+    BYTE* const oend = ostart + dstCapacity;
+    BYTE* op = ostart;
+    size_t remainingSize = srcSize;
+
+    /* check */
+    if (srcSize < ZSTD_frameHeaderSize_min+ZSTD_blockHeaderSize) return ERROR(srcSize_wrong);
+
+    /* Frame Header */
+    {   size_t const frameHeaderSize = ZSTD_frameHeaderSize(src, ZSTD_frameHeaderSize_prefix);
+        if (ZSTD_isError(frameHeaderSize)) return frameHeaderSize;
+        if (srcSize < frameHeaderSize+ZSTD_blockHeaderSize) return ERROR(srcSize_wrong);
+        CHECK_F(ZSTD_decodeFrameHeader(dctx, src, frameHeaderSize));
+        ip += frameHeaderSize; remainingSize -= frameHeaderSize;
+    }
+
+    /* Loop on each block */
+    while (1) {
+        size_t decodedSize;
+        blockProperties_t blockProperties;
+        size_t const cBlockSize = ZSTD_getcBlockSize(ip, remainingSize, &blockProperties);
+        if (ZSTD_isError(cBlockSize)) return cBlockSize;
+
+        ip += ZSTD_blockHeaderSize;
+        remainingSize -= ZSTD_blockHeaderSize;
+        if (cBlockSize > remainingSize) return ERROR(srcSize_wrong);
+
+        switch(blockProperties.blockType)
+        {
+        case bt_compressed:
+            decodedSize = ZSTD_decompressBlock_internal(dctx, op, oend-op, ip, cBlockSize);
+            break;
+        case bt_raw :
+            decodedSize = ZSTD_copyRawBlock(op, oend-op, ip, cBlockSize);
+            break;
+        case bt_rle :
+            decodedSize = ZSTD_generateNxBytes(op, oend-op, *ip, blockProperties.origSize);
+            break;
+        case bt_reserved :
+        default:
+            return ERROR(corruption_detected);
+        }
+
+        if (ZSTD_isError(decodedSize)) return decodedSize;
+        if (dctx->fParams.checksumFlag) XXH64_update(&dctx->xxhState, op, decodedSize);
+        op += decodedSize;
+        ip += cBlockSize;
+        remainingSize -= cBlockSize;
+        if (blockProperties.lastBlock) break;
+    }
+
+    if (dctx->fParams.checksumFlag) {   /* Frame content checksum verification */
+        U32 const checkCalc = (U32)XXH64_digest(&dctx->xxhState);
+        U32 checkRead;
+        if (remainingSize<4) return ERROR(checksum_wrong);
+        checkRead = MEM_readLE32(ip);
+        if (checkRead != checkCalc) return ERROR(checksum_wrong);
+        remainingSize -= 4;
+    }
+
+    if (remainingSize) return ERROR(srcSize_wrong);
+    return op-ostart;
+}
+
+
+size_t ZSTD_decompress_usingDict(ZSTD_DCtx* dctx,
+                                 void* dst, size_t dstCapacity,
+                           const void* src, size_t srcSize,
+                           const void* dict, size_t dictSize)
+{
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT==1)
+    if (ZSTD_isLegacy(src, srcSize)) return ZSTD_decompressLegacy(dst, dstCapacity, src, srcSize, dict, dictSize);
+#endif
+    ZSTD_decompressBegin_usingDict(dctx, dict, dictSize);
+    ZSTD_checkContinuity(dctx, dst);
+    return ZSTD_decompressFrame(dctx, dst, dstCapacity, src, srcSize);
+}
+
+
+size_t ZSTD_decompressDCtx(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+    return ZSTD_decompress_usingDict(dctx, dst, dstCapacity, src, srcSize, NULL, 0);
+}
+
+
+size_t ZSTD_decompress(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+#if defined(ZSTD_HEAPMODE) && (ZSTD_HEAPMODE==1)
+    size_t regenSize;
+    ZSTD_DCtx* const dctx = ZSTD_createDCtx();
+    if (dctx==NULL) return ERROR(memory_allocation);
+    regenSize = ZSTD_decompressDCtx(dctx, dst, dstCapacity, src, srcSize);
+    ZSTD_freeDCtx(dctx);
+    return regenSize;
+#else   /* stack mode */
+    ZSTD_DCtx dctx;
+    return ZSTD_decompressDCtx(&dctx, dst, dstCapacity, src, srcSize);
+#endif
+}
+
+
+/*-**************************************
+*   Advanced Streaming Decompression API
+*   Bufferless and synchronous
+****************************************/
+size_t ZSTD_nextSrcSizeToDecompress(ZSTD_DCtx* dctx) { return dctx->expected; }
+
+ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx) {
+    switch(dctx->stage)
+    {
+    default:   /* should not happen */
+    case ZSTDds_getFrameHeaderSize:
+    case ZSTDds_decodeFrameHeader:
+        return ZSTDnit_frameHeader;
+    case ZSTDds_decodeBlockHeader:
+        return ZSTDnit_blockHeader;
+    case ZSTDds_decompressBlock:
+        return ZSTDnit_block;
+    case ZSTDds_decompressLastBlock:
+        return ZSTDnit_lastBlock;
+    case ZSTDds_checkChecksum:
+        return ZSTDnit_checksum;
+    case ZSTDds_decodeSkippableHeader:
+    case ZSTDds_skipFrame:
+        return ZSTDnit_skippableFrame;
+    }
+}
+
+int ZSTD_isSkipFrame(ZSTD_DCtx* dctx) { return dctx->stage == ZSTDds_skipFrame; }   /* for zbuff */
+
+/** ZSTD_decompressContinue() :
+*   @return : nb of bytes generated into `dst` (necessarily <= `dstCapacity)
+*             or an error code, which can be tested using ZSTD_isError() */
+size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+    /* Sanity check */
+    if (srcSize != dctx->expected) return ERROR(srcSize_wrong);
+    if (dstCapacity) ZSTD_checkContinuity(dctx, dst);
+
+    switch (dctx->stage)
+    {
+    case ZSTDds_getFrameHeaderSize :
+        if (srcSize != ZSTD_frameHeaderSize_prefix) return ERROR(srcSize_wrong);      /* impossible */
+        if ((MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {        /* skippable frame */
+            memcpy(dctx->headerBuffer, src, ZSTD_frameHeaderSize_prefix);
+            dctx->expected = ZSTD_skippableHeaderSize - ZSTD_frameHeaderSize_prefix;  /* magic number + skippable frame length */
+            dctx->stage = ZSTDds_decodeSkippableHeader;
+            return 0;
+        }
+        dctx->headerSize = ZSTD_frameHeaderSize(src, ZSTD_frameHeaderSize_prefix);
+        if (ZSTD_isError(dctx->headerSize)) return dctx->headerSize;
+        memcpy(dctx->headerBuffer, src, ZSTD_frameHeaderSize_prefix);
+        if (dctx->headerSize > ZSTD_frameHeaderSize_prefix) {
+            dctx->expected = dctx->headerSize - ZSTD_frameHeaderSize_prefix;
+            dctx->stage = ZSTDds_decodeFrameHeader;
+            return 0;
+        }
+        dctx->expected = 0;   /* not necessary to copy more */
+
+    case ZSTDds_decodeFrameHeader:
+        memcpy(dctx->headerBuffer + ZSTD_frameHeaderSize_prefix, src, dctx->expected);
+        CHECK_F(ZSTD_decodeFrameHeader(dctx, dctx->headerBuffer, dctx->headerSize));
+        dctx->expected = ZSTD_blockHeaderSize;
+        dctx->stage = ZSTDds_decodeBlockHeader;
+        return 0;
+
+    case ZSTDds_decodeBlockHeader:
+        {   blockProperties_t bp;
+            size_t const cBlockSize = ZSTD_getcBlockSize(src, ZSTD_blockHeaderSize, &bp);
+            if (ZSTD_isError(cBlockSize)) return cBlockSize;
+            dctx->expected = cBlockSize;
+            dctx->bType = bp.blockType;
+            dctx->rleSize = bp.origSize;
+            if (cBlockSize) {
+                dctx->stage = bp.lastBlock ? ZSTDds_decompressLastBlock : ZSTDds_decompressBlock;
+                return 0;
+            }
+            /* empty block */
+            if (bp.lastBlock) {
+                if (dctx->fParams.checksumFlag) {
+                    dctx->expected = 4;
+                    dctx->stage = ZSTDds_checkChecksum;
+                } else {
+                    dctx->expected = 0; /* end of frame */
+                    dctx->stage = ZSTDds_getFrameHeaderSize;
+                }
+            } else {
+                dctx->expected = 3;  /* go directly to next header */
+                dctx->stage = ZSTDds_decodeBlockHeader;
+            }
+            return 0;
+        }
+    case ZSTDds_decompressLastBlock:
+    case ZSTDds_decompressBlock:
+        {   size_t rSize;
+            switch(dctx->bType)
+            {
+            case bt_compressed:
+                rSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize);
+                break;
+            case bt_raw :
+                rSize = ZSTD_copyRawBlock(dst, dstCapacity, src, srcSize);
+                break;
+            case bt_rle :
+                rSize = ZSTD_setRleBlock(dst, dstCapacity, src, srcSize, dctx->rleSize);
+                break;
+            case bt_reserved :   /* should never happen */
+            default:
+                return ERROR(corruption_detected);
+            }
+            if (ZSTD_isError(rSize)) return rSize;
+            if (dctx->fParams.checksumFlag) XXH64_update(&dctx->xxhState, dst, rSize);
+
+            if (dctx->stage == ZSTDds_decompressLastBlock) {   /* end of frame */
+                if (dctx->fParams.checksumFlag) {  /* another round for frame checksum */
+                    dctx->expected = 4;
+                    dctx->stage = ZSTDds_checkChecksum;
+                } else {
+                    dctx->expected = 0;   /* ends here */
+                    dctx->stage = ZSTDds_getFrameHeaderSize;
+                }
+            } else {
+                dctx->stage = ZSTDds_decodeBlockHeader;
+                dctx->expected = ZSTD_blockHeaderSize;
+                dctx->previousDstEnd = (char*)dst + rSize;
+            }
+            return rSize;
+        }
+    case ZSTDds_checkChecksum:
+        {   U32 const h32 = (U32)XXH64_digest(&dctx->xxhState);
+            U32 const check32 = MEM_readLE32(src);   /* srcSize == 4, guaranteed by dctx->expected */
+            if (check32 != h32) return ERROR(checksum_wrong);
+            dctx->expected = 0;
+            dctx->stage = ZSTDds_getFrameHeaderSize;
+            return 0;
+        }
+    case ZSTDds_decodeSkippableHeader:
+        {   memcpy(dctx->headerBuffer + ZSTD_frameHeaderSize_prefix, src, dctx->expected);
+            dctx->expected = MEM_readLE32(dctx->headerBuffer + 4);
+            dctx->stage = ZSTDds_skipFrame;
+            return 0;
+        }
+    case ZSTDds_skipFrame:
+        {   dctx->expected = 0;
+            dctx->stage = ZSTDds_getFrameHeaderSize;
+            return 0;
+        }
+    default:
+        return ERROR(GENERIC);   /* impossible */
+    }
+}
+
+
+static size_t ZSTD_refDictContent(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
+{
+    dctx->dictEnd = dctx->previousDstEnd;
+    dctx->vBase = (const char*)dict - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->base));
+    dctx->base = dict;
+    dctx->previousDstEnd = (const char*)dict + dictSize;
+    return 0;
+}
+
+static size_t ZSTD_loadEntropy(ZSTD_DCtx* dctx, const void* const dict, size_t const dictSize)
+{
+    const BYTE* dictPtr = (const BYTE*)dict;
+    const BYTE* const dictEnd = dictPtr + dictSize;
+
+    {   size_t const hSize = HUF_readDTableX4(dctx->hufTable, dict, dictSize);
+        if (HUF_isError(hSize)) return ERROR(dictionary_corrupted);
+        dictPtr += hSize;
+    }
+
+    {   short offcodeNCount[MaxOff+1];
+        U32 offcodeMaxValue=MaxOff, offcodeLog;
+        size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr);
+        if (FSE_isError(offcodeHeaderSize)) return ERROR(dictionary_corrupted);
+        if (offcodeLog > OffFSELog) return ERROR(dictionary_corrupted);
+        CHECK_E(FSE_buildDTable(dctx->OFTable, offcodeNCount, offcodeMaxValue, offcodeLog), dictionary_corrupted);
+        dictPtr += offcodeHeaderSize;
+    }
+
+    {   short matchlengthNCount[MaxML+1];
+        unsigned matchlengthMaxValue = MaxML, matchlengthLog;
+        size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, dictEnd-dictPtr);
+        if (FSE_isError(matchlengthHeaderSize)) return ERROR(dictionary_corrupted);
+        if (matchlengthLog > MLFSELog) return ERROR(dictionary_corrupted);
+        CHECK_E(FSE_buildDTable(dctx->MLTable, matchlengthNCount, matchlengthMaxValue, matchlengthLog), dictionary_corrupted);
+        dictPtr += matchlengthHeaderSize;
+    }
+
+    {   short litlengthNCount[MaxLL+1];
+        unsigned litlengthMaxValue = MaxLL, litlengthLog;
+        size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, dictEnd-dictPtr);
+        if (FSE_isError(litlengthHeaderSize)) return ERROR(dictionary_corrupted);
+        if (litlengthLog > LLFSELog) return ERROR(dictionary_corrupted);
+        CHECK_E(FSE_buildDTable(dctx->LLTable, litlengthNCount, litlengthMaxValue, litlengthLog), dictionary_corrupted);
+        dictPtr += litlengthHeaderSize;
+    }
+
+    if (dictPtr+12 > dictEnd) return ERROR(dictionary_corrupted);
+    dctx->rep[0] = MEM_readLE32(dictPtr+0); if (dctx->rep[0] >= dictSize) return ERROR(dictionary_corrupted);
+    dctx->rep[1] = MEM_readLE32(dictPtr+4); if (dctx->rep[1] >= dictSize) return ERROR(dictionary_corrupted);
+    dctx->rep[2] = MEM_readLE32(dictPtr+8); if (dctx->rep[2] >= dictSize) return ERROR(dictionary_corrupted);
+    dictPtr += 12;
+
+    dctx->litEntropy = dctx->fseEntropy = 1;
+    return dictPtr - (const BYTE*)dict;
+}
+
+static size_t ZSTD_decompress_insertDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
+{
+    if (dictSize < 8) return ZSTD_refDictContent(dctx, dict, dictSize);
+    {   U32 const magic = MEM_readLE32(dict);
+        if (magic != ZSTD_DICT_MAGIC) {
+            return ZSTD_refDictContent(dctx, dict, dictSize);   /* pure content mode */
+    }   }
+    dctx->dictID = MEM_readLE32((const char*)dict + 4);
+
+    /* load entropy tables */
+    dict = (const char*)dict + 8;
+    dictSize -= 8;
+    {   size_t const eSize = ZSTD_loadEntropy(dctx, dict, dictSize);
+        if (ZSTD_isError(eSize)) return ERROR(dictionary_corrupted);
+        dict = (const char*)dict + eSize;
+        dictSize -= eSize;
+    }
+
+    /* reference dictionary content */
+    return ZSTD_refDictContent(dctx, dict, dictSize);
+}
+
+size_t ZSTD_decompressBegin_usingDict(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
+{
+    CHECK_F(ZSTD_decompressBegin(dctx));
+    if (dict && dictSize) CHECK_E(ZSTD_decompress_insertDictionary(dctx, dict, dictSize), dictionary_corrupted);
+    return 0;
+}
+
+
+/* ======   ZSTD_DDict   ====== */
+
+struct ZSTD_DDict_s {
+    void* dict;
+    size_t dictSize;
+    ZSTD_DCtx* refContext;
+};  /* typedef'd to ZSTD_DDict within "zstd.h" */
+
+ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictSize, ZSTD_customMem customMem)
+{
+    if (!customMem.customAlloc && !customMem.customFree) customMem = defaultCustomMem;
+    if (!customMem.customAlloc || !customMem.customFree) return NULL;
+
+    {   ZSTD_DDict* const ddict = (ZSTD_DDict*) ZSTD_malloc(sizeof(ZSTD_DDict), customMem);
+        void* const dictContent = ZSTD_malloc(dictSize, customMem);
+        ZSTD_DCtx* const dctx = ZSTD_createDCtx_advanced(customMem);
+
+        if (!dictContent || !ddict || !dctx) {
+            ZSTD_free(dictContent, customMem);
+            ZSTD_free(ddict, customMem);
+            ZSTD_free(dctx, customMem);
+            return NULL;
+        }
+
+        if (dictSize) {
+            memcpy(dictContent, dict, dictSize);
+        }
+        {   size_t const errorCode = ZSTD_decompressBegin_usingDict(dctx, dictContent, dictSize);
+            if (ZSTD_isError(errorCode)) {
+                ZSTD_free(dictContent, customMem);
+                ZSTD_free(ddict, customMem);
+                ZSTD_free(dctx, customMem);
+                return NULL;
+        }   }
+
+        ddict->dict = dictContent;
+        ddict->dictSize = dictSize;
+        ddict->refContext = dctx;
+        return ddict;
+    }
+}
+
+/*! ZSTD_createDDict() :
+*   Create a digested dictionary, ready to start decompression without startup delay.
+*   `dict` can be released after `ZSTD_DDict` creation */
+ZSTD_DDict* ZSTD_createDDict(const void* dict, size_t dictSize)
+{
+    ZSTD_customMem const allocator = { NULL, NULL, NULL };
+    return ZSTD_createDDict_advanced(dict, dictSize, allocator);
+}
+
+size_t ZSTD_freeDDict(ZSTD_DDict* ddict)
+{
+    if (ddict==NULL) return 0;   /* support free on NULL */
+    {   ZSTD_customMem const cMem = ddict->refContext->customMem;
+        ZSTD_freeDCtx(ddict->refContext);
+        ZSTD_free(ddict->dict, cMem);
+        ZSTD_free(ddict, cMem);
+        return 0;
+    }
+}
+
+size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict)
+{
+    if (ddict==NULL) return 0;   /* support sizeof on NULL */
+    return sizeof(*ddict) + sizeof(ddict->refContext) + ddict->dictSize;
+}
+
+/*! ZSTD_getDictID_fromDict() :
+ *  Provides the dictID stored within dictionary.
+ *  if @return == 0, the dictionary is not conformant with Zstandard specification.
+ *  It can still be loaded, but as a content-only dictionary. */
+unsigned ZSTD_getDictID_fromDict(const void* dict, size_t dictSize)
+{
+    if (dictSize < 8) return 0;
+    if (MEM_readLE32(dict) != ZSTD_DICT_MAGIC) return 0;
+    return MEM_readLE32((const char*)dict + 4);
+}
+
+/*! ZSTD_getDictID_fromDDict() :
+ *  Provides the dictID of the dictionary loaded into `ddict`.
+ *  If @return == 0, the dictionary is not conformant to Zstandard specification, or empty.
+ *  Non-conformant dictionaries can still be loaded, but as content-only dictionaries. */
+unsigned ZSTD_getDictID_fromDDict(const ZSTD_DDict* ddict)
+{
+    if (ddict==NULL) return 0;
+    return ZSTD_getDictID_fromDict(ddict->dict, ddict->dictSize);
+}
+
+/*! ZSTD_getDictID_fromFrame() :
+ *  Provides the dictID required to decompressed the frame stored within `src`.
+ *  If @return == 0, the dictID could not be decoded.
+ *  This could for one of the following reasons :
+ *  - The frame does not require a dictionary to be decoded (most common case).
+ *  - The frame was built with dictID intentionally removed. Whatever dictionary is necessary is a hidden information.
+ *    Note : this use case also happens when using a non-conformant dictionary.
+ *  - `srcSize` is too small, and as a result, the frame header could not be decoded (only possible if `srcSize < ZSTD_FRAMEHEADERSIZE_MAX`).
+ *  - This is not a Zstandard frame.
+ *  When identifying the exact failure cause, it's possible to used ZSTD_getFrameParams(), which will provide a more precise error code. */
+unsigned ZSTD_getDictID_fromFrame(const void* src, size_t srcSize)
+{
+    ZSTD_frameParams zfp = { 0 , 0 , 0 , 0 };
+    size_t const hError = ZSTD_getFrameParams(&zfp, src, srcSize);
+    if (ZSTD_isError(hError)) return 0;
+    return zfp.dictID;
+}
+
+
+/*! ZSTD_decompress_usingDDict() :
+*   Decompression using a pre-digested Dictionary
+*   Use dictionary without significant overhead. */
+size_t ZSTD_decompress_usingDDict(ZSTD_DCtx* dctx,
+                                  void* dst, size_t dstCapacity,
+                            const void* src, size_t srcSize,
+                            const ZSTD_DDict* ddict)
+{
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT==1)
+    if (ZSTD_isLegacy(src, srcSize)) return ZSTD_decompressLegacy(dst, dstCapacity, src, srcSize, ddict->dict, ddict->dictSize);
+#endif
+    ZSTD_refDCtx(dctx, ddict->refContext);
+    ZSTD_checkContinuity(dctx, dst);
+    return ZSTD_decompressFrame(dctx, dst, dstCapacity, src, srcSize);
+}
+
+
+/*=====================================
+*   Streaming decompression
+*====================================*/
+
+typedef enum { zdss_init, zdss_loadHeader,
+               zdss_read, zdss_load, zdss_flush } ZSTD_dStreamStage;
+
+/* *** Resource management *** */
+struct ZSTD_DStream_s {
+    ZSTD_DCtx* dctx;
+    ZSTD_DDict* ddictLocal;
+    const ZSTD_DDict* ddict;
+    ZSTD_frameParams fParams;
+    ZSTD_dStreamStage stage;
+    char*  inBuff;
+    size_t inBuffSize;
+    size_t inPos;
+    size_t maxWindowSize;
+    char*  outBuff;
+    size_t outBuffSize;
+    size_t outStart;
+    size_t outEnd;
+    size_t blockSize;
+    BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX];   /* tmp buffer to store frame header */
+    size_t lhSize;
+    ZSTD_customMem customMem;
+    void* legacyContext;
+    U32 previousLegacyVersion;
+    U32 legacyVersion;
+    U32 hostageByte;
+};   /* typedef'd to ZSTD_DStream within "zstd.h" */
+
+
+ZSTD_DStream* ZSTD_createDStream(void)
+{
+    return ZSTD_createDStream_advanced(defaultCustomMem);
+}
+
+ZSTD_DStream* ZSTD_createDStream_advanced(ZSTD_customMem customMem)
+{
+    ZSTD_DStream* zds;
+
+    if (!customMem.customAlloc && !customMem.customFree) customMem = defaultCustomMem;
+    if (!customMem.customAlloc || !customMem.customFree) return NULL;
+
+    zds = (ZSTD_DStream*) ZSTD_malloc(sizeof(ZSTD_DStream), customMem);
+    if (zds==NULL) return NULL;
+    memset(zds, 0, sizeof(ZSTD_DStream));
+    memcpy(&zds->customMem, &customMem, sizeof(ZSTD_customMem));
+    zds->dctx = ZSTD_createDCtx_advanced(customMem);
+    if (zds->dctx == NULL) { ZSTD_freeDStream(zds); return NULL; }
+    zds->stage = zdss_init;
+    zds->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
+    return zds;
+}
+
+size_t ZSTD_freeDStream(ZSTD_DStream* zds)
+{
+    if (zds==NULL) return 0;   /* support free on null */
+    {   ZSTD_customMem const cMem = zds->customMem;
+        ZSTD_freeDCtx(zds->dctx);
+        ZSTD_freeDDict(zds->ddictLocal);
+        ZSTD_free(zds->inBuff, cMem);
+        ZSTD_free(zds->outBuff, cMem);
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
+        if (zds->legacyContext)
+            ZSTD_freeLegacyStreamContext(zds->legacyContext, zds->previousLegacyVersion);
+#endif
+        ZSTD_free(zds, cMem);
+        return 0;
+    }
+}
+
+
+/* *** Initialization *** */
+
+size_t ZSTD_DStreamInSize(void)  { return ZSTD_BLOCKSIZE_ABSOLUTEMAX + ZSTD_blockHeaderSize; }
+size_t ZSTD_DStreamOutSize(void) { return ZSTD_BLOCKSIZE_ABSOLUTEMAX; }
+
+size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize)
+{
+    zds->stage = zdss_loadHeader;
+    zds->lhSize = zds->inPos = zds->outStart = zds->outEnd = 0;
+    ZSTD_freeDDict(zds->ddictLocal);
+    if (dict) {
+        zds->ddictLocal = ZSTD_createDDict(dict, dictSize);
+        if (zds->ddictLocal == NULL) return ERROR(memory_allocation);
+    } else zds->ddictLocal = NULL;
+    zds->ddict = zds->ddictLocal;
+    zds->legacyVersion = 0;
+    zds->hostageByte = 0;
+    return ZSTD_frameHeaderSize_prefix;
+}
+
+size_t ZSTD_initDStream(ZSTD_DStream* zds)
+{
+    return ZSTD_initDStream_usingDict(zds, NULL, 0);
+}
+
+size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* zds, const ZSTD_DDict* ddict)  /**< note : ddict will just be referenced, and must outlive decompression session */
+{
+    size_t const initResult = ZSTD_initDStream(zds);
+    zds->ddict = ddict;
+    return initResult;
+}
+
+size_t ZSTD_resetDStream(ZSTD_DStream* zds)
+{
+    zds->stage = zdss_loadHeader;
+    zds->lhSize = zds->inPos = zds->outStart = zds->outEnd = 0;
+    zds->legacyVersion = 0;
+    zds->hostageByte = 0;
+    return ZSTD_frameHeaderSize_prefix;
+}
+
+size_t ZSTD_setDStreamParameter(ZSTD_DStream* zds,
+                                ZSTD_DStreamParameter_e paramType, unsigned paramValue)
+{
+    switch(paramType)
+    {
+        default : return ERROR(parameter_unknown);
+        case ZSTDdsp_maxWindowSize : zds->maxWindowSize = paramValue ? paramValue : (U32)(-1); break;
+    }
+    return 0;
+}
+
+
+size_t ZSTD_sizeof_DStream(const ZSTD_DStream* zds)
+{
+    if (zds==NULL) return 0;   /* support sizeof on NULL */
+    return sizeof(*zds) + ZSTD_sizeof_DCtx(zds->dctx) + ZSTD_sizeof_DDict(zds->ddictLocal) + zds->inBuffSize + zds->outBuffSize;
+}
+
+
+/* *****   Decompression   ***** */
+
+MEM_STATIC size_t ZSTD_limitCopy(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+    size_t const length = MIN(dstCapacity, srcSize);
+    memcpy(dst, src, length);
+    return length;
+}
+
+
+size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
+{
+    const char* const istart = (const char*)(input->src) + input->pos;
+    const char* const iend = (const char*)(input->src) + input->size;
+    const char* ip = istart;
+    char* const ostart = (char*)(output->dst) + output->pos;
+    char* const oend = (char*)(output->dst) + output->size;
+    char* op = ostart;
+    U32 someMoreWork = 1;
+
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
+    if (zds->legacyVersion)
+        return ZSTD_decompressLegacyStream(zds->legacyContext, zds->legacyVersion, output, input);
+#endif
+
+    while (someMoreWork) {
+        switch(zds->stage)
+        {
+        case zdss_init :
+            ZSTD_resetDStream(zds);   /* transparent reset on starting decoding a new frame */
+            /* fall-through */
+
+        case zdss_loadHeader :
+            {   size_t const hSize = ZSTD_getFrameParams(&zds->fParams, zds->headerBuffer, zds->lhSize);
+                if (ZSTD_isError(hSize))
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
+                {   U32 const legacyVersion = ZSTD_isLegacy(istart, iend-istart);
+                    if (legacyVersion) {
+                        const void* const dict = zds->ddict ? zds->ddict->dict : NULL;
+                        size_t const dictSize = zds->ddict ? zds->ddict->dictSize : 0;
+                        CHECK_F(ZSTD_initLegacyStream(&zds->legacyContext, zds->previousLegacyVersion, legacyVersion,
+                                                       dict, dictSize));
+                        zds->legacyVersion = zds->previousLegacyVersion = legacyVersion;
+                        return ZSTD_decompressLegacyStream(zds->legacyContext, zds->legacyVersion, output, input);
+                    } else {
+                        return hSize; /* error */
+                }   }
+#else
+                return hSize;
+#endif
+                if (hSize != 0) {   /* need more input */
+                    size_t const toLoad = hSize - zds->lhSize;   /* if hSize!=0, hSize > zds->lhSize */
+                    if (toLoad > (size_t)(iend-ip)) {   /* not enough input to load full header */
+                        memcpy(zds->headerBuffer + zds->lhSize, ip, iend-ip);
+                        zds->lhSize += iend-ip;
+                        input->pos = input->size;
+                        return (MAX(ZSTD_frameHeaderSize_min, hSize) - zds->lhSize) + ZSTD_blockHeaderSize;   /* remaining header bytes + next block header */
+                    }
+                    memcpy(zds->headerBuffer + zds->lhSize, ip, toLoad); zds->lhSize = hSize; ip += toLoad;
+                    break;
+            }   }
+
+            /* Consume header */
+            {   const ZSTD_DCtx* refContext = zds->ddict ? zds->ddict->refContext : NULL;
+                ZSTD_refDCtx(zds->dctx, refContext);
+            }
+            {   size_t const h1Size = ZSTD_nextSrcSizeToDecompress(zds->dctx);  /* == ZSTD_frameHeaderSize_prefix */
+                CHECK_F(ZSTD_decompressContinue(zds->dctx, NULL, 0, zds->headerBuffer, h1Size));
+                {   size_t const h2Size = ZSTD_nextSrcSizeToDecompress(zds->dctx);
+                    CHECK_F(ZSTD_decompressContinue(zds->dctx, NULL, 0, zds->headerBuffer+h1Size, h2Size));
+            }   }
+
+            zds->fParams.windowSize = MAX(zds->fParams.windowSize, 1U << ZSTD_WINDOWLOG_ABSOLUTEMIN);
+            if (zds->fParams.windowSize > zds->maxWindowSize) return ERROR(frameParameter_windowTooLarge);
+
+            /* Adapt buffer sizes to frame header instructions */
+            {   size_t const blockSize = MIN(zds->fParams.windowSize, ZSTD_BLOCKSIZE_ABSOLUTEMAX);
+                size_t const neededOutSize = zds->fParams.windowSize + blockSize;
+                zds->blockSize = blockSize;
+                if (zds->inBuffSize < blockSize) {
+                    ZSTD_free(zds->inBuff, zds->customMem);
+                    zds->inBuffSize = blockSize;
+                    zds->inBuff = (char*)ZSTD_malloc(blockSize, zds->customMem);
+                    if (zds->inBuff == NULL) return ERROR(memory_allocation);
+                }
+                if (zds->outBuffSize < neededOutSize) {
+                    ZSTD_free(zds->outBuff, zds->customMem);
+                    zds->outBuffSize = neededOutSize;
+                    zds->outBuff = (char*)ZSTD_malloc(neededOutSize, zds->customMem);
+                    if (zds->outBuff == NULL) return ERROR(memory_allocation);
+            }   }
+            zds->stage = zdss_read;
+            /* pass-through */
+
+        case zdss_read:
+            {   size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zds->dctx);
+                if (neededInSize==0) {  /* end of frame */
+                    zds->stage = zdss_init;
+                    someMoreWork = 0;
+                    break;
+                }
+                if ((size_t)(iend-ip) >= neededInSize) {  /* decode directly from src */
+                    const int isSkipFrame = ZSTD_isSkipFrame(zds->dctx);
+                    size_t const decodedSize = ZSTD_decompressContinue(zds->dctx,
+                        zds->outBuff + zds->outStart, (isSkipFrame ? 0 : zds->outBuffSize - zds->outStart),
+                        ip, neededInSize);
+                    if (ZSTD_isError(decodedSize)) return decodedSize;
+                    ip += neededInSize;
+                    if (!decodedSize && !isSkipFrame) break;   /* this was just a header */
+                    zds->outEnd = zds->outStart + decodedSize;
+                    zds->stage = zdss_flush;
+                    break;
+                }
+                if (ip==iend) { someMoreWork = 0; break; }   /* no more input */
+                zds->stage = zdss_load;
+                /* pass-through */
+            }
+
+        case zdss_load:
+            {   size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zds->dctx);
+                size_t const toLoad = neededInSize - zds->inPos;   /* should always be <= remaining space within inBuff */
+                size_t loadedSize;
+                if (toLoad > zds->inBuffSize - zds->inPos) return ERROR(corruption_detected);   /* should never happen */
+                loadedSize = ZSTD_limitCopy(zds->inBuff + zds->inPos, toLoad, ip, iend-ip);
+                ip += loadedSize;
+                zds->inPos += loadedSize;
+                if (loadedSize < toLoad) { someMoreWork = 0; break; }   /* not enough input, wait for more */
+
+                /* decode loaded input */
+                {  const int isSkipFrame = ZSTD_isSkipFrame(zds->dctx);
+                   size_t const decodedSize = ZSTD_decompressContinue(zds->dctx,
+                        zds->outBuff + zds->outStart, zds->outBuffSize - zds->outStart,
+                        zds->inBuff, neededInSize);
+                    if (ZSTD_isError(decodedSize)) return decodedSize;
+                    zds->inPos = 0;   /* input is consumed */
+                    if (!decodedSize && !isSkipFrame) { zds->stage = zdss_read; break; }   /* this was just a header */
+                    zds->outEnd = zds->outStart +  decodedSize;
+                    zds->stage = zdss_flush;
+                    /* pass-through */
+            }   }
+
+        case zdss_flush:
+            {   size_t const toFlushSize = zds->outEnd - zds->outStart;
+                size_t const flushedSize = ZSTD_limitCopy(op, oend-op, zds->outBuff + zds->outStart, toFlushSize);
+                op += flushedSize;
+                zds->outStart += flushedSize;
+                if (flushedSize == toFlushSize) {  /* flush completed */
+                    zds->stage = zdss_read;
+                    if (zds->outStart + zds->blockSize > zds->outBuffSize)
+                        zds->outStart = zds->outEnd = 0;
+                    break;
+                }
+                /* cannot complete flush */
+                someMoreWork = 0;
+                break;
+            }
+        default: return ERROR(GENERIC);   /* impossible */
+    }   }
+
+    /* result */
+    input->pos += (size_t)(ip-istart);
+    output->pos += (size_t)(op-ostart);
+    {   size_t nextSrcSizeHint = ZSTD_nextSrcSizeToDecompress(zds->dctx);
+        if (!nextSrcSizeHint) {   /* frame fully decoded */
+            if (zds->outEnd == zds->outStart) {  /* output fully flushed */
+                if (zds->hostageByte) {
+                    if (input->pos >= input->size) { zds->stage = zdss_read; return 1; }  /* can't release hostage (not present) */
+                    input->pos++;  /* release hostage */
+                }
+                return 0;
+            }
+            if (!zds->hostageByte) { /* output not fully flushed; keep last byte as hostage; will be released when all output is flushed */
+                input->pos--;   /* note : pos > 0, otherwise, impossible to finish reading last block */
+                zds->hostageByte=1;
+            }
+            return 1;
+        }
+        nextSrcSizeHint += ZSTD_blockHeaderSize * (ZSTD_nextInputType(zds->dctx) == ZSTDnit_block);   /* preload header of next block */
+        if (zds->inPos > nextSrcSizeHint) return ERROR(GENERIC);   /* should never happen */
+        nextSrcSizeHint -= zds->inPos;   /* already loaded*/
+        return nextSrcSizeHint;
+    }
+}
diff --git a/lib/deprecated/zbuff.h b/lib/deprecated/zbuff.h
new file mode 100644
index 0000000..85f9735
--- /dev/null
+++ b/lib/deprecated/zbuff.h
@@ -0,0 +1,210 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+/* ***************************************************************
+*  NOTES/WARNINGS
+******************************************************************/
+/* The streaming API defined here is deprecated.
+ * Consider migrating towards ZSTD_compressStream() API in `zstd.h`
+ * See 'lib/README.md'.
+ *****************************************************************/
+
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+#ifndef ZSTD_BUFFERED_H_23987
+#define ZSTD_BUFFERED_H_23987
+
+/* *************************************
+*  Dependencies
+***************************************/
+#include       /* size_t */
+#include "zstd.h"        /* ZSTD_CStream, ZSTD_DStream, ZSTDLIB_API */
+
+
+/* ***************************************************************
+*  Compiler specifics
+*****************************************************************/
+/* Deprecation warnings */
+/* Should these warnings be a problem,
+   it is generally possible to disable them,
+   typically with -Wno-deprecated-declarations for gcc
+   or _CRT_SECURE_NO_WARNINGS in Visual.
+   Otherwise, it's also possible to define ZBUFF_DISABLE_DEPRECATE_WARNINGS */
+#ifdef ZBUFF_DISABLE_DEPRECATE_WARNINGS
+#  define ZBUFF_DEPRECATED(message) ZSTDLIB_API  /* disable deprecation warnings */
+#else
+#  if (defined(__GNUC__) && (__GNUC__ >= 5)) || defined(__clang__)
+#    define ZBUFF_DEPRECATED(message) ZSTDLIB_API __attribute__((deprecated(message)))
+#  elif defined(__GNUC__) && (__GNUC__ >= 3)
+#    define ZBUFF_DEPRECATED(message) ZSTDLIB_API __attribute__((deprecated))
+#  elif defined(_MSC_VER)
+#    define ZBUFF_DEPRECATED(message) ZSTDLIB_API __declspec(deprecated(message))
+#  else
+#    pragma message("WARNING: You need to implement ZBUFF_DEPRECATED for this compiler")
+#    define ZBUFF_DEPRECATED(message) ZSTDLIB_API
+#  endif
+#endif /* ZBUFF_DISABLE_DEPRECATE_WARNINGS */
+
+
+/* *************************************
+*  Streaming functions
+***************************************/
+/* This is the easier "buffered" streaming API,
+*  using an internal buffer to lift all restrictions on user-provided buffers
+*  which can be any size, any place, for both input and output.
+*  ZBUFF and ZSTD are 100% interoperable,
+*  frames created by one can be decoded by the other one */
+
+typedef ZSTD_CStream ZBUFF_CCtx;
+ZBUFF_DEPRECATED("use ZSTD_createCStream") ZBUFF_CCtx* ZBUFF_createCCtx(void);
+ZBUFF_DEPRECATED("use ZSTD_freeCStream")   size_t      ZBUFF_freeCCtx(ZBUFF_CCtx* cctx);
+
+ZBUFF_DEPRECATED("use ZSTD_initCStream")           size_t ZBUFF_compressInit(ZBUFF_CCtx* cctx, int compressionLevel);
+ZBUFF_DEPRECATED("use ZSTD_initCStream_usingDict") size_t ZBUFF_compressInitDictionary(ZBUFF_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel);
+
+ZBUFF_DEPRECATED("use ZSTD_compressStream") size_t ZBUFF_compressContinue(ZBUFF_CCtx* cctx, void* dst, size_t* dstCapacityPtr, const void* src, size_t* srcSizePtr);
+ZBUFF_DEPRECATED("use ZSTD_flushStream")    size_t ZBUFF_compressFlush(ZBUFF_CCtx* cctx, void* dst, size_t* dstCapacityPtr);
+ZBUFF_DEPRECATED("use ZSTD_endStream")      size_t ZBUFF_compressEnd(ZBUFF_CCtx* cctx, void* dst, size_t* dstCapacityPtr);
+
+/*-*************************************************
+*  Streaming compression - howto
+*
+*  A ZBUFF_CCtx object is required to track streaming operation.
+*  Use ZBUFF_createCCtx() and ZBUFF_freeCCtx() to create/release resources.
+*  ZBUFF_CCtx objects can be reused multiple times.
+*
+*  Start by initializing ZBUF_CCtx.
+*  Use ZBUFF_compressInit() to start a new compression operation.
+*  Use ZBUFF_compressInitDictionary() for a compression which requires a dictionary.
+*
+*  Use ZBUFF_compressContinue() repetitively to consume input stream.
+*  *srcSizePtr and *dstCapacityPtr can be any size.
+*  The function will report how many bytes were read or written within *srcSizePtr and *dstCapacityPtr.
+*  Note that it may not consume the entire input, in which case it's up to the caller to present again remaining data.
+*  The content of `dst` will be overwritten (up to *dstCapacityPtr) at each call, so save its content if it matters or change @dst .
+*  @return : a hint to preferred nb of bytes to use as input for next function call (it's just a hint, to improve latency)
+*            or an error code, which can be tested using ZBUFF_isError().
+*
+*  At any moment, it's possible to flush whatever data remains within buffer, using ZBUFF_compressFlush().
+*  The nb of bytes written into `dst` will be reported into *dstCapacityPtr.
+*  Note that the function cannot output more than *dstCapacityPtr,
+*  therefore, some content might still be left into internal buffer if *dstCapacityPtr is too small.
+*  @return : nb of bytes still present into internal buffer (0 if it's empty)
+*            or an error code, which can be tested using ZBUFF_isError().
+*
+*  ZBUFF_compressEnd() instructs to finish a frame.
+*  It will perform a flush and write frame epilogue.
+*  The epilogue is required for decoders to consider a frame completed.
+*  Similar to ZBUFF_compressFlush(), it may not be able to output the entire internal buffer content if *dstCapacityPtr is too small.
+*  In which case, call again ZBUFF_compressFlush() to complete the flush.
+*  @return : nb of bytes still present into internal buffer (0 if it's empty)
+*            or an error code, which can be tested using ZBUFF_isError().
+*
+*  Hint : _recommended buffer_ sizes (not compulsory) : ZBUFF_recommendedCInSize() / ZBUFF_recommendedCOutSize()
+*  input : ZBUFF_recommendedCInSize==128 KB block size is the internal unit, use this value to reduce intermediate stages (better latency)
+*  output : ZBUFF_recommendedCOutSize==ZSTD_compressBound(128 KB) + 3 + 3 : ensures it's always possible to write/flush/end a full block. Skip some buffering.
+*  By using both, it ensures that input will be entirely consumed, and output will always contain the result, reducing intermediate buffering.
+* **************************************************/
+
+
+typedef ZSTD_DStream ZBUFF_DCtx;
+ZBUFF_DEPRECATED("use ZSTD_createDStream") ZBUFF_DCtx* ZBUFF_createDCtx(void);
+ZBUFF_DEPRECATED("use ZSTD_freeDStream")   size_t      ZBUFF_freeDCtx(ZBUFF_DCtx* dctx);
+
+ZBUFF_DEPRECATED("use ZSTD_initDStream")           size_t ZBUFF_decompressInit(ZBUFF_DCtx* dctx);
+ZBUFF_DEPRECATED("use ZSTD_initDStream_usingDict") size_t ZBUFF_decompressInitDictionary(ZBUFF_DCtx* dctx, const void* dict, size_t dictSize);
+
+ZBUFF_DEPRECATED("use ZSTD_decompressStream") size_t ZBUFF_decompressContinue(ZBUFF_DCtx* dctx,
+                                            void* dst, size_t* dstCapacityPtr,
+                                      const void* src, size_t* srcSizePtr);
+
+/*-***************************************************************************
+*  Streaming decompression howto
+*
+*  A ZBUFF_DCtx object is required to track streaming operations.
+*  Use ZBUFF_createDCtx() and ZBUFF_freeDCtx() to create/release resources.
+*  Use ZBUFF_decompressInit() to start a new decompression operation,
+*   or ZBUFF_decompressInitDictionary() if decompression requires a dictionary.
+*  Note that ZBUFF_DCtx objects can be re-init multiple times.
+*
+*  Use ZBUFF_decompressContinue() repetitively to consume your input.
+*  *srcSizePtr and *dstCapacityPtr can be any size.
+*  The function will report how many bytes were read or written by modifying *srcSizePtr and *dstCapacityPtr.
+*  Note that it may not consume the entire input, in which case it's up to the caller to present remaining input again.
+*  The content of `dst` will be overwritten (up to *dstCapacityPtr) at each function call, so save its content if it matters, or change `dst`.
+*  @return : 0 when a frame is completely decoded and fully flushed,
+*            1 when there is still some data left within internal buffer to flush,
+*            >1 when more data is expected, with value being a suggested next input size (it's just a hint, which helps latency),
+*            or an error code, which can be tested using ZBUFF_isError().
+*
+*  Hint : recommended buffer sizes (not compulsory) : ZBUFF_recommendedDInSize() and ZBUFF_recommendedDOutSize()
+*  output : ZBUFF_recommendedDOutSize== 128 KB block size is the internal unit, it ensures it's always possible to write a full block when decoded.
+*  input  : ZBUFF_recommendedDInSize == 128KB + 3;
+*           just follow indications from ZBUFF_decompressContinue() to minimize latency. It should always be <= 128 KB + 3 .
+* *******************************************************************************/
+
+
+/* *************************************
+*  Tool functions
+***************************************/
+ZBUFF_DEPRECATED("use ZSTD_isError")      unsigned ZBUFF_isError(size_t errorCode);
+ZBUFF_DEPRECATED("use ZSTD_getErrorName") const char* ZBUFF_getErrorName(size_t errorCode);
+
+/** Functions below provide recommended buffer sizes for Compression or Decompression operations.
+*   These sizes are just hints, they tend to offer better latency */
+ZBUFF_DEPRECATED("use ZSTD_CStreamInSize")  size_t ZBUFF_recommendedCInSize(void);
+ZBUFF_DEPRECATED("use ZSTD_CStreamOutSize") size_t ZBUFF_recommendedCOutSize(void);
+ZBUFF_DEPRECATED("use ZSTD_DStreamInSize")  size_t ZBUFF_recommendedDInSize(void);
+ZBUFF_DEPRECATED("use ZSTD_DStreamOutSize") size_t ZBUFF_recommendedDOutSize(void);
+
+#endif  /* ZSTD_BUFFERED_H_23987 */
+
+
+#ifdef ZBUFF_STATIC_LINKING_ONLY
+#ifndef ZBUFF_STATIC_H_30298098432
+#define ZBUFF_STATIC_H_30298098432
+
+/* ====================================================================================
+ * The definitions in this section are considered experimental.
+ * They should never be used in association with a dynamic library, as they may change in the future.
+ * They are provided for advanced usages.
+ * Use them only in association with static linking.
+ * ==================================================================================== */
+
+/*--- Dependency ---*/
+#define ZSTD_STATIC_LINKING_ONLY   /* ZSTD_parameters, ZSTD_customMem */
+#include "zstd.h"
+
+
+/*--- Custom memory allocator ---*/
+/*! ZBUFF_createCCtx_advanced() :
+ *  Create a ZBUFF compression context using external alloc and free functions */
+ZBUFF_DEPRECATED("use ZSTD_createCStream_advanced") ZBUFF_CCtx* ZBUFF_createCCtx_advanced(ZSTD_customMem customMem);
+
+/*! ZBUFF_createDCtx_advanced() :
+ *  Create a ZBUFF decompression context using external alloc and free functions */
+ZBUFF_DEPRECATED("use ZSTD_createDStream_advanced") ZBUFF_DCtx* ZBUFF_createDCtx_advanced(ZSTD_customMem customMem);
+
+
+/*--- Advanced Streaming Initialization ---*/
+ZBUFF_DEPRECATED("use ZSTD_initDStream_usingDict") size_t ZBUFF_compressInit_advanced(ZBUFF_CCtx* zbc,
+                                               const void* dict, size_t dictSize,
+                                               ZSTD_parameters params, unsigned long long pledgedSrcSize);
+
+
+#endif    /* ZBUFF_STATIC_H_30298098432 */
+#endif    /* ZBUFF_STATIC_LINKING_ONLY */
+
+
+#if defined (__cplusplus)
+}
+#endif
diff --git a/lib/deprecated/zbuff_compress.c b/lib/deprecated/zbuff_compress.c
new file mode 100644
index 0000000..5a37a00
--- /dev/null
+++ b/lib/deprecated/zbuff_compress.c
@@ -0,0 +1,145 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+
+
+/* *************************************
+*  Dependencies
+***************************************/
+#define ZBUFF_STATIC_LINKING_ONLY
+#include "zbuff.h"
+
+
+/*-***********************************************************
+*  Streaming compression
+*
+*  A ZBUFF_CCtx object is required to track streaming operation.
+*  Use ZBUFF_createCCtx() and ZBUFF_freeCCtx() to create/release resources.
+*  Use ZBUFF_compressInit() to start a new compression operation.
+*  ZBUFF_CCtx objects can be reused multiple times.
+*
+*  Use ZBUFF_compressContinue() repetitively to consume your input.
+*  *srcSizePtr and *dstCapacityPtr can be any size.
+*  The function will report how many bytes were read or written by modifying *srcSizePtr and *dstCapacityPtr.
+*  Note that it may not consume the entire input, in which case it's up to the caller to call again the function with remaining input.
+*  The content of dst will be overwritten (up to *dstCapacityPtr) at each function call, so save its content if it matters or change dst .
+*  @return : a hint to preferred nb of bytes to use as input for next function call (it's only a hint, to improve latency)
+*            or an error code, which can be tested using ZBUFF_isError().
+*
+*  ZBUFF_compressFlush() can be used to instruct ZBUFF to compress and output whatever remains within its buffer.
+*  Note that it will not output more than *dstCapacityPtr.
+*  Therefore, some content might still be left into its internal buffer if dst buffer is too small.
+*  @return : nb of bytes still present into internal buffer (0 if it's empty)
+*            or an error code, which can be tested using ZBUFF_isError().
+*
+*  ZBUFF_compressEnd() instructs to finish a frame.
+*  It will perform a flush and write frame epilogue.
+*  Similar to ZBUFF_compressFlush(), it may not be able to output the entire internal buffer content if *dstCapacityPtr is too small.
+*  @return : nb of bytes still present into internal buffer (0 if it's empty)
+*            or an error code, which can be tested using ZBUFF_isError().
+*
+*  Hint : recommended buffer sizes (not compulsory)
+*  input : ZSTD_BLOCKSIZE_MAX (128 KB), internal unit size, it improves latency to use this value.
+*  output : ZSTD_compressBound(ZSTD_BLOCKSIZE_MAX) + ZSTD_blockHeaderSize + ZBUFF_endFrameSize : ensures it's always possible to write/flush/end a full block at best speed.
+* ***********************************************************/
+
+ZBUFF_CCtx* ZBUFF_createCCtx(void)
+{
+    return ZSTD_createCStream();
+}
+
+ZBUFF_CCtx* ZBUFF_createCCtx_advanced(ZSTD_customMem customMem)
+{
+    return ZSTD_createCStream_advanced(customMem);
+}
+
+size_t ZBUFF_freeCCtx(ZBUFF_CCtx* zbc)
+{
+    return ZSTD_freeCStream(zbc);
+}
+
+
+/* ======   Initialization   ====== */
+
+size_t ZBUFF_compressInit_advanced(ZBUFF_CCtx* zbc,
+                                   const void* dict, size_t dictSize,
+                                   ZSTD_parameters params, unsigned long long pledgedSrcSize)
+{
+    return ZSTD_initCStream_advanced(zbc, dict, dictSize, params, pledgedSrcSize);
+}
+
+
+size_t ZBUFF_compressInitDictionary(ZBUFF_CCtx* zbc, const void* dict, size_t dictSize, int compressionLevel)
+{
+    return ZSTD_initCStream_usingDict(zbc, dict, dictSize, compressionLevel);
+}
+
+size_t ZBUFF_compressInit(ZBUFF_CCtx* zbc, int compressionLevel)
+{
+    return ZSTD_initCStream(zbc, compressionLevel);
+}
+
+/* ======   Compression   ====== */
+
+
+size_t ZBUFF_compressContinue(ZBUFF_CCtx* zbc,
+                              void* dst, size_t* dstCapacityPtr,
+                        const void* src, size_t* srcSizePtr)
+{
+    size_t result;
+    ZSTD_outBuffer outBuff;
+    ZSTD_inBuffer inBuff;
+    outBuff.dst = dst;
+    outBuff.pos = 0;
+    outBuff.size = *dstCapacityPtr;
+    inBuff.src = src;
+    inBuff.pos = 0;
+    inBuff.size = *srcSizePtr;
+    result = ZSTD_compressStream(zbc, &outBuff, &inBuff);
+    *dstCapacityPtr = outBuff.pos;
+    *srcSizePtr = inBuff.pos;
+    return result;
+}
+
+
+
+/* ======   Finalize   ====== */
+
+size_t ZBUFF_compressFlush(ZBUFF_CCtx* zbc, void* dst, size_t* dstCapacityPtr)
+{
+    size_t result;
+    ZSTD_outBuffer outBuff;
+    outBuff.dst = dst;
+    outBuff.pos = 0;
+    outBuff.size = *dstCapacityPtr;
+    result = ZSTD_flushStream(zbc, &outBuff);
+    *dstCapacityPtr = outBuff.pos;
+    return result;
+}
+
+
+size_t ZBUFF_compressEnd(ZBUFF_CCtx* zbc, void* dst, size_t* dstCapacityPtr)
+{
+    size_t result;
+    ZSTD_outBuffer outBuff;
+    outBuff.dst = dst;
+    outBuff.pos = 0;
+    outBuff.size = *dstCapacityPtr;
+    result = ZSTD_endStream(zbc, &outBuff);
+    *dstCapacityPtr = outBuff.pos;
+    return result;
+}
+
+
+
+/* *************************************
+*  Tool functions
+***************************************/
+size_t ZBUFF_recommendedCInSize(void)  { return ZSTD_CStreamInSize(); }
+size_t ZBUFF_recommendedCOutSize(void) { return ZSTD_CStreamOutSize(); }
diff --git a/lib/deprecated/zbuff_decompress.c b/lib/deprecated/zbuff_decompress.c
new file mode 100644
index 0000000..d9c155e
--- /dev/null
+++ b/lib/deprecated/zbuff_decompress.c
@@ -0,0 +1,74 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+
+
+/* *************************************
+*  Dependencies
+***************************************/
+#define ZBUFF_STATIC_LINKING_ONLY
+#include "zbuff.h"
+
+
+ZBUFF_DCtx* ZBUFF_createDCtx(void)
+{
+    return ZSTD_createDStream();
+}
+
+ZBUFF_DCtx* ZBUFF_createDCtx_advanced(ZSTD_customMem customMem)
+{
+    return ZSTD_createDStream_advanced(customMem);
+}
+
+size_t ZBUFF_freeDCtx(ZBUFF_DCtx* zbd)
+{
+    return ZSTD_freeDStream(zbd);
+}
+
+
+/* *** Initialization *** */
+
+size_t ZBUFF_decompressInitDictionary(ZBUFF_DCtx* zbd, const void* dict, size_t dictSize)
+{
+    return ZSTD_initDStream_usingDict(zbd, dict, dictSize);
+}
+
+size_t ZBUFF_decompressInit(ZBUFF_DCtx* zbd)
+{
+    return ZSTD_initDStream(zbd);
+}
+
+
+/* *** Decompression *** */
+
+size_t ZBUFF_decompressContinue(ZBUFF_DCtx* zbd,
+                                void* dst, size_t* dstCapacityPtr,
+                          const void* src, size_t* srcSizePtr)
+{
+    ZSTD_outBuffer outBuff;
+    ZSTD_inBuffer inBuff;
+    size_t result;
+    outBuff.dst  = dst;
+    outBuff.pos  = 0;
+    outBuff.size = *dstCapacityPtr;
+    inBuff.src  = src;
+    inBuff.pos  = 0;
+    inBuff.size = *srcSizePtr;
+    result = ZSTD_decompressStream(zbd, &outBuff, &inBuff);
+    *dstCapacityPtr = outBuff.pos;
+    *srcSizePtr = inBuff.pos;
+    return result;
+}
+
+
+/* *************************************
+*  Tool functions
+***************************************/
+size_t ZBUFF_recommendedDInSize(void)  { return ZSTD_DStreamInSize(); }
+size_t ZBUFF_recommendedDOutSize(void) { return ZSTD_DStreamOutSize(); }
diff --git a/lib/dictBuilder/divsufsort.c b/lib/dictBuilder/divsufsort.c
new file mode 100644
index 0000000..60cceb0
--- /dev/null
+++ b/lib/dictBuilder/divsufsort.c
@@ -0,0 +1,1913 @@
+/*
+ * divsufsort.c for libdivsufsort-lite
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*- Compiler specifics -*/
+#ifdef __clang__
+#pragma clang diagnostic ignored "-Wshorten-64-to-32"
+#endif
+
+#if defined(_MSC_VER)
+#  pragma warning(disable : 4244)
+#  pragma warning(disable : 4127)    /* C4127 : Condition expression is constant */
+#endif
+
+
+/*- Dependencies -*/
+#include 
+#include 
+#include 
+
+#include "divsufsort.h"
+
+/*- Constants -*/
+#if defined(INLINE)
+# undef INLINE
+#endif
+#if !defined(INLINE)
+# define INLINE __inline
+#endif
+#if defined(ALPHABET_SIZE) && (ALPHABET_SIZE < 1)
+# undef ALPHABET_SIZE
+#endif
+#if !defined(ALPHABET_SIZE)
+# define ALPHABET_SIZE (256)
+#endif
+#define BUCKET_A_SIZE (ALPHABET_SIZE)
+#define BUCKET_B_SIZE (ALPHABET_SIZE * ALPHABET_SIZE)
+#if defined(SS_INSERTIONSORT_THRESHOLD)
+# if SS_INSERTIONSORT_THRESHOLD < 1
+#  undef SS_INSERTIONSORT_THRESHOLD
+#  define SS_INSERTIONSORT_THRESHOLD (1)
+# endif
+#else
+# define SS_INSERTIONSORT_THRESHOLD (8)
+#endif
+#if defined(SS_BLOCKSIZE)
+# if SS_BLOCKSIZE < 0
+#  undef SS_BLOCKSIZE
+#  define SS_BLOCKSIZE (0)
+# elif 32768 <= SS_BLOCKSIZE
+#  undef SS_BLOCKSIZE
+#  define SS_BLOCKSIZE (32767)
+# endif
+#else
+# define SS_BLOCKSIZE (1024)
+#endif
+/* minstacksize = log(SS_BLOCKSIZE) / log(3) * 2 */
+#if SS_BLOCKSIZE == 0
+# define SS_MISORT_STACKSIZE (96)
+#elif SS_BLOCKSIZE <= 4096
+# define SS_MISORT_STACKSIZE (16)
+#else
+# define SS_MISORT_STACKSIZE (24)
+#endif
+#define SS_SMERGE_STACKSIZE (32)
+#define TR_INSERTIONSORT_THRESHOLD (8)
+#define TR_STACKSIZE (64)
+
+
+/*- Macros -*/
+#ifndef SWAP
+# define SWAP(_a, _b) do { t = (_a); (_a) = (_b); (_b) = t; } while(0)
+#endif /* SWAP */
+#ifndef MIN
+# define MIN(_a, _b) (((_a) < (_b)) ? (_a) : (_b))
+#endif /* MIN */
+#ifndef MAX
+# define MAX(_a, _b) (((_a) > (_b)) ? (_a) : (_b))
+#endif /* MAX */
+#define STACK_PUSH(_a, _b, _c, _d)\
+  do {\
+    assert(ssize < STACK_SIZE);\
+    stack[ssize].a = (_a), stack[ssize].b = (_b),\
+    stack[ssize].c = (_c), stack[ssize++].d = (_d);\
+  } while(0)
+#define STACK_PUSH5(_a, _b, _c, _d, _e)\
+  do {\
+    assert(ssize < STACK_SIZE);\
+    stack[ssize].a = (_a), stack[ssize].b = (_b),\
+    stack[ssize].c = (_c), stack[ssize].d = (_d), stack[ssize++].e = (_e);\
+  } while(0)
+#define STACK_POP(_a, _b, _c, _d)\
+  do {\
+    assert(0 <= ssize);\
+    if(ssize == 0) { return; }\
+    (_a) = stack[--ssize].a, (_b) = stack[ssize].b,\
+    (_c) = stack[ssize].c, (_d) = stack[ssize].d;\
+  } while(0)
+#define STACK_POP5(_a, _b, _c, _d, _e)\
+  do {\
+    assert(0 <= ssize);\
+    if(ssize == 0) { return; }\
+    (_a) = stack[--ssize].a, (_b) = stack[ssize].b,\
+    (_c) = stack[ssize].c, (_d) = stack[ssize].d, (_e) = stack[ssize].e;\
+  } while(0)
+#define BUCKET_A(_c0) bucket_A[(_c0)]
+#if ALPHABET_SIZE == 256
+#define BUCKET_B(_c0, _c1) (bucket_B[((_c1) << 8) | (_c0)])
+#define BUCKET_BSTAR(_c0, _c1) (bucket_B[((_c0) << 8) | (_c1)])
+#else
+#define BUCKET_B(_c0, _c1) (bucket_B[(_c1) * ALPHABET_SIZE + (_c0)])
+#define BUCKET_BSTAR(_c0, _c1) (bucket_B[(_c0) * ALPHABET_SIZE + (_c1)])
+#endif
+
+
+/*- Private Functions -*/
+
+static const int lg_table[256]= {
+ -1,0,1,1,2,2,2,2,3,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,
+  5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,
+  6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
+  6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
+  7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
+  7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
+  7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
+  7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7
+};
+
+#if (SS_BLOCKSIZE == 0) || (SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE)
+
+static INLINE
+int
+ss_ilg(int n) {
+#if SS_BLOCKSIZE == 0
+  return (n & 0xffff0000) ?
+          ((n & 0xff000000) ?
+            24 + lg_table[(n >> 24) & 0xff] :
+            16 + lg_table[(n >> 16) & 0xff]) :
+          ((n & 0x0000ff00) ?
+             8 + lg_table[(n >>  8) & 0xff] :
+             0 + lg_table[(n >>  0) & 0xff]);
+#elif SS_BLOCKSIZE < 256
+  return lg_table[n];
+#else
+  return (n & 0xff00) ?
+          8 + lg_table[(n >> 8) & 0xff] :
+          0 + lg_table[(n >> 0) & 0xff];
+#endif
+}
+
+#endif /* (SS_BLOCKSIZE == 0) || (SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE) */
+
+#if SS_BLOCKSIZE != 0
+
+static const int sqq_table[256] = {
+  0,  16,  22,  27,  32,  35,  39,  42,  45,  48,  50,  53,  55,  57,  59,  61,
+ 64,  65,  67,  69,  71,  73,  75,  76,  78,  80,  81,  83,  84,  86,  87,  89,
+ 90,  91,  93,  94,  96,  97,  98,  99, 101, 102, 103, 104, 106, 107, 108, 109,
+110, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126,
+128, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142,
+143, 144, 144, 145, 146, 147, 148, 149, 150, 150, 151, 152, 153, 154, 155, 155,
+156, 157, 158, 159, 160, 160, 161, 162, 163, 163, 164, 165, 166, 167, 167, 168,
+169, 170, 170, 171, 172, 173, 173, 174, 175, 176, 176, 177, 178, 178, 179, 180,
+181, 181, 182, 183, 183, 184, 185, 185, 186, 187, 187, 188, 189, 189, 190, 191,
+192, 192, 193, 193, 194, 195, 195, 196, 197, 197, 198, 199, 199, 200, 201, 201,
+202, 203, 203, 204, 204, 205, 206, 206, 207, 208, 208, 209, 209, 210, 211, 211,
+212, 212, 213, 214, 214, 215, 215, 216, 217, 217, 218, 218, 219, 219, 220, 221,
+221, 222, 222, 223, 224, 224, 225, 225, 226, 226, 227, 227, 228, 229, 229, 230,
+230, 231, 231, 232, 232, 233, 234, 234, 235, 235, 236, 236, 237, 237, 238, 238,
+239, 240, 240, 241, 241, 242, 242, 243, 243, 244, 244, 245, 245, 246, 246, 247,
+247, 248, 248, 249, 249, 250, 250, 251, 251, 252, 252, 253, 253, 254, 254, 255
+};
+
+static INLINE
+int
+ss_isqrt(int x) {
+  int y, e;
+
+  if(x >= (SS_BLOCKSIZE * SS_BLOCKSIZE)) { return SS_BLOCKSIZE; }
+  e = (x & 0xffff0000) ?
+        ((x & 0xff000000) ?
+          24 + lg_table[(x >> 24) & 0xff] :
+          16 + lg_table[(x >> 16) & 0xff]) :
+        ((x & 0x0000ff00) ?
+           8 + lg_table[(x >>  8) & 0xff] :
+           0 + lg_table[(x >>  0) & 0xff]);
+
+  if(e >= 16) {
+    y = sqq_table[x >> ((e - 6) - (e & 1))] << ((e >> 1) - 7);
+    if(e >= 24) { y = (y + 1 + x / y) >> 1; }
+    y = (y + 1 + x / y) >> 1;
+  } else if(e >= 8) {
+    y = (sqq_table[x >> ((e - 6) - (e & 1))] >> (7 - (e >> 1))) + 1;
+  } else {
+    return sqq_table[x] >> 4;
+  }
+
+  return (x < (y * y)) ? y - 1 : y;
+}
+
+#endif /* SS_BLOCKSIZE != 0 */
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Compares two suffixes. */
+static INLINE
+int
+ss_compare(const unsigned char *T,
+           const int *p1, const int *p2,
+           int depth) {
+  const unsigned char *U1, *U2, *U1n, *U2n;
+
+  for(U1 = T + depth + *p1,
+      U2 = T + depth + *p2,
+      U1n = T + *(p1 + 1) + 2,
+      U2n = T + *(p2 + 1) + 2;
+      (U1 < U1n) && (U2 < U2n) && (*U1 == *U2);
+      ++U1, ++U2) {
+  }
+
+  return U1 < U1n ?
+        (U2 < U2n ? *U1 - *U2 : 1) :
+        (U2 < U2n ? -1 : 0);
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+#if (SS_BLOCKSIZE != 1) && (SS_INSERTIONSORT_THRESHOLD != 1)
+
+/* Insertionsort for small size groups */
+static
+void
+ss_insertionsort(const unsigned char *T, const int *PA,
+                 int *first, int *last, int depth) {
+  int *i, *j;
+  int t;
+  int r;
+
+  for(i = last - 2; first <= i; --i) {
+    for(t = *i, j = i + 1; 0 < (r = ss_compare(T, PA + t, PA + *j, depth));) {
+      do { *(j - 1) = *j; } while((++j < last) && (*j < 0));
+      if(last <= j) { break; }
+    }
+    if(r == 0) { *j = ~*j; }
+    *(j - 1) = t;
+  }
+}
+
+#endif /* (SS_BLOCKSIZE != 1) && (SS_INSERTIONSORT_THRESHOLD != 1) */
+
+
+/*---------------------------------------------------------------------------*/
+
+#if (SS_BLOCKSIZE == 0) || (SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE)
+
+static INLINE
+void
+ss_fixdown(const unsigned char *Td, const int *PA,
+           int *SA, int i, int size) {
+  int j, k;
+  int v;
+  int c, d, e;
+
+  for(v = SA[i], c = Td[PA[v]]; (j = 2 * i + 1) < size; SA[i] = SA[k], i = k) {
+    d = Td[PA[SA[k = j++]]];
+    if(d < (e = Td[PA[SA[j]]])) { k = j; d = e; }
+    if(d <= c) { break; }
+  }
+  SA[i] = v;
+}
+
+/* Simple top-down heapsort. */
+static
+void
+ss_heapsort(const unsigned char *Td, const int *PA, int *SA, int size) {
+  int i, m;
+  int t;
+
+  m = size;
+  if((size % 2) == 0) {
+    m--;
+    if(Td[PA[SA[m / 2]]] < Td[PA[SA[m]]]) { SWAP(SA[m], SA[m / 2]); }
+  }
+
+  for(i = m / 2 - 1; 0 <= i; --i) { ss_fixdown(Td, PA, SA, i, m); }
+  if((size % 2) == 0) { SWAP(SA[0], SA[m]); ss_fixdown(Td, PA, SA, 0, m); }
+  for(i = m - 1; 0 < i; --i) {
+    t = SA[0], SA[0] = SA[i];
+    ss_fixdown(Td, PA, SA, 0, i);
+    SA[i] = t;
+  }
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Returns the median of three elements. */
+static INLINE
+int *
+ss_median3(const unsigned char *Td, const int *PA,
+           int *v1, int *v2, int *v3) {
+  int *t;
+  if(Td[PA[*v1]] > Td[PA[*v2]]) { SWAP(v1, v2); }
+  if(Td[PA[*v2]] > Td[PA[*v3]]) {
+    if(Td[PA[*v1]] > Td[PA[*v3]]) { return v1; }
+    else { return v3; }
+  }
+  return v2;
+}
+
+/* Returns the median of five elements. */
+static INLINE
+int *
+ss_median5(const unsigned char *Td, const int *PA,
+           int *v1, int *v2, int *v3, int *v4, int *v5) {
+  int *t;
+  if(Td[PA[*v2]] > Td[PA[*v3]]) { SWAP(v2, v3); }
+  if(Td[PA[*v4]] > Td[PA[*v5]]) { SWAP(v4, v5); }
+  if(Td[PA[*v2]] > Td[PA[*v4]]) { SWAP(v2, v4); SWAP(v3, v5); }
+  if(Td[PA[*v1]] > Td[PA[*v3]]) { SWAP(v1, v3); }
+  if(Td[PA[*v1]] > Td[PA[*v4]]) { SWAP(v1, v4); SWAP(v3, v5); }
+  if(Td[PA[*v3]] > Td[PA[*v4]]) { return v4; }
+  return v3;
+}
+
+/* Returns the pivot element. */
+static INLINE
+int *
+ss_pivot(const unsigned char *Td, const int *PA, int *first, int *last) {
+  int *middle;
+  int t;
+
+  t = last - first;
+  middle = first + t / 2;
+
+  if(t <= 512) {
+    if(t <= 32) {
+      return ss_median3(Td, PA, first, middle, last - 1);
+    } else {
+      t >>= 2;
+      return ss_median5(Td, PA, first, first + t, middle, last - 1 - t, last - 1);
+    }
+  }
+  t >>= 3;
+  first  = ss_median3(Td, PA, first, first + t, first + (t << 1));
+  middle = ss_median3(Td, PA, middle - t, middle, middle + t);
+  last   = ss_median3(Td, PA, last - 1 - (t << 1), last - 1 - t, last - 1);
+  return ss_median3(Td, PA, first, middle, last);
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Binary partition for substrings. */
+static INLINE
+int *
+ss_partition(const int *PA,
+                    int *first, int *last, int depth) {
+  int *a, *b;
+  int t;
+  for(a = first - 1, b = last;;) {
+    for(; (++a < b) && ((PA[*a] + depth) >= (PA[*a + 1] + 1));) { *a = ~*a; }
+    for(; (a < --b) && ((PA[*b] + depth) <  (PA[*b + 1] + 1));) { }
+    if(b <= a) { break; }
+    t = ~*b;
+    *b = *a;
+    *a = t;
+  }
+  if(first < a) { *first = ~*first; }
+  return a;
+}
+
+/* Multikey introsort for medium size groups. */
+static
+void
+ss_mintrosort(const unsigned char *T, const int *PA,
+              int *first, int *last,
+              int depth) {
+#define STACK_SIZE SS_MISORT_STACKSIZE
+  struct { int *a, *b, c; int d; } stack[STACK_SIZE];
+  const unsigned char *Td;
+  int *a, *b, *c, *d, *e, *f;
+  int s, t;
+  int ssize;
+  int limit;
+  int v, x = 0;
+
+  for(ssize = 0, limit = ss_ilg(last - first);;) {
+
+    if((last - first) <= SS_INSERTIONSORT_THRESHOLD) {
+#if 1 < SS_INSERTIONSORT_THRESHOLD
+      if(1 < (last - first)) { ss_insertionsort(T, PA, first, last, depth); }
+#endif
+      STACK_POP(first, last, depth, limit);
+      continue;
+    }
+
+    Td = T + depth;
+    if(limit-- == 0) { ss_heapsort(Td, PA, first, last - first); }
+    if(limit < 0) {
+      for(a = first + 1, v = Td[PA[*first]]; a < last; ++a) {
+        if((x = Td[PA[*a]]) != v) {
+          if(1 < (a - first)) { break; }
+          v = x;
+          first = a;
+        }
+      }
+      if(Td[PA[*first] - 1] < v) {
+        first = ss_partition(PA, first, a, depth);
+      }
+      if((a - first) <= (last - a)) {
+        if(1 < (a - first)) {
+          STACK_PUSH(a, last, depth, -1);
+          last = a, depth += 1, limit = ss_ilg(a - first);
+        } else {
+          first = a, limit = -1;
+        }
+      } else {
+        if(1 < (last - a)) {
+          STACK_PUSH(first, a, depth + 1, ss_ilg(a - first));
+          first = a, limit = -1;
+        } else {
+          last = a, depth += 1, limit = ss_ilg(a - first);
+        }
+      }
+      continue;
+    }
+
+    /* choose pivot */
+    a = ss_pivot(Td, PA, first, last);
+    v = Td[PA[*a]];
+    SWAP(*first, *a);
+
+    /* partition */
+    for(b = first; (++b < last) && ((x = Td[PA[*b]]) == v);) { }
+    if(((a = b) < last) && (x < v)) {
+      for(; (++b < last) && ((x = Td[PA[*b]]) <= v);) {
+        if(x == v) { SWAP(*b, *a); ++a; }
+      }
+    }
+    for(c = last; (b < --c) && ((x = Td[PA[*c]]) == v);) { }
+    if((b < (d = c)) && (x > v)) {
+      for(; (b < --c) && ((x = Td[PA[*c]]) >= v);) {
+        if(x == v) { SWAP(*c, *d); --d; }
+      }
+    }
+    for(; b < c;) {
+      SWAP(*b, *c);
+      for(; (++b < c) && ((x = Td[PA[*b]]) <= v);) {
+        if(x == v) { SWAP(*b, *a); ++a; }
+      }
+      for(; (b < --c) && ((x = Td[PA[*c]]) >= v);) {
+        if(x == v) { SWAP(*c, *d); --d; }
+      }
+    }
+
+    if(a <= d) {
+      c = b - 1;
+
+      if((s = a - first) > (t = b - a)) { s = t; }
+      for(e = first, f = b - s; 0 < s; --s, ++e, ++f) { SWAP(*e, *f); }
+      if((s = d - c) > (t = last - d - 1)) { s = t; }
+      for(e = b, f = last - s; 0 < s; --s, ++e, ++f) { SWAP(*e, *f); }
+
+      a = first + (b - a), c = last - (d - c);
+      b = (v <= Td[PA[*a] - 1]) ? a : ss_partition(PA, a, c, depth);
+
+      if((a - first) <= (last - c)) {
+        if((last - c) <= (c - b)) {
+          STACK_PUSH(b, c, depth + 1, ss_ilg(c - b));
+          STACK_PUSH(c, last, depth, limit);
+          last = a;
+        } else if((a - first) <= (c - b)) {
+          STACK_PUSH(c, last, depth, limit);
+          STACK_PUSH(b, c, depth + 1, ss_ilg(c - b));
+          last = a;
+        } else {
+          STACK_PUSH(c, last, depth, limit);
+          STACK_PUSH(first, a, depth, limit);
+          first = b, last = c, depth += 1, limit = ss_ilg(c - b);
+        }
+      } else {
+        if((a - first) <= (c - b)) {
+          STACK_PUSH(b, c, depth + 1, ss_ilg(c - b));
+          STACK_PUSH(first, a, depth, limit);
+          first = c;
+        } else if((last - c) <= (c - b)) {
+          STACK_PUSH(first, a, depth, limit);
+          STACK_PUSH(b, c, depth + 1, ss_ilg(c - b));
+          first = c;
+        } else {
+          STACK_PUSH(first, a, depth, limit);
+          STACK_PUSH(c, last, depth, limit);
+          first = b, last = c, depth += 1, limit = ss_ilg(c - b);
+        }
+      }
+    } else {
+      limit += 1;
+      if(Td[PA[*first] - 1] < v) {
+        first = ss_partition(PA, first, last, depth);
+        limit = ss_ilg(last - first);
+      }
+      depth += 1;
+    }
+  }
+#undef STACK_SIZE
+}
+
+#endif /* (SS_BLOCKSIZE == 0) || (SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE) */
+
+
+/*---------------------------------------------------------------------------*/
+
+#if SS_BLOCKSIZE != 0
+
+static INLINE
+void
+ss_blockswap(int *a, int *b, int n) {
+  int t;
+  for(; 0 < n; --n, ++a, ++b) {
+    t = *a, *a = *b, *b = t;
+  }
+}
+
+static INLINE
+void
+ss_rotate(int *first, int *middle, int *last) {
+  int *a, *b, t;
+  int l, r;
+  l = middle - first, r = last - middle;
+  for(; (0 < l) && (0 < r);) {
+    if(l == r) { ss_blockswap(first, middle, l); break; }
+    if(l < r) {
+      a = last - 1, b = middle - 1;
+      t = *a;
+      do {
+        *a-- = *b, *b-- = *a;
+        if(b < first) {
+          *a = t;
+          last = a;
+          if((r -= l + 1) <= l) { break; }
+          a -= 1, b = middle - 1;
+          t = *a;
+        }
+      } while(1);
+    } else {
+      a = first, b = middle;
+      t = *a;
+      do {
+        *a++ = *b, *b++ = *a;
+        if(last <= b) {
+          *a = t;
+          first = a + 1;
+          if((l -= r + 1) <= r) { break; }
+          a += 1, b = middle;
+          t = *a;
+        }
+      } while(1);
+    }
+  }
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+static
+void
+ss_inplacemerge(const unsigned char *T, const int *PA,
+                int *first, int *middle, int *last,
+                int depth) {
+  const int *p;
+  int *a, *b;
+  int len, half;
+  int q, r;
+  int x;
+
+  for(;;) {
+    if(*(last - 1) < 0) { x = 1; p = PA + ~*(last - 1); }
+    else                { x = 0; p = PA +  *(last - 1); }
+    for(a = first, len = middle - first, half = len >> 1, r = -1;
+        0 < len;
+        len = half, half >>= 1) {
+      b = a + half;
+      q = ss_compare(T, PA + ((0 <= *b) ? *b : ~*b), p, depth);
+      if(q < 0) {
+        a = b + 1;
+        half -= (len & 1) ^ 1;
+      } else {
+        r = q;
+      }
+    }
+    if(a < middle) {
+      if(r == 0) { *a = ~*a; }
+      ss_rotate(a, middle, last);
+      last -= middle - a;
+      middle = a;
+      if(first == middle) { break; }
+    }
+    --last;
+    if(x != 0) { while(*--last < 0) { } }
+    if(middle == last) { break; }
+  }
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Merge-forward with internal buffer. */
+static
+void
+ss_mergeforward(const unsigned char *T, const int *PA,
+                int *first, int *middle, int *last,
+                int *buf, int depth) {
+  int *a, *b, *c, *bufend;
+  int t;
+  int r;
+
+  bufend = buf + (middle - first) - 1;
+  ss_blockswap(buf, first, middle - first);
+
+  for(t = *(a = first), b = buf, c = middle;;) {
+    r = ss_compare(T, PA + *b, PA + *c, depth);
+    if(r < 0) {
+      do {
+        *a++ = *b;
+        if(bufend <= b) { *bufend = t; return; }
+        *b++ = *a;
+      } while(*b < 0);
+    } else if(r > 0) {
+      do {
+        *a++ = *c, *c++ = *a;
+        if(last <= c) {
+          while(b < bufend) { *a++ = *b, *b++ = *a; }
+          *a = *b, *b = t;
+          return;
+        }
+      } while(*c < 0);
+    } else {
+      *c = ~*c;
+      do {
+        *a++ = *b;
+        if(bufend <= b) { *bufend = t; return; }
+        *b++ = *a;
+      } while(*b < 0);
+
+      do {
+        *a++ = *c, *c++ = *a;
+        if(last <= c) {
+          while(b < bufend) { *a++ = *b, *b++ = *a; }
+          *a = *b, *b = t;
+          return;
+        }
+      } while(*c < 0);
+    }
+  }
+}
+
+/* Merge-backward with internal buffer. */
+static
+void
+ss_mergebackward(const unsigned char *T, const int *PA,
+                 int *first, int *middle, int *last,
+                 int *buf, int depth) {
+  const int *p1, *p2;
+  int *a, *b, *c, *bufend;
+  int t;
+  int r;
+  int x;
+
+  bufend = buf + (last - middle) - 1;
+  ss_blockswap(buf, middle, last - middle);
+
+  x = 0;
+  if(*bufend < 0)       { p1 = PA + ~*bufend; x |= 1; }
+  else                  { p1 = PA +  *bufend; }
+  if(*(middle - 1) < 0) { p2 = PA + ~*(middle - 1); x |= 2; }
+  else                  { p2 = PA +  *(middle - 1); }
+  for(t = *(a = last - 1), b = bufend, c = middle - 1;;) {
+    r = ss_compare(T, p1, p2, depth);
+    if(0 < r) {
+      if(x & 1) { do { *a-- = *b, *b-- = *a; } while(*b < 0); x ^= 1; }
+      *a-- = *b;
+      if(b <= buf) { *buf = t; break; }
+      *b-- = *a;
+      if(*b < 0) { p1 = PA + ~*b; x |= 1; }
+      else       { p1 = PA +  *b; }
+    } else if(r < 0) {
+      if(x & 2) { do { *a-- = *c, *c-- = *a; } while(*c < 0); x ^= 2; }
+      *a-- = *c, *c-- = *a;
+      if(c < first) {
+        while(buf < b) { *a-- = *b, *b-- = *a; }
+        *a = *b, *b = t;
+        break;
+      }
+      if(*c < 0) { p2 = PA + ~*c; x |= 2; }
+      else       { p2 = PA +  *c; }
+    } else {
+      if(x & 1) { do { *a-- = *b, *b-- = *a; } while(*b < 0); x ^= 1; }
+      *a-- = ~*b;
+      if(b <= buf) { *buf = t; break; }
+      *b-- = *a;
+      if(x & 2) { do { *a-- = *c, *c-- = *a; } while(*c < 0); x ^= 2; }
+      *a-- = *c, *c-- = *a;
+      if(c < first) {
+        while(buf < b) { *a-- = *b, *b-- = *a; }
+        *a = *b, *b = t;
+        break;
+      }
+      if(*b < 0) { p1 = PA + ~*b; x |= 1; }
+      else       { p1 = PA +  *b; }
+      if(*c < 0) { p2 = PA + ~*c; x |= 2; }
+      else       { p2 = PA +  *c; }
+    }
+  }
+}
+
+/* D&C based merge. */
+static
+void
+ss_swapmerge(const unsigned char *T, const int *PA,
+             int *first, int *middle, int *last,
+             int *buf, int bufsize, int depth) {
+#define STACK_SIZE SS_SMERGE_STACKSIZE
+#define GETIDX(a) ((0 <= (a)) ? (a) : (~(a)))
+#define MERGE_CHECK(a, b, c)\
+  do {\
+    if(((c) & 1) ||\
+       (((c) & 2) && (ss_compare(T, PA + GETIDX(*((a) - 1)), PA + *(a), depth) == 0))) {\
+      *(a) = ~*(a);\
+    }\
+    if(((c) & 4) && ((ss_compare(T, PA + GETIDX(*((b) - 1)), PA + *(b), depth) == 0))) {\
+      *(b) = ~*(b);\
+    }\
+  } while(0)
+  struct { int *a, *b, *c; int d; } stack[STACK_SIZE];
+  int *l, *r, *lm, *rm;
+  int m, len, half;
+  int ssize;
+  int check, next;
+
+  for(check = 0, ssize = 0;;) {
+    if((last - middle) <= bufsize) {
+      if((first < middle) && (middle < last)) {
+        ss_mergebackward(T, PA, first, middle, last, buf, depth);
+      }
+      MERGE_CHECK(first, last, check);
+      STACK_POP(first, middle, last, check);
+      continue;
+    }
+
+    if((middle - first) <= bufsize) {
+      if(first < middle) {
+        ss_mergeforward(T, PA, first, middle, last, buf, depth);
+      }
+      MERGE_CHECK(first, last, check);
+      STACK_POP(first, middle, last, check);
+      continue;
+    }
+
+    for(m = 0, len = MIN(middle - first, last - middle), half = len >> 1;
+        0 < len;
+        len = half, half >>= 1) {
+      if(ss_compare(T, PA + GETIDX(*(middle + m + half)),
+                       PA + GETIDX(*(middle - m - half - 1)), depth) < 0) {
+        m += half + 1;
+        half -= (len & 1) ^ 1;
+      }
+    }
+
+    if(0 < m) {
+      lm = middle - m, rm = middle + m;
+      ss_blockswap(lm, middle, m);
+      l = r = middle, next = 0;
+      if(rm < last) {
+        if(*rm < 0) {
+          *rm = ~*rm;
+          if(first < lm) { for(; *--l < 0;) { } next |= 4; }
+          next |= 1;
+        } else if(first < lm) {
+          for(; *r < 0; ++r) { }
+          next |= 2;
+        }
+      }
+
+      if((l - first) <= (last - r)) {
+        STACK_PUSH(r, rm, last, (next & 3) | (check & 4));
+        middle = lm, last = l, check = (check & 3) | (next & 4);
+      } else {
+        if((next & 2) && (r == middle)) { next ^= 6; }
+        STACK_PUSH(first, lm, l, (check & 3) | (next & 4));
+        first = r, middle = rm, check = (next & 3) | (check & 4);
+      }
+    } else {
+      if(ss_compare(T, PA + GETIDX(*(middle - 1)), PA + *middle, depth) == 0) {
+        *middle = ~*middle;
+      }
+      MERGE_CHECK(first, last, check);
+      STACK_POP(first, middle, last, check);
+    }
+  }
+#undef STACK_SIZE
+}
+
+#endif /* SS_BLOCKSIZE != 0 */
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Substring sort */
+static
+void
+sssort(const unsigned char *T, const int *PA,
+       int *first, int *last,
+       int *buf, int bufsize,
+       int depth, int n, int lastsuffix) {
+  int *a;
+#if SS_BLOCKSIZE != 0
+  int *b, *middle, *curbuf;
+  int j, k, curbufsize, limit;
+#endif
+  int i;
+
+  if(lastsuffix != 0) { ++first; }
+
+#if SS_BLOCKSIZE == 0
+  ss_mintrosort(T, PA, first, last, depth);
+#else
+  if((bufsize < SS_BLOCKSIZE) &&
+      (bufsize < (last - first)) &&
+      (bufsize < (limit = ss_isqrt(last - first)))) {
+    if(SS_BLOCKSIZE < limit) { limit = SS_BLOCKSIZE; }
+    buf = middle = last - limit, bufsize = limit;
+  } else {
+    middle = last, limit = 0;
+  }
+  for(a = first, i = 0; SS_BLOCKSIZE < (middle - a); a += SS_BLOCKSIZE, ++i) {
+#if SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE
+    ss_mintrosort(T, PA, a, a + SS_BLOCKSIZE, depth);
+#elif 1 < SS_BLOCKSIZE
+    ss_insertionsort(T, PA, a, a + SS_BLOCKSIZE, depth);
+#endif
+    curbufsize = last - (a + SS_BLOCKSIZE);
+    curbuf = a + SS_BLOCKSIZE;
+    if(curbufsize <= bufsize) { curbufsize = bufsize, curbuf = buf; }
+    for(b = a, k = SS_BLOCKSIZE, j = i; j & 1; b -= k, k <<= 1, j >>= 1) {
+      ss_swapmerge(T, PA, b - k, b, b + k, curbuf, curbufsize, depth);
+    }
+  }
+#if SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE
+  ss_mintrosort(T, PA, a, middle, depth);
+#elif 1 < SS_BLOCKSIZE
+  ss_insertionsort(T, PA, a, middle, depth);
+#endif
+  for(k = SS_BLOCKSIZE; i != 0; k <<= 1, i >>= 1) {
+    if(i & 1) {
+      ss_swapmerge(T, PA, a - k, a, middle, buf, bufsize, depth);
+      a -= k;
+    }
+  }
+  if(limit != 0) {
+#if SS_INSERTIONSORT_THRESHOLD < SS_BLOCKSIZE
+    ss_mintrosort(T, PA, middle, last, depth);
+#elif 1 < SS_BLOCKSIZE
+    ss_insertionsort(T, PA, middle, last, depth);
+#endif
+    ss_inplacemerge(T, PA, first, middle, last, depth);
+  }
+#endif
+
+  if(lastsuffix != 0) {
+    /* Insert last type B* suffix. */
+    int PAi[2]; PAi[0] = PA[*(first - 1)], PAi[1] = n - 2;
+    for(a = first, i = *(first - 1);
+        (a < last) && ((*a < 0) || (0 < ss_compare(T, &(PAi[0]), PA + *a, depth)));
+        ++a) {
+      *(a - 1) = *a;
+    }
+    *(a - 1) = i;
+  }
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+static INLINE
+int
+tr_ilg(int n) {
+  return (n & 0xffff0000) ?
+          ((n & 0xff000000) ?
+            24 + lg_table[(n >> 24) & 0xff] :
+            16 + lg_table[(n >> 16) & 0xff]) :
+          ((n & 0x0000ff00) ?
+             8 + lg_table[(n >>  8) & 0xff] :
+             0 + lg_table[(n >>  0) & 0xff]);
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Simple insertionsort for small size groups. */
+static
+void
+tr_insertionsort(const int *ISAd, int *first, int *last) {
+  int *a, *b;
+  int t, r;
+
+  for(a = first + 1; a < last; ++a) {
+    for(t = *a, b = a - 1; 0 > (r = ISAd[t] - ISAd[*b]);) {
+      do { *(b + 1) = *b; } while((first <= --b) && (*b < 0));
+      if(b < first) { break; }
+    }
+    if(r == 0) { *b = ~*b; }
+    *(b + 1) = t;
+  }
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+static INLINE
+void
+tr_fixdown(const int *ISAd, int *SA, int i, int size) {
+  int j, k;
+  int v;
+  int c, d, e;
+
+  for(v = SA[i], c = ISAd[v]; (j = 2 * i + 1) < size; SA[i] = SA[k], i = k) {
+    d = ISAd[SA[k = j++]];
+    if(d < (e = ISAd[SA[j]])) { k = j; d = e; }
+    if(d <= c) { break; }
+  }
+  SA[i] = v;
+}
+
+/* Simple top-down heapsort. */
+static
+void
+tr_heapsort(const int *ISAd, int *SA, int size) {
+  int i, m;
+  int t;
+
+  m = size;
+  if((size % 2) == 0) {
+    m--;
+    if(ISAd[SA[m / 2]] < ISAd[SA[m]]) { SWAP(SA[m], SA[m / 2]); }
+  }
+
+  for(i = m / 2 - 1; 0 <= i; --i) { tr_fixdown(ISAd, SA, i, m); }
+  if((size % 2) == 0) { SWAP(SA[0], SA[m]); tr_fixdown(ISAd, SA, 0, m); }
+  for(i = m - 1; 0 < i; --i) {
+    t = SA[0], SA[0] = SA[i];
+    tr_fixdown(ISAd, SA, 0, i);
+    SA[i] = t;
+  }
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Returns the median of three elements. */
+static INLINE
+int *
+tr_median3(const int *ISAd, int *v1, int *v2, int *v3) {
+  int *t;
+  if(ISAd[*v1] > ISAd[*v2]) { SWAP(v1, v2); }
+  if(ISAd[*v2] > ISAd[*v3]) {
+    if(ISAd[*v1] > ISAd[*v3]) { return v1; }
+    else { return v3; }
+  }
+  return v2;
+}
+
+/* Returns the median of five elements. */
+static INLINE
+int *
+tr_median5(const int *ISAd,
+           int *v1, int *v2, int *v3, int *v4, int *v5) {
+  int *t;
+  if(ISAd[*v2] > ISAd[*v3]) { SWAP(v2, v3); }
+  if(ISAd[*v4] > ISAd[*v5]) { SWAP(v4, v5); }
+  if(ISAd[*v2] > ISAd[*v4]) { SWAP(v2, v4); SWAP(v3, v5); }
+  if(ISAd[*v1] > ISAd[*v3]) { SWAP(v1, v3); }
+  if(ISAd[*v1] > ISAd[*v4]) { SWAP(v1, v4); SWAP(v3, v5); }
+  if(ISAd[*v3] > ISAd[*v4]) { return v4; }
+  return v3;
+}
+
+/* Returns the pivot element. */
+static INLINE
+int *
+tr_pivot(const int *ISAd, int *first, int *last) {
+  int *middle;
+  int t;
+
+  t = last - first;
+  middle = first + t / 2;
+
+  if(t <= 512) {
+    if(t <= 32) {
+      return tr_median3(ISAd, first, middle, last - 1);
+    } else {
+      t >>= 2;
+      return tr_median5(ISAd, first, first + t, middle, last - 1 - t, last - 1);
+    }
+  }
+  t >>= 3;
+  first  = tr_median3(ISAd, first, first + t, first + (t << 1));
+  middle = tr_median3(ISAd, middle - t, middle, middle + t);
+  last   = tr_median3(ISAd, last - 1 - (t << 1), last - 1 - t, last - 1);
+  return tr_median3(ISAd, first, middle, last);
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+typedef struct _trbudget_t trbudget_t;
+struct _trbudget_t {
+  int chance;
+  int remain;
+  int incval;
+  int count;
+};
+
+static INLINE
+void
+trbudget_init(trbudget_t *budget, int chance, int incval) {
+  budget->chance = chance;
+  budget->remain = budget->incval = incval;
+}
+
+static INLINE
+int
+trbudget_check(trbudget_t *budget, int size) {
+  if(size <= budget->remain) { budget->remain -= size; return 1; }
+  if(budget->chance == 0) { budget->count += size; return 0; }
+  budget->remain += budget->incval - size;
+  budget->chance -= 1;
+  return 1;
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+static INLINE
+void
+tr_partition(const int *ISAd,
+             int *first, int *middle, int *last,
+             int **pa, int **pb, int v) {
+  int *a, *b, *c, *d, *e, *f;
+  int t, s;
+  int x = 0;
+
+  for(b = middle - 1; (++b < last) && ((x = ISAd[*b]) == v);) { }
+  if(((a = b) < last) && (x < v)) {
+    for(; (++b < last) && ((x = ISAd[*b]) <= v);) {
+      if(x == v) { SWAP(*b, *a); ++a; }
+    }
+  }
+  for(c = last; (b < --c) && ((x = ISAd[*c]) == v);) { }
+  if((b < (d = c)) && (x > v)) {
+    for(; (b < --c) && ((x = ISAd[*c]) >= v);) {
+      if(x == v) { SWAP(*c, *d); --d; }
+    }
+  }
+  for(; b < c;) {
+    SWAP(*b, *c);
+    for(; (++b < c) && ((x = ISAd[*b]) <= v);) {
+      if(x == v) { SWAP(*b, *a); ++a; }
+    }
+    for(; (b < --c) && ((x = ISAd[*c]) >= v);) {
+      if(x == v) { SWAP(*c, *d); --d; }
+    }
+  }
+
+  if(a <= d) {
+    c = b - 1;
+    if((s = a - first) > (t = b - a)) { s = t; }
+    for(e = first, f = b - s; 0 < s; --s, ++e, ++f) { SWAP(*e, *f); }
+    if((s = d - c) > (t = last - d - 1)) { s = t; }
+    for(e = b, f = last - s; 0 < s; --s, ++e, ++f) { SWAP(*e, *f); }
+    first += (b - a), last -= (d - c);
+  }
+  *pa = first, *pb = last;
+}
+
+static
+void
+tr_copy(int *ISA, const int *SA,
+        int *first, int *a, int *b, int *last,
+        int depth) {
+  /* sort suffixes of middle partition
+     by using sorted order of suffixes of left and right partition. */
+  int *c, *d, *e;
+  int s, v;
+
+  v = b - SA - 1;
+  for(c = first, d = a - 1; c <= d; ++c) {
+    if((0 <= (s = *c - depth)) && (ISA[s] == v)) {
+      *++d = s;
+      ISA[s] = d - SA;
+    }
+  }
+  for(c = last - 1, e = d + 1, d = b; e < d; --c) {
+    if((0 <= (s = *c - depth)) && (ISA[s] == v)) {
+      *--d = s;
+      ISA[s] = d - SA;
+    }
+  }
+}
+
+static
+void
+tr_partialcopy(int *ISA, const int *SA,
+               int *first, int *a, int *b, int *last,
+               int depth) {
+  int *c, *d, *e;
+  int s, v;
+  int rank, lastrank, newrank = -1;
+
+  v = b - SA - 1;
+  lastrank = -1;
+  for(c = first, d = a - 1; c <= d; ++c) {
+    if((0 <= (s = *c - depth)) && (ISA[s] == v)) {
+      *++d = s;
+      rank = ISA[s + depth];
+      if(lastrank != rank) { lastrank = rank; newrank = d - SA; }
+      ISA[s] = newrank;
+    }
+  }
+
+  lastrank = -1;
+  for(e = d; first <= e; --e) {
+    rank = ISA[*e];
+    if(lastrank != rank) { lastrank = rank; newrank = e - SA; }
+    if(newrank != rank) { ISA[*e] = newrank; }
+  }
+
+  lastrank = -1;
+  for(c = last - 1, e = d + 1, d = b; e < d; --c) {
+    if((0 <= (s = *c - depth)) && (ISA[s] == v)) {
+      *--d = s;
+      rank = ISA[s + depth];
+      if(lastrank != rank) { lastrank = rank; newrank = d - SA; }
+      ISA[s] = newrank;
+    }
+  }
+}
+
+static
+void
+tr_introsort(int *ISA, const int *ISAd,
+             int *SA, int *first, int *last,
+             trbudget_t *budget) {
+#define STACK_SIZE TR_STACKSIZE
+  struct { const int *a; int *b, *c; int d, e; }stack[STACK_SIZE];
+  int *a, *b, *c;
+  int t;
+  int v, x = 0;
+  int incr = ISAd - ISA;
+  int limit, next;
+  int ssize, trlink = -1;
+
+  for(ssize = 0, limit = tr_ilg(last - first);;) {
+
+    if(limit < 0) {
+      if(limit == -1) {
+        /* tandem repeat partition */
+        tr_partition(ISAd - incr, first, first, last, &a, &b, last - SA - 1);
+
+        /* update ranks */
+        if(a < last) {
+          for(c = first, v = a - SA - 1; c < a; ++c) { ISA[*c] = v; }
+        }
+        if(b < last) {
+          for(c = a, v = b - SA - 1; c < b; ++c) { ISA[*c] = v; }
+        }
+
+        /* push */
+        if(1 < (b - a)) {
+          STACK_PUSH5(NULL, a, b, 0, 0);
+          STACK_PUSH5(ISAd - incr, first, last, -2, trlink);
+          trlink = ssize - 2;
+        }
+        if((a - first) <= (last - b)) {
+          if(1 < (a - first)) {
+            STACK_PUSH5(ISAd, b, last, tr_ilg(last - b), trlink);
+            last = a, limit = tr_ilg(a - first);
+          } else if(1 < (last - b)) {
+            first = b, limit = tr_ilg(last - b);
+          } else {
+            STACK_POP5(ISAd, first, last, limit, trlink);
+          }
+        } else {
+          if(1 < (last - b)) {
+            STACK_PUSH5(ISAd, first, a, tr_ilg(a - first), trlink);
+            first = b, limit = tr_ilg(last - b);
+          } else if(1 < (a - first)) {
+            last = a, limit = tr_ilg(a - first);
+          } else {
+            STACK_POP5(ISAd, first, last, limit, trlink);
+          }
+        }
+      } else if(limit == -2) {
+        /* tandem repeat copy */
+        a = stack[--ssize].b, b = stack[ssize].c;
+        if(stack[ssize].d == 0) {
+          tr_copy(ISA, SA, first, a, b, last, ISAd - ISA);
+        } else {
+          if(0 <= trlink) { stack[trlink].d = -1; }
+          tr_partialcopy(ISA, SA, first, a, b, last, ISAd - ISA);
+        }
+        STACK_POP5(ISAd, first, last, limit, trlink);
+      } else {
+        /* sorted partition */
+        if(0 <= *first) {
+          a = first;
+          do { ISA[*a] = a - SA; } while((++a < last) && (0 <= *a));
+          first = a;
+        }
+        if(first < last) {
+          a = first; do { *a = ~*a; } while(*++a < 0);
+          next = (ISA[*a] != ISAd[*a]) ? tr_ilg(a - first + 1) : -1;
+          if(++a < last) { for(b = first, v = a - SA - 1; b < a; ++b) { ISA[*b] = v; } }
+
+          /* push */
+          if(trbudget_check(budget, a - first)) {
+            if((a - first) <= (last - a)) {
+              STACK_PUSH5(ISAd, a, last, -3, trlink);
+              ISAd += incr, last = a, limit = next;
+            } else {
+              if(1 < (last - a)) {
+                STACK_PUSH5(ISAd + incr, first, a, next, trlink);
+                first = a, limit = -3;
+              } else {
+                ISAd += incr, last = a, limit = next;
+              }
+            }
+          } else {
+            if(0 <= trlink) { stack[trlink].d = -1; }
+            if(1 < (last - a)) {
+              first = a, limit = -3;
+            } else {
+              STACK_POP5(ISAd, first, last, limit, trlink);
+            }
+          }
+        } else {
+          STACK_POP5(ISAd, first, last, limit, trlink);
+        }
+      }
+      continue;
+    }
+
+    if((last - first) <= TR_INSERTIONSORT_THRESHOLD) {
+      tr_insertionsort(ISAd, first, last);
+      limit = -3;
+      continue;
+    }
+
+    if(limit-- == 0) {
+      tr_heapsort(ISAd, first, last - first);
+      for(a = last - 1; first < a; a = b) {
+        for(x = ISAd[*a], b = a - 1; (first <= b) && (ISAd[*b] == x); --b) { *b = ~*b; }
+      }
+      limit = -3;
+      continue;
+    }
+
+    /* choose pivot */
+    a = tr_pivot(ISAd, first, last);
+    SWAP(*first, *a);
+    v = ISAd[*first];
+
+    /* partition */
+    tr_partition(ISAd, first, first + 1, last, &a, &b, v);
+    if((last - first) != (b - a)) {
+      next = (ISA[*a] != v) ? tr_ilg(b - a) : -1;
+
+      /* update ranks */
+      for(c = first, v = a - SA - 1; c < a; ++c) { ISA[*c] = v; }
+      if(b < last) { for(c = a, v = b - SA - 1; c < b; ++c) { ISA[*c] = v; } }
+
+      /* push */
+      if((1 < (b - a)) && (trbudget_check(budget, b - a))) {
+        if((a - first) <= (last - b)) {
+          if((last - b) <= (b - a)) {
+            if(1 < (a - first)) {
+              STACK_PUSH5(ISAd + incr, a, b, next, trlink);
+              STACK_PUSH5(ISAd, b, last, limit, trlink);
+              last = a;
+            } else if(1 < (last - b)) {
+              STACK_PUSH5(ISAd + incr, a, b, next, trlink);
+              first = b;
+            } else {
+              ISAd += incr, first = a, last = b, limit = next;
+            }
+          } else if((a - first) <= (b - a)) {
+            if(1 < (a - first)) {
+              STACK_PUSH5(ISAd, b, last, limit, trlink);
+              STACK_PUSH5(ISAd + incr, a, b, next, trlink);
+              last = a;
+            } else {
+              STACK_PUSH5(ISAd, b, last, limit, trlink);
+              ISAd += incr, first = a, last = b, limit = next;
+            }
+          } else {
+            STACK_PUSH5(ISAd, b, last, limit, trlink);
+            STACK_PUSH5(ISAd, first, a, limit, trlink);
+            ISAd += incr, first = a, last = b, limit = next;
+          }
+        } else {
+          if((a - first) <= (b - a)) {
+            if(1 < (last - b)) {
+              STACK_PUSH5(ISAd + incr, a, b, next, trlink);
+              STACK_PUSH5(ISAd, first, a, limit, trlink);
+              first = b;
+            } else if(1 < (a - first)) {
+              STACK_PUSH5(ISAd + incr, a, b, next, trlink);
+              last = a;
+            } else {
+              ISAd += incr, first = a, last = b, limit = next;
+            }
+          } else if((last - b) <= (b - a)) {
+            if(1 < (last - b)) {
+              STACK_PUSH5(ISAd, first, a, limit, trlink);
+              STACK_PUSH5(ISAd + incr, a, b, next, trlink);
+              first = b;
+            } else {
+              STACK_PUSH5(ISAd, first, a, limit, trlink);
+              ISAd += incr, first = a, last = b, limit = next;
+            }
+          } else {
+            STACK_PUSH5(ISAd, first, a, limit, trlink);
+            STACK_PUSH5(ISAd, b, last, limit, trlink);
+            ISAd += incr, first = a, last = b, limit = next;
+          }
+        }
+      } else {
+        if((1 < (b - a)) && (0 <= trlink)) { stack[trlink].d = -1; }
+        if((a - first) <= (last - b)) {
+          if(1 < (a - first)) {
+            STACK_PUSH5(ISAd, b, last, limit, trlink);
+            last = a;
+          } else if(1 < (last - b)) {
+            first = b;
+          } else {
+            STACK_POP5(ISAd, first, last, limit, trlink);
+          }
+        } else {
+          if(1 < (last - b)) {
+            STACK_PUSH5(ISAd, first, a, limit, trlink);
+            first = b;
+          } else if(1 < (a - first)) {
+            last = a;
+          } else {
+            STACK_POP5(ISAd, first, last, limit, trlink);
+          }
+        }
+      }
+    } else {
+      if(trbudget_check(budget, last - first)) {
+        limit = tr_ilg(last - first), ISAd += incr;
+      } else {
+        if(0 <= trlink) { stack[trlink].d = -1; }
+        STACK_POP5(ISAd, first, last, limit, trlink);
+      }
+    }
+  }
+#undef STACK_SIZE
+}
+
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Tandem repeat sort */
+static
+void
+trsort(int *ISA, int *SA, int n, int depth) {
+  int *ISAd;
+  int *first, *last;
+  trbudget_t budget;
+  int t, skip, unsorted;
+
+  trbudget_init(&budget, tr_ilg(n) * 2 / 3, n);
+/*  trbudget_init(&budget, tr_ilg(n) * 3 / 4, n); */
+  for(ISAd = ISA + depth; -n < *SA; ISAd += ISAd - ISA) {
+    first = SA;
+    skip = 0;
+    unsorted = 0;
+    do {
+      if((t = *first) < 0) { first -= t; skip += t; }
+      else {
+        if(skip != 0) { *(first + skip) = skip; skip = 0; }
+        last = SA + ISA[t] + 1;
+        if(1 < (last - first)) {
+          budget.count = 0;
+          tr_introsort(ISA, ISAd, SA, first, last, &budget);
+          if(budget.count != 0) { unsorted += budget.count; }
+          else { skip = first - last; }
+        } else if((last - first) == 1) {
+          skip = -1;
+        }
+        first = last;
+      }
+    } while(first < (SA + n));
+    if(skip != 0) { *(first + skip) = skip; }
+    if(unsorted == 0) { break; }
+  }
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+/* Sorts suffixes of type B*. */
+static
+int
+sort_typeBstar(const unsigned char *T, int *SA,
+               int *bucket_A, int *bucket_B,
+               int n, int openMP) {
+  int *PAb, *ISAb, *buf;
+#ifdef LIBBSC_OPENMP
+  int *curbuf;
+  int l;
+#endif
+  int i, j, k, t, m, bufsize;
+  int c0, c1;
+#ifdef LIBBSC_OPENMP
+  int d0, d1;
+#endif
+  (void)openMP;
+
+  /* Initialize bucket arrays. */
+  for(i = 0; i < BUCKET_A_SIZE; ++i) { bucket_A[i] = 0; }
+  for(i = 0; i < BUCKET_B_SIZE; ++i) { bucket_B[i] = 0; }
+
+  /* Count the number of occurrences of the first one or two characters of each
+     type A, B and B* suffix. Moreover, store the beginning position of all
+     type B* suffixes into the array SA. */
+  for(i = n - 1, m = n, c0 = T[n - 1]; 0 <= i;) {
+    /* type A suffix. */
+    do { ++BUCKET_A(c1 = c0); } while((0 <= --i) && ((c0 = T[i]) >= c1));
+    if(0 <= i) {
+      /* type B* suffix. */
+      ++BUCKET_BSTAR(c0, c1);
+      SA[--m] = i;
+      /* type B suffix. */
+      for(--i, c1 = c0; (0 <= i) && ((c0 = T[i]) <= c1); --i, c1 = c0) {
+        ++BUCKET_B(c0, c1);
+      }
+    }
+  }
+  m = n - m;
+/*
+note:
+  A type B* suffix is lexicographically smaller than a type B suffix that
+  begins with the same first two characters.
+*/
+
+  /* Calculate the index of start/end point of each bucket. */
+  for(c0 = 0, i = 0, j = 0; c0 < ALPHABET_SIZE; ++c0) {
+    t = i + BUCKET_A(c0);
+    BUCKET_A(c0) = i + j; /* start point */
+    i = t + BUCKET_B(c0, c0);
+    for(c1 = c0 + 1; c1 < ALPHABET_SIZE; ++c1) {
+      j += BUCKET_BSTAR(c0, c1);
+      BUCKET_BSTAR(c0, c1) = j; /* end point */
+      i += BUCKET_B(c0, c1);
+    }
+  }
+
+  if(0 < m) {
+    /* Sort the type B* suffixes by their first two characters. */
+    PAb = SA + n - m; ISAb = SA + m;
+    for(i = m - 2; 0 <= i; --i) {
+      t = PAb[i], c0 = T[t], c1 = T[t + 1];
+      SA[--BUCKET_BSTAR(c0, c1)] = i;
+    }
+    t = PAb[m - 1], c0 = T[t], c1 = T[t + 1];
+    SA[--BUCKET_BSTAR(c0, c1)] = m - 1;
+
+    /* Sort the type B* substrings using sssort. */
+#ifdef LIBBSC_OPENMP
+    if (openMP)
+    {
+        buf = SA + m;
+        c0 = ALPHABET_SIZE - 2, c1 = ALPHABET_SIZE - 1, j = m;
+#pragma omp parallel default(shared) private(bufsize, curbuf, k, l, d0, d1)
+        {
+          bufsize = (n - (2 * m)) / omp_get_num_threads();
+          curbuf = buf + omp_get_thread_num() * bufsize;
+          k = 0;
+          for(;;) {
+            #pragma omp critical(sssort_lock)
+            {
+              if(0 < (l = j)) {
+                d0 = c0, d1 = c1;
+                do {
+                  k = BUCKET_BSTAR(d0, d1);
+                  if(--d1 <= d0) {
+                    d1 = ALPHABET_SIZE - 1;
+                    if(--d0 < 0) { break; }
+                  }
+                } while(((l - k) <= 1) && (0 < (l = k)));
+                c0 = d0, c1 = d1, j = k;
+              }
+            }
+            if(l == 0) { break; }
+            sssort(T, PAb, SA + k, SA + l,
+                   curbuf, bufsize, 2, n, *(SA + k) == (m - 1));
+          }
+        }
+    }
+    else
+    {
+        buf = SA + m, bufsize = n - (2 * m);
+        for(c0 = ALPHABET_SIZE - 2, j = m; 0 < j; --c0) {
+          for(c1 = ALPHABET_SIZE - 1; c0 < c1; j = i, --c1) {
+            i = BUCKET_BSTAR(c0, c1);
+            if(1 < (j - i)) {
+              sssort(T, PAb, SA + i, SA + j,
+                     buf, bufsize, 2, n, *(SA + i) == (m - 1));
+            }
+          }
+        }
+    }
+#else
+    buf = SA + m, bufsize = n - (2 * m);
+    for(c0 = ALPHABET_SIZE - 2, j = m; 0 < j; --c0) {
+      for(c1 = ALPHABET_SIZE - 1; c0 < c1; j = i, --c1) {
+        i = BUCKET_BSTAR(c0, c1);
+        if(1 < (j - i)) {
+          sssort(T, PAb, SA + i, SA + j,
+                 buf, bufsize, 2, n, *(SA + i) == (m - 1));
+        }
+      }
+    }
+#endif
+
+    /* Compute ranks of type B* substrings. */
+    for(i = m - 1; 0 <= i; --i) {
+      if(0 <= SA[i]) {
+        j = i;
+        do { ISAb[SA[i]] = i; } while((0 <= --i) && (0 <= SA[i]));
+        SA[i + 1] = i - j;
+        if(i <= 0) { break; }
+      }
+      j = i;
+      do { ISAb[SA[i] = ~SA[i]] = j; } while(SA[--i] < 0);
+      ISAb[SA[i]] = j;
+    }
+
+    /* Construct the inverse suffix array of type B* suffixes using trsort. */
+    trsort(ISAb, SA, m, 1);
+
+    /* Set the sorted order of tyoe B* suffixes. */
+    for(i = n - 1, j = m, c0 = T[n - 1]; 0 <= i;) {
+      for(--i, c1 = c0; (0 <= i) && ((c0 = T[i]) >= c1); --i, c1 = c0) { }
+      if(0 <= i) {
+        t = i;
+        for(--i, c1 = c0; (0 <= i) && ((c0 = T[i]) <= c1); --i, c1 = c0) { }
+        SA[ISAb[--j]] = ((t == 0) || (1 < (t - i))) ? t : ~t;
+      }
+    }
+
+    /* Calculate the index of start/end point of each bucket. */
+    BUCKET_B(ALPHABET_SIZE - 1, ALPHABET_SIZE - 1) = n; /* end point */
+    for(c0 = ALPHABET_SIZE - 2, k = m - 1; 0 <= c0; --c0) {
+      i = BUCKET_A(c0 + 1) - 1;
+      for(c1 = ALPHABET_SIZE - 1; c0 < c1; --c1) {
+        t = i - BUCKET_B(c0, c1);
+        BUCKET_B(c0, c1) = i; /* end point */
+
+        /* Move all type B* suffixes to the correct position. */
+        for(i = t, j = BUCKET_BSTAR(c0, c1);
+            j <= k;
+            --i, --k) { SA[i] = SA[k]; }
+      }
+      BUCKET_BSTAR(c0, c0 + 1) = i - BUCKET_B(c0, c0) + 1; /* start point */
+      BUCKET_B(c0, c0) = i; /* end point */
+    }
+  }
+
+  return m;
+}
+
+/* Constructs the suffix array by using the sorted order of type B* suffixes. */
+static
+void
+construct_SA(const unsigned char *T, int *SA,
+             int *bucket_A, int *bucket_B,
+             int n, int m) {
+  int *i, *j, *k;
+  int s;
+  int c0, c1, c2;
+
+  if(0 < m) {
+    /* Construct the sorted order of type B suffixes by using
+       the sorted order of type B* suffixes. */
+    for(c1 = ALPHABET_SIZE - 2; 0 <= c1; --c1) {
+      /* Scan the suffix array from right to left. */
+      for(i = SA + BUCKET_BSTAR(c1, c1 + 1),
+          j = SA + BUCKET_A(c1 + 1) - 1, k = NULL, c2 = -1;
+          i <= j;
+          --j) {
+        if(0 < (s = *j)) {
+          assert(T[s] == c1);
+          assert(((s + 1) < n) && (T[s] <= T[s + 1]));
+          assert(T[s - 1] <= T[s]);
+          *j = ~s;
+          c0 = T[--s];
+          if((0 < s) && (T[s - 1] > c0)) { s = ~s; }
+          if(c0 != c2) {
+            if(0 <= c2) { BUCKET_B(c2, c1) = k - SA; }
+            k = SA + BUCKET_B(c2 = c0, c1);
+          }
+          assert(k < j);
+          *k-- = s;
+        } else {
+          assert(((s == 0) && (T[s] == c1)) || (s < 0));
+          *j = ~s;
+        }
+      }
+    }
+  }
+
+  /* Construct the suffix array by using
+     the sorted order of type B suffixes. */
+  k = SA + BUCKET_A(c2 = T[n - 1]);
+  *k++ = (T[n - 2] < c2) ? ~(n - 1) : (n - 1);
+  /* Scan the suffix array from left to right. */
+  for(i = SA, j = SA + n; i < j; ++i) {
+    if(0 < (s = *i)) {
+      assert(T[s - 1] >= T[s]);
+      c0 = T[--s];
+      if((s == 0) || (T[s - 1] < c0)) { s = ~s; }
+      if(c0 != c2) {
+        BUCKET_A(c2) = k - SA;
+        k = SA + BUCKET_A(c2 = c0);
+      }
+      assert(i < k);
+      *k++ = s;
+    } else {
+      assert(s < 0);
+      *i = ~s;
+    }
+  }
+}
+
+/* Constructs the burrows-wheeler transformed string directly
+   by using the sorted order of type B* suffixes. */
+static
+int
+construct_BWT(const unsigned char *T, int *SA,
+              int *bucket_A, int *bucket_B,
+              int n, int m) {
+  int *i, *j, *k, *orig;
+  int s;
+  int c0, c1, c2;
+
+  if(0 < m) {
+    /* Construct the sorted order of type B suffixes by using
+       the sorted order of type B* suffixes. */
+    for(c1 = ALPHABET_SIZE - 2; 0 <= c1; --c1) {
+      /* Scan the suffix array from right to left. */
+      for(i = SA + BUCKET_BSTAR(c1, c1 + 1),
+          j = SA + BUCKET_A(c1 + 1) - 1, k = NULL, c2 = -1;
+          i <= j;
+          --j) {
+        if(0 < (s = *j)) {
+          assert(T[s] == c1);
+          assert(((s + 1) < n) && (T[s] <= T[s + 1]));
+          assert(T[s - 1] <= T[s]);
+          c0 = T[--s];
+          *j = ~((int)c0);
+          if((0 < s) && (T[s - 1] > c0)) { s = ~s; }
+          if(c0 != c2) {
+            if(0 <= c2) { BUCKET_B(c2, c1) = k - SA; }
+            k = SA + BUCKET_B(c2 = c0, c1);
+          }
+          assert(k < j);
+          *k-- = s;
+        } else if(s != 0) {
+          *j = ~s;
+#ifndef NDEBUG
+        } else {
+          assert(T[s] == c1);
+#endif
+        }
+      }
+    }
+  }
+
+  /* Construct the BWTed string by using
+     the sorted order of type B suffixes. */
+  k = SA + BUCKET_A(c2 = T[n - 1]);
+  *k++ = (T[n - 2] < c2) ? ~((int)T[n - 2]) : (n - 1);
+  /* Scan the suffix array from left to right. */
+  for(i = SA, j = SA + n, orig = SA; i < j; ++i) {
+    if(0 < (s = *i)) {
+      assert(T[s - 1] >= T[s]);
+      c0 = T[--s];
+      *i = c0;
+      if((0 < s) && (T[s - 1] < c0)) { s = ~((int)T[s - 1]); }
+      if(c0 != c2) {
+        BUCKET_A(c2) = k - SA;
+        k = SA + BUCKET_A(c2 = c0);
+      }
+      assert(i < k);
+      *k++ = s;
+    } else if(s != 0) {
+      *i = ~s;
+    } else {
+      orig = i;
+    }
+  }
+
+  return orig - SA;
+}
+
+/* Constructs the burrows-wheeler transformed string directly
+   by using the sorted order of type B* suffixes. */
+static
+int
+construct_BWT_indexes(const unsigned char *T, int *SA,
+                      int *bucket_A, int *bucket_B,
+                      int n, int m,
+                      unsigned char * num_indexes, int * indexes) {
+  int *i, *j, *k, *orig;
+  int s;
+  int c0, c1, c2;
+
+  int mod = n / 8;
+  {
+      mod |= mod >> 1;  mod |= mod >> 2;
+      mod |= mod >> 4;  mod |= mod >> 8;
+      mod |= mod >> 16; mod >>= 1;
+
+      *num_indexes = (unsigned char)((n - 1) / (mod + 1));
+  }
+
+  if(0 < m) {
+    /* Construct the sorted order of type B suffixes by using
+       the sorted order of type B* suffixes. */
+    for(c1 = ALPHABET_SIZE - 2; 0 <= c1; --c1) {
+      /* Scan the suffix array from right to left. */
+      for(i = SA + BUCKET_BSTAR(c1, c1 + 1),
+          j = SA + BUCKET_A(c1 + 1) - 1, k = NULL, c2 = -1;
+          i <= j;
+          --j) {
+        if(0 < (s = *j)) {
+          assert(T[s] == c1);
+          assert(((s + 1) < n) && (T[s] <= T[s + 1]));
+          assert(T[s - 1] <= T[s]);
+
+          if ((s & mod) == 0) indexes[s / (mod + 1) - 1] = j - SA;
+
+          c0 = T[--s];
+          *j = ~((int)c0);
+          if((0 < s) && (T[s - 1] > c0)) { s = ~s; }
+          if(c0 != c2) {
+            if(0 <= c2) { BUCKET_B(c2, c1) = k - SA; }
+            k = SA + BUCKET_B(c2 = c0, c1);
+          }
+          assert(k < j);
+          *k-- = s;
+        } else if(s != 0) {
+          *j = ~s;
+#ifndef NDEBUG
+        } else {
+          assert(T[s] == c1);
+#endif
+        }
+      }
+    }
+  }
+
+  /* Construct the BWTed string by using
+     the sorted order of type B suffixes. */
+  k = SA + BUCKET_A(c2 = T[n - 1]);
+  if (T[n - 2] < c2) {
+    if (((n - 1) & mod) == 0) indexes[(n - 1) / (mod + 1) - 1] = k - SA;
+    *k++ = ~((int)T[n - 2]);
+  }
+  else {
+    *k++ = n - 1;
+  }
+
+  /* Scan the suffix array from left to right. */
+  for(i = SA, j = SA + n, orig = SA; i < j; ++i) {
+    if(0 < (s = *i)) {
+      assert(T[s - 1] >= T[s]);
+
+      if ((s & mod) == 0) indexes[s / (mod + 1) - 1] = i - SA;
+
+      c0 = T[--s];
+      *i = c0;
+      if(c0 != c2) {
+        BUCKET_A(c2) = k - SA;
+        k = SA + BUCKET_A(c2 = c0);
+      }
+      assert(i < k);
+      if((0 < s) && (T[s - 1] < c0)) {
+          if ((s & mod) == 0) indexes[s / (mod + 1) - 1] = k - SA;
+          *k++ = ~((int)T[s - 1]);
+      } else
+        *k++ = s;
+    } else if(s != 0) {
+      *i = ~s;
+    } else {
+      orig = i;
+    }
+  }
+
+  return orig - SA;
+}
+
+
+/*---------------------------------------------------------------------------*/
+
+/*- Function -*/
+
+int
+divsufsort(const unsigned char *T, int *SA, int n, int openMP) {
+  int *bucket_A, *bucket_B;
+  int m;
+  int err = 0;
+
+  /* Check arguments. */
+  if((T == NULL) || (SA == NULL) || (n < 0)) { return -1; }
+  else if(n == 0) { return 0; }
+  else if(n == 1) { SA[0] = 0; return 0; }
+  else if(n == 2) { m = (T[0] < T[1]); SA[m ^ 1] = 0, SA[m] = 1; return 0; }
+
+  bucket_A = (int *)malloc(BUCKET_A_SIZE * sizeof(int));
+  bucket_B = (int *)malloc(BUCKET_B_SIZE * sizeof(int));
+
+  /* Suffixsort. */
+  if((bucket_A != NULL) && (bucket_B != NULL)) {
+    m = sort_typeBstar(T, SA, bucket_A, bucket_B, n, openMP);
+    construct_SA(T, SA, bucket_A, bucket_B, n, m);
+  } else {
+    err = -2;
+  }
+
+  free(bucket_B);
+  free(bucket_A);
+
+  return err;
+}
+
+int
+divbwt(const unsigned char *T, unsigned char *U, int *A, int n, unsigned char * num_indexes, int * indexes, int openMP) {
+  int *B;
+  int *bucket_A, *bucket_B;
+  int m, pidx, i;
+
+  /* Check arguments. */
+  if((T == NULL) || (U == NULL) || (n < 0)) { return -1; }
+  else if(n <= 1) { if(n == 1) { U[0] = T[0]; } return n; }
+
+  if((B = A) == NULL) { B = (int *)malloc((size_t)(n + 1) * sizeof(int)); }
+  bucket_A = (int *)malloc(BUCKET_A_SIZE * sizeof(int));
+  bucket_B = (int *)malloc(BUCKET_B_SIZE * sizeof(int));
+
+  /* Burrows-Wheeler Transform. */
+  if((B != NULL) && (bucket_A != NULL) && (bucket_B != NULL)) {
+    m = sort_typeBstar(T, B, bucket_A, bucket_B, n, openMP);
+
+    if (num_indexes == NULL || indexes == NULL) {
+        pidx = construct_BWT(T, B, bucket_A, bucket_B, n, m);
+    } else {
+        pidx = construct_BWT_indexes(T, B, bucket_A, bucket_B, n, m, num_indexes, indexes);
+    }
+
+    /* Copy to output string. */
+    U[0] = T[n - 1];
+    for(i = 0; i < pidx; ++i) { U[i + 1] = (unsigned char)B[i]; }
+    for(i += 1; i < n; ++i) { U[i] = (unsigned char)B[i]; }
+    pidx += 1;
+  } else {
+    pidx = -2;
+  }
+
+  free(bucket_B);
+  free(bucket_A);
+  if(A == NULL) { free(B); }
+
+  return pidx;
+}
diff --git a/lib/dictBuilder/divsufsort.h b/lib/dictBuilder/divsufsort.h
new file mode 100644
index 0000000..5440994
--- /dev/null
+++ b/lib/dictBuilder/divsufsort.h
@@ -0,0 +1,67 @@
+/*
+ * divsufsort.h for libdivsufsort-lite
+ * Copyright (c) 2003-2008 Yuta Mori All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _DIVSUFSORT_H
+#define _DIVSUFSORT_H 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif /* __cplusplus */
+
+
+/*- Prototypes -*/
+
+/**
+ * Constructs the suffix array of a given string.
+ * @param T [0..n-1] The input string.
+ * @param SA [0..n-1] The output array of suffixes.
+ * @param n The length of the given string.
+ * @param openMP enables OpenMP optimization.
+ * @return 0 if no error occurred, -1 or -2 otherwise.
+ */
+int
+divsufsort(const unsigned char *T, int *SA, int n, int openMP);
+
+/**
+ * Constructs the burrows-wheeler transformed string of a given string.
+ * @param T [0..n-1] The input string.
+ * @param U [0..n-1] The output string. (can be T)
+ * @param A [0..n-1] The temporary array. (can be NULL)
+ * @param n The length of the given string.
+ * @param num_indexes The length of secondary indexes array. (can be NULL)
+ * @param indexes The secondary indexes array. (can be NULL)
+ * @param openMP enables OpenMP optimization.
+ * @return The primary index if no error occurred, -1 or -2 otherwise.
+ */
+int
+divbwt(const unsigned char *T, unsigned char *U, int *A, int n, unsigned char * num_indexes, int * indexes, int openMP);
+
+
+#ifdef __cplusplus
+} /* extern "C" */
+#endif /* __cplusplus */
+
+#endif /* _DIVSUFSORT_H */
diff --git a/lib/dictBuilder/zdict.c b/lib/dictBuilder/zdict.c
new file mode 100644
index 0000000..ea50b54
--- /dev/null
+++ b/lib/dictBuilder/zdict.c
@@ -0,0 +1,1012 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+
+/*-**************************************
+*  Tuning parameters
+****************************************/
+#define ZDICT_MAX_SAMPLES_SIZE (2000U << 20)
+#define ZDICT_MIN_SAMPLES_SIZE 512
+
+
+/*-**************************************
+*  Compiler Options
+****************************************/
+/* Unix Large Files support (>4GB) */
+#define _FILE_OFFSET_BITS 64
+#if (defined(__sun__) && (!defined(__LP64__)))   /* Sun Solaris 32-bits requires specific definitions */
+#  define _LARGEFILE_SOURCE
+#elif ! defined(__LP64__)                        /* No point defining Large file for 64 bit */
+#  define _LARGEFILE64_SOURCE
+#endif
+
+
+/*-*************************************
+*  Dependencies
+***************************************/
+#include         /* malloc, free */
+#include         /* memset */
+#include          /* fprintf, fopen, ftello64 */
+#include           /* clock */
+
+#include "mem.h"           /* read */
+#include "error_private.h"
+#include "fse.h"           /* FSE_normalizeCount, FSE_writeNCount */
+#define HUF_STATIC_LINKING_ONLY
+#include "huf.h"
+#include "zstd_internal.h" /* includes zstd.h */
+#include "xxhash.h"
+#include "divsufsort.h"
+#ifndef ZDICT_STATIC_LINKING_ONLY
+#  define ZDICT_STATIC_LINKING_ONLY
+#endif
+#include "zdict.h"
+
+
+/*-*************************************
+*  Constants
+***************************************/
+#define KB *(1 <<10)
+#define MB *(1 <<20)
+#define GB *(1U<<30)
+
+#define DICTLISTSIZE_DEFAULT 10000
+
+#define NOISELENGTH 32
+
+#define MINRATIO 4
+static const int g_compressionLevel_default = 5;
+static const U32 g_selectivity_default = 9;
+static const size_t g_provision_entropySize = 200;
+static const size_t g_min_fast_dictContent = 192;
+
+
+/*-*************************************
+*  Console display
+***************************************/
+#define DISPLAY(...)         { fprintf(stderr, __VA_ARGS__); fflush( stderr ); }
+#define DISPLAYLEVEL(l, ...) if (notificationLevel>=l) { DISPLAY(__VA_ARGS__); }    /* 0 : no display;   1: errors;   2: default;  3: details;  4: debug */
+
+static clock_t ZDICT_clockSpan(clock_t nPrevious) { return clock() - nPrevious; }
+
+static void ZDICT_printHex(const void* ptr, size_t length)
+{
+    const BYTE* const b = (const BYTE*)ptr;
+    size_t u;
+    for (u=0; u126) c = '.';   /* non-printable char */
+        DISPLAY("%c", c);
+    }
+}
+
+
+/*-********************************************************
+*  Helper functions
+**********************************************************/
+unsigned ZDICT_isError(size_t errorCode) { return ERR_isError(errorCode); }
+
+const char* ZDICT_getErrorName(size_t errorCode) { return ERR_getErrorName(errorCode); }
+
+unsigned ZDICT_getDictID(const void* dictBuffer, size_t dictSize)
+{
+    if (dictSize < 8) return 0;
+    if (MEM_readLE32(dictBuffer) != ZSTD_DICT_MAGIC) return 0;
+    return MEM_readLE32((const char*)dictBuffer + 4);
+}
+
+
+/*-********************************************************
+*  Dictionary training functions
+**********************************************************/
+static unsigned ZDICT_NbCommonBytes (register size_t val)
+{
+    if (MEM_isLittleEndian()) {
+        if (MEM_64bits()) {
+#       if defined(_MSC_VER) && defined(_WIN64)
+            unsigned long r = 0;
+            _BitScanForward64( &r, (U64)val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 3)
+            return (__builtin_ctzll((U64)val) >> 3);
+#       else
+            static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2, 0, 3, 1, 3, 1, 4, 2, 7, 0, 2, 3, 6, 1, 5, 3, 5, 1, 3, 4, 4, 2, 5, 6, 7, 7, 0, 1, 2, 3, 3, 4, 6, 2, 6, 5, 5, 3, 4, 5, 6, 7, 1, 2, 4, 6, 4, 4, 5, 7, 2, 6, 5, 7, 6, 7, 7 };
+            return DeBruijnBytePos[((U64)((val & -(long long)val) * 0x0218A392CDABBD3FULL)) >> 58];
+#       endif
+        } else { /* 32 bits */
+#       if defined(_MSC_VER)
+            unsigned long r=0;
+            _BitScanForward( &r, (U32)val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 3)
+            return (__builtin_ctz((U32)val) >> 3);
+#       else
+            static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0, 3, 2, 2, 1, 3, 2, 0, 1, 3, 3, 1, 2, 2, 2, 2, 0, 3, 1, 2, 0, 1, 0, 1, 1 };
+            return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27];
+#       endif
+        }
+    } else {  /* Big Endian CPU */
+        if (MEM_64bits()) {
+#       if defined(_MSC_VER) && defined(_WIN64)
+            unsigned long r = 0;
+            _BitScanReverse64( &r, val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 3)
+            return (__builtin_clzll(val) >> 3);
+#       else
+            unsigned r;
+            const unsigned n32 = sizeof(size_t)*4;   /* calculate this way due to compiler complaining in 32-bits mode */
+            if (!(val>>n32)) { r=4; } else { r=0; val>>=n32; }
+            if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; }
+            r += (!val);
+            return r;
+#       endif
+        } else { /* 32 bits */
+#       if defined(_MSC_VER)
+            unsigned long r = 0;
+            _BitScanReverse( &r, (unsigned long)val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 3)
+            return (__builtin_clz((U32)val) >> 3);
+#       else
+            unsigned r;
+            if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; }
+            r += (!val);
+            return r;
+#       endif
+    }   }
+}
+
+
+/*! ZDICT_count() :
+    Count the nb of common bytes between 2 pointers.
+    Note : this function presumes end of buffer followed by noisy guard band.
+*/
+static size_t ZDICT_count(const void* pIn, const void* pMatch)
+{
+    const char* const pStart = (const char*)pIn;
+    for (;;) {
+        size_t const diff = MEM_readST(pMatch) ^ MEM_readST(pIn);
+        if (!diff) {
+            pIn = (const char*)pIn+sizeof(size_t);
+            pMatch = (const char*)pMatch+sizeof(size_t);
+            continue;
+        }
+        pIn = (const char*)pIn+ZDICT_NbCommonBytes(diff);
+        return (size_t)((const char*)pIn - pStart);
+    }
+}
+
+
+typedef struct {
+    U32 pos;
+    U32 length;
+    U32 savings;
+} dictItem;
+
+static void ZDICT_initDictItem(dictItem* d)
+{
+    d->pos = 1;
+    d->length = 0;
+    d->savings = (U32)(-1);
+}
+
+
+#define LLIMIT 64          /* heuristic determined experimentally */
+#define MINMATCHLENGTH 7   /* heuristic determined experimentally */
+static dictItem ZDICT_analyzePos(
+                       BYTE* doneMarks,
+                       const int* suffix, U32 start,
+                       const void* buffer, U32 minRatio, U32 notificationLevel)
+{
+    U32 lengthList[LLIMIT] = {0};
+    U32 cumulLength[LLIMIT] = {0};
+    U32 savings[LLIMIT] = {0};
+    const BYTE* b = (const BYTE*)buffer;
+    size_t length;
+    size_t maxLength = LLIMIT;
+    size_t pos = suffix[start];
+    U32 end = start;
+    dictItem solution;
+
+    /* init */
+    memset(&solution, 0, sizeof(solution));
+    doneMarks[pos] = 1;
+
+    /* trivial repetition cases */
+    if ( (MEM_read16(b+pos+0) == MEM_read16(b+pos+2))
+       ||(MEM_read16(b+pos+1) == MEM_read16(b+pos+3))
+       ||(MEM_read16(b+pos+2) == MEM_read16(b+pos+4)) ) {
+        /* skip and mark segment */
+        U16 u16 = MEM_read16(b+pos+4);
+        U32 u, e = 6;
+        while (MEM_read16(b+pos+e) == u16) e+=2 ;
+        if (b[pos+e] == b[pos+e-1]) e++;
+        for (u=1; u=MINMATCHLENGTH);
+
+    /* look backward */
+    do {
+        length = ZDICT_count(b + pos, b + *(suffix+start-1));
+        if (length >=MINMATCHLENGTH) start--;
+    } while(length >= MINMATCHLENGTH);
+
+    /* exit if not found a minimum nb of repetitions */
+    if (end-start < minRatio) {
+        U32 idx;
+        for(idx=start; idx= %i at pos %7u  ", (U32)(end-start), MINMATCHLENGTH, (U32)pos);
+        DISPLAYLEVEL(4, "\n");
+
+        for (searchLength = MINMATCHLENGTH ; ; searchLength++) {
+            BYTE currentChar = 0;
+            U32 currentCount = 0;
+            U32 currentID = refinedStart;
+            U32 id;
+            U32 selectedCount = 0;
+            U32 selectedID = currentID;
+            for (id =refinedStart; id < refinedEnd; id++) {
+                if (b[ suffix[id] + searchLength] != currentChar) {
+                    if (currentCount > selectedCount) {
+                        selectedCount = currentCount;
+                        selectedID = currentID;
+                    }
+                    currentID = id;
+                    currentChar = b[ suffix[id] + searchLength];
+                    currentCount = 0;
+                }
+                currentCount ++;
+            }
+            if (currentCount > selectedCount) {  /* for last */
+                selectedCount = currentCount;
+                selectedID = currentID;
+            }
+
+            if (selectedCount < minRatio)
+                break;
+            refinedStart = selectedID;
+            refinedEnd = refinedStart + selectedCount;
+        }
+
+        /* evaluate gain based on new ref */
+        start = refinedStart;
+        pos = suffix[refinedStart];
+        end = start;
+        memset(lengthList, 0, sizeof(lengthList));
+
+        /* look forward */
+        do {
+            end++;
+            length = ZDICT_count(b + pos, b + suffix[end]);
+            if (length >= LLIMIT) length = LLIMIT-1;
+            lengthList[length]++;
+        } while (length >=MINMATCHLENGTH);
+
+        /* look backward */
+		length = MINMATCHLENGTH;
+		while ((length >= MINMATCHLENGTH) & (start > 0)) {
+			length = ZDICT_count(b + pos, b + suffix[start - 1]);
+			if (length >= LLIMIT) length = LLIMIT - 1;
+			lengthList[length]++;
+			if (length >= MINMATCHLENGTH) start--;
+		}
+
+        /* largest useful length */
+        memset(cumulLength, 0, sizeof(cumulLength));
+        cumulLength[maxLength-1] = lengthList[maxLength-1];
+        for (i=(int)(maxLength-2); i>=0; i--)
+            cumulLength[i] = cumulLength[i+1] + lengthList[i];
+
+        for (i=LLIMIT-1; i>=MINMATCHLENGTH; i--) if (cumulLength[i]>=minRatio) break;
+        maxLength = i;
+
+        /* reduce maxLength in case of final into repetitive data */
+        {   U32 l = (U32)maxLength;
+            BYTE const c = b[pos + maxLength-1];
+            while (b[pos+l-2]==c) l--;
+            maxLength = l;
+        }
+        if (maxLength < MINMATCHLENGTH) return solution;   /* skip : no long-enough solution */
+
+        /* calculate savings */
+        savings[5] = 0;
+        for (i=MINMATCHLENGTH; i<=(int)maxLength; i++)
+            savings[i] = savings[i-1] + (lengthList[i] * (i-3));
+
+        DISPLAYLEVEL(4, "Selected ref at position %u, of length %u : saves %u (ratio: %.2f)  \n",
+                     (U32)pos, (U32)maxLength, savings[maxLength], (double)savings[maxLength] / maxLength);
+
+        solution.pos = (U32)pos;
+        solution.length = (U32)maxLength;
+        solution.savings = savings[maxLength];
+
+        /* mark positions done */
+        {   U32 id;
+            for (id=start; id solution.length) length = solution.length;
+                }
+                pEnd = (U32)(testedPos + length);
+                for (p=testedPos; ppos;
+    const U32 eltEnd = elt.pos + elt.length;
+
+    /* tail overlap */
+    U32 u; for (u=1; u elt.pos) && (table[u].pos <= eltEnd)) {  /* overlap, existing > new */
+            /* append */
+            U32 addedLength = table[u].pos - elt.pos;
+            table[u].length += addedLength;
+            table[u].pos = elt.pos;
+            table[u].savings += elt.savings * addedLength / elt.length;   /* rough approx */
+            table[u].savings += elt.length / 8;    /* rough approx bonus */
+            elt = table[u];
+            /* sort : improve rank */
+            while ((u>1) && (table[u-1].savings < elt.savings))
+            table[u] = table[u-1], u--;
+            table[u] = elt;
+            return u;
+    }   }
+
+    /* front overlap */
+    for (u=1; u= elt.pos) && (table[u].pos < elt.pos)) {  /* overlap, existing < new */
+            /* append */
+            int addedLength = (int)eltEnd - (table[u].pos + table[u].length);
+            table[u].savings += elt.length / 8;    /* rough approx bonus */
+            if (addedLength > 0) {   /* otherwise, elt fully included into existing */
+                table[u].length += addedLength;
+                table[u].savings += elt.savings * addedLength / elt.length;   /* rough approx */
+            }
+            /* sort : improve rank */
+            elt = table[u];
+            while ((u>1) && (table[u-1].savings < elt.savings))
+                table[u] = table[u-1], u--;
+            table[u] = elt;
+            return u;
+    }   }
+
+    return 0;
+}
+
+
+static void ZDICT_removeDictItem(dictItem* table, U32 id)
+{
+    /* convention : first element is nb of elts */
+    U32 const max = table->pos;
+    U32 u;
+    if (!id) return;   /* protection, should never happen */
+    for (u=id; upos--;
+}
+
+
+static void ZDICT_insertDictItem(dictItem* table, U32 maxSize, dictItem elt)
+{
+    /* merge if possible */
+    U32 mergeId = ZDICT_checkMerge(table, elt, 0);
+    if (mergeId) {
+        U32 newMerge = 1;
+        while (newMerge) {
+            newMerge = ZDICT_checkMerge(table, table[mergeId], mergeId);
+            if (newMerge) ZDICT_removeDictItem(table, mergeId);
+            mergeId = newMerge;
+        }
+        return;
+    }
+
+    /* insert */
+    {   U32 current;
+        U32 nextElt = table->pos;
+        if (nextElt >= maxSize) nextElt = maxSize-1;
+        current = nextElt-1;
+        while (table[current].savings < elt.savings) {
+            table[current+1] = table[current];
+            current--;
+        }
+        table[current+1] = elt;
+        table->pos = nextElt+1;
+    }
+}
+
+
+static U32 ZDICT_dictSize(const dictItem* dictList)
+{
+    U32 u, dictSize = 0;
+    for (u=1; u=l) { \
+            if (ZDICT_clockSpan(displayClock) > refreshRate)  \
+            { displayClock = clock(); DISPLAY(__VA_ARGS__); \
+            if (notificationLevel>=4) fflush(stdout); } }
+
+    /* init */
+    DISPLAYLEVEL(2, "\r%70s\r", "");   /* clean display line */
+    if (!suffix0 || !reverseSuffix || !doneMarks || !filePos) {
+        result = ERROR(memory_allocation);
+        goto _cleanup;
+    }
+    if (minRatio < MINRATIO) minRatio = MINRATIO;
+    memset(doneMarks, 0, bufferSize+16);
+
+    /* limit sample set size (divsufsort limitation)*/
+    if (bufferSize > ZDICT_MAX_SAMPLES_SIZE) DISPLAYLEVEL(3, "sample set too large : reduced to %u MB ...\n", (U32)(ZDICT_MAX_SAMPLES_SIZE>>20));
+    while (bufferSize > ZDICT_MAX_SAMPLES_SIZE) bufferSize -= fileSizes[--nbFiles];
+
+    /* sort */
+    DISPLAYLEVEL(2, "sorting %u files of total size %u MB ...\n", nbFiles, (U32)(bufferSize>>20));
+    {   int const divSuftSortResult = divsufsort((const unsigned char*)buffer, suffix, (int)bufferSize, 0);
+        if (divSuftSortResult != 0) { result = ERROR(GENERIC); goto _cleanup; }
+    }
+    suffix[bufferSize] = (int)bufferSize;   /* leads into noise */
+    suffix0[0] = (int)bufferSize;           /* leads into noise */
+    /* build reverse suffix sort */
+    {   size_t pos;
+        for (pos=0; pos < bufferSize; pos++)
+            reverseSuffix[suffix[pos]] = (U32)pos;
+        /* note filePos tracks borders between samples.
+           It's not used at this stage, but planned to become useful in a later update */
+        filePos[0] = 0;
+        for (pos=1; pos> 21);
+    }
+}
+
+
+typedef struct
+{
+    ZSTD_CCtx* ref;
+    ZSTD_CCtx* zc;
+    void* workPlace;   /* must be ZSTD_BLOCKSIZE_ABSOLUTEMAX allocated */
+} EStats_ress_t;
+
+#define MAXREPOFFSET 1024
+
+static void ZDICT_countEStats(EStats_ress_t esr, ZSTD_parameters params,
+                            U32* countLit, U32* offsetcodeCount, U32* matchlengthCount, U32* litlengthCount, U32* repOffsets,
+                            const void* src, size_t srcSize, U32 notificationLevel)
+{
+    size_t const blockSizeMax = MIN (ZSTD_BLOCKSIZE_ABSOLUTEMAX, 1 << params.cParams.windowLog);
+    size_t cSize;
+
+    if (srcSize > blockSizeMax) srcSize = blockSizeMax;   /* protection vs large samples */
+    {  size_t const errorCode = ZSTD_copyCCtx(esr.zc, esr.ref, 0);
+            if (ZSTD_isError(errorCode)) { DISPLAYLEVEL(1, "warning : ZSTD_copyCCtx failed \n"); return; }
+    }
+    cSize = ZSTD_compressBlock(esr.zc, esr.workPlace, ZSTD_BLOCKSIZE_ABSOLUTEMAX, src, srcSize);
+    if (ZSTD_isError(cSize)) { DISPLAYLEVEL(1, "warning : could not compress sample size %u \n", (U32)srcSize); return; }
+
+    if (cSize) {  /* if == 0; block is not compressible */
+        const seqStore_t* seqStorePtr = ZSTD_getSeqStore(esr.zc);
+
+        /* literals stats */
+        {   const BYTE* bytePtr;
+            for(bytePtr = seqStorePtr->litStart; bytePtr < seqStorePtr->lit; bytePtr++)
+                countLit[*bytePtr]++;
+        }
+
+        /* seqStats */
+        {   U32 const nbSeq = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
+            ZSTD_seqToCodes(seqStorePtr);
+
+            {   const BYTE* codePtr = seqStorePtr->ofCode;
+                U32 u;
+                for (u=0; umlCode;
+                U32 u;
+                for (u=0; ullCode;
+                U32 u;
+                for (u=0; u= 2) { /* rep offsets */
+                const seqDef* const seq = seqStorePtr->sequencesStart;
+                U32 offset1 = seq[0].offset - 3;
+                U32 offset2 = seq[1].offset - 3;
+                if (offset1 >= MAXREPOFFSET) offset1 = 0;
+                if (offset2 >= MAXREPOFFSET) offset2 = 0;
+                repOffsets[offset1] += 3;
+                repOffsets[offset2] += 1;
+    }   }   }
+}
+
+/*
+static size_t ZDICT_maxSampleSize(const size_t* fileSizes, unsigned nbFiles)
+{
+    unsigned u;
+    size_t max=0;
+    for (u=0; u0; u--) {
+        offsetCount_t tmp;
+        if (table[u-1].count >= table[u].count) break;
+        tmp = table[u-1];
+        table[u-1] = table[u];
+        table[u] = tmp;
+    }
+}
+
+
+#define OFFCODE_MAX 30  /* only applicable to first block */
+static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
+                                   unsigned compressionLevel,
+                             const void*  srcBuffer, const size_t* fileSizes, unsigned nbFiles,
+                             const void* dictBuffer, size_t  dictBufferSize,
+                                   unsigned notificationLevel)
+{
+    U32 countLit[256];
+    HUF_CREATE_STATIC_CTABLE(hufTable, 255);
+    U32 offcodeCount[OFFCODE_MAX+1];
+    short offcodeNCount[OFFCODE_MAX+1];
+    U32 offcodeMax = ZSTD_highbit32((U32)(dictBufferSize + 128 KB));
+    U32 matchLengthCount[MaxML+1];
+    short matchLengthNCount[MaxML+1];
+    U32 litLengthCount[MaxLL+1];
+    short litLengthNCount[MaxLL+1];
+    U32 repOffset[MAXREPOFFSET];
+    offsetCount_t bestRepOffset[ZSTD_REP_NUM+1];
+    EStats_ress_t esr;
+    ZSTD_parameters params;
+    U32 u, huffLog = 11, Offlog = OffFSELog, mlLog = MLFSELog, llLog = LLFSELog, total;
+    size_t pos = 0, errorCode;
+    size_t eSize = 0;
+    size_t const totalSrcSize = ZDICT_totalSampleSize(fileSizes, nbFiles);
+    size_t const averageSampleSize = totalSrcSize / (nbFiles + !nbFiles);
+    BYTE* dstPtr = (BYTE*)dstBuffer;
+
+    /* init */
+    esr.ref = ZSTD_createCCtx();
+    esr.zc = ZSTD_createCCtx();
+    esr.workPlace = malloc(ZSTD_BLOCKSIZE_ABSOLUTEMAX);
+    if (!esr.ref || !esr.zc || !esr.workPlace) {
+        eSize = ERROR(memory_allocation);
+        DISPLAYLEVEL(1, "Not enough memory \n");
+        goto _cleanup;
+    }
+    if (offcodeMax>OFFCODE_MAX) { eSize = ERROR(dictionary_wrong); goto _cleanup; }   /* too large dictionary */
+    for (u=0; u<256; u++) countLit[u]=1;   /* any character must be described */
+    for (u=0; u<=offcodeMax; u++) offcodeCount[u]=1;
+    for (u=0; u<=MaxML; u++) matchLengthCount[u]=1;
+    for (u=0; u<=MaxLL; u++) litLengthCount[u]=1;
+    memset(repOffset, 0, sizeof(repOffset));
+    repOffset[1] = repOffset[4] = repOffset[8] = 1;
+    memset(bestRepOffset, 0, sizeof(bestRepOffset));
+    if (compressionLevel==0) compressionLevel=g_compressionLevel_default;
+    params = ZSTD_getParams(compressionLevel, averageSampleSize, dictBufferSize);
+    {   size_t const beginResult = ZSTD_compressBegin_advanced(esr.ref, dictBuffer, dictBufferSize, params, 0);
+            if (ZSTD_isError(beginResult)) {
+            eSize = ERROR(GENERIC);
+            DISPLAYLEVEL(1, "error : ZSTD_compressBegin_advanced failed \n");
+            goto _cleanup;
+    }   }
+
+    /* collect stats on all files */
+    for (u=0; u 30) ? MINRATIO : nbSamples >> selectivity;
+    size_t const targetDictSize = maxDictSize;
+    size_t const samplesBuffSize = ZDICT_totalSampleSize(samplesSizes, nbSamples);
+    size_t dictSize = 0;
+    U32 const notificationLevel = params.notificationLevel;
+
+    /* checks */
+    if (!dictList) return ERROR(memory_allocation);
+    if (maxDictSize <= g_provision_entropySize + g_min_fast_dictContent) { free(dictList); return ERROR(dstSize_tooSmall); }
+    if (samplesBuffSize < ZDICT_MIN_SAMPLES_SIZE) { free(dictList); return 0; }   /* not enough source to create dictionary */
+
+    /* init */
+    ZDICT_initDictItem(dictList);
+
+    /* build dictionary */
+    ZDICT_trainBuffer(dictList, dictListSize,
+                    samplesBuffer, samplesBuffSize,
+                    samplesSizes, nbSamples,
+                    minRep, notificationLevel);
+
+    /* display best matches */
+    if (params.notificationLevel>= 3) {
+        U32 const nb = MIN(25, dictList[0].pos);
+        U32 const dictContentSize = ZDICT_dictSize(dictList);
+        U32 u;
+        DISPLAYLEVEL(3, "\n %u segments found, of total size %u \n", dictList[0].pos-1, dictContentSize);
+        DISPLAYLEVEL(3, "list %u best segments \n", nb-1);
+        for (u=1; u samplesBuffSize) || ((pos + length) > samplesBuffSize))
+                return ERROR(GENERIC);   /* should never happen */
+            DISPLAYLEVEL(3, "%3u:%3u bytes at pos %8u, savings %7u bytes |",
+                         u, length, pos, dictList[u].savings);
+            ZDICT_printHex((const char*)samplesBuffer+pos, printedLength);
+            DISPLAYLEVEL(3, "| \n");
+    }   }
+
+
+    /* create dictionary */
+    {   U32 dictContentSize = ZDICT_dictSize(dictList);
+        if (dictContentSize < targetDictSize/3) {
+            DISPLAYLEVEL(2, "!  warning : selected content significantly smaller than requested (%u < %u) \n", dictContentSize, (U32)maxDictSize);
+            if (minRep > MINRATIO) {
+                DISPLAYLEVEL(2, "!  consider increasing selectivity to produce larger dictionary (-s%u) \n", selectivity+1);
+                DISPLAYLEVEL(2, "!  note : larger dictionaries are not necessarily better, test its efficiency on samples \n");
+            }
+            if (samplesBuffSize < 10 * targetDictSize)
+                DISPLAYLEVEL(2, "!  consider increasing the number of samples (total size : %u MB)\n", (U32)(samplesBuffSize>>20));
+        }
+
+        if ((dictContentSize > targetDictSize*3) && (nbSamples > 2*MINRATIO) && (selectivity>1)) {
+            U32 proposedSelectivity = selectivity-1;
+            while ((nbSamples >> proposedSelectivity) <= MINRATIO) { proposedSelectivity--; }
+            DISPLAYLEVEL(2, "!  note : calculated dictionary significantly larger than requested (%u > %u) \n", dictContentSize, (U32)maxDictSize);
+            DISPLAYLEVEL(2, "!  consider increasing dictionary size, or produce denser dictionary (-s%u) \n", proposedSelectivity);
+            DISPLAYLEVEL(2, "!  always test dictionary efficiency on samples \n");
+        }
+
+        /* limit dictionary size */
+        {   U32 const max = dictList->pos;   /* convention : nb of useful elts within dictList */
+            U32 currentSize = 0;
+            U32 n; for (n=1; n targetDictSize) { currentSize -= dictList[n].length; break; }
+            }
+            dictList->pos = n;
+            dictContentSize = currentSize;
+        }
+
+        /* build dict content */
+        {   U32 u;
+            BYTE* ptr = (BYTE*)dictBuffer + maxDictSize;
+            for (u=1; upos; u++) {
+                U32 l = dictList[u].length;
+                ptr -= l;
+                if (ptr<(BYTE*)dictBuffer) { free(dictList); return ERROR(GENERIC); }   /* should not happen */
+                memcpy(ptr, (const char*)samplesBuffer+dictList[u].pos, l);
+        }   }
+
+        dictSize = ZDICT_addEntropyTablesFromBuffer_advanced(dictBuffer, dictContentSize, maxDictSize,
+                                                             samplesBuffer, samplesSizes, nbSamples,
+                                                             params);
+    }
+
+    /* clean up */
+    free(dictList);
+    return dictSize;
+}
+
+
+/* issue : samplesBuffer need to be followed by a noisy guard band.
+*  work around : duplicate the buffer, and add the noise */
+size_t ZDICT_trainFromBuffer_advanced(void* dictBuffer, size_t dictBufferCapacity,
+                                      const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
+                                      ZDICT_params_t params)
+{
+    size_t result;
+    void* newBuff;
+    size_t const sBuffSize = ZDICT_totalSampleSize(samplesSizes, nbSamples);
+    if (sBuffSize < ZDICT_MIN_SAMPLES_SIZE) return 0;   /* not enough content => no dictionary */
+
+    newBuff = malloc(sBuffSize + NOISELENGTH);
+    if (!newBuff) return ERROR(memory_allocation);
+
+    memcpy(newBuff, samplesBuffer, sBuffSize);
+    ZDICT_fillNoise((char*)newBuff + sBuffSize, NOISELENGTH);   /* guard band, for end of buffer condition */
+
+    result = ZDICT_trainFromBuffer_unsafe(
+                                        dictBuffer, dictBufferCapacity,
+                                        newBuff, samplesSizes, nbSamples,
+                                        params);
+    free(newBuff);
+    return result;
+}
+
+
+size_t ZDICT_trainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
+                             const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples)
+{
+    ZDICT_params_t params;
+    memset(¶ms, 0, sizeof(params));
+    return ZDICT_trainFromBuffer_advanced(dictBuffer, dictBufferCapacity,
+                                          samplesBuffer, samplesSizes, nbSamples,
+                                          params);
+}
+
+size_t ZDICT_addEntropyTablesFromBuffer(void* dictBuffer, size_t dictContentSize, size_t dictBufferCapacity,
+                                        const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples)
+{
+    ZDICT_params_t params;
+    memset(¶ms, 0, sizeof(params));
+    return ZDICT_addEntropyTablesFromBuffer_advanced(dictBuffer, dictContentSize, dictBufferCapacity,
+                                                     samplesBuffer, samplesSizes, nbSamples,
+                                                     params);
+}
diff --git a/lib/dictBuilder/zdict.h b/lib/dictBuilder/zdict.h
new file mode 100644
index 0000000..642a435
--- /dev/null
+++ b/lib/dictBuilder/zdict.h
@@ -0,0 +1,111 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+#ifndef DICTBUILDER_H_001
+#define DICTBUILDER_H_001
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+
+/*======  Dependencies  ======*/
+#include   /* size_t */
+
+
+/*======  Export for Windows  ======*/
+/*!
+*  ZSTD_DLL_EXPORT :
+*  Enable exporting of functions when building a Windows DLL
+*/
+#if defined(_WIN32) && defined(ZSTD_DLL_EXPORT) && (ZSTD_DLL_EXPORT==1)
+#  define ZDICTLIB_API __declspec(dllexport)
+#else
+#  define ZDICTLIB_API
+#endif
+
+
+/*! ZDICT_trainFromBuffer() :
+    Train a dictionary from an array of samples.
+    Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
+    supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
+    The resulting dictionary will be saved into `dictBuffer`.
+    @return : size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
+              or an error code, which can be tested with ZDICT_isError().
+    Tips : In general, a reasonable dictionary has a size of ~ 100 KB.
+           It's obviously possible to target smaller or larger ones, just by specifying different `dictBufferCapacity`.
+           In general, it's recommended to provide a few thousands samples, but this can vary a lot.
+           It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
+*/
+ZDICTLIB_API size_t ZDICT_trainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
+                       const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples);
+
+
+/*======   Helper functions   ======*/
+ZDICTLIB_API unsigned ZDICT_getDictID(const void* dictBuffer, size_t dictSize);  /**< extracts dictID; @return zero if error (not a valid dictionary) */
+ZDICTLIB_API unsigned ZDICT_isError(size_t errorCode);
+ZDICTLIB_API const char* ZDICT_getErrorName(size_t errorCode);
+
+
+
+#ifdef ZDICT_STATIC_LINKING_ONLY
+
+/* ====================================================================================
+ * The definitions in this section are considered experimental.
+ * They should never be used with a dynamic library, as they may change in the future.
+ * They are provided for advanced usages.
+ * Use them only in association with static linking.
+ * ==================================================================================== */
+
+typedef struct {
+    unsigned selectivityLevel;   /* 0 means default; larger => select more => larger dictionary */
+    int      compressionLevel;   /* 0 means default; target a specific zstd compression level */
+    unsigned notificationLevel;  /* Write to stderr; 0 = none (default); 1 = errors; 2 = progression; 3 = details; 4 = debug; */
+    unsigned dictID;             /* 0 means auto mode (32-bits random value); other : force dictID value */
+    unsigned reserved[2];        /* reserved space for future parameters */
+} ZDICT_params_t;
+
+
+/*! ZDICT_trainFromBuffer_advanced() :
+    Same as ZDICT_trainFromBuffer() with control over more parameters.
+    `parameters` is optional and can be provided with values set to 0 to mean "default".
+    @return : size of dictionary stored into `dictBuffer` (<= `dictBufferSize`),
+              or an error code, which can be tested by ZDICT_isError().
+    note : ZDICT_trainFromBuffer_advanced() will send notifications into stderr if instructed to, using notificationLevel>0.
+*/
+size_t ZDICT_trainFromBuffer_advanced(void* dictBuffer, size_t dictBufferCapacity,
+                                const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
+                                ZDICT_params_t parameters);
+
+
+/*! ZDICT_addEntropyTablesFromBuffer() :
+
+    Given a content-only dictionary (built using any 3rd party algorithm),
+    add entropy tables computed from an array of samples.
+    Samples must be stored concatenated in a flat buffer `samplesBuffer`,
+    supplied with an array of sizes `samplesSizes`, providing the size of each sample in order.
+
+    The input dictionary content must be stored *at the end* of `dictBuffer`.
+    Its size is `dictContentSize`.
+    The resulting dictionary with added entropy tables will be *written back to `dictBuffer`*,
+    starting from its beginning.
+    @return : size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`).
+*/
+size_t ZDICT_addEntropyTablesFromBuffer(void* dictBuffer, size_t dictContentSize, size_t dictBufferCapacity,
+                                        const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples);
+
+
+
+#endif   /* ZDICT_STATIC_LINKING_ONLY */
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif   /* DICTBUILDER_H_001 */
diff --git a/lib/dll/example/Makefile b/lib/dll/example/Makefile
new file mode 100644
index 0000000..36041a0
--- /dev/null
+++ b/lib/dll/example/Makefile
@@ -0,0 +1,47 @@
+# ##########################################################################
+# Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree. An additional grant
+# of patent rights can be found in the PATENTS file in the same directory.
+# ##########################################################################
+
+VOID    := /dev/null
+ZSTDDIR  := ../include
+LIBDIR  := ../static
+DLLDIR  := ../dll
+
+CFLAGS  ?= -O3   # can select custom flags. For example : CFLAGS="-O2 -g" make
+CFLAGS  += -Wall -Wextra -Wundef -Wcast-qual -Wcast-align -Wshadow -Wswitch-enum \
+           -Wdeclaration-after-statement -Wstrict-prototypes \
+           -Wpointer-arith -Wstrict-aliasing=1
+CFLAGS  += $(MOREFLAGS)
+CPPFLAGS:= -I$(ZSTDDIR) -DXXH_NAMESPACE=ZSTD_
+FLAGS   := $(CFLAGS) $(CPPFLAGS) $(LDFLAGS)
+
+
+# Define *.exe as extension for Windows systems
+ifneq (,$(filter Windows%,$(OS)))
+EXT =.exe
+else
+EXT =
+endif
+
+.PHONY: default fullbench-dll fullbench-lib
+
+
+default: all
+
+all: fullbench-dll fullbench-lib
+
+
+fullbench-lib: fullbench.c datagen.c
+	$(CC) $(FLAGS) $^ -o $@$(EXT) $(LIBDIR)/libzstd_static.lib
+
+fullbench-dll: fullbench.c datagen.c
+	$(CC) $(FLAGS) $^ -o $@$(EXT) -DZSTD_DLL_IMPORT=1 $(DLLDIR)/libzstd.dll
+
+clean:
+	@$(RM) fullbench-dll$(EXT) fullbench-lib$(EXT) \
+	@echo Cleaning completed
diff --git a/lib/dll/example/README.md b/lib/dll/example/README.md
new file mode 100644
index 0000000..957a29f
--- /dev/null
+++ b/lib/dll/example/README.md
@@ -0,0 +1,69 @@
+ZSTD Windows binary package
+====================================
+
+#### The package contents
+
+- `zstd.exe`                  : Command Line Utility, supporting gzip-like arguments
+- `dll\libzstd.dll`           : The DLL of ZSTD library
+- `dll\libzstd.lib`           : The import library of ZSTD library for Visual C++
+- `example\`                  : The example of usage of ZSTD library
+- `include\`                  : Header files required with ZSTD library
+- `static\libzstd_static.lib` : The static ZSTD library
+
+
+#### Usage of Command Line Interface
+
+Command Line Interface (CLI) supports gzip-like arguments.
+By default CLI takes an input file and compresses it to an output file:
+```
+    Usage: zstd [arg] [input] [output]
+```
+The full list of commands for CLI can be obtained with `-h` or `-H`. The ratio can
+be improved with commands from `-3` to `-16` but higher levels also have slower
+compression. CLI includes in-memory compression benchmark module with compression
+levels starting from `-b` and ending with `-e` with iteration time of `-i` seconds.
+CLI supports aggregation of parameters i.e. `-b1`, `-e18`, and `-i1` can be joined
+into `-b1e18i1`.
+
+
+#### The example of usage of static and dynamic ZSTD libraries with gcc/MinGW
+
+Use `cd example` and `make` to build `fullbench-dll` and `fullbench-lib`.
+`fullbench-dll` uses a dynamic ZSTD library from the `dll` directory.
+`fullbench-lib` uses a static ZSTD library from the `lib` directory.
+
+
+#### Using ZSTD DLL with gcc/MinGW
+
+The header files from `include\` and the dynamic library `dll\libzstd.dll`
+are required to compile a project using gcc/MinGW.
+The dynamic library has to be added to linking options.
+It means that if a project that uses ZSTD consists of a single `test-dll.c`
+file it should be linked with `dll\libzstd.dll`. For example:
+```
+    gcc $(CFLAGS) -Iinclude\ test-dll.c -o test-dll dll\libzstd.dll
+```
+The compiled executable will require ZSTD DLL which is available at `dll\libzstd.dll`.
+
+
+#### The example of usage of static and dynamic ZSTD libraries with Visual C++
+
+Open `example\fullbench-dll.sln` to compile `fullbench-dll` that uses a
+dynamic ZSTD library from the `dll` directory. The solution works with Visual C++
+2010 or newer. When one will open the solution with Visual C++ newer than 2010
+then the solution will upgraded to the current version.
+
+
+#### Using ZSTD DLL with Visual C++
+
+The header files from `include\` and the import library `dll\libzstd.lib`
+are required to compile a project using Visual C++.
+
+1. The path to header files should be added to `Additional Include Directories` that can
+   be found in project properties `C/C++` then `General`.
+2. The import library has to be added to `Additional Dependencies` that can
+   be found in project properties `Linker` then `Input`.
+   If one will provide only the name `libzstd.lib` without a full path to the library
+   the directory has to be added to `Linker\General\Additional Library Directories`.
+
+The compiled executable will require ZSTD DLL which is available at `dll\libzstd.dll`.
diff --git a/lib/dll/example/build_package.bat b/lib/dll/example/build_package.bat
new file mode 100644
index 0000000..a7a7ab0
--- /dev/null
+++ b/lib/dll/example/build_package.bat
@@ -0,0 +1,17 @@
+@ECHO OFF
+MKDIR bin\dll bin\static bin\example bin\include
+COPY tests\fullbench.c bin\example\
+COPY programs\datagen.c bin\example\
+COPY programs\datagen.h bin\example\
+COPY programs\util.h bin\example\
+COPY lib\common\mem.h bin\example\
+COPY lib\common\zstd_errors.h bin\example\
+COPY lib\common\zstd_internal.h bin\example\
+COPY lib\common\error_private.h bin\example\
+COPY lib\zstd.h bin\include\
+COPY lib\libzstd.a bin\static\libzstd_static.lib
+COPY lib\dll\libzstd.* bin\dll\
+COPY lib\dll\example\Makefile bin\example\
+COPY lib\dll\example\fullbench-dll.* bin\example\
+COPY lib\dll\example\README.md bin\
+COPY programs\zstd.exe bin\zstd.exe
diff --git a/lib/dll/example/fullbench-dll.sln b/lib/dll/example/fullbench-dll.sln
new file mode 100644
index 0000000..ef8d4c0
--- /dev/null
+++ b/lib/dll/example/fullbench-dll.sln
@@ -0,0 +1,25 @@
+Microsoft Visual Studio Solution File, Format Version 12.00
+# Visual Studio Express 2012 for Windows Desktop
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "fullbench-dll", "fullbench-dll.vcxproj", "{13992FD2-077E-4954-B065-A428198201A9}"
+EndProject
+Global
+	GlobalSection(SolutionConfigurationPlatforms) = preSolution
+		Debug|Win32 = Debug|Win32
+		Debug|x64 = Debug|x64
+		Release|Win32 = Release|Win32
+		Release|x64 = Release|x64
+	EndGlobalSection
+	GlobalSection(ProjectConfigurationPlatforms) = postSolution
+		{13992FD2-077E-4954-B065-A428198201A9}.Debug|Win32.ActiveCfg = Debug|Win32
+		{13992FD2-077E-4954-B065-A428198201A9}.Debug|Win32.Build.0 = Debug|Win32
+		{13992FD2-077E-4954-B065-A428198201A9}.Debug|x64.ActiveCfg = Debug|x64
+		{13992FD2-077E-4954-B065-A428198201A9}.Debug|x64.Build.0 = Debug|x64
+		{13992FD2-077E-4954-B065-A428198201A9}.Release|Win32.ActiveCfg = Release|Win32
+		{13992FD2-077E-4954-B065-A428198201A9}.Release|Win32.Build.0 = Release|Win32
+		{13992FD2-077E-4954-B065-A428198201A9}.Release|x64.ActiveCfg = Release|x64
+		{13992FD2-077E-4954-B065-A428198201A9}.Release|x64.Build.0 = Release|x64
+	EndGlobalSection
+	GlobalSection(SolutionProperties) = preSolution
+		HideSolutionNode = FALSE
+	EndGlobalSection
+EndGlobal
diff --git a/lib/dll/example/fullbench-dll.vcxproj b/lib/dll/example/fullbench-dll.vcxproj
new file mode 100644
index 0000000..4e67662
--- /dev/null
+++ b/lib/dll/example/fullbench-dll.vcxproj
@@ -0,0 +1,179 @@
+
+
+  
+    
+      Debug
+      Win32
+    
+    
+      Debug
+      x64
+    
+    
+      Release
+      Win32
+    
+    
+      Release
+      x64
+    
+  
+  
+    {00000000-1CC8-4FD7-9281-6B8DBB9D3DF8}
+    Win32Proj
+    fullbench-dll
+    $(SolutionDir)bin\$(Platform)_$(Configuration)\
+    $(SolutionDir)bin\obj\$(RootNamespace)_$(Platform)_$(Configuration)\
+  
+  
+  
+    Application
+    true
+    MultiByte
+  
+  
+    Application
+    true
+    MultiByte
+  
+  
+    Application
+    false
+    true
+    MultiByte
+  
+  
+    Application
+    false
+    true
+    MultiByte
+  
+  
+  
+  
+  
+    
+  
+  
+    
+  
+  
+    
+  
+  
+    
+  
+  
+  
+    true
+    $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath);
+    false
+  
+  
+    true
+    $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath);
+    false
+  
+  
+    false
+    $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath);
+    false
+  
+  
+    false
+    $(IncludePath);$(SolutionDir)..\..\lib;$(SolutionDir)..\..\programs;$(SolutionDir)..\..\lib\legacy;$(SolutionDir)..\..\lib\common;$(UniversalCRT_IncludePath);
+    false
+  
+  
+    
+      
+      
+      Level4
+      Disabled
+      WIN32;_DEBUG;_CONSOLE;ZSTD_DLL_IMPORT=1;%(PreprocessorDefinitions)
+      true
+      false
+      ..\include
+    
+    
+      Console
+      true
+      $(SolutionDir)..\dll;%(AdditionalLibraryDirectories)
+      libzstd.lib;%(AdditionalDependencies)
+    
+  
+  
+    
+      
+      
+      Level4
+      Disabled
+      WIN32;_DEBUG;_CONSOLE;ZSTD_DLL_IMPORT=1;%(PreprocessorDefinitions)
+      true
+      false
+      ..\include
+    
+    
+      Console
+      true
+      $(SolutionDir)..\dll;%(AdditionalLibraryDirectories)
+      libzstd.lib;%(AdditionalDependencies)
+    
+  
+  
+    
+      Level4
+      
+      
+      MaxSpeed
+      true
+      true
+      WIN32;_DEBUG;_CONSOLE;ZSTD_DLL_IMPORT=1;%(PreprocessorDefinitions)
+      false
+      ..\include
+      false
+      MultiThreaded
+    
+    
+      Console
+      true
+      true
+      true
+      $(SolutionDir)..\dll;%(AdditionalLibraryDirectories)
+      libzstd.lib;%(AdditionalDependencies)
+    
+  
+  
+    
+      Level4
+      
+      
+      MaxSpeed
+      true
+      true
+      WIN32;_DEBUG;_CONSOLE;ZSTD_DLL_IMPORT=1;%(PreprocessorDefinitions)
+      false
+      false
+      ..\include
+      MultiThreaded
+    
+    
+      Console
+      true
+      true
+      true
+      $(SolutionDir)..\dll;%(AdditionalLibraryDirectories)
+      libzstd.lib;%(AdditionalDependencies)
+    
+  
+  
+    
+    
+  
+  
+    
+  
+  
+  
+  
+
\ No newline at end of file
diff --git a/lib/dll/libzstd.def b/lib/dll/libzstd.def
new file mode 100644
index 0000000..0a3259e
--- /dev/null
+++ b/lib/dll/libzstd.def
@@ -0,0 +1,86 @@
+LIBRARY libzstd.dll
+EXPORTS
+    ZDICT_getDictID
+    ZDICT_getErrorName
+    ZDICT_isError
+    ZDICT_trainFromBuffer
+    ZSTD_CStreamInSize
+    ZSTD_CStreamOutSize
+    ZSTD_DStreamInSize
+    ZSTD_DStreamOutSize
+    ZSTD_adjustCParams
+    ZSTD_checkCParams
+    ZSTD_compress
+    ZSTD_compressBegin
+    ZSTD_compressBegin_advanced
+    ZSTD_compressBegin_usingDict
+    ZSTD_compressBlock
+    ZSTD_compressBound
+    ZSTD_compressCCtx
+    ZSTD_compressContinue
+    ZSTD_compressEnd
+    ZSTD_compressStream
+    ZSTD_compress_advanced
+    ZSTD_compress_usingCDict
+    ZSTD_compress_usingDict
+    ZSTD_copyCCtx
+    ZSTD_copyDCtx
+    ZSTD_createCCtx
+    ZSTD_createCCtx_advanced
+    ZSTD_createCDict
+    ZSTD_createCDict_advanced
+    ZSTD_createCStream
+    ZSTD_createCStream_advanced
+    ZSTD_createDCtx
+    ZSTD_createDCtx_advanced
+    ZSTD_createDDict
+    ZSTD_createDStream
+    ZSTD_createDStream_advanced
+    ZSTD_decompress
+    ZSTD_decompressBegin
+    ZSTD_decompressBegin_usingDict
+    ZSTD_decompressBlock
+    ZSTD_decompressContinue
+    ZSTD_decompressDCtx
+    ZSTD_decompressStream
+    ZSTD_decompress_usingDDict
+    ZSTD_decompress_usingDict
+    ZSTD_endStream
+    ZSTD_estimateCCtxSize
+    ZSTD_estimateDCtxSize
+    ZSTD_flushStream
+    ZSTD_freeCCtx
+    ZSTD_freeCDict
+    ZSTD_freeCStream
+    ZSTD_freeDCtx
+    ZSTD_freeDDict
+    ZSTD_freeDStream
+    ZSTD_getBlockSizeMax
+    ZSTD_getCParams
+    ZSTD_getDecompressedSize
+    ZSTD_getErrorName
+    ZSTD_getFrameParams
+    ZSTD_getParams
+    ZSTD_initCStream
+    ZSTD_initCStream_advanced
+    ZSTD_initCStream_usingCDict
+    ZSTD_initCStream_usingDict
+    ZSTD_initDStream
+    ZSTD_initDStream_usingDDict
+    ZSTD_initDStream_usingDict
+    ZSTD_insertBlock
+    ZSTD_isError
+    ZSTD_isFrame
+    ZSTD_maxCLevel
+    ZSTD_nextInputType
+    ZSTD_nextSrcSizeToDecompress
+    ZSTD_resetCStream
+    ZSTD_resetDStream
+    ZSTD_setDStreamParameter
+    ZSTD_sizeof_CCtx
+    ZSTD_sizeof_CDict
+    ZSTD_sizeof_CStream
+    ZSTD_sizeof_DCtx
+    ZSTD_sizeof_DDict
+    ZSTD_sizeof_DStream
+    ZSTD_versionNumber
diff --git a/lib/legacy/zstd_legacy.h b/lib/legacy/zstd_legacy.h
new file mode 100644
index 0000000..2a9f36a
--- /dev/null
+++ b/lib/legacy/zstd_legacy.h
@@ -0,0 +1,259 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+#ifndef ZSTD_LEGACY_H
+#define ZSTD_LEGACY_H
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/* *************************************
+*  Includes
+***************************************/
+#include "mem.h"            /* MEM_STATIC */
+#include "error_private.h"  /* ERROR */
+#include "zstd.h"           /* ZSTD_inBuffer, ZSTD_outBuffer */
+#include "zstd_v01.h"
+#include "zstd_v02.h"
+#include "zstd_v03.h"
+#include "zstd_v04.h"
+#include "zstd_v05.h"
+#include "zstd_v06.h"
+#include "zstd_v07.h"
+
+
+/** ZSTD_isLegacy() :
+    @return : > 0 if supported by legacy decoder. 0 otherwise.
+              return value is the version.
+*/
+MEM_STATIC unsigned ZSTD_isLegacy(const void* src, size_t srcSize)
+{
+    U32 magicNumberLE;
+    if (srcSize<4) return 0;
+    magicNumberLE = MEM_readLE32(src);
+    switch(magicNumberLE)
+    {
+        case ZSTDv01_magicNumberLE:return 1;
+        case ZSTDv02_magicNumber : return 2;
+        case ZSTDv03_magicNumber : return 3;
+        case ZSTDv04_magicNumber : return 4;
+        case ZSTDv05_MAGICNUMBER : return 5;
+        case ZSTDv06_MAGICNUMBER : return 6;
+        case ZSTDv07_MAGICNUMBER : return 7;
+        default : return 0;
+    }
+}
+
+
+MEM_STATIC unsigned long long ZSTD_getDecompressedSize_legacy(const void* src, size_t srcSize)
+{
+    U32 const version = ZSTD_isLegacy(src, srcSize);
+    if (version < 5) return 0;  /* no decompressed size in frame header, or not a legacy format */
+    if (version==5) {
+        ZSTDv05_parameters fParams;
+        size_t const frResult = ZSTDv05_getFrameParams(&fParams, src, srcSize);
+        if (frResult != 0) return 0;
+        return fParams.srcSize;
+    }
+    if (version==6) {
+        ZSTDv06_frameParams fParams;
+        size_t const frResult = ZSTDv06_getFrameParams(&fParams, src, srcSize);
+        if (frResult != 0) return 0;
+        return fParams.frameContentSize;
+    }
+    if (version==7) {
+        ZSTDv07_frameParams fParams;
+        size_t const frResult = ZSTDv07_getFrameParams(&fParams, src, srcSize);
+        if (frResult != 0) return 0;
+        return fParams.frameContentSize;
+    }
+    return 0;   /* should not be possible */
+}
+
+
+MEM_STATIC size_t ZSTD_decompressLegacy(
+                     void* dst, size_t dstCapacity,
+               const void* src, size_t compressedSize,
+               const void* dict,size_t dictSize)
+{
+    U32 const version = ZSTD_isLegacy(src, compressedSize);
+    switch(version)
+    {
+        case 1 :
+            return ZSTDv01_decompress(dst, dstCapacity, src, compressedSize);
+        case 2 :
+            return ZSTDv02_decompress(dst, dstCapacity, src, compressedSize);
+        case 3 :
+            return ZSTDv03_decompress(dst, dstCapacity, src, compressedSize);
+        case 4 :
+            return ZSTDv04_decompress(dst, dstCapacity, src, compressedSize);
+        case 5 :
+            {   size_t result;
+                ZSTDv05_DCtx* const zd = ZSTDv05_createDCtx();
+                if (zd==NULL) return ERROR(memory_allocation);
+                result = ZSTDv05_decompress_usingDict(zd, dst, dstCapacity, src, compressedSize, dict, dictSize);
+                ZSTDv05_freeDCtx(zd);
+                return result;
+            }
+        case 6 :
+            {   size_t result;
+                ZSTDv06_DCtx* const zd = ZSTDv06_createDCtx();
+                if (zd==NULL) return ERROR(memory_allocation);
+                result = ZSTDv06_decompress_usingDict(zd, dst, dstCapacity, src, compressedSize, dict, dictSize);
+                ZSTDv06_freeDCtx(zd);
+                return result;
+            }
+        case 7 :
+            {   size_t result;
+                ZSTDv07_DCtx* const zd = ZSTDv07_createDCtx();
+                if (zd==NULL) return ERROR(memory_allocation);
+                result = ZSTDv07_decompress_usingDict(zd, dst, dstCapacity, src, compressedSize, dict, dictSize);
+                ZSTDv07_freeDCtx(zd);
+                return result;
+            }
+        default :
+            return ERROR(prefix_unknown);
+    }
+}
+
+
+MEM_STATIC size_t ZSTD_freeLegacyStreamContext(void* legacyContext, U32 version)
+{
+    switch(version)
+    {
+        default :
+        case 1 :
+        case 2 :
+        case 3 :
+            return ERROR(version_unsupported);
+        case 4 : return ZBUFFv04_freeDCtx((ZBUFFv04_DCtx*)legacyContext);
+        case 5 : return ZBUFFv05_freeDCtx((ZBUFFv05_DCtx*)legacyContext);
+        case 6 : return ZBUFFv06_freeDCtx((ZBUFFv06_DCtx*)legacyContext);
+        case 7 : return ZBUFFv07_freeDCtx((ZBUFFv07_DCtx*)legacyContext);
+    }
+}
+
+
+MEM_STATIC size_t ZSTD_initLegacyStream(void** legacyContext, U32 prevVersion, U32 newVersion,
+                                        const void* dict, size_t dictSize)
+{
+    if (prevVersion != newVersion) ZSTD_freeLegacyStreamContext(*legacyContext, prevVersion);
+    switch(newVersion)
+    {
+        default :
+        case 1 :
+        case 2 :
+        case 3 :
+            return 0;
+        case 4 :
+        {
+            ZBUFFv04_DCtx* dctx = (prevVersion != newVersion) ? ZBUFFv04_createDCtx() : (ZBUFFv04_DCtx*)*legacyContext;
+            if (dctx==NULL) return ERROR(memory_allocation);
+            ZBUFFv04_decompressInit(dctx);
+            ZBUFFv04_decompressWithDictionary(dctx, dict, dictSize);
+            *legacyContext = dctx;
+            return 0;
+        }
+        case 5 :
+        {
+            ZBUFFv05_DCtx* dctx = (prevVersion != newVersion) ? ZBUFFv05_createDCtx() : (ZBUFFv05_DCtx*)*legacyContext;
+            if (dctx==NULL) return ERROR(memory_allocation);
+            ZBUFFv05_decompressInitDictionary(dctx, dict, dictSize);
+            *legacyContext = dctx;
+            return 0;
+        }
+        case 6 :
+        {
+            ZBUFFv06_DCtx* dctx = (prevVersion != newVersion) ? ZBUFFv06_createDCtx() : (ZBUFFv06_DCtx*)*legacyContext;
+            if (dctx==NULL) return ERROR(memory_allocation);
+            ZBUFFv06_decompressInitDictionary(dctx, dict, dictSize);
+            *legacyContext = dctx;
+            return 0;
+        }
+        case 7 :
+        {
+            ZBUFFv07_DCtx* dctx = (prevVersion != newVersion) ? ZBUFFv07_createDCtx() : (ZBUFFv07_DCtx*)*legacyContext;
+            if (dctx==NULL) return ERROR(memory_allocation);
+            ZBUFFv07_decompressInitDictionary(dctx, dict, dictSize);
+            *legacyContext = dctx;
+            return 0;
+        }
+    }
+}
+
+
+
+MEM_STATIC size_t ZSTD_decompressLegacyStream(void* legacyContext, U32 version,
+                                              ZSTD_outBuffer* output, ZSTD_inBuffer* input)
+{
+    switch(version)
+    {
+        default :
+        case 1 :
+        case 2 :
+        case 3 :
+            return ERROR(version_unsupported);
+        case 4 :
+            {
+                ZBUFFv04_DCtx* dctx = (ZBUFFv04_DCtx*) legacyContext;
+                const void* src = (const char*)input->src + input->pos;
+                size_t readSize = input->size - input->pos;
+                void* dst = (char*)output->dst + output->pos;
+                size_t decodedSize = output->size - output->pos;
+                size_t const hintSize = ZBUFFv04_decompressContinue(dctx, dst, &decodedSize, src, &readSize);
+                output->pos += decodedSize;
+                input->pos += readSize;
+                return hintSize;
+            }
+        case 5 :
+            {
+                ZBUFFv05_DCtx* dctx = (ZBUFFv05_DCtx*) legacyContext;
+                const void* src = (const char*)input->src + input->pos;
+                size_t readSize = input->size - input->pos;
+                void* dst = (char*)output->dst + output->pos;
+                size_t decodedSize = output->size - output->pos;
+                size_t const hintSize = ZBUFFv05_decompressContinue(dctx, dst, &decodedSize, src, &readSize);
+                output->pos += decodedSize;
+                input->pos += readSize;
+                return hintSize;
+            }
+        case 6 :
+            {
+                ZBUFFv06_DCtx* dctx = (ZBUFFv06_DCtx*) legacyContext;
+                const void* src = (const char*)input->src + input->pos;
+                size_t readSize = input->size - input->pos;
+                void* dst = (char*)output->dst + output->pos;
+                size_t decodedSize = output->size - output->pos;
+                size_t const hintSize = ZBUFFv06_decompressContinue(dctx, dst, &decodedSize, src, &readSize);
+                output->pos += decodedSize;
+                input->pos += readSize;
+                return hintSize;
+            }
+        case 7 :
+            {
+                ZBUFFv07_DCtx* dctx = (ZBUFFv07_DCtx*) legacyContext;
+                const void* src = (const char*)input->src + input->pos;
+                size_t readSize = input->size - input->pos;
+                void* dst = (char*)output->dst + output->pos;
+                size_t decodedSize = output->size - output->pos;
+                size_t const hintSize = ZBUFFv07_decompressContinue(dctx, dst, &decodedSize, src, &readSize);
+                output->pos += decodedSize;
+                input->pos += readSize;
+                return hintSize;
+            }
+    }
+}
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif   /* ZSTD_LEGACY_H */
diff --git a/lib/legacy/zstd_v01.c b/lib/legacy/zstd_v01.c
new file mode 100644
index 0000000..6fd30c0
--- /dev/null
+++ b/lib/legacy/zstd_v01.c
@@ -0,0 +1,2095 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+
+/******************************************
+*  Includes
+******************************************/
+#include     /* size_t, ptrdiff_t */
+#include "zstd_v01.h"
+#include "error_private.h"
+
+
+/******************************************
+*  Static allocation
+******************************************/
+/* You can statically allocate FSE CTable/DTable as a table of unsigned using below macro */
+#define FSE_DTABLE_SIZE_U32(maxTableLog)                   (1 + (1<2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; etc.)
+*  Increasing memory usage improves compression ratio
+*  Reduced memory usage can improve speed, due to cache effect
+*  Recommended max value is 14, for 16KB, which nicely fits into Intel x86 L1 cache */
+#define FSE_MAX_MEMORY_USAGE 14
+#define FSE_DEFAULT_MEMORY_USAGE 13
+
+/* FSE_MAX_SYMBOL_VALUE :
+*  Maximum symbol value authorized.
+*  Required for proper stack allocation */
+#define FSE_MAX_SYMBOL_VALUE 255
+
+
+/****************************************************************
+*  template functions type & suffix
+****************************************************************/
+#define FSE_FUNCTION_TYPE BYTE
+#define FSE_FUNCTION_EXTENSION
+
+
+/****************************************************************
+*  Byte symbol type
+****************************************************************/
+typedef struct
+{
+    unsigned short newState;
+    unsigned char  symbol;
+    unsigned char  nbBits;
+} FSE_decode_t;   /* size == U32 */
+
+
+
+/****************************************************************
+*  Compiler specifics
+****************************************************************/
+#ifdef _MSC_VER    /* Visual Studio */
+#  define FORCE_INLINE static __forceinline
+#  include                     /* For Visual 2005 */
+#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
+#  pragma warning(disable : 4214)        /* disable: C4214: non-int bitfields */
+#else
+#  define GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
+#  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+#    ifdef __GNUC__
+#      define FORCE_INLINE static inline __attribute__((always_inline))
+#    else
+#      define FORCE_INLINE static inline
+#    endif
+#  else
+#    define FORCE_INLINE static
+#  endif /* __STDC_VERSION__ */
+#endif
+
+
+/****************************************************************
+*  Includes
+****************************************************************/
+#include      /* malloc, free, qsort */
+#include      /* memcpy, memset */
+#include       /* printf (debug) */
+
+
+#ifndef MEM_ACCESS_MODULE
+#define MEM_ACCESS_MODULE
+/****************************************************************
+*  Basic Types
+*****************************************************************/
+#if defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+# include 
+typedef  uint8_t BYTE;
+typedef uint16_t U16;
+typedef  int16_t S16;
+typedef uint32_t U32;
+typedef  int32_t S32;
+typedef uint64_t U64;
+typedef  int64_t S64;
+#else
+typedef unsigned char       BYTE;
+typedef unsigned short      U16;
+typedef   signed short      S16;
+typedef unsigned int        U32;
+typedef   signed int        S32;
+typedef unsigned long long  U64;
+typedef   signed long long  S64;
+#endif
+
+#endif   /* MEM_ACCESS_MODULE */
+
+/****************************************************************
+*  Memory I/O
+*****************************************************************/
+/* FSE_FORCE_MEMORY_ACCESS
+ * By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
+ * Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
+ * The below switch allow to select different access method for improved performance.
+ * Method 0 (default) : use `memcpy()`. Safe and portable.
+ * Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
+ *            This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
+ * Method 2 : direct access. This method is portable but violate C standard.
+ *            It can generate buggy code on targets generating assembly depending on alignment.
+ *            But in some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
+ * See http://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
+ * Prefer these methods in priority order (0 > 1 > 2)
+ */
+#ifndef FSE_FORCE_MEMORY_ACCESS   /* can be defined externally, on command line for example */
+#  if defined(__GNUC__) && ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) )
+#    define FSE_FORCE_MEMORY_ACCESS 2
+#  elif (defined(__INTEL_COMPILER) && !defined(WIN32)) || \
+  (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7S__) ))
+#    define FSE_FORCE_MEMORY_ACCESS 1
+#  endif
+#endif
+
+
+static unsigned FSE_32bits(void)
+{
+    return sizeof(void*)==4;
+}
+
+static unsigned FSE_isLittleEndian(void)
+{
+    const union { U32 i; BYTE c[4]; } one = { 1 };   /* don't use static : performance detrimental  */
+    return one.c[0];
+}
+
+#if defined(FSE_FORCE_MEMORY_ACCESS) && (FSE_FORCE_MEMORY_ACCESS==2)
+
+static U16 FSE_read16(const void* memPtr) { return *(const U16*) memPtr; }
+static U32 FSE_read32(const void* memPtr) { return *(const U32*) memPtr; }
+static U64 FSE_read64(const void* memPtr) { return *(const U64*) memPtr; }
+
+#elif defined(FSE_FORCE_MEMORY_ACCESS) && (FSE_FORCE_MEMORY_ACCESS==1)
+
+/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
+/* currently only defined for gcc and icc */
+typedef union { U16 u16; U32 u32; U64 u64; } __attribute__((packed)) unalign;
+
+static U16 FSE_read16(const void* ptr) { return ((const unalign*)ptr)->u16; }
+static U32 FSE_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
+static U64 FSE_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
+
+#else
+
+static U16 FSE_read16(const void* memPtr)
+{
+    U16 val; memcpy(&val, memPtr, sizeof(val)); return val;
+}
+
+static U32 FSE_read32(const void* memPtr)
+{
+    U32 val; memcpy(&val, memPtr, sizeof(val)); return val;
+}
+
+static U64 FSE_read64(const void* memPtr)
+{
+    U64 val; memcpy(&val, memPtr, sizeof(val)); return val;
+}
+
+#endif // FSE_FORCE_MEMORY_ACCESS
+
+static U16 FSE_readLE16(const void* memPtr)
+{
+    if (FSE_isLittleEndian())
+        return FSE_read16(memPtr);
+    else
+    {
+        const BYTE* p = (const BYTE*)memPtr;
+        return (U16)(p[0] + (p[1]<<8));
+    }
+}
+
+static U32 FSE_readLE32(const void* memPtr)
+{
+    if (FSE_isLittleEndian())
+        return FSE_read32(memPtr);
+    else
+    {
+        const BYTE* p = (const BYTE*)memPtr;
+        return (U32)((U32)p[0] + ((U32)p[1]<<8) + ((U32)p[2]<<16) + ((U32)p[3]<<24));
+    }
+}
+
+
+static U64 FSE_readLE64(const void* memPtr)
+{
+    if (FSE_isLittleEndian())
+        return FSE_read64(memPtr);
+    else
+    {
+        const BYTE* p = (const BYTE*)memPtr;
+        return (U64)((U64)p[0] + ((U64)p[1]<<8) + ((U64)p[2]<<16) + ((U64)p[3]<<24)
+                     + ((U64)p[4]<<32) + ((U64)p[5]<<40) + ((U64)p[6]<<48) + ((U64)p[7]<<56));
+    }
+}
+
+static size_t FSE_readLEST(const void* memPtr)
+{
+    if (FSE_32bits())
+        return (size_t)FSE_readLE32(memPtr);
+    else
+        return (size_t)FSE_readLE64(memPtr);
+}
+
+
+
+/****************************************************************
+*  Constants
+*****************************************************************/
+#define FSE_MAX_TABLELOG  (FSE_MAX_MEMORY_USAGE-2)
+#define FSE_MAX_TABLESIZE (1U< FSE_TABLELOG_ABSOLUTE_MAX
+#error "FSE_MAX_TABLELOG > FSE_TABLELOG_ABSOLUTE_MAX is not supported"
+#endif
+
+
+/****************************************************************
+*  Error Management
+****************************************************************/
+#define FSE_STATIC_ASSERT(c) { enum { FSE_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
+
+
+/****************************************************************
+*  Complex types
+****************************************************************/
+typedef struct
+{
+    int deltaFindState;
+    U32 deltaNbBits;
+} FSE_symbolCompressionTransform; /* total 8 bytes */
+
+typedef U32 DTable_max_t[FSE_DTABLE_SIZE_U32(FSE_MAX_TABLELOG)];
+
+/****************************************************************
+*  Internal functions
+****************************************************************/
+FORCE_INLINE unsigned FSE_highbit32 (register U32 val)
+{
+#   if defined(_MSC_VER)   /* Visual */
+    unsigned long r;
+    _BitScanReverse ( &r, val );
+    return (unsigned) r;
+#   elif defined(__GNUC__) && (GCC_VERSION >= 304)   /* GCC Intrinsic */
+    return 31 - __builtin_clz (val);
+#   else   /* Software version */
+    static const unsigned DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 };
+    U32 v = val;
+    unsigned r;
+    v |= v >> 1;
+    v |= v >> 2;
+    v |= v >> 4;
+    v |= v >> 8;
+    v |= v >> 16;
+    r = DeBruijnClz[ (U32) (v * 0x07C4ACDDU) >> 27];
+    return r;
+#   endif
+}
+
+
+/****************************************************************
+*  Templates
+****************************************************************/
+/*
+  designed to be included
+  for type-specific functions (template emulation in C)
+  Objective is to write these functions only once, for improved maintenance
+*/
+
+/* safety checks */
+#ifndef FSE_FUNCTION_EXTENSION
+#  error "FSE_FUNCTION_EXTENSION must be defined"
+#endif
+#ifndef FSE_FUNCTION_TYPE
+#  error "FSE_FUNCTION_TYPE must be defined"
+#endif
+
+/* Function names */
+#define FSE_CAT(X,Y) X##Y
+#define FSE_FUNCTION_NAME(X,Y) FSE_CAT(X,Y)
+#define FSE_TYPE_NAME(X,Y) FSE_CAT(X,Y)
+
+
+
+static U32 FSE_tableStep(U32 tableSize) { return (tableSize>>1) + (tableSize>>3) + 3; }
+
+#define FSE_DECODE_TYPE FSE_decode_t
+
+
+typedef struct {
+    U16 tableLog;
+    U16 fastMode;
+} FSE_DTableHeader;   /* sizeof U32 */
+
+static size_t FSE_buildDTable
+(FSE_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
+{
+    void* ptr = dt;
+    FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)ptr;
+    FSE_DECODE_TYPE* const tableDecode = (FSE_DECODE_TYPE*)(ptr) + 1;   /* because dt is unsigned, 32-bits aligned on 32-bits */
+    const U32 tableSize = 1 << tableLog;
+    const U32 tableMask = tableSize-1;
+    const U32 step = FSE_tableStep(tableSize);
+    U16 symbolNext[FSE_MAX_SYMBOL_VALUE+1];
+    U32 position = 0;
+    U32 highThreshold = tableSize-1;
+    const S16 largeLimit= (S16)(1 << (tableLog-1));
+    U32 noLarge = 1;
+    U32 s;
+
+    /* Sanity Checks */
+    if (maxSymbolValue > FSE_MAX_SYMBOL_VALUE) return (size_t)-FSE_ERROR_maxSymbolValue_tooLarge;
+    if (tableLog > FSE_MAX_TABLELOG) return (size_t)-FSE_ERROR_tableLog_tooLarge;
+
+    /* Init, lay down lowprob symbols */
+    DTableH[0].tableLog = (U16)tableLog;
+    for (s=0; s<=maxSymbolValue; s++)
+    {
+        if (normalizedCounter[s]==-1)
+        {
+            tableDecode[highThreshold--].symbol = (FSE_FUNCTION_TYPE)s;
+            symbolNext[s] = 1;
+        }
+        else
+        {
+            if (normalizedCounter[s] >= largeLimit) noLarge=0;
+            symbolNext[s] = normalizedCounter[s];
+        }
+    }
+
+    /* Spread symbols */
+    for (s=0; s<=maxSymbolValue; s++)
+    {
+        int i;
+        for (i=0; i highThreshold) position = (position + step) & tableMask;   /* lowprob area */
+        }
+    }
+
+    if (position!=0) return (size_t)-FSE_ERROR_GENERIC;   /* position must reach all cells once, otherwise normalizedCounter is incorrect */
+
+    /* Build Decoding table */
+    {
+        U32 i;
+        for (i=0; ifastMode = (U16)noLarge;
+    return 0;
+}
+
+
+/******************************************
+*  FSE byte symbol
+******************************************/
+#ifndef FSE_COMMONDEFS_ONLY
+
+static unsigned FSE_isError(size_t code) { return (code > (size_t)(-FSE_ERROR_maxCode)); }
+
+static short FSE_abs(short a)
+{
+    return a<0? -a : a;
+}
+
+
+/****************************************************************
+*  Header bitstream management
+****************************************************************/
+static size_t FSE_readNCount (short* normalizedCounter, unsigned* maxSVPtr, unsigned* tableLogPtr,
+                 const void* headerBuffer, size_t hbSize)
+{
+    const BYTE* const istart = (const BYTE*) headerBuffer;
+    const BYTE* const iend = istart + hbSize;
+    const BYTE* ip = istart;
+    int nbBits;
+    int remaining;
+    int threshold;
+    U32 bitStream;
+    int bitCount;
+    unsigned charnum = 0;
+    int previous0 = 0;
+
+    if (hbSize < 4) return (size_t)-FSE_ERROR_srcSize_wrong;
+    bitStream = FSE_readLE32(ip);
+    nbBits = (bitStream & 0xF) + FSE_MIN_TABLELOG;   /* extract tableLog */
+    if (nbBits > FSE_TABLELOG_ABSOLUTE_MAX) return (size_t)-FSE_ERROR_tableLog_tooLarge;
+    bitStream >>= 4;
+    bitCount = 4;
+    *tableLogPtr = nbBits;
+    remaining = (1<1) && (charnum<=*maxSVPtr))
+    {
+        if (previous0)
+        {
+            unsigned n0 = charnum;
+            while ((bitStream & 0xFFFF) == 0xFFFF)
+            {
+                n0+=24;
+                if (ip < iend-5)
+                {
+                    ip+=2;
+                    bitStream = FSE_readLE32(ip) >> bitCount;
+                }
+                else
+                {
+                    bitStream >>= 16;
+                    bitCount+=16;
+                }
+            }
+            while ((bitStream & 3) == 3)
+            {
+                n0+=3;
+                bitStream>>=2;
+                bitCount+=2;
+            }
+            n0 += bitStream & 3;
+            bitCount += 2;
+            if (n0 > *maxSVPtr) return (size_t)-FSE_ERROR_maxSymbolValue_tooSmall;
+            while (charnum < n0) normalizedCounter[charnum++] = 0;
+            if ((ip <= iend-7) || (ip + (bitCount>>3) <= iend-4))
+            {
+                ip += bitCount>>3;
+                bitCount &= 7;
+                bitStream = FSE_readLE32(ip) >> bitCount;
+            }
+            else
+                bitStream >>= 2;
+        }
+        {
+            const short max = (short)((2*threshold-1)-remaining);
+            short count;
+
+            if ((bitStream & (threshold-1)) < (U32)max)
+            {
+                count = (short)(bitStream & (threshold-1));
+                bitCount   += nbBits-1;
+            }
+            else
+            {
+                count = (short)(bitStream & (2*threshold-1));
+                if (count >= threshold) count -= max;
+                bitCount   += nbBits;
+            }
+
+            count--;   /* extra accuracy */
+            remaining -= FSE_abs(count);
+            normalizedCounter[charnum++] = count;
+            previous0 = !count;
+            while (remaining < threshold)
+            {
+                nbBits--;
+                threshold >>= 1;
+            }
+
+            {
+                if ((ip <= iend-7) || (ip + (bitCount>>3) <= iend-4))
+                {
+                    ip += bitCount>>3;
+                    bitCount &= 7;
+                }
+                else
+                {
+                    bitCount -= (int)(8 * (iend - 4 - ip));
+                    ip = iend - 4;
+                }
+                bitStream = FSE_readLE32(ip) >> (bitCount & 31);
+            }
+        }
+    }
+    if (remaining != 1) return (size_t)-FSE_ERROR_GENERIC;
+    *maxSVPtr = charnum-1;
+
+    ip += (bitCount+7)>>3;
+    if ((size_t)(ip-istart) > hbSize) return (size_t)-FSE_ERROR_srcSize_wrong;
+    return ip-istart;
+}
+
+
+/*********************************************************
+*  Decompression (Byte symbols)
+*********************************************************/
+static size_t FSE_buildDTable_rle (FSE_DTable* dt, BYTE symbolValue)
+{
+    void* ptr = dt;
+    FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)ptr;
+    FSE_decode_t* const cell = (FSE_decode_t*)(ptr) + 1;   /* because dt is unsigned */
+
+    DTableH->tableLog = 0;
+    DTableH->fastMode = 0;
+
+    cell->newState = 0;
+    cell->symbol = symbolValue;
+    cell->nbBits = 0;
+
+    return 0;
+}
+
+
+static size_t FSE_buildDTable_raw (FSE_DTable* dt, unsigned nbBits)
+{
+    void* ptr = dt;
+    FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)ptr;
+    FSE_decode_t* const dinfo = (FSE_decode_t*)(ptr) + 1;   /* because dt is unsigned */
+    const unsigned tableSize = 1 << nbBits;
+    const unsigned tableMask = tableSize - 1;
+    const unsigned maxSymbolValue = tableMask;
+    unsigned s;
+
+    /* Sanity checks */
+    if (nbBits < 1) return (size_t)-FSE_ERROR_GENERIC;             /* min size */
+
+    /* Build Decoding Table */
+    DTableH->tableLog = (U16)nbBits;
+    DTableH->fastMode = 1;
+    for (s=0; s<=maxSymbolValue; s++)
+    {
+        dinfo[s].newState = 0;
+        dinfo[s].symbol = (BYTE)s;
+        dinfo[s].nbBits = (BYTE)nbBits;
+    }
+
+    return 0;
+}
+
+
+/* FSE_initDStream
+ * Initialize a FSE_DStream_t.
+ * srcBuffer must point at the beginning of an FSE block.
+ * The function result is the size of the FSE_block (== srcSize).
+ * If srcSize is too small, the function will return an errorCode;
+ */
+static size_t FSE_initDStream(FSE_DStream_t* bitD, const void* srcBuffer, size_t srcSize)
+{
+    if (srcSize < 1) return (size_t)-FSE_ERROR_srcSize_wrong;
+
+    if (srcSize >=  sizeof(size_t))
+    {
+        U32 contain32;
+        bitD->start = (const char*)srcBuffer;
+        bitD->ptr   = (const char*)srcBuffer + srcSize - sizeof(size_t);
+        bitD->bitContainer = FSE_readLEST(bitD->ptr);
+        contain32 = ((const BYTE*)srcBuffer)[srcSize-1];
+        if (contain32 == 0) return (size_t)-FSE_ERROR_GENERIC;   /* stop bit not present */
+        bitD->bitsConsumed = 8 - FSE_highbit32(contain32);
+    }
+    else
+    {
+        U32 contain32;
+        bitD->start = (const char*)srcBuffer;
+        bitD->ptr   = bitD->start;
+        bitD->bitContainer = *(const BYTE*)(bitD->start);
+        switch(srcSize)
+        {
+            case 7: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[6]) << (sizeof(size_t)*8 - 16);
+            case 6: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[5]) << (sizeof(size_t)*8 - 24);
+            case 5: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[4]) << (sizeof(size_t)*8 - 32);
+            case 4: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[3]) << 24;
+            case 3: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[2]) << 16;
+            case 2: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[1]) <<  8;
+            default:;
+        }
+        contain32 = ((const BYTE*)srcBuffer)[srcSize-1];
+        if (contain32 == 0) return (size_t)-FSE_ERROR_GENERIC;   /* stop bit not present */
+        bitD->bitsConsumed = 8 - FSE_highbit32(contain32);
+        bitD->bitsConsumed += (U32)(sizeof(size_t) - srcSize)*8;
+    }
+
+    return srcSize;
+}
+
+
+/*!FSE_lookBits
+ * Provides next n bits from the bitContainer.
+ * bitContainer is not modified (bits are still present for next read/look)
+ * On 32-bits, maxNbBits==25
+ * On 64-bits, maxNbBits==57
+ * return : value extracted.
+ */
+static size_t FSE_lookBits(FSE_DStream_t* bitD, U32 nbBits)
+{
+    const U32 bitMask = sizeof(bitD->bitContainer)*8 - 1;
+    return ((bitD->bitContainer << (bitD->bitsConsumed & bitMask)) >> 1) >> ((bitMask-nbBits) & bitMask);
+}
+
+static size_t FSE_lookBitsFast(FSE_DStream_t* bitD, U32 nbBits)   /* only if nbBits >= 1 !! */
+{
+    const U32 bitMask = sizeof(bitD->bitContainer)*8 - 1;
+    return (bitD->bitContainer << (bitD->bitsConsumed & bitMask)) >> (((bitMask+1)-nbBits) & bitMask);
+}
+
+static void FSE_skipBits(FSE_DStream_t* bitD, U32 nbBits)
+{
+    bitD->bitsConsumed += nbBits;
+}
+
+
+/*!FSE_readBits
+ * Read next n bits from the bitContainer.
+ * On 32-bits, don't read more than maxNbBits==25
+ * On 64-bits, don't read more than maxNbBits==57
+ * Use the fast variant *only* if n >= 1.
+ * return : value extracted.
+ */
+static size_t FSE_readBits(FSE_DStream_t* bitD, U32 nbBits)
+{
+    size_t value = FSE_lookBits(bitD, nbBits);
+    FSE_skipBits(bitD, nbBits);
+    return value;
+}
+
+static size_t FSE_readBitsFast(FSE_DStream_t* bitD, U32 nbBits)   /* only if nbBits >= 1 !! */
+{
+    size_t value = FSE_lookBitsFast(bitD, nbBits);
+    FSE_skipBits(bitD, nbBits);
+    return value;
+}
+
+static unsigned FSE_reloadDStream(FSE_DStream_t* bitD)
+{
+    if (bitD->bitsConsumed > (sizeof(bitD->bitContainer)*8))  /* should never happen */
+        return FSE_DStream_tooFar;
+
+    if (bitD->ptr >= bitD->start + sizeof(bitD->bitContainer))
+    {
+        bitD->ptr -= bitD->bitsConsumed >> 3;
+        bitD->bitsConsumed &= 7;
+        bitD->bitContainer = FSE_readLEST(bitD->ptr);
+        return FSE_DStream_unfinished;
+    }
+    if (bitD->ptr == bitD->start)
+    {
+        if (bitD->bitsConsumed < sizeof(bitD->bitContainer)*8) return FSE_DStream_endOfBuffer;
+        return FSE_DStream_completed;
+    }
+    {
+        U32 nbBytes = bitD->bitsConsumed >> 3;
+        U32 result = FSE_DStream_unfinished;
+        if (bitD->ptr - nbBytes < bitD->start)
+        {
+            nbBytes = (U32)(bitD->ptr - bitD->start);  /* ptr > start */
+            result = FSE_DStream_endOfBuffer;
+        }
+        bitD->ptr -= nbBytes;
+        bitD->bitsConsumed -= nbBytes*8;
+        bitD->bitContainer = FSE_readLEST(bitD->ptr);   /* reminder : srcSize > sizeof(bitD) */
+        return result;
+    }
+}
+
+
+static void FSE_initDState(FSE_DState_t* DStatePtr, FSE_DStream_t* bitD, const FSE_DTable* dt)
+{
+    const void* ptr = dt;
+    const FSE_DTableHeader* const DTableH = (const FSE_DTableHeader*)ptr;
+    DStatePtr->state = FSE_readBits(bitD, DTableH->tableLog);
+    FSE_reloadDStream(bitD);
+    DStatePtr->table = dt + 1;
+}
+
+static BYTE FSE_decodeSymbol(FSE_DState_t* DStatePtr, FSE_DStream_t* bitD)
+{
+    const FSE_decode_t DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
+    const U32  nbBits = DInfo.nbBits;
+    BYTE symbol = DInfo.symbol;
+    size_t lowBits = FSE_readBits(bitD, nbBits);
+
+    DStatePtr->state = DInfo.newState + lowBits;
+    return symbol;
+}
+
+static BYTE FSE_decodeSymbolFast(FSE_DState_t* DStatePtr, FSE_DStream_t* bitD)
+{
+    const FSE_decode_t DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
+    const U32 nbBits = DInfo.nbBits;
+    BYTE symbol = DInfo.symbol;
+    size_t lowBits = FSE_readBitsFast(bitD, nbBits);
+
+    DStatePtr->state = DInfo.newState + lowBits;
+    return symbol;
+}
+
+/* FSE_endOfDStream
+   Tells if bitD has reached end of bitStream or not */
+
+static unsigned FSE_endOfDStream(const FSE_DStream_t* bitD)
+{
+    return ((bitD->ptr == bitD->start) && (bitD->bitsConsumed == sizeof(bitD->bitContainer)*8));
+}
+
+static unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr)
+{
+    return DStatePtr->state == 0;
+}
+
+
+FORCE_INLINE size_t FSE_decompress_usingDTable_generic(
+          void* dst, size_t maxDstSize,
+    const void* cSrc, size_t cSrcSize,
+    const FSE_DTable* dt, const unsigned fast)
+{
+    BYTE* const ostart = (BYTE*) dst;
+    BYTE* op = ostart;
+    BYTE* const omax = op + maxDstSize;
+    BYTE* const olimit = omax-3;
+
+    FSE_DStream_t bitD;
+    FSE_DState_t state1;
+    FSE_DState_t state2;
+    size_t errorCode;
+
+    /* Init */
+    errorCode = FSE_initDStream(&bitD, cSrc, cSrcSize);   /* replaced last arg by maxCompressed Size */
+    if (FSE_isError(errorCode)) return errorCode;
+
+    FSE_initDState(&state1, &bitD, dt);
+    FSE_initDState(&state2, &bitD, dt);
+
+#define FSE_GETSYMBOL(statePtr) fast ? FSE_decodeSymbolFast(statePtr, &bitD) : FSE_decodeSymbol(statePtr, &bitD)
+
+    /* 4 symbols per loop */
+    for ( ; (FSE_reloadDStream(&bitD)==FSE_DStream_unfinished) && (op sizeof(bitD.bitContainer)*8)    /* This test must be static */
+            FSE_reloadDStream(&bitD);
+
+        op[1] = FSE_GETSYMBOL(&state2);
+
+        if (FSE_MAX_TABLELOG*4+7 > sizeof(bitD.bitContainer)*8)    /* This test must be static */
+            { if (FSE_reloadDStream(&bitD) > FSE_DStream_unfinished) { op+=2; break; } }
+
+        op[2] = FSE_GETSYMBOL(&state1);
+
+        if (FSE_MAX_TABLELOG*2+7 > sizeof(bitD.bitContainer)*8)    /* This test must be static */
+            FSE_reloadDStream(&bitD);
+
+        op[3] = FSE_GETSYMBOL(&state2);
+    }
+
+    /* tail */
+    /* note : FSE_reloadDStream(&bitD) >= FSE_DStream_partiallyFilled; Ends at exactly FSE_DStream_completed */
+    while (1)
+    {
+        if ( (FSE_reloadDStream(&bitD)>FSE_DStream_completed) || (op==omax) || (FSE_endOfDStream(&bitD) && (fast || FSE_endOfDState(&state1))) )
+            break;
+
+        *op++ = FSE_GETSYMBOL(&state1);
+
+        if ( (FSE_reloadDStream(&bitD)>FSE_DStream_completed) || (op==omax) || (FSE_endOfDStream(&bitD) && (fast || FSE_endOfDState(&state2))) )
+            break;
+
+        *op++ = FSE_GETSYMBOL(&state2);
+    }
+
+    /* end ? */
+    if (FSE_endOfDStream(&bitD) && FSE_endOfDState(&state1) && FSE_endOfDState(&state2))
+        return op-ostart;
+
+    if (op==omax) return (size_t)-FSE_ERROR_dstSize_tooSmall;   /* dst buffer is full, but cSrc unfinished */
+
+    return (size_t)-FSE_ERROR_corruptionDetected;
+}
+
+
+static size_t FSE_decompress_usingDTable(void* dst, size_t originalSize,
+                            const void* cSrc, size_t cSrcSize,
+                            const FSE_DTable* dt)
+{
+    FSE_DTableHeader DTableH;
+    memcpy(&DTableH, dt, sizeof(DTableH));   /* memcpy() into local variable, to avoid strict aliasing warning */
+
+    /* select fast mode (static) */
+    if (DTableH.fastMode) return FSE_decompress_usingDTable_generic(dst, originalSize, cSrc, cSrcSize, dt, 1);
+    return FSE_decompress_usingDTable_generic(dst, originalSize, cSrc, cSrcSize, dt, 0);
+}
+
+
+static size_t FSE_decompress(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize)
+{
+    const BYTE* const istart = (const BYTE*)cSrc;
+    const BYTE* ip = istart;
+    short counting[FSE_MAX_SYMBOL_VALUE+1];
+    DTable_max_t dt;   /* Static analyzer seems unable to understand this table will be properly initialized later */
+    unsigned tableLog;
+    unsigned maxSymbolValue = FSE_MAX_SYMBOL_VALUE;
+    size_t errorCode;
+
+    if (cSrcSize<2) return (size_t)-FSE_ERROR_srcSize_wrong;   /* too small input size */
+
+    /* normal FSE decoding mode */
+    errorCode = FSE_readNCount (counting, &maxSymbolValue, &tableLog, istart, cSrcSize);
+    if (FSE_isError(errorCode)) return errorCode;
+    if (errorCode >= cSrcSize) return (size_t)-FSE_ERROR_srcSize_wrong;   /* too small input size */
+    ip += errorCode;
+    cSrcSize -= errorCode;
+
+    errorCode = FSE_buildDTable (dt, counting, maxSymbolValue, tableLog);
+    if (FSE_isError(errorCode)) return errorCode;
+
+    /* always return, even if it is an error code */
+    return FSE_decompress_usingDTable (dst, maxDstSize, ip, cSrcSize, dt);
+}
+
+
+
+/* *******************************************************
+*  Huff0 : Huffman block compression
+*********************************************************/
+#define HUF_MAX_SYMBOL_VALUE 255
+#define HUF_DEFAULT_TABLELOG  12       /* used by default, when not specified */
+#define HUF_MAX_TABLELOG  12           /* max possible tableLog; for allocation purpose; can be modified */
+#define HUF_ABSOLUTEMAX_TABLELOG  16   /* absolute limit of HUF_MAX_TABLELOG. Beyond that value, code does not work */
+#if (HUF_MAX_TABLELOG > HUF_ABSOLUTEMAX_TABLELOG)
+#  error "HUF_MAX_TABLELOG is too large !"
+#endif
+
+typedef struct HUF_CElt_s {
+  U16  val;
+  BYTE nbBits;
+} HUF_CElt ;
+
+typedef struct nodeElt_s {
+    U32 count;
+    U16 parent;
+    BYTE byte;
+    BYTE nbBits;
+} nodeElt;
+
+
+/* *******************************************************
+*  Huff0 : Huffman block decompression
+*********************************************************/
+typedef struct {
+    BYTE byte;
+    BYTE nbBits;
+} HUF_DElt;
+
+static size_t HUF_readDTable (U16* DTable, const void* src, size_t srcSize)
+{
+    BYTE huffWeight[HUF_MAX_SYMBOL_VALUE + 1];
+    U32 rankVal[HUF_ABSOLUTEMAX_TABLELOG + 1];  /* large enough for values from 0 to 16 */
+    U32 weightTotal;
+    U32 maxBits;
+    const BYTE* ip = (const BYTE*) src;
+    size_t iSize;
+    size_t oSize;
+    U32 n;
+    U32 nextRankStart;
+    void* ptr = DTable+1;
+    HUF_DElt* const dt = (HUF_DElt*)ptr;
+
+    if (!srcSize) return (size_t)-FSE_ERROR_srcSize_wrong;
+    iSize = ip[0];
+
+    FSE_STATIC_ASSERT(sizeof(HUF_DElt) == sizeof(U16));   /* if compilation fails here, assertion is false */
+    //memset(huffWeight, 0, sizeof(huffWeight));   /* should not be necessary, but some analyzer complain ... */
+    if (iSize >= 128)  /* special header */
+    {
+        if (iSize >= (242))   /* RLE */
+        {
+            static int l[14] = { 1, 2, 3, 4, 7, 8, 15, 16, 31, 32, 63, 64, 127, 128 };
+            oSize = l[iSize-242];
+            memset(huffWeight, 1, sizeof(huffWeight));
+            iSize = 0;
+        }
+        else   /* Incompressible */
+        {
+            oSize = iSize - 127;
+            iSize = ((oSize+1)/2);
+            if (iSize+1 > srcSize) return (size_t)-FSE_ERROR_srcSize_wrong;
+            ip += 1;
+            for (n=0; n> 4;
+                huffWeight[n+1] = ip[n/2] & 15;
+            }
+        }
+    }
+    else  /* header compressed with FSE (normal case) */
+    {
+        if (iSize+1 > srcSize) return (size_t)-FSE_ERROR_srcSize_wrong;
+        oSize = FSE_decompress(huffWeight, HUF_MAX_SYMBOL_VALUE, ip+1, iSize);   /* max 255 values decoded, last one is implied */
+        if (FSE_isError(oSize)) return oSize;
+    }
+
+    /* collect weight stats */
+    memset(rankVal, 0, sizeof(rankVal));
+    weightTotal = 0;
+    for (n=0; n= HUF_ABSOLUTEMAX_TABLELOG) return (size_t)-FSE_ERROR_corruptionDetected;
+        rankVal[huffWeight[n]]++;
+        weightTotal += (1 << huffWeight[n]) >> 1;
+    }
+    if (weightTotal == 0) return (size_t)-FSE_ERROR_corruptionDetected;
+
+    /* get last non-null symbol weight (implied, total must be 2^n) */
+    maxBits = FSE_highbit32(weightTotal) + 1;
+    if (maxBits > DTable[0]) return (size_t)-FSE_ERROR_tableLog_tooLarge;   /* DTable is too small */
+    DTable[0] = (U16)maxBits;
+    {
+        U32 total = 1 << maxBits;
+        U32 rest = total - weightTotal;
+        U32 verif = 1 << FSE_highbit32(rest);
+        U32 lastWeight = FSE_highbit32(rest) + 1;
+        if (verif != rest) return (size_t)-FSE_ERROR_corruptionDetected;    /* last value must be a clean power of 2 */
+        huffWeight[oSize] = (BYTE)lastWeight;
+        rankVal[lastWeight]++;
+    }
+
+    /* check tree construction validity */
+    if ((rankVal[1] < 2) || (rankVal[1] & 1)) return (size_t)-FSE_ERROR_corruptionDetected;   /* by construction : at least 2 elts of rank 1, must be even */
+
+    /* Prepare ranks */
+    nextRankStart = 0;
+    for (n=1; n<=maxBits; n++)
+    {
+        U32 current = nextRankStart;
+        nextRankStart += (rankVal[n] << (n-1));
+        rankVal[n] = current;
+    }
+
+    /* fill DTable */
+    for (n=0; n<=oSize; n++)
+    {
+        const U32 w = huffWeight[n];
+        const U32 length = (1 << w) >> 1;
+        U32 i;
+        HUF_DElt D;
+        D.byte = (BYTE)n; D.nbBits = (BYTE)(maxBits + 1 - w);
+        for (i = rankVal[w]; i < rankVal[w] + length; i++)
+            dt[i] = D;
+        rankVal[w] += length;
+    }
+
+    return iSize+1;
+}
+
+
+static BYTE HUF_decodeSymbol(FSE_DStream_t* Dstream, const HUF_DElt* dt, const U32 dtLog)
+{
+        const size_t val = FSE_lookBitsFast(Dstream, dtLog); /* note : dtLog >= 1 */
+        const BYTE c = dt[val].byte;
+        FSE_skipBits(Dstream, dt[val].nbBits);
+        return c;
+}
+
+static size_t HUF_decompress_usingDTable(   /* -3% slower when non static */
+          void* dst, size_t maxDstSize,
+    const void* cSrc, size_t cSrcSize,
+    const U16* DTable)
+{
+    BYTE* const ostart = (BYTE*) dst;
+    BYTE* op = ostart;
+    BYTE* const omax = op + maxDstSize;
+    BYTE* const olimit = omax-15;
+
+    const void* ptr = DTable;
+    const HUF_DElt* const dt = (const HUF_DElt*)(ptr)+1;
+    const U32 dtLog = DTable[0];
+    size_t errorCode;
+    U32 reloadStatus;
+
+    /* Init */
+
+    const U16* jumpTable = (const U16*)cSrc;
+    const size_t length1 = FSE_readLE16(jumpTable);
+    const size_t length2 = FSE_readLE16(jumpTable+1);
+    const size_t length3 = FSE_readLE16(jumpTable+2);
+    const size_t length4 = cSrcSize - 6 - length1 - length2 - length3;   // check coherency !!
+    const char* const start1 = (const char*)(cSrc) + 6;
+    const char* const start2 = start1 + length1;
+    const char* const start3 = start2 + length2;
+    const char* const start4 = start3 + length3;
+    FSE_DStream_t bitD1, bitD2, bitD3, bitD4;
+
+    if (length1+length2+length3+6 >= cSrcSize) return (size_t)-FSE_ERROR_srcSize_wrong;
+
+    errorCode = FSE_initDStream(&bitD1, start1, length1);
+    if (FSE_isError(errorCode)) return errorCode;
+    errorCode = FSE_initDStream(&bitD2, start2, length2);
+    if (FSE_isError(errorCode)) return errorCode;
+    errorCode = FSE_initDStream(&bitD3, start3, length3);
+    if (FSE_isError(errorCode)) return errorCode;
+    errorCode = FSE_initDStream(&bitD4, start4, length4);
+    if (FSE_isError(errorCode)) return errorCode;
+
+    reloadStatus=FSE_reloadDStream(&bitD2);
+
+    /* 16 symbols per loop */
+    for ( ; (reloadStatus12)) FSE_reloadDStream(&Dstream)
+
+#define HUF_DECODE_SYMBOL_2(n, Dstream) \
+        op[n] = HUF_decodeSymbol(&Dstream, dt, dtLog); \
+        if (FSE_32bits()) FSE_reloadDStream(&Dstream)
+
+        HUF_DECODE_SYMBOL_1( 0, bitD1);
+        HUF_DECODE_SYMBOL_1( 1, bitD2);
+        HUF_DECODE_SYMBOL_1( 2, bitD3);
+        HUF_DECODE_SYMBOL_1( 3, bitD4);
+        HUF_DECODE_SYMBOL_2( 4, bitD1);
+        HUF_DECODE_SYMBOL_2( 5, bitD2);
+        HUF_DECODE_SYMBOL_2( 6, bitD3);
+        HUF_DECODE_SYMBOL_2( 7, bitD4);
+        HUF_DECODE_SYMBOL_1( 8, bitD1);
+        HUF_DECODE_SYMBOL_1( 9, bitD2);
+        HUF_DECODE_SYMBOL_1(10, bitD3);
+        HUF_DECODE_SYMBOL_1(11, bitD4);
+        HUF_DECODE_SYMBOL_0(12, bitD1);
+        HUF_DECODE_SYMBOL_0(13, bitD2);
+        HUF_DECODE_SYMBOL_0(14, bitD3);
+        HUF_DECODE_SYMBOL_0(15, bitD4);
+    }
+
+    if (reloadStatus!=FSE_DStream_completed)   /* not complete : some bitStream might be FSE_DStream_unfinished */
+        return (size_t)-FSE_ERROR_corruptionDetected;
+
+    /* tail */
+    {
+        // bitTail = bitD1;   // *much* slower : -20% !??!
+        FSE_DStream_t bitTail;
+        bitTail.ptr = bitD1.ptr;
+        bitTail.bitsConsumed = bitD1.bitsConsumed;
+        bitTail.bitContainer = bitD1.bitContainer;   // required in case of FSE_DStream_endOfBuffer
+        bitTail.start = start1;
+        for ( ; (FSE_reloadDStream(&bitTail) < FSE_DStream_completed) && (op= cSrcSize) return (size_t)-FSE_ERROR_srcSize_wrong;
+    ip += errorCode;
+    cSrcSize -= errorCode;
+
+    return HUF_decompress_usingDTable (dst, maxDstSize, ip, cSrcSize, DTable);
+}
+
+
+#endif   /* FSE_COMMONDEFS_ONLY */
+
+/*
+    zstd - standard compression library
+    Copyright (C) 2014-2015, Yann Collet.
+
+    BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following conditions are
+    met:
+    * Redistributions of source code must retain the above copyright
+    notice, this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above
+    copyright notice, this list of conditions and the following disclaimer
+    in the documentation and/or other materials provided with the
+    distribution.
+    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+    "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+    A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+    OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+    SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+    LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+    DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+    THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+    (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+    OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+    You can contact the author at :
+    - zstd source repository : https://github.com/Cyan4973/zstd
+    - ztsd public forum : https://groups.google.com/forum/#!forum/lz4c
+*/
+
+/****************************************************************
+*  Tuning parameters
+*****************************************************************/
+/* MEMORY_USAGE :
+*  Memory usage formula : N->2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; etc.)
+*  Increasing memory usage improves compression ratio
+*  Reduced memory usage can improve speed, due to cache effect */
+#define ZSTD_MEMORY_USAGE 17
+
+
+/**************************************
+   CPU Feature Detection
+**************************************/
+/*
+ * Automated efficient unaligned memory access detection
+ * Based on known hardware architectures
+ * This list will be updated thanks to feedbacks
+ */
+#if defined(CPU_HAS_EFFICIENT_UNALIGNED_MEMORY_ACCESS) \
+    || defined(__ARM_FEATURE_UNALIGNED) \
+    || defined(__i386__) || defined(__x86_64__) \
+    || defined(_M_IX86) || defined(_M_X64) \
+    || defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_8__) \
+    || (defined(_M_ARM) && (_M_ARM >= 7))
+#  define ZSTD_UNALIGNED_ACCESS 1
+#else
+#  define ZSTD_UNALIGNED_ACCESS 0
+#endif
+
+
+/********************************************************
+*  Includes
+*********************************************************/
+#include       /* calloc */
+#include       /* memcpy, memmove */
+#include        /* debug : printf */
+
+
+/********************************************************
+*  Compiler specifics
+*********************************************************/
+#ifdef __AVX2__
+#  include    /* AVX2 intrinsics */
+#endif
+
+#ifdef _MSC_VER    /* Visual Studio */
+#  include                     /* For Visual 2005 */
+#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
+#  pragma warning(disable : 4324)        /* disable: C4324: padded structure */
+#endif
+
+
+#ifndef MEM_ACCESS_MODULE
+#define MEM_ACCESS_MODULE
+/********************************************************
+*  Basic Types
+*********************************************************/
+#if defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+# include 
+typedef  uint8_t BYTE;
+typedef uint16_t U16;
+typedef  int16_t S16;
+typedef uint32_t U32;
+typedef  int32_t S32;
+typedef uint64_t U64;
+#else
+typedef unsigned char       BYTE;
+typedef unsigned short      U16;
+typedef   signed short      S16;
+typedef unsigned int        U32;
+typedef   signed int        S32;
+typedef unsigned long long  U64;
+#endif
+
+#endif   /* MEM_ACCESS_MODULE */
+
+
+/********************************************************
+*  Constants
+*********************************************************/
+static const U32 ZSTD_magicNumber = 0xFD2FB51E;   /* 3rd version : seqNb header */
+
+#define HASH_LOG (ZSTD_MEMORY_USAGE - 2)
+#define HASH_TABLESIZE (1 << HASH_LOG)
+#define HASH_MASK (HASH_TABLESIZE - 1)
+
+#define KNUTH 2654435761
+
+#define BIT7 128
+#define BIT6  64
+#define BIT5  32
+#define BIT4  16
+
+#define KB *(1 <<10)
+#define MB *(1 <<20)
+#define GB *(1U<<30)
+
+#define BLOCKSIZE (128 KB)                 /* define, for static allocation */
+
+#define WORKPLACESIZE (BLOCKSIZE*3)
+#define MINMATCH 4
+#define MLbits   7
+#define LLbits   6
+#define Offbits  5
+#define MaxML  ((1<>3];
+#else
+    U32 hashTable[HASH_TABLESIZE];
+#endif
+	BYTE buffer[WORKPLACESIZE];
+} cctxi_t;
+
+
+
+
+/**************************************
+*  Error Management
+**************************************/
+/* published entry point */
+unsigned ZSTDv01_isError(size_t code) { return ERR_isError(code); }
+
+
+/**************************************
+*  Tool functions
+**************************************/
+#define ZSTD_VERSION_MAJOR    0    /* for breaking interface changes  */
+#define ZSTD_VERSION_MINOR    1    /* for new (non-breaking) interface capabilities */
+#define ZSTD_VERSION_RELEASE  3    /* for tweaks, bug-fixes, or development */
+#define ZSTD_VERSION_NUMBER  (ZSTD_VERSION_MAJOR *100*100 + ZSTD_VERSION_MINOR *100 + ZSTD_VERSION_RELEASE)
+
+/**************************************************************
+*   Decompression code
+**************************************************************/
+
+size_t ZSTDv01_getcBlockSize(const void* src, size_t srcSize, blockProperties_t* bpPtr)
+{
+    const BYTE* const in = (const BYTE* const)src;
+    BYTE headerFlags;
+    U32 cSize;
+
+    if (srcSize < 3) return ERROR(srcSize_wrong);
+
+    headerFlags = *in;
+    cSize = in[2] + (in[1]<<8) + ((in[0] & 7)<<16);
+
+    bpPtr->blockType = (blockType_t)(headerFlags >> 6);
+    bpPtr->origSize = (bpPtr->blockType == bt_rle) ? cSize : 0;
+
+    if (bpPtr->blockType == bt_end) return 0;
+    if (bpPtr->blockType == bt_rle) return 1;
+    return cSize;
+}
+
+
+static size_t ZSTD_copyUncompressedBlock(void* dst, size_t maxDstSize, const void* src, size_t srcSize)
+{
+    if (srcSize > maxDstSize) return ERROR(dstSize_tooSmall);
+    memcpy(dst, src, srcSize);
+    return srcSize;
+}
+
+
+static size_t ZSTD_decompressLiterals(void* ctx,
+                                      void* dst, size_t maxDstSize,
+                                const void* src, size_t srcSize)
+{
+    BYTE* op = (BYTE*)dst;
+    BYTE* const oend = op + maxDstSize;
+    const BYTE* ip = (const BYTE*)src;
+    size_t errorCode;
+    size_t litSize;
+
+    /* check : minimum 2, for litSize, +1, for content */
+    if (srcSize <= 3) return ERROR(corruption_detected);
+
+    litSize = ip[1] + (ip[0]<<8);
+    litSize += ((ip[-3] >> 3) & 7) << 16;   // mmmmh....
+    op = oend - litSize;
+
+    (void)ctx;
+    if (litSize > maxDstSize) return ERROR(dstSize_tooSmall);
+    errorCode = HUF_decompress(op, litSize, ip+2, srcSize-2);
+    if (FSE_isError(errorCode)) return ERROR(GENERIC);
+    return litSize;
+}
+
+
+size_t ZSTDv01_decodeLiteralsBlock(void* ctx,
+                                void* dst, size_t maxDstSize,
+                          const BYTE** litStart, size_t* litSize,
+                          const void* src, size_t srcSize)
+{
+    const BYTE* const istart = (const BYTE* const)src;
+    const BYTE* ip = istart;
+    BYTE* const ostart = (BYTE* const)dst;
+    BYTE* const oend = ostart + maxDstSize;
+    blockProperties_t litbp;
+
+    size_t litcSize = ZSTDv01_getcBlockSize(src, srcSize, &litbp);
+    if (ZSTDv01_isError(litcSize)) return litcSize;
+    if (litcSize > srcSize - ZSTD_blockHeaderSize) return ERROR(srcSize_wrong);
+    ip += ZSTD_blockHeaderSize;
+
+    switch(litbp.blockType)
+    {
+    case bt_raw:
+        *litStart = ip;
+        ip += litcSize;
+        *litSize = litcSize;
+        break;
+    case bt_rle:
+        {
+            size_t rleSize = litbp.origSize;
+            if (rleSize>maxDstSize) return ERROR(dstSize_tooSmall);
+            if (!srcSize) return ERROR(srcSize_wrong);
+            memset(oend - rleSize, *ip, rleSize);
+            *litStart = oend - rleSize;
+            *litSize = rleSize;
+            ip++;
+            break;
+        }
+    case bt_compressed:
+        {
+            size_t decodedLitSize = ZSTD_decompressLiterals(ctx, dst, maxDstSize, ip, litcSize);
+            if (ZSTDv01_isError(decodedLitSize)) return decodedLitSize;
+            *litStart = oend - decodedLitSize;
+            *litSize = decodedLitSize;
+            ip += litcSize;
+            break;
+        }
+    case bt_end:
+    default:
+        return ERROR(GENERIC);
+    }
+
+    return ip-istart;
+}
+
+
+size_t ZSTDv01_decodeSeqHeaders(int* nbSeq, const BYTE** dumpsPtr, size_t* dumpsLengthPtr,
+                         FSE_DTable* DTableLL, FSE_DTable* DTableML, FSE_DTable* DTableOffb,
+                         const void* src, size_t srcSize)
+{
+    const BYTE* const istart = (const BYTE* const)src;
+    const BYTE* ip = istart;
+    const BYTE* const iend = istart + srcSize;
+    U32 LLtype, Offtype, MLtype;
+    U32 LLlog, Offlog, MLlog;
+    size_t dumpsLength;
+
+    /* check */
+    if (srcSize < 5) return ERROR(srcSize_wrong);
+
+    /* SeqHead */
+    *nbSeq = ZSTD_readLE16(ip); ip+=2;
+    LLtype  = *ip >> 6;
+    Offtype = (*ip >> 4) & 3;
+    MLtype  = (*ip >> 2) & 3;
+    if (*ip & 2)
+    {
+        dumpsLength  = ip[2];
+        dumpsLength += ip[1] << 8;
+        ip += 3;
+    }
+    else
+    {
+        dumpsLength  = ip[1];
+        dumpsLength += (ip[0] & 1) << 8;
+        ip += 2;
+    }
+    *dumpsPtr = ip;
+    ip += dumpsLength;
+    *dumpsLengthPtr = dumpsLength;
+
+    /* check */
+    if (ip > iend-3) return ERROR(srcSize_wrong); /* min : all 3 are "raw", hence no header, but at least xxLog bits per type */
+
+    /* sequences */
+    {
+        S16 norm[MaxML+1];    /* assumption : MaxML >= MaxLL and MaxOff */
+        size_t headerSize;
+
+        /* Build DTables */
+        switch(LLtype)
+        {
+        case bt_rle :
+            LLlog = 0;
+            FSE_buildDTable_rle(DTableLL, *ip++); break;
+        case bt_raw :
+            LLlog = LLbits;
+            FSE_buildDTable_raw(DTableLL, LLbits); break;
+        default :
+            {   U32 max = MaxLL;
+                headerSize = FSE_readNCount(norm, &max, &LLlog, ip, iend-ip);
+                if (FSE_isError(headerSize)) return ERROR(GENERIC);
+                if (LLlog > LLFSELog) return ERROR(corruption_detected);
+                ip += headerSize;
+                FSE_buildDTable(DTableLL, norm, max, LLlog);
+        }   }
+
+        switch(Offtype)
+        {
+        case bt_rle :
+            Offlog = 0;
+            if (ip > iend-2) return ERROR(srcSize_wrong); /* min : "raw", hence no header, but at least xxLog bits */
+            FSE_buildDTable_rle(DTableOffb, *ip++); break;
+        case bt_raw :
+            Offlog = Offbits;
+            FSE_buildDTable_raw(DTableOffb, Offbits); break;
+        default :
+            {   U32 max = MaxOff;
+                headerSize = FSE_readNCount(norm, &max, &Offlog, ip, iend-ip);
+                if (FSE_isError(headerSize)) return ERROR(GENERIC);
+                if (Offlog > OffFSELog) return ERROR(corruption_detected);
+                ip += headerSize;
+                FSE_buildDTable(DTableOffb, norm, max, Offlog);
+        }   }
+
+        switch(MLtype)
+        {
+        case bt_rle :
+            MLlog = 0;
+            if (ip > iend-2) return ERROR(srcSize_wrong); /* min : "raw", hence no header, but at least xxLog bits */
+            FSE_buildDTable_rle(DTableML, *ip++); break;
+        case bt_raw :
+            MLlog = MLbits;
+            FSE_buildDTable_raw(DTableML, MLbits); break;
+        default :
+            {   U32 max = MaxML;
+                headerSize = FSE_readNCount(norm, &max, &MLlog, ip, iend-ip);
+                if (FSE_isError(headerSize)) return ERROR(GENERIC);
+                if (MLlog > MLFSELog) return ERROR(corruption_detected);
+                ip += headerSize;
+                FSE_buildDTable(DTableML, norm, max, MLlog);
+    }   }   }
+
+    return ip-istart;
+}
+
+
+typedef struct {
+    size_t litLength;
+    size_t offset;
+    size_t matchLength;
+} seq_t;
+
+typedef struct {
+    FSE_DStream_t DStream;
+    FSE_DState_t stateLL;
+    FSE_DState_t stateOffb;
+    FSE_DState_t stateML;
+    size_t prevOffset;
+    const BYTE* dumps;
+    const BYTE* dumpsEnd;
+} seqState_t;
+
+
+static void ZSTD_decodeSequence(seq_t* seq, seqState_t* seqState)
+{
+    size_t litLength;
+    size_t prevOffset;
+    size_t offset;
+    size_t matchLength;
+    const BYTE* dumps = seqState->dumps;
+    const BYTE* const de = seqState->dumpsEnd;
+
+    /* Literal length */
+    litLength = FSE_decodeSymbol(&(seqState->stateLL), &(seqState->DStream));
+    prevOffset = litLength ? seq->offset : seqState->prevOffset;
+    seqState->prevOffset = seq->offset;
+    if (litLength == MaxLL)
+    {
+        U32 add = dumps 1 byte */
+                dumps += 3;
+            }
+        }
+    }
+
+    /* Offset */
+    {
+        U32 offsetCode, nbBits;
+        offsetCode = FSE_decodeSymbol(&(seqState->stateOffb), &(seqState->DStream));
+        if (ZSTD_32bits()) FSE_reloadDStream(&(seqState->DStream));
+        nbBits = offsetCode - 1;
+        if (offsetCode==0) nbBits = 0;   /* cmove */
+        offset = ((size_t)1 << (nbBits & ((sizeof(offset)*8)-1))) + FSE_readBits(&(seqState->DStream), nbBits);
+        if (ZSTD_32bits()) FSE_reloadDStream(&(seqState->DStream));
+        if (offsetCode==0) offset = prevOffset;
+    }
+
+    /* MatchLength */
+    matchLength = FSE_decodeSymbol(&(seqState->stateML), &(seqState->DStream));
+    if (matchLength == MaxML)
+    {
+        U32 add = dumps 1 byte */
+                dumps += 3;
+            }
+        }
+    }
+    matchLength += MINMATCH;
+
+    /* save result */
+    seq->litLength = litLength;
+    seq->offset = offset;
+    seq->matchLength = matchLength;
+    seqState->dumps = dumps;
+}
+
+
+static size_t ZSTD_execSequence(BYTE* op,
+                                seq_t sequence,
+                                const BYTE** litPtr, const BYTE* const litLimit,
+                                BYTE* const base, BYTE* const oend)
+{
+    static const int dec32table[] = {0, 1, 2, 1, 4, 4, 4, 4};   /* added */
+    static const int dec64table[] = {8, 8, 8, 7, 8, 9,10,11};   /* substracted */
+    const BYTE* const ostart = op;
+    const size_t litLength = sequence.litLength;
+    BYTE* const endMatch = op + litLength + sequence.matchLength;    /* risk : address space overflow (32-bits) */
+    const BYTE* const litEnd = *litPtr + litLength;
+
+    /* check */
+    if (endMatch > oend) return ERROR(dstSize_tooSmall);   /* overwrite beyond dst buffer */
+    if (litEnd > litLimit) return ERROR(corruption_detected);
+    if (sequence.matchLength > (size_t)(*litPtr-op))  return ERROR(dstSize_tooSmall);    /* overwrite literal segment */
+
+    /* copy Literals */
+    if (((size_t)(*litPtr - op) < 8) || ((size_t)(oend-litEnd) < 8) || (op+litLength > oend-8))
+        memmove(op, *litPtr, litLength);   /* overwrite risk */
+    else
+        ZSTD_wildcopy(op, *litPtr, litLength);
+    op += litLength;
+    *litPtr = litEnd;   /* update for next sequence */
+
+    /* check : last match must be at a minimum distance of 8 from end of dest buffer */
+    if (oend-op < 8) return ERROR(dstSize_tooSmall);
+
+    /* copy Match */
+    {
+        const U32 overlapRisk = (((size_t)(litEnd - endMatch)) < 12);
+        const BYTE* match = op - sequence.offset;            /* possible underflow at op - offset ? */
+        size_t qutt = 12;
+        U64 saved[2];
+
+        /* check */
+        if (match < base) return ERROR(corruption_detected);
+        if (sequence.offset > (size_t)base) return ERROR(corruption_detected);
+
+        /* save beginning of literal sequence, in case of write overlap */
+        if (overlapRisk)
+        {
+            if ((endMatch + qutt) > oend) qutt = oend-endMatch;
+            memcpy(saved, endMatch, qutt);
+        }
+
+        if (sequence.offset < 8)
+        {
+            const int dec64 = dec64table[sequence.offset];
+            op[0] = match[0];
+            op[1] = match[1];
+            op[2] = match[2];
+            op[3] = match[3];
+            match += dec32table[sequence.offset];
+            ZSTD_copy4(op+4, match);
+            match -= dec64;
+        } else { ZSTD_copy8(op, match); }
+        op += 8; match += 8;
+
+        if (endMatch > oend-(16-MINMATCH))
+        {
+            if (op < oend-8)
+            {
+                ZSTD_wildcopy(op, match, (oend-8) - op);
+                match += (oend-8) - op;
+                op = oend-8;
+            }
+            while (opLLTable;
+    U32* DTableML = dctx->MLTable;
+    U32* DTableOffb = dctx->OffTable;
+    BYTE* const base = (BYTE*) (dctx->base);
+
+    /* Build Decoding Tables */
+    errorCode = ZSTDv01_decodeSeqHeaders(&nbSeq, &dumps, &dumpsLength,
+                                      DTableLL, DTableML, DTableOffb,
+                                      ip, iend-ip);
+    if (ZSTDv01_isError(errorCode)) return errorCode;
+    ip += errorCode;
+
+    /* Regen sequences */
+    {
+        seq_t sequence;
+        seqState_t seqState;
+
+        memset(&sequence, 0, sizeof(sequence));
+        seqState.dumps = dumps;
+        seqState.dumpsEnd = dumps + dumpsLength;
+        seqState.prevOffset = 1;
+        errorCode = FSE_initDStream(&(seqState.DStream), ip, iend-ip);
+        if (FSE_isError(errorCode)) return ERROR(corruption_detected);
+        FSE_initDState(&(seqState.stateLL), &(seqState.DStream), DTableLL);
+        FSE_initDState(&(seqState.stateOffb), &(seqState.DStream), DTableOffb);
+        FSE_initDState(&(seqState.stateML), &(seqState.DStream), DTableML);
+
+        for ( ; (FSE_reloadDStream(&(seqState.DStream)) <= FSE_DStream_completed) && (nbSeq>0) ; )
+        {
+            size_t oneSeqSize;
+            nbSeq--;
+            ZSTD_decodeSequence(&sequence, &seqState);
+            oneSeqSize = ZSTD_execSequence(op, sequence, &litPtr, litEnd, base, oend);
+            if (ZSTDv01_isError(oneSeqSize)) return oneSeqSize;
+            op += oneSeqSize;
+        }
+
+        /* check if reached exact end */
+        if ( !FSE_endOfDStream(&(seqState.DStream)) ) return ERROR(corruption_detected);   /* requested too much : data is corrupted */
+        if (nbSeq<0) return ERROR(corruption_detected);   /* requested too many sequences : data is corrupted */
+
+        /* last literal segment */
+        {
+            size_t lastLLSize = litEnd - litPtr;
+            if (op+lastLLSize > oend) return ERROR(dstSize_tooSmall);
+            if (op != litPtr) memmove(op, litPtr, lastLLSize);
+            op += lastLLSize;
+        }
+    }
+
+    return op-ostart;
+}
+
+
+static size_t ZSTD_decompressBlock(
+                            void* ctx,
+                            void* dst, size_t maxDstSize,
+                      const void* src, size_t srcSize)
+{
+    /* blockType == blockCompressed, srcSize is trusted */
+    const BYTE* ip = (const BYTE*)src;
+    const BYTE* litPtr = NULL;
+    size_t litSize = 0;
+    size_t errorCode;
+
+    /* Decode literals sub-block */
+    errorCode = ZSTDv01_decodeLiteralsBlock(ctx, dst, maxDstSize, &litPtr, &litSize, src, srcSize);
+    if (ZSTDv01_isError(errorCode)) return errorCode;
+    ip += errorCode;
+    srcSize -= errorCode;
+
+    return ZSTD_decompressSequences(ctx, dst, maxDstSize, ip, srcSize, litPtr, litSize);
+}
+
+
+size_t ZSTDv01_decompressDCtx(void* ctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize)
+{
+    const BYTE* ip = (const BYTE*)src;
+    const BYTE* iend = ip + srcSize;
+    BYTE* const ostart = (BYTE* const)dst;
+    BYTE* op = ostart;
+    BYTE* const oend = ostart + maxDstSize;
+    size_t remainingSize = srcSize;
+    U32 magicNumber;
+    size_t errorCode=0;
+    blockProperties_t blockProperties;
+
+    /* Frame Header */
+    if (srcSize < ZSTD_frameHeaderSize+ZSTD_blockHeaderSize) return ERROR(srcSize_wrong);
+    magicNumber = ZSTD_readBE32(src);
+    if (magicNumber != ZSTD_magicNumber) return ERROR(prefix_unknown);
+    ip += ZSTD_frameHeaderSize; remainingSize -= ZSTD_frameHeaderSize;
+
+    /* Loop on each block */
+    while (1)
+    {
+        size_t blockSize = ZSTDv01_getcBlockSize(ip, iend-ip, &blockProperties);
+        if (ZSTDv01_isError(blockSize)) return blockSize;
+
+        ip += ZSTD_blockHeaderSize;
+        remainingSize -= ZSTD_blockHeaderSize;
+        if (blockSize > remainingSize) return ERROR(srcSize_wrong);
+
+        switch(blockProperties.blockType)
+        {
+        case bt_compressed:
+            errorCode = ZSTD_decompressBlock(ctx, op, oend-op, ip, blockSize);
+            break;
+        case bt_raw :
+            errorCode = ZSTD_copyUncompressedBlock(op, oend-op, ip, blockSize);
+            break;
+        case bt_rle :
+            return ERROR(GENERIC);   /* not yet supported */
+            break;
+        case bt_end :
+            /* end of frame */
+            if (remainingSize) return ERROR(srcSize_wrong);
+            break;
+        default:
+            return ERROR(GENERIC);
+        }
+        if (blockSize == 0) break;   /* bt_end */
+
+        if (ZSTDv01_isError(errorCode)) return errorCode;
+        op += errorCode;
+        ip += blockSize;
+        remainingSize -= blockSize;
+    }
+
+    return op-ostart;
+}
+
+size_t ZSTDv01_decompress(void* dst, size_t maxDstSize, const void* src, size_t srcSize)
+{
+    dctx_t ctx;
+    ctx.base = dst;
+    return ZSTDv01_decompressDCtx(&ctx, dst, maxDstSize, src, srcSize);
+}
+
+
+/*******************************
+*  Streaming Decompression API
+*******************************/
+
+size_t ZSTDv01_resetDCtx(ZSTDv01_Dctx* dctx)
+{
+    dctx->expected = ZSTD_frameHeaderSize;
+    dctx->phase = 0;
+    dctx->previousDstEnd = NULL;
+    dctx->base = NULL;
+    return 0;
+}
+
+ZSTDv01_Dctx* ZSTDv01_createDCtx(void)
+{
+    ZSTDv01_Dctx* dctx = (ZSTDv01_Dctx*)malloc(sizeof(ZSTDv01_Dctx));
+    if (dctx==NULL) return NULL;
+    ZSTDv01_resetDCtx(dctx);
+    return dctx;
+}
+
+size_t ZSTDv01_freeDCtx(ZSTDv01_Dctx* dctx)
+{
+    free(dctx);
+    return 0;
+}
+
+size_t ZSTDv01_nextSrcSizeToDecompress(ZSTDv01_Dctx* dctx)
+{
+    return ((dctx_t*)dctx)->expected;
+}
+
+size_t ZSTDv01_decompressContinue(ZSTDv01_Dctx* dctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize)
+{
+    dctx_t* ctx = (dctx_t*)dctx;
+
+    /* Sanity check */
+    if (srcSize != ctx->expected) return ERROR(srcSize_wrong);
+    if (dst != ctx->previousDstEnd)  /* not contiguous */
+        ctx->base = dst;
+
+    /* Decompress : frame header */
+    if (ctx->phase == 0)
+    {
+        /* Check frame magic header */
+        U32 magicNumber = ZSTD_readBE32(src);
+        if (magicNumber != ZSTD_magicNumber) return ERROR(prefix_unknown);
+        ctx->phase = 1;
+        ctx->expected = ZSTD_blockHeaderSize;
+        return 0;
+    }
+
+    /* Decompress : block header */
+    if (ctx->phase == 1)
+    {
+        blockProperties_t bp;
+        size_t blockSize = ZSTDv01_getcBlockSize(src, ZSTD_blockHeaderSize, &bp);
+        if (ZSTDv01_isError(blockSize)) return blockSize;
+        if (bp.blockType == bt_end)
+        {
+            ctx->expected = 0;
+            ctx->phase = 0;
+        }
+        else
+        {
+            ctx->expected = blockSize;
+            ctx->bType = bp.blockType;
+            ctx->phase = 2;
+        }
+
+        return 0;
+    }
+
+    /* Decompress : block content */
+    {
+        size_t rSize;
+        switch(ctx->bType)
+        {
+        case bt_compressed:
+            rSize = ZSTD_decompressBlock(ctx, dst, maxDstSize, src, srcSize);
+            break;
+        case bt_raw :
+            rSize = ZSTD_copyUncompressedBlock(dst, maxDstSize, src, srcSize);
+            break;
+        case bt_rle :
+            return ERROR(GENERIC);   /* not yet handled */
+            break;
+        case bt_end :   /* should never happen (filtered at phase 1) */
+            rSize = 0;
+            break;
+        default:
+            return ERROR(GENERIC);
+        }
+        ctx->phase = 1;
+        ctx->expected = ZSTD_blockHeaderSize;
+        ctx->previousDstEnd = (void*)( ((char*)dst) + rSize);
+        return rSize;
+    }
+
+}
diff --git a/lib/legacy/zstd_v01.h b/lib/legacy/zstd_v01.h
new file mode 100644
index 0000000..0f2323d
--- /dev/null
+++ b/lib/legacy/zstd_v01.h
@@ -0,0 +1,80 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+#ifndef ZSTD_V01_H_28739879432
+#define ZSTD_V01_H_28739879432
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/* *************************************
+*  Includes
+***************************************/
+#include    /* size_t */
+
+
+/* *************************************
+*  Simple one-step function
+***************************************/
+/**
+ZSTDv01_decompress() : decompress ZSTD frames compliant with v0.1.x format
+    compressedSize : is the exact source size
+    maxOriginalSize : is the size of the 'dst' buffer, which must be already allocated.
+                      It must be equal or larger than originalSize, otherwise decompression will fail.
+    return : the number of bytes decompressed into destination buffer (originalSize)
+             or an errorCode if it fails (which can be tested using ZSTDv01_isError())
+*/
+size_t ZSTDv01_decompress( void* dst, size_t maxOriginalSize,
+                     const void* src, size_t compressedSize);
+
+/**
+ZSTDv01_isError() : tells if the result of ZSTDv01_decompress() is an error
+*/
+unsigned ZSTDv01_isError(size_t code);
+
+
+/* *************************************
+*  Advanced functions
+***************************************/
+typedef struct ZSTDv01_Dctx_s ZSTDv01_Dctx;
+ZSTDv01_Dctx* ZSTDv01_createDCtx(void);
+size_t ZSTDv01_freeDCtx(ZSTDv01_Dctx* dctx);
+
+size_t ZSTDv01_decompressDCtx(void* ctx,
+                              void* dst, size_t maxOriginalSize,
+                        const void* src, size_t compressedSize);
+
+/* *************************************
+*  Streaming functions
+***************************************/
+size_t ZSTDv01_resetDCtx(ZSTDv01_Dctx* dctx);
+
+size_t ZSTDv01_nextSrcSizeToDecompress(ZSTDv01_Dctx* dctx);
+size_t ZSTDv01_decompressContinue(ZSTDv01_Dctx* dctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize);
+/**
+  Use above functions alternatively.
+  ZSTD_nextSrcSizeToDecompress() tells how much bytes to provide as 'srcSize' to ZSTD_decompressContinue().
+  ZSTD_decompressContinue() will use previous data blocks to improve compression if they are located prior to current block.
+  Result is the number of bytes regenerated within 'dst'.
+  It can be zero, which is not an error; it just means ZSTD_decompressContinue() has decoded some header.
+*/
+
+/* *************************************
+*  Prefix - version detection
+***************************************/
+#define ZSTDv01_magicNumber   0xFD2FB51E   /* Big Endian version */
+#define ZSTDv01_magicNumberLE 0x1EB52FFD   /* Little Endian version */
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_V01_H_28739879432 */
diff --git a/lib/legacy/zstd_v02.c b/lib/legacy/zstd_v02.c
new file mode 100644
index 0000000..b8a12ab
--- /dev/null
+++ b/lib/legacy/zstd_v02.c
@@ -0,0 +1,3518 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+
+#include     /* size_t, ptrdiff_t */
+#include "zstd_v02.h"
+#include "error_private.h"
+
+
+/******************************************
+*  Compiler-specific
+******************************************/
+#if defined(_MSC_VER)   /* Visual Studio */
+#   include   /* _byteswap_ulong */
+#   include   /* _byteswap_* */
+#endif
+
+
+/* ******************************************************************
+   mem.h
+   low-level memory access routines
+   Copyright (C) 2013-2015, Yann Collet.
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+    You can contact the author at :
+    - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
+    - Public forum : https://groups.google.com/forum/#!forum/lz4c
+****************************************************************** */
+#ifndef MEM_H_MODULE
+#define MEM_H_MODULE
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/******************************************
+*  Includes
+******************************************/
+#include     /* size_t, ptrdiff_t */
+#include     /* memcpy */
+
+
+/******************************************
+*  Compiler-specific
+******************************************/
+#if defined(__GNUC__)
+#  define MEM_STATIC static __attribute__((unused))
+#elif defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
+#  define MEM_STATIC static inline
+#elif defined(_MSC_VER)
+#  define MEM_STATIC static __inline
+#else
+#  define MEM_STATIC static  /* this version may generate warnings for unused static functions; disable the relevant warning */
+#endif
+
+
+/****************************************************************
+*  Basic Types
+*****************************************************************/
+#if defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
+# include 
+  typedef  uint8_t BYTE;
+  typedef uint16_t U16;
+  typedef  int16_t S16;
+  typedef uint32_t U32;
+  typedef  int32_t S32;
+  typedef uint64_t U64;
+  typedef  int64_t S64;
+#else
+  typedef unsigned char       BYTE;
+  typedef unsigned short      U16;
+  typedef   signed short      S16;
+  typedef unsigned int        U32;
+  typedef   signed int        S32;
+  typedef unsigned long long  U64;
+  typedef   signed long long  S64;
+#endif
+
+
+/****************************************************************
+*  Memory I/O
+*****************************************************************/
+/* MEM_FORCE_MEMORY_ACCESS
+ * By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
+ * Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
+ * The below switch allow to select different access method for improved performance.
+ * Method 0 (default) : use `memcpy()`. Safe and portable.
+ * Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
+ *            This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
+ * Method 2 : direct access. This method is portable but violate C standard.
+ *            It can generate buggy code on targets generating assembly depending on alignment.
+ *            But in some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
+ * See http://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
+ * Prefer these methods in priority order (0 > 1 > 2)
+ */
+#ifndef MEM_FORCE_MEMORY_ACCESS   /* can be defined externally, on command line for example */
+#  if defined(__GNUC__) && ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) )
+#    define MEM_FORCE_MEMORY_ACCESS 2
+#  elif (defined(__INTEL_COMPILER) && !defined(WIN32)) || \
+  (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7S__) ))
+#    define MEM_FORCE_MEMORY_ACCESS 1
+#  endif
+#endif
+
+MEM_STATIC unsigned MEM_32bits(void) { return sizeof(void*)==4; }
+MEM_STATIC unsigned MEM_64bits(void) { return sizeof(void*)==8; }
+
+MEM_STATIC unsigned MEM_isLittleEndian(void)
+{
+    const union { U32 u; BYTE c[4]; } one = { 1 };   /* don't use static : performance detrimental  */
+    return one.c[0];
+}
+
+#if defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==2)
+
+/* violates C standard on structure alignment.
+Only use if no other choice to achieve best performance on target platform */
+MEM_STATIC U16 MEM_read16(const void* memPtr) { return *(const U16*) memPtr; }
+MEM_STATIC U32 MEM_read32(const void* memPtr) { return *(const U32*) memPtr; }
+MEM_STATIC U64 MEM_read64(const void* memPtr) { return *(const U64*) memPtr; }
+
+MEM_STATIC void MEM_write16(void* memPtr, U16 value) { *(U16*)memPtr = value; }
+
+#elif defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==1)
+
+/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
+/* currently only defined for gcc and icc */
+typedef union { U16 u16; U32 u32; U64 u64; } __attribute__((packed)) unalign;
+
+MEM_STATIC U16 MEM_read16(const void* ptr) { return ((const unalign*)ptr)->u16; }
+MEM_STATIC U32 MEM_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
+MEM_STATIC U64 MEM_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
+
+MEM_STATIC void MEM_write16(void* memPtr, U16 value) { ((unalign*)memPtr)->u16 = value; }
+
+#else
+
+/* default method, safe and standard.
+   can sometimes prove slower */
+
+MEM_STATIC U16 MEM_read16(const void* memPtr)
+{
+    U16 val; memcpy(&val, memPtr, sizeof(val)); return val;
+}
+
+MEM_STATIC U32 MEM_read32(const void* memPtr)
+{
+    U32 val; memcpy(&val, memPtr, sizeof(val)); return val;
+}
+
+MEM_STATIC U64 MEM_read64(const void* memPtr)
+{
+    U64 val; memcpy(&val, memPtr, sizeof(val)); return val;
+}
+
+MEM_STATIC void MEM_write16(void* memPtr, U16 value)
+{
+    memcpy(memPtr, &value, sizeof(value));
+}
+
+#endif // MEM_FORCE_MEMORY_ACCESS
+
+
+MEM_STATIC U16 MEM_readLE16(const void* memPtr)
+{
+    if (MEM_isLittleEndian())
+        return MEM_read16(memPtr);
+    else
+    {
+        const BYTE* p = (const BYTE*)memPtr;
+        return (U16)(p[0] + (p[1]<<8));
+    }
+}
+
+MEM_STATIC void MEM_writeLE16(void* memPtr, U16 val)
+{
+    if (MEM_isLittleEndian())
+    {
+        MEM_write16(memPtr, val);
+    }
+    else
+    {
+        BYTE* p = (BYTE*)memPtr;
+        p[0] = (BYTE)val;
+        p[1] = (BYTE)(val>>8);
+    }
+}
+
+MEM_STATIC U32 MEM_readLE32(const void* memPtr)
+{
+    if (MEM_isLittleEndian())
+        return MEM_read32(memPtr);
+    else
+    {
+        const BYTE* p = (const BYTE*)memPtr;
+        return (U32)((U32)p[0] + ((U32)p[1]<<8) + ((U32)p[2]<<16) + ((U32)p[3]<<24));
+    }
+}
+
+
+MEM_STATIC U64 MEM_readLE64(const void* memPtr)
+{
+    if (MEM_isLittleEndian())
+        return MEM_read64(memPtr);
+    else
+    {
+        const BYTE* p = (const BYTE*)memPtr;
+        return (U64)((U64)p[0] + ((U64)p[1]<<8) + ((U64)p[2]<<16) + ((U64)p[3]<<24)
+                     + ((U64)p[4]<<32) + ((U64)p[5]<<40) + ((U64)p[6]<<48) + ((U64)p[7]<<56));
+    }
+}
+
+
+MEM_STATIC size_t MEM_readLEST(const void* memPtr)
+{
+    if (MEM_32bits())
+        return (size_t)MEM_readLE32(memPtr);
+    else
+        return (size_t)MEM_readLE64(memPtr);
+}
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* MEM_H_MODULE */
+
+
+/* ******************************************************************
+   bitstream
+   Part of NewGen Entropy library
+   header file (to include)
+   Copyright (C) 2013-2015, Yann Collet.
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+   You can contact the author at :
+   - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+   - Public forum : https://groups.google.com/forum/#!forum/lz4c
+****************************************************************** */
+#ifndef BITSTREAM_H_MODULE
+#define BITSTREAM_H_MODULE
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+
+/*
+*  This API consists of small unitary functions, which highly benefit from being inlined.
+*  Since link-time-optimization is not available for all compilers,
+*  these functions are defined into a .h to be included.
+*/
+
+
+/**********************************************
+*  bitStream decompression API (read backward)
+**********************************************/
+typedef struct
+{
+    size_t   bitContainer;
+    unsigned bitsConsumed;
+    const char* ptr;
+    const char* start;
+} BIT_DStream_t;
+
+typedef enum { BIT_DStream_unfinished = 0,
+               BIT_DStream_endOfBuffer = 1,
+               BIT_DStream_completed = 2,
+               BIT_DStream_overflow = 3 } BIT_DStream_status;  /* result of BIT_reloadDStream() */
+               /* 1,2,4,8 would be better for bitmap combinations, but slows down performance a bit ... :( */
+
+MEM_STATIC size_t   BIT_initDStream(BIT_DStream_t* bitD, const void* srcBuffer, size_t srcSize);
+MEM_STATIC size_t   BIT_readBits(BIT_DStream_t* bitD, unsigned nbBits);
+MEM_STATIC BIT_DStream_status BIT_reloadDStream(BIT_DStream_t* bitD);
+MEM_STATIC unsigned BIT_endOfDStream(const BIT_DStream_t* bitD);
+
+
+/*
+* Start by invoking BIT_initDStream().
+* A chunk of the bitStream is then stored into a local register.
+* Local register size is 64-bits on 64-bits systems, 32-bits on 32-bits systems (size_t).
+* You can then retrieve bitFields stored into the local register, **in reverse order**.
+* Local register is manually filled from memory by the BIT_reloadDStream() method.
+* A reload guarantee a minimum of ((8*sizeof(size_t))-7) bits when its result is BIT_DStream_unfinished.
+* Otherwise, it can be less than that, so proceed accordingly.
+* Checking if DStream has reached its end can be performed with BIT_endOfDStream()
+*/
+
+
+/******************************************
+*  unsafe API
+******************************************/
+MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, unsigned nbBits);
+/* faster, but works only if nbBits >= 1 */
+
+
+
+/****************************************************************
+*  Helper functions
+****************************************************************/
+MEM_STATIC unsigned BIT_highbit32 (register U32 val)
+{
+#   if defined(_MSC_VER)   /* Visual */
+    unsigned long r=0;
+    _BitScanReverse ( &r, val );
+    return (unsigned) r;
+#   elif defined(__GNUC__) && (__GNUC__ >= 3)   /* Use GCC Intrinsic */
+    return 31 - __builtin_clz (val);
+#   else   /* Software version */
+    static const unsigned DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 };
+    U32 v = val;
+    unsigned r;
+    v |= v >> 1;
+    v |= v >> 2;
+    v |= v >> 4;
+    v |= v >> 8;
+    v |= v >> 16;
+    r = DeBruijnClz[ (U32) (v * 0x07C4ACDDU) >> 27];
+    return r;
+#   endif
+}
+
+
+
+/**********************************************************
+* bitStream decoding
+**********************************************************/
+
+/*!BIT_initDStream
+*  Initialize a BIT_DStream_t.
+*  @bitD : a pointer to an already allocated BIT_DStream_t structure
+*  @srcBuffer must point at the beginning of a bitStream
+*  @srcSize must be the exact size of the bitStream
+*  @result : size of stream (== srcSize) or an errorCode if a problem is detected
+*/
+MEM_STATIC size_t BIT_initDStream(BIT_DStream_t* bitD, const void* srcBuffer, size_t srcSize)
+{
+    if (srcSize < 1) { memset(bitD, 0, sizeof(*bitD)); return ERROR(srcSize_wrong); }
+
+    if (srcSize >=  sizeof(size_t))   /* normal case */
+    {
+        U32 contain32;
+        bitD->start = (const char*)srcBuffer;
+        bitD->ptr   = (const char*)srcBuffer + srcSize - sizeof(size_t);
+        bitD->bitContainer = MEM_readLEST(bitD->ptr);
+        contain32 = ((const BYTE*)srcBuffer)[srcSize-1];
+        if (contain32 == 0) return ERROR(GENERIC);   /* endMark not present */
+        bitD->bitsConsumed = 8 - BIT_highbit32(contain32);
+    }
+    else
+    {
+        U32 contain32;
+        bitD->start = (const char*)srcBuffer;
+        bitD->ptr   = bitD->start;
+        bitD->bitContainer = *(const BYTE*)(bitD->start);
+        switch(srcSize)
+        {
+            case 7: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[6]) << (sizeof(size_t)*8 - 16);
+            case 6: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[5]) << (sizeof(size_t)*8 - 24);
+            case 5: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[4]) << (sizeof(size_t)*8 - 32);
+            case 4: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[3]) << 24;
+            case 3: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[2]) << 16;
+            case 2: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[1]) <<  8;
+            default:;
+        }
+        contain32 = ((const BYTE*)srcBuffer)[srcSize-1];
+        if (contain32 == 0) return ERROR(GENERIC);   /* endMark not present */
+        bitD->bitsConsumed = 8 - BIT_highbit32(contain32);
+        bitD->bitsConsumed += (U32)(sizeof(size_t) - srcSize)*8;
+    }
+
+    return srcSize;
+}
+
+/*!BIT_lookBits
+ * Provides next n bits from local register
+ * local register is not modified (bits are still present for next read/look)
+ * On 32-bits, maxNbBits==25
+ * On 64-bits, maxNbBits==57
+ * @return : value extracted
+ */
+MEM_STATIC size_t BIT_lookBits(BIT_DStream_t* bitD, U32 nbBits)
+{
+    const U32 bitMask = sizeof(bitD->bitContainer)*8 - 1;
+    return ((bitD->bitContainer << (bitD->bitsConsumed & bitMask)) >> 1) >> ((bitMask-nbBits) & bitMask);
+}
+
+/*! BIT_lookBitsFast :
+*   unsafe version; only works only if nbBits >= 1 */
+MEM_STATIC size_t BIT_lookBitsFast(BIT_DStream_t* bitD, U32 nbBits)
+{
+    const U32 bitMask = sizeof(bitD->bitContainer)*8 - 1;
+    return (bitD->bitContainer << (bitD->bitsConsumed & bitMask)) >> (((bitMask+1)-nbBits) & bitMask);
+}
+
+MEM_STATIC void BIT_skipBits(BIT_DStream_t* bitD, U32 nbBits)
+{
+    bitD->bitsConsumed += nbBits;
+}
+
+/*!BIT_readBits
+ * Read next n bits from local register.
+ * pay attention to not read more than nbBits contained into local register.
+ * @return : extracted value.
+ */
+MEM_STATIC size_t BIT_readBits(BIT_DStream_t* bitD, U32 nbBits)
+{
+    size_t value = BIT_lookBits(bitD, nbBits);
+    BIT_skipBits(bitD, nbBits);
+    return value;
+}
+
+/*!BIT_readBitsFast :
+*  unsafe version; only works only if nbBits >= 1 */
+MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, U32 nbBits)
+{
+    size_t value = BIT_lookBitsFast(bitD, nbBits);
+    BIT_skipBits(bitD, nbBits);
+    return value;
+}
+
+MEM_STATIC BIT_DStream_status BIT_reloadDStream(BIT_DStream_t* bitD)
+{
+	if (bitD->bitsConsumed > (sizeof(bitD->bitContainer)*8))  /* should never happen */
+		return BIT_DStream_overflow;
+
+    if (bitD->ptr >= bitD->start + sizeof(bitD->bitContainer))
+    {
+        bitD->ptr -= bitD->bitsConsumed >> 3;
+        bitD->bitsConsumed &= 7;
+        bitD->bitContainer = MEM_readLEST(bitD->ptr);
+        return BIT_DStream_unfinished;
+    }
+    if (bitD->ptr == bitD->start)
+    {
+        if (bitD->bitsConsumed < sizeof(bitD->bitContainer)*8) return BIT_DStream_endOfBuffer;
+        return BIT_DStream_completed;
+    }
+    {
+        U32 nbBytes = bitD->bitsConsumed >> 3;
+        BIT_DStream_status result = BIT_DStream_unfinished;
+        if (bitD->ptr - nbBytes < bitD->start)
+        {
+            nbBytes = (U32)(bitD->ptr - bitD->start);  /* ptr > start */
+            result = BIT_DStream_endOfBuffer;
+        }
+        bitD->ptr -= nbBytes;
+        bitD->bitsConsumed -= nbBytes*8;
+        bitD->bitContainer = MEM_readLEST(bitD->ptr);   /* reminder : srcSize > sizeof(bitD) */
+        return result;
+    }
+}
+
+/*! BIT_endOfDStream
+*   @return Tells if DStream has reached its exact end
+*/
+MEM_STATIC unsigned BIT_endOfDStream(const BIT_DStream_t* DStream)
+{
+    return ((DStream->ptr == DStream->start) && (DStream->bitsConsumed == sizeof(DStream->bitContainer)*8));
+}
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* BITSTREAM_H_MODULE */
+/* ******************************************************************
+   Error codes and messages
+   Copyright (C) 2013-2015, Yann Collet
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+   You can contact the author at :
+   - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+   - Public forum : https://groups.google.com/forum/#!forum/lz4c
+****************************************************************** */
+#ifndef ERROR_H_MODULE
+#define ERROR_H_MODULE
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+
+/******************************************
+*  Compiler-specific
+******************************************/
+#if defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
+#  define ERR_STATIC static inline
+#elif defined(_MSC_VER)
+#  define ERR_STATIC static __inline
+#elif defined(__GNUC__)
+#  define ERR_STATIC static __attribute__((unused))
+#else
+#  define ERR_STATIC static  /* this version may generate warnings for unused static functions; disable the relevant warning */
+#endif
+
+
+/******************************************
+*  Error Management
+******************************************/
+#define PREFIX(name) ZSTD_error_##name
+
+#define ERROR(name) (size_t)-PREFIX(name)
+
+#define ERROR_LIST(ITEM) \
+        ITEM(PREFIX(No_Error)) ITEM(PREFIX(GENERIC)) \
+        ITEM(PREFIX(dstSize_tooSmall)) ITEM(PREFIX(srcSize_wrong)) \
+        ITEM(PREFIX(prefix_unknown)) ITEM(PREFIX(corruption_detected)) \
+        ITEM(PREFIX(tableLog_tooLarge)) ITEM(PREFIX(maxSymbolValue_tooLarge)) ITEM(PREFIX(maxSymbolValue_tooSmall)) \
+        ITEM(PREFIX(maxCode))
+
+#define ERROR_GENERATE_ENUM(ENUM) ENUM,
+typedef enum { ERROR_LIST(ERROR_GENERATE_ENUM) } ERR_codes;  /* enum is exposed, to detect & handle specific errors; compare function result to -enum value */
+
+#define ERROR_CONVERTTOSTRING(STRING) #STRING,
+#define ERROR_GENERATE_STRING(EXPR) ERROR_CONVERTTOSTRING(EXPR)
+static const char* ERR_strings[] = { ERROR_LIST(ERROR_GENERATE_STRING) };
+
+ERR_STATIC unsigned ERR_isError(size_t code) { return (code > ERROR(maxCode)); }
+
+ERR_STATIC const char* ERR_getErrorName(size_t code)
+{
+    static const char* codeError = "Unspecified error code";
+    if (ERR_isError(code)) return ERR_strings[-(int)(code)];
+    return codeError;
+}
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ERROR_H_MODULE */
+/*
+Constructor and Destructor of type FSE_CTable
+    Note that its size depends on 'tableLog' and 'maxSymbolValue' */
+typedef unsigned FSE_CTable;   /* don't allocate that. It's just a way to be more restrictive than void* */
+typedef unsigned FSE_DTable;   /* don't allocate that. It's just a way to be more restrictive than void* */
+
+
+/* ******************************************************************
+   FSE : Finite State Entropy coder
+   header file for static linking (only)
+   Copyright (C) 2013-2015, Yann Collet
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+   You can contact the author at :
+   - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+   - Public forum : https://groups.google.com/forum/#!forum/lz4c
+****************************************************************** */
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+
+/******************************************
+*  Static allocation
+******************************************/
+/* FSE buffer bounds */
+#define FSE_NCOUNTBOUND 512
+#define FSE_BLOCKBOUND(size) (size + (size>>7))
+#define FSE_COMPRESSBOUND(size) (FSE_NCOUNTBOUND + FSE_BLOCKBOUND(size))   /* Macro version, useful for static allocation */
+
+/* You can statically allocate FSE CTable/DTable as a table of unsigned using below macro */
+#define FSE_CTABLE_SIZE_U32(maxTableLog, maxSymbolValue)   (1 + (1<<(maxTableLog-1)) + ((maxSymbolValue+1)*2))
+#define FSE_DTABLE_SIZE_U32(maxTableLog)                   (1 + (1<= BIT_DStream_completed
+
+When it's done, verify decompression is fully completed, by checking both DStream and the relevant states.
+Checking if DStream has reached its end is performed by :
+    BIT_endOfDStream(&DStream);
+Check also the states. There might be some symbols left there, if some high probability ones (>50%) are possible.
+    FSE_endOfDState(&DState);
+*/
+
+
+/******************************************
+*  FSE unsafe API
+******************************************/
+static unsigned char FSE_decodeSymbolFast(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD);
+/* faster, but works only if nbBits is always >= 1 (otherwise, result will be corrupted) */
+
+
+/******************************************
+*  Implementation of inline functions
+******************************************/
+
+/* decompression */
+
+typedef struct {
+    U16 tableLog;
+    U16 fastMode;
+} FSE_DTableHeader;   /* sizeof U32 */
+
+typedef struct
+{
+    unsigned short newState;
+    unsigned char  symbol;
+    unsigned char  nbBits;
+} FSE_decode_t;   /* size == U32 */
+
+MEM_STATIC void FSE_initDState(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD, const FSE_DTable* dt)
+{
+    FSE_DTableHeader DTableH;
+    memcpy(&DTableH, dt, sizeof(DTableH));
+    DStatePtr->state = BIT_readBits(bitD, DTableH.tableLog);
+    BIT_reloadDStream(bitD);
+    DStatePtr->table = dt + 1;
+}
+
+MEM_STATIC BYTE FSE_decodeSymbol(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD)
+{
+    const FSE_decode_t DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
+    const U32  nbBits = DInfo.nbBits;
+    BYTE symbol = DInfo.symbol;
+    size_t lowBits = BIT_readBits(bitD, nbBits);
+
+    DStatePtr->state = DInfo.newState + lowBits;
+    return symbol;
+}
+
+MEM_STATIC BYTE FSE_decodeSymbolFast(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD)
+{
+    const FSE_decode_t DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
+    const U32 nbBits = DInfo.nbBits;
+    BYTE symbol = DInfo.symbol;
+    size_t lowBits = BIT_readBitsFast(bitD, nbBits);
+
+    DStatePtr->state = DInfo.newState + lowBits;
+    return symbol;
+}
+
+MEM_STATIC unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr)
+{
+    return DStatePtr->state == 0;
+}
+
+
+#if defined (__cplusplus)
+}
+#endif
+/* ******************************************************************
+   Huff0 : Huffman coder, part of New Generation Entropy library
+   header file for static linking (only)
+   Copyright (C) 2013-2015, Yann Collet
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+   You can contact the author at :
+   - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+   - Public forum : https://groups.google.com/forum/#!forum/lz4c
+****************************************************************** */
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/******************************************
+*  Static allocation macros
+******************************************/
+/* Huff0 buffer bounds */
+#define HUF_CTABLEBOUND 129
+#define HUF_BLOCKBOUND(size) (size + (size>>8) + 8)   /* only true if incompressible pre-filtered with fast heuristic */
+#define HUF_COMPRESSBOUND(size) (HUF_CTABLEBOUND + HUF_BLOCKBOUND(size))   /* Macro version, useful for static allocation */
+
+/* static allocation of Huff0's DTable */
+#define HUF_DTABLE_SIZE(maxTableLog)   (1 + (1<   /* size_t */
+
+
+/* *************************************
+*  Version
+***************************************/
+#define ZSTD_VERSION_MAJOR    0    /* for breaking interface changes  */
+#define ZSTD_VERSION_MINOR    2    /* for new (non-breaking) interface capabilities */
+#define ZSTD_VERSION_RELEASE  2    /* for tweaks, bug-fixes, or development */
+#define ZSTD_VERSION_NUMBER  (ZSTD_VERSION_MAJOR *100*100 + ZSTD_VERSION_MINOR *100 + ZSTD_VERSION_RELEASE)
+
+
+/* *************************************
+*  Advanced functions
+***************************************/
+typedef struct ZSTD_CCtx_s ZSTD_CCtx;   /* incomplete type */
+
+#if defined (__cplusplus)
+}
+#endif
+/*
+    zstd - standard compression library
+    Header File for static linking only
+    Copyright (C) 2014-2015, Yann Collet.
+
+    BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following conditions are
+    met:
+    * Redistributions of source code must retain the above copyright
+    notice, this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above
+    copyright notice, this list of conditions and the following disclaimer
+    in the documentation and/or other materials provided with the
+    distribution.
+    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+    "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+    A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+    OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+    SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+    LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+    DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+    THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+    (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+    OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+    You can contact the author at :
+    - zstd source repository : https://github.com/Cyan4973/zstd
+    - ztsd public forum : https://groups.google.com/forum/#!forum/lz4c
+*/
+
+/* The objects defined into this file should be considered experimental.
+ * They are not labelled stable, as their prototype may change in the future.
+ * You can use them for tests, provide feedback, or if you can endure risk of future changes.
+ */
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/* *************************************
+*  Streaming functions
+***************************************/
+
+typedef struct ZSTD_DCtx_s ZSTD_DCtx;
+
+/*
+  Use above functions alternatively.
+  ZSTD_nextSrcSizeToDecompress() tells how much bytes to provide as 'srcSize' to ZSTD_decompressContinue().
+  ZSTD_decompressContinue() will use previous data blocks to improve compression if they are located prior to current block.
+  Result is the number of bytes regenerated within 'dst'.
+  It can be zero, which is not an error; it just means ZSTD_decompressContinue() has decoded some header.
+*/
+
+/* *************************************
+*  Prefix - version detection
+***************************************/
+#define ZSTD_magicNumber 0xFD2FB522   /* v0.2 (current)*/
+
+
+#if defined (__cplusplus)
+}
+#endif
+/* ******************************************************************
+   FSE : Finite State Entropy coder
+   Copyright (C) 2013-2015, Yann Collet.
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+    You can contact the author at :
+    - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
+    - Public forum : https://groups.google.com/forum/#!forum/lz4c
+****************************************************************** */
+
+#ifndef FSE_COMMONDEFS_ONLY
+
+/****************************************************************
+*  Tuning parameters
+****************************************************************/
+/* MEMORY_USAGE :
+*  Memory usage formula : N->2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; etc.)
+*  Increasing memory usage improves compression ratio
+*  Reduced memory usage can improve speed, due to cache effect
+*  Recommended max value is 14, for 16KB, which nicely fits into Intel x86 L1 cache */
+#define FSE_MAX_MEMORY_USAGE 14
+#define FSE_DEFAULT_MEMORY_USAGE 13
+
+/* FSE_MAX_SYMBOL_VALUE :
+*  Maximum symbol value authorized.
+*  Required for proper stack allocation */
+#define FSE_MAX_SYMBOL_VALUE 255
+
+
+/****************************************************************
+*  template functions type & suffix
+****************************************************************/
+#define FSE_FUNCTION_TYPE BYTE
+#define FSE_FUNCTION_EXTENSION
+
+
+/****************************************************************
+*  Byte symbol type
+****************************************************************/
+#endif   /* !FSE_COMMONDEFS_ONLY */
+
+
+/****************************************************************
+*  Compiler specifics
+****************************************************************/
+#ifdef _MSC_VER    /* Visual Studio */
+#  define FORCE_INLINE static __forceinline
+#  include                     /* For Visual 2005 */
+#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
+#  pragma warning(disable : 4214)        /* disable: C4214: non-int bitfields */
+#else
+#  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+#    ifdef __GNUC__
+#      define FORCE_INLINE static inline __attribute__((always_inline))
+#    else
+#      define FORCE_INLINE static inline
+#    endif
+#  else
+#    define FORCE_INLINE static
+#  endif /* __STDC_VERSION__ */
+#endif
+
+
+/****************************************************************
+*  Includes
+****************************************************************/
+#include      /* malloc, free, qsort */
+#include      /* memcpy, memset */
+#include       /* printf (debug) */
+
+/****************************************************************
+*  Constants
+*****************************************************************/
+#define FSE_MAX_TABLELOG  (FSE_MAX_MEMORY_USAGE-2)
+#define FSE_MAX_TABLESIZE (1U< FSE_TABLELOG_ABSOLUTE_MAX
+#error "FSE_MAX_TABLELOG > FSE_TABLELOG_ABSOLUTE_MAX is not supported"
+#endif
+
+
+/****************************************************************
+*  Error Management
+****************************************************************/
+#define FSE_STATIC_ASSERT(c) { enum { FSE_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
+
+
+/****************************************************************
+*  Complex types
+****************************************************************/
+typedef U32 DTable_max_t[FSE_DTABLE_SIZE_U32(FSE_MAX_TABLELOG)];
+
+
+/****************************************************************
+*  Templates
+****************************************************************/
+/*
+  designed to be included
+  for type-specific functions (template emulation in C)
+  Objective is to write these functions only once, for improved maintenance
+*/
+
+/* safety checks */
+#ifndef FSE_FUNCTION_EXTENSION
+#  error "FSE_FUNCTION_EXTENSION must be defined"
+#endif
+#ifndef FSE_FUNCTION_TYPE
+#  error "FSE_FUNCTION_TYPE must be defined"
+#endif
+
+/* Function names */
+#define FSE_CAT(X,Y) X##Y
+#define FSE_FUNCTION_NAME(X,Y) FSE_CAT(X,Y)
+#define FSE_TYPE_NAME(X,Y) FSE_CAT(X,Y)
+
+
+/* Function templates */
+
+#define FSE_DECODE_TYPE FSE_decode_t
+
+static U32 FSE_tableStep(U32 tableSize) { return (tableSize>>1) + (tableSize>>3) + 3; }
+
+static size_t FSE_buildDTable
+(FSE_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
+{
+    void* ptr = dt+1;
+    FSE_DECODE_TYPE* const tableDecode = (FSE_DECODE_TYPE*)ptr;
+    FSE_DTableHeader DTableH;
+    const U32 tableSize = 1 << tableLog;
+    const U32 tableMask = tableSize-1;
+    const U32 step = FSE_tableStep(tableSize);
+    U16 symbolNext[FSE_MAX_SYMBOL_VALUE+1];
+    U32 position = 0;
+    U32 highThreshold = tableSize-1;
+    const S16 largeLimit= (S16)(1 << (tableLog-1));
+    U32 noLarge = 1;
+    U32 s;
+
+    /* Sanity Checks */
+    if (maxSymbolValue > FSE_MAX_SYMBOL_VALUE) return ERROR(maxSymbolValue_tooLarge);
+    if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge);
+
+    /* Init, lay down lowprob symbols */
+    DTableH.tableLog = (U16)tableLog;
+    for (s=0; s<=maxSymbolValue; s++)
+    {
+        if (normalizedCounter[s]==-1)
+        {
+            tableDecode[highThreshold--].symbol = (FSE_FUNCTION_TYPE)s;
+            symbolNext[s] = 1;
+        }
+        else
+        {
+            if (normalizedCounter[s] >= largeLimit) noLarge=0;
+            symbolNext[s] = normalizedCounter[s];
+        }
+    }
+
+    /* Spread symbols */
+    for (s=0; s<=maxSymbolValue; s++)
+    {
+        int i;
+        for (i=0; i highThreshold) position = (position + step) & tableMask;   /* lowprob area */
+        }
+    }
+
+    if (position!=0) return ERROR(GENERIC);   /* position must reach all cells once, otherwise normalizedCounter is incorrect */
+
+    /* Build Decoding table */
+    {
+        U32 i;
+        for (i=0; i FSE_TABLELOG_ABSOLUTE_MAX) return ERROR(tableLog_tooLarge);
+    bitStream >>= 4;
+    bitCount = 4;
+    *tableLogPtr = nbBits;
+    remaining = (1<1) && (charnum<=*maxSVPtr))
+    {
+        if (previous0)
+        {
+            unsigned n0 = charnum;
+            while ((bitStream & 0xFFFF) == 0xFFFF)
+            {
+                n0+=24;
+                if (ip < iend-5)
+                {
+                    ip+=2;
+                    bitStream = MEM_readLE32(ip) >> bitCount;
+                }
+                else
+                {
+                    bitStream >>= 16;
+                    bitCount+=16;
+                }
+            }
+            while ((bitStream & 3) == 3)
+            {
+                n0+=3;
+                bitStream>>=2;
+                bitCount+=2;
+            }
+            n0 += bitStream & 3;
+            bitCount += 2;
+            if (n0 > *maxSVPtr) return ERROR(maxSymbolValue_tooSmall);
+            while (charnum < n0) normalizedCounter[charnum++] = 0;
+            if ((ip <= iend-7) || (ip + (bitCount>>3) <= iend-4))
+            {
+                ip += bitCount>>3;
+                bitCount &= 7;
+                bitStream = MEM_readLE32(ip) >> bitCount;
+            }
+            else
+                bitStream >>= 2;
+        }
+        {
+            const short max = (short)((2*threshold-1)-remaining);
+            short count;
+
+            if ((bitStream & (threshold-1)) < (U32)max)
+            {
+                count = (short)(bitStream & (threshold-1));
+                bitCount   += nbBits-1;
+            }
+            else
+            {
+                count = (short)(bitStream & (2*threshold-1));
+                if (count >= threshold) count -= max;
+                bitCount   += nbBits;
+            }
+
+            count--;   /* extra accuracy */
+            remaining -= FSE_abs(count);
+            normalizedCounter[charnum++] = count;
+            previous0 = !count;
+            while (remaining < threshold)
+            {
+                nbBits--;
+                threshold >>= 1;
+            }
+
+            {
+                if ((ip <= iend-7) || (ip + (bitCount>>3) <= iend-4))
+                {
+                    ip += bitCount>>3;
+                    bitCount &= 7;
+                }
+                else
+                {
+                    bitCount -= (int)(8 * (iend - 4 - ip));
+					ip = iend - 4;
+				}
+                bitStream = MEM_readLE32(ip) >> (bitCount & 31);
+            }
+        }
+    }
+    if (remaining != 1) return ERROR(GENERIC);
+    *maxSVPtr = charnum-1;
+
+    ip += (bitCount+7)>>3;
+    if ((size_t)(ip-istart) > hbSize) return ERROR(srcSize_wrong);
+    return ip-istart;
+}
+
+
+/*********************************************************
+*  Decompression (Byte symbols)
+*********************************************************/
+static size_t FSE_buildDTable_rle (FSE_DTable* dt, BYTE symbolValue)
+{
+    void* ptr = dt;
+    FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)ptr;
+    FSE_decode_t* const cell = (FSE_decode_t*)(ptr) + 1;   /* because dt is unsigned */
+
+    DTableH->tableLog = 0;
+    DTableH->fastMode = 0;
+
+    cell->newState = 0;
+    cell->symbol = symbolValue;
+    cell->nbBits = 0;
+
+    return 0;
+}
+
+
+static size_t FSE_buildDTable_raw (FSE_DTable* dt, unsigned nbBits)
+{
+    void* ptr = dt;
+    FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)ptr;
+    FSE_decode_t* const dinfo = (FSE_decode_t*)(ptr) + 1;   /* because dt is unsigned */
+    const unsigned tableSize = 1 << nbBits;
+    const unsigned tableMask = tableSize - 1;
+    const unsigned maxSymbolValue = tableMask;
+    unsigned s;
+
+    /* Sanity checks */
+    if (nbBits < 1) return ERROR(GENERIC);         /* min size */
+
+    /* Build Decoding Table */
+    DTableH->tableLog = (U16)nbBits;
+    DTableH->fastMode = 1;
+    for (s=0; s<=maxSymbolValue; s++)
+    {
+        dinfo[s].newState = 0;
+        dinfo[s].symbol = (BYTE)s;
+        dinfo[s].nbBits = (BYTE)nbBits;
+    }
+
+    return 0;
+}
+
+FORCE_INLINE size_t FSE_decompress_usingDTable_generic(
+          void* dst, size_t maxDstSize,
+    const void* cSrc, size_t cSrcSize,
+    const FSE_DTable* dt, const unsigned fast)
+{
+    BYTE* const ostart = (BYTE*) dst;
+    BYTE* op = ostart;
+    BYTE* const omax = op + maxDstSize;
+    BYTE* const olimit = omax-3;
+
+    BIT_DStream_t bitD;
+    FSE_DState_t state1;
+    FSE_DState_t state2;
+    size_t errorCode;
+
+    /* Init */
+    errorCode = BIT_initDStream(&bitD, cSrc, cSrcSize);   /* replaced last arg by maxCompressed Size */
+    if (FSE_isError(errorCode)) return errorCode;
+
+    FSE_initDState(&state1, &bitD, dt);
+    FSE_initDState(&state2, &bitD, dt);
+
+#define FSE_GETSYMBOL(statePtr) fast ? FSE_decodeSymbolFast(statePtr, &bitD) : FSE_decodeSymbol(statePtr, &bitD)
+
+    /* 4 symbols per loop */
+    for ( ; (BIT_reloadDStream(&bitD)==BIT_DStream_unfinished) && (op sizeof(bitD.bitContainer)*8)    /* This test must be static */
+            BIT_reloadDStream(&bitD);
+
+        op[1] = FSE_GETSYMBOL(&state2);
+
+        if (FSE_MAX_TABLELOG*4+7 > sizeof(bitD.bitContainer)*8)    /* This test must be static */
+            { if (BIT_reloadDStream(&bitD) > BIT_DStream_unfinished) { op+=2; break; } }
+
+        op[2] = FSE_GETSYMBOL(&state1);
+
+        if (FSE_MAX_TABLELOG*2+7 > sizeof(bitD.bitContainer)*8)    /* This test must be static */
+            BIT_reloadDStream(&bitD);
+
+        op[3] = FSE_GETSYMBOL(&state2);
+    }
+
+    /* tail */
+    /* note : BIT_reloadDStream(&bitD) >= FSE_DStream_partiallyFilled; Ends at exactly BIT_DStream_completed */
+    while (1)
+    {
+        if ( (BIT_reloadDStream(&bitD)>BIT_DStream_completed) || (op==omax) || (BIT_endOfDStream(&bitD) && (fast || FSE_endOfDState(&state1))) )
+            break;
+
+        *op++ = FSE_GETSYMBOL(&state1);
+
+        if ( (BIT_reloadDStream(&bitD)>BIT_DStream_completed) || (op==omax) || (BIT_endOfDStream(&bitD) && (fast || FSE_endOfDState(&state2))) )
+            break;
+
+        *op++ = FSE_GETSYMBOL(&state2);
+    }
+
+    /* end ? */
+    if (BIT_endOfDStream(&bitD) && FSE_endOfDState(&state1) && FSE_endOfDState(&state2))
+        return op-ostart;
+
+    if (op==omax) return ERROR(dstSize_tooSmall);   /* dst buffer is full, but cSrc unfinished */
+
+    return ERROR(corruption_detected);
+}
+
+
+static size_t FSE_decompress_usingDTable(void* dst, size_t originalSize,
+                            const void* cSrc, size_t cSrcSize,
+                            const FSE_DTable* dt)
+{
+    FSE_DTableHeader DTableH;
+    memcpy(&DTableH, dt, sizeof(DTableH));
+
+    /* select fast mode (static) */
+    if (DTableH.fastMode) return FSE_decompress_usingDTable_generic(dst, originalSize, cSrc, cSrcSize, dt, 1);
+    return FSE_decompress_usingDTable_generic(dst, originalSize, cSrc, cSrcSize, dt, 0);
+}
+
+
+static size_t FSE_decompress(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize)
+{
+    const BYTE* const istart = (const BYTE*)cSrc;
+    const BYTE* ip = istart;
+    short counting[FSE_MAX_SYMBOL_VALUE+1];
+    DTable_max_t dt;   /* Static analyzer seems unable to understand this table will be properly initialized later */
+    unsigned tableLog;
+    unsigned maxSymbolValue = FSE_MAX_SYMBOL_VALUE;
+    size_t errorCode;
+
+    if (cSrcSize<2) return ERROR(srcSize_wrong);   /* too small input size */
+
+    /* normal FSE decoding mode */
+    errorCode = FSE_readNCount (counting, &maxSymbolValue, &tableLog, istart, cSrcSize);
+    if (FSE_isError(errorCode)) return errorCode;
+    if (errorCode >= cSrcSize) return ERROR(srcSize_wrong);   /* too small input size */
+    ip += errorCode;
+    cSrcSize -= errorCode;
+
+    errorCode = FSE_buildDTable (dt, counting, maxSymbolValue, tableLog);
+    if (FSE_isError(errorCode)) return errorCode;
+
+    /* always return, even if it is an error code */
+    return FSE_decompress_usingDTable (dst, maxDstSize, ip, cSrcSize, dt);
+}
+
+
+
+#endif   /* FSE_COMMONDEFS_ONLY */
+/* ******************************************************************
+   Huff0 : Huffman coder, part of New Generation Entropy library
+   Copyright (C) 2013-2015, Yann Collet.
+
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+    You can contact the author at :
+    - FSE+Huff0 source repository : https://github.com/Cyan4973/FiniteStateEntropy
+    - Public forum : https://groups.google.com/forum/#!forum/lz4c
+****************************************************************** */
+
+/****************************************************************
+*  Compiler specifics
+****************************************************************/
+#if defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
+/* inline is defined */
+#elif defined(_MSC_VER)
+#  define inline __inline
+#else
+#  define inline /* disable inline */
+#endif
+
+
+#ifdef _MSC_VER    /* Visual Studio */
+#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
+#endif
+
+
+/****************************************************************
+*  Includes
+****************************************************************/
+#include      /* malloc, free, qsort */
+#include      /* memcpy, memset */
+#include       /* printf (debug) */
+
+/****************************************************************
+*  Error Management
+****************************************************************/
+#define HUF_STATIC_ASSERT(c) { enum { HUF_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
+
+
+/******************************************
+*  Helper functions
+******************************************/
+static unsigned HUF_isError(size_t code) { return ERR_isError(code); }
+
+#define HUF_ABSOLUTEMAX_TABLELOG  16   /* absolute limit of HUF_MAX_TABLELOG. Beyond that value, code does not work */
+#define HUF_MAX_TABLELOG  12           /* max configured tableLog (for static allocation); can be modified up to HUF_ABSOLUTEMAX_TABLELOG */
+#define HUF_DEFAULT_TABLELOG  HUF_MAX_TABLELOG   /* tableLog by default, when not specified */
+#define HUF_MAX_SYMBOL_VALUE 255
+#if (HUF_MAX_TABLELOG > HUF_ABSOLUTEMAX_TABLELOG)
+#  error "HUF_MAX_TABLELOG is too large !"
+#endif
+
+
+
+/*********************************************************
+*  Huff0 : Huffman block decompression
+*********************************************************/
+typedef struct { BYTE byte; BYTE nbBits; } HUF_DEltX2;   /* single-symbol decoding */
+
+typedef struct { U16 sequence; BYTE nbBits; BYTE length; } HUF_DEltX4;  /* double-symbols decoding */
+
+typedef struct { BYTE symbol; BYTE weight; } sortedSymbol_t;
+
+/*! HUF_readStats
+    Read compact Huffman tree, saved by HUF_writeCTable
+    @huffWeight : destination buffer
+    @return : size read from `src`
+*/
+static size_t HUF_readStats(BYTE* huffWeight, size_t hwSize, U32* rankStats,
+                            U32* nbSymbolsPtr, U32* tableLogPtr,
+                            const void* src, size_t srcSize)
+{
+    U32 weightTotal;
+    U32 tableLog;
+    const BYTE* ip = (const BYTE*) src;
+    size_t iSize;
+    size_t oSize;
+    U32 n;
+
+    if (!srcSize) return ERROR(srcSize_wrong);
+    iSize = ip[0];
+    //memset(huffWeight, 0, hwSize);   /* is not necessary, even though some analyzer complain ... */
+
+    if (iSize >= 128)  /* special header */
+    {
+        if (iSize >= (242))   /* RLE */
+        {
+            static int l[14] = { 1, 2, 3, 4, 7, 8, 15, 16, 31, 32, 63, 64, 127, 128 };
+            oSize = l[iSize-242];
+            memset(huffWeight, 1, hwSize);
+            iSize = 0;
+        }
+        else   /* Incompressible */
+        {
+            oSize = iSize - 127;
+            iSize = ((oSize+1)/2);
+            if (iSize+1 > srcSize) return ERROR(srcSize_wrong);
+            if (oSize >= hwSize) return ERROR(corruption_detected);
+            ip += 1;
+            for (n=0; n> 4;
+                huffWeight[n+1] = ip[n/2] & 15;
+            }
+        }
+    }
+    else  /* header compressed with FSE (normal case) */
+    {
+        if (iSize+1 > srcSize) return ERROR(srcSize_wrong);
+        oSize = FSE_decompress(huffWeight, hwSize-1, ip+1, iSize);   /* max (hwSize-1) values decoded, as last one is implied */
+        if (FSE_isError(oSize)) return oSize;
+    }
+
+    /* collect weight stats */
+    memset(rankStats, 0, (HUF_ABSOLUTEMAX_TABLELOG + 1) * sizeof(U32));
+    weightTotal = 0;
+    for (n=0; n= HUF_ABSOLUTEMAX_TABLELOG) return ERROR(corruption_detected);
+        rankStats[huffWeight[n]]++;
+        weightTotal += (1 << huffWeight[n]) >> 1;
+    }
+    if (weightTotal == 0) return ERROR(corruption_detected);
+
+    /* get last non-null symbol weight (implied, total must be 2^n) */
+    tableLog = BIT_highbit32(weightTotal) + 1;
+    if (tableLog > HUF_ABSOLUTEMAX_TABLELOG) return ERROR(corruption_detected);
+    {
+        U32 total = 1 << tableLog;
+        U32 rest = total - weightTotal;
+        U32 verif = 1 << BIT_highbit32(rest);
+        U32 lastWeight = BIT_highbit32(rest) + 1;
+        if (verif != rest) return ERROR(corruption_detected);    /* last value must be a clean power of 2 */
+        huffWeight[oSize] = (BYTE)lastWeight;
+        rankStats[lastWeight]++;
+    }
+
+    /* check tree construction validity */
+    if ((rankStats[1] < 2) || (rankStats[1] & 1)) return ERROR(corruption_detected);   /* by construction : at least 2 elts of rank 1, must be even */
+
+    /* results */
+    *nbSymbolsPtr = (U32)(oSize+1);
+    *tableLogPtr = tableLog;
+    return iSize+1;
+}
+
+
+/**************************/
+/* single-symbol decoding */
+/**************************/
+
+static size_t HUF_readDTableX2 (U16* DTable, const void* src, size_t srcSize)
+{
+    BYTE huffWeight[HUF_MAX_SYMBOL_VALUE + 1];
+    U32 rankVal[HUF_ABSOLUTEMAX_TABLELOG + 1];   /* large enough for values from 0 to 16 */
+    U32 tableLog = 0;
+    const BYTE* ip = (const BYTE*) src;
+    size_t iSize = ip[0];
+    U32 nbSymbols = 0;
+    U32 n;
+    U32 nextRankStart;
+    void* ptr = DTable+1;
+    HUF_DEltX2* const dt = (HUF_DEltX2*)ptr;
+
+    HUF_STATIC_ASSERT(sizeof(HUF_DEltX2) == sizeof(U16));   /* if compilation fails here, assertion is false */
+    //memset(huffWeight, 0, sizeof(huffWeight));   /* is not necessary, even though some analyzer complain ... */
+
+    iSize = HUF_readStats(huffWeight, HUF_MAX_SYMBOL_VALUE + 1, rankVal, &nbSymbols, &tableLog, src, srcSize);
+    if (HUF_isError(iSize)) return iSize;
+
+    /* check result */
+    if (tableLog > DTable[0]) return ERROR(tableLog_tooLarge);   /* DTable is too small */
+    DTable[0] = (U16)tableLog;   /* maybe should separate sizeof DTable, as allocated, from used size of DTable, in case of DTable re-use */
+
+    /* Prepare ranks */
+    nextRankStart = 0;
+    for (n=1; n<=tableLog; n++)
+    {
+        U32 current = nextRankStart;
+        nextRankStart += (rankVal[n] << (n-1));
+        rankVal[n] = current;
+    }
+
+    /* fill DTable */
+    for (n=0; n> 1;
+        U32 i;
+        HUF_DEltX2 D;
+        D.byte = (BYTE)n; D.nbBits = (BYTE)(tableLog + 1 - w);
+        for (i = rankVal[w]; i < rankVal[w] + length; i++)
+            dt[i] = D;
+        rankVal[w] += length;
+    }
+
+    return iSize;
+}
+
+static BYTE HUF_decodeSymbolX2(BIT_DStream_t* Dstream, const HUF_DEltX2* dt, const U32 dtLog)
+{
+        const size_t val = BIT_lookBitsFast(Dstream, dtLog); /* note : dtLog >= 1 */
+        const BYTE c = dt[val].byte;
+        BIT_skipBits(Dstream, dt[val].nbBits);
+        return c;
+}
+
+#define HUF_DECODE_SYMBOLX2_0(ptr, DStreamPtr) \
+    *ptr++ = HUF_decodeSymbolX2(DStreamPtr, dt, dtLog)
+
+#define HUF_DECODE_SYMBOLX2_1(ptr, DStreamPtr) \
+    if (MEM_64bits() || (HUF_MAX_TABLELOG<=12)) \
+        HUF_DECODE_SYMBOLX2_0(ptr, DStreamPtr)
+
+#define HUF_DECODE_SYMBOLX2_2(ptr, DStreamPtr) \
+    if (MEM_64bits()) \
+        HUF_DECODE_SYMBOLX2_0(ptr, DStreamPtr)
+
+static inline size_t HUF_decodeStreamX2(BYTE* p, BIT_DStream_t* const bitDPtr, BYTE* const pEnd, const HUF_DEltX2* const dt, const U32 dtLog)
+{
+    BYTE* const pStart = p;
+
+    /* up to 4 symbols at a time */
+    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) && (p <= pEnd-4))
+    {
+        HUF_DECODE_SYMBOLX2_2(p, bitDPtr);
+        HUF_DECODE_SYMBOLX2_1(p, bitDPtr);
+        HUF_DECODE_SYMBOLX2_2(p, bitDPtr);
+        HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+    }
+
+    /* closer to the end */
+    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) && (p < pEnd))
+        HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+
+    /* no more data to retrieve from bitstream, hence no need to reload */
+    while (p < pEnd)
+        HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+
+    return pEnd-pStart;
+}
+
+
+static size_t HUF_decompress4X2_usingDTable(
+          void* dst,  size_t dstSize,
+    const void* cSrc, size_t cSrcSize,
+    const U16* DTable)
+{
+    if (cSrcSize < 10) return ERROR(corruption_detected);   /* strict minimum : jump table + 1 byte per stream */
+
+    {
+        const BYTE* const istart = (const BYTE*) cSrc;
+        BYTE* const ostart = (BYTE*) dst;
+        BYTE* const oend = ostart + dstSize;
+
+        const void* ptr = DTable;
+        const HUF_DEltX2* const dt = ((const HUF_DEltX2*)ptr) +1;
+        const U32 dtLog = DTable[0];
+        size_t errorCode;
+
+        /* Init */
+        BIT_DStream_t bitD1;
+        BIT_DStream_t bitD2;
+        BIT_DStream_t bitD3;
+        BIT_DStream_t bitD4;
+        const size_t length1 = MEM_readLE16(istart);
+        const size_t length2 = MEM_readLE16(istart+2);
+        const size_t length3 = MEM_readLE16(istart+4);
+        size_t length4;
+        const BYTE* const istart1 = istart + 6;  /* jumpTable */
+        const BYTE* const istart2 = istart1 + length1;
+        const BYTE* const istart3 = istart2 + length2;
+        const BYTE* const istart4 = istart3 + length3;
+        const size_t segmentSize = (dstSize+3) / 4;
+        BYTE* const opStart2 = ostart + segmentSize;
+        BYTE* const opStart3 = opStart2 + segmentSize;
+        BYTE* const opStart4 = opStart3 + segmentSize;
+        BYTE* op1 = ostart;
+        BYTE* op2 = opStart2;
+        BYTE* op3 = opStart3;
+        BYTE* op4 = opStart4;
+        U32 endSignal;
+
+        length4 = cSrcSize - (length1 + length2 + length3 + 6);
+        if (length4 > cSrcSize) return ERROR(corruption_detected);   /* overflow */
+        errorCode = BIT_initDStream(&bitD1, istart1, length1);
+        if (HUF_isError(errorCode)) return errorCode;
+        errorCode = BIT_initDStream(&bitD2, istart2, length2);
+        if (HUF_isError(errorCode)) return errorCode;
+        errorCode = BIT_initDStream(&bitD3, istart3, length3);
+        if (HUF_isError(errorCode)) return errorCode;
+        errorCode = BIT_initDStream(&bitD4, istart4, length4);
+        if (HUF_isError(errorCode)) return errorCode;
+
+        /* 16-32 symbols per loop (4-8 symbols per stream) */
+        endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
+        for ( ; (endSignal==BIT_DStream_unfinished) && (op4<(oend-7)) ; )
+        {
+            HUF_DECODE_SYMBOLX2_2(op1, &bitD1);
+            HUF_DECODE_SYMBOLX2_2(op2, &bitD2);
+            HUF_DECODE_SYMBOLX2_2(op3, &bitD3);
+            HUF_DECODE_SYMBOLX2_2(op4, &bitD4);
+            HUF_DECODE_SYMBOLX2_1(op1, &bitD1);
+            HUF_DECODE_SYMBOLX2_1(op2, &bitD2);
+            HUF_DECODE_SYMBOLX2_1(op3, &bitD3);
+            HUF_DECODE_SYMBOLX2_1(op4, &bitD4);
+            HUF_DECODE_SYMBOLX2_2(op1, &bitD1);
+            HUF_DECODE_SYMBOLX2_2(op2, &bitD2);
+            HUF_DECODE_SYMBOLX2_2(op3, &bitD3);
+            HUF_DECODE_SYMBOLX2_2(op4, &bitD4);
+            HUF_DECODE_SYMBOLX2_0(op1, &bitD1);
+            HUF_DECODE_SYMBOLX2_0(op2, &bitD2);
+            HUF_DECODE_SYMBOLX2_0(op3, &bitD3);
+            HUF_DECODE_SYMBOLX2_0(op4, &bitD4);
+
+            endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
+        }
+
+        /* check corruption */
+        if (op1 > opStart2) return ERROR(corruption_detected);
+        if (op2 > opStart3) return ERROR(corruption_detected);
+        if (op3 > opStart4) return ERROR(corruption_detected);
+        /* note : op4 supposed already verified within main loop */
+
+        /* finish bitStreams one by one */
+        HUF_decodeStreamX2(op1, &bitD1, opStart2, dt, dtLog);
+        HUF_decodeStreamX2(op2, &bitD2, opStart3, dt, dtLog);
+        HUF_decodeStreamX2(op3, &bitD3, opStart4, dt, dtLog);
+        HUF_decodeStreamX2(op4, &bitD4, oend,     dt, dtLog);
+
+        /* check */
+        endSignal = BIT_endOfDStream(&bitD1) & BIT_endOfDStream(&bitD2) & BIT_endOfDStream(&bitD3) & BIT_endOfDStream(&bitD4);
+        if (!endSignal) return ERROR(corruption_detected);
+
+        /* decoded size */
+        return dstSize;
+    }
+}
+
+
+static size_t HUF_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_MAX_TABLELOG);
+    const BYTE* ip = (const BYTE*) cSrc;
+    size_t errorCode;
+
+    errorCode = HUF_readDTableX2 (DTable, cSrc, cSrcSize);
+    if (HUF_isError(errorCode)) return errorCode;
+    if (errorCode >= cSrcSize) return ERROR(srcSize_wrong);
+    ip += errorCode;
+    cSrcSize -= errorCode;
+
+    return HUF_decompress4X2_usingDTable (dst, dstSize, ip, cSrcSize, DTable);
+}
+
+
+/***************************/
+/* double-symbols decoding */
+/***************************/
+
+static void HUF_fillDTableX4Level2(HUF_DEltX4* DTable, U32 sizeLog, const U32 consumed,
+                           const U32* rankValOrigin, const int minWeight,
+                           const sortedSymbol_t* sortedSymbols, const U32 sortedListSize,
+                           U32 nbBitsBaseline, U16 baseSeq)
+{
+    HUF_DEltX4 DElt;
+    U32 rankVal[HUF_ABSOLUTEMAX_TABLELOG + 1];
+    U32 s;
+
+    /* get pre-calculated rankVal */
+    memcpy(rankVal, rankValOrigin, sizeof(rankVal));
+
+    /* fill skipped values */
+    if (minWeight>1)
+    {
+        U32 i, skipSize = rankVal[minWeight];
+        MEM_writeLE16(&(DElt.sequence), baseSeq);
+        DElt.nbBits   = (BYTE)(consumed);
+        DElt.length   = 1;
+        for (i = 0; i < skipSize; i++)
+            DTable[i] = DElt;
+    }
+
+    /* fill DTable */
+    for (s=0; s= 1 */
+
+        rankVal[weight] += length;
+    }
+}
+
+typedef U32 rankVal_t[HUF_ABSOLUTEMAX_TABLELOG][HUF_ABSOLUTEMAX_TABLELOG + 1];
+
+static void HUF_fillDTableX4(HUF_DEltX4* DTable, const U32 targetLog,
+                           const sortedSymbol_t* sortedList, const U32 sortedListSize,
+                           const U32* rankStart, rankVal_t rankValOrigin, const U32 maxWeight,
+                           const U32 nbBitsBaseline)
+{
+    U32 rankVal[HUF_ABSOLUTEMAX_TABLELOG + 1];
+    const int scaleLog = nbBitsBaseline - targetLog;   /* note : targetLog >= srcLog, hence scaleLog <= 1 */
+    const U32 minBits  = nbBitsBaseline - maxWeight;
+    U32 s;
+
+    memcpy(rankVal, rankValOrigin, sizeof(rankVal));
+
+    /* fill DTable */
+    for (s=0; s= minBits)   /* enough room for a second symbol */
+        {
+            U32 sortedRank;
+            int minWeight = nbBits + scaleLog;
+            if (minWeight < 1) minWeight = 1;
+            sortedRank = rankStart[minWeight];
+            HUF_fillDTableX4Level2(DTable+start, targetLog-nbBits, nbBits,
+                           rankValOrigin[nbBits], minWeight,
+                           sortedList+sortedRank, sortedListSize-sortedRank,
+                           nbBitsBaseline, symbol);
+        }
+        else
+        {
+            U32 i;
+            const U32 end = start + length;
+            HUF_DEltX4 DElt;
+
+            MEM_writeLE16(&(DElt.sequence), symbol);
+            DElt.nbBits   = (BYTE)(nbBits);
+            DElt.length   = 1;
+            for (i = start; i < end; i++)
+                DTable[i] = DElt;
+        }
+        rankVal[weight] += length;
+    }
+}
+
+static size_t HUF_readDTableX4 (U32* DTable, const void* src, size_t srcSize)
+{
+    BYTE weightList[HUF_MAX_SYMBOL_VALUE + 1];
+    sortedSymbol_t sortedSymbol[HUF_MAX_SYMBOL_VALUE + 1];
+    U32 rankStats[HUF_ABSOLUTEMAX_TABLELOG + 1] = { 0 };
+    U32 rankStart0[HUF_ABSOLUTEMAX_TABLELOG + 2] = { 0 };
+    U32* const rankStart = rankStart0+1;
+    rankVal_t rankVal;
+    U32 tableLog, maxW, sizeOfSort, nbSymbols;
+    const U32 memLog = DTable[0];
+    const BYTE* ip = (const BYTE*) src;
+    size_t iSize = ip[0];
+    void* ptr = DTable;
+    HUF_DEltX4* const dt = ((HUF_DEltX4*)ptr) + 1;
+
+    HUF_STATIC_ASSERT(sizeof(HUF_DEltX4) == sizeof(U32));   /* if compilation fails here, assertion is false */
+    if (memLog > HUF_ABSOLUTEMAX_TABLELOG) return ERROR(tableLog_tooLarge);
+    //memset(weightList, 0, sizeof(weightList));   /* is not necessary, even though some analyzer complain ... */
+
+    iSize = HUF_readStats(weightList, HUF_MAX_SYMBOL_VALUE + 1, rankStats, &nbSymbols, &tableLog, src, srcSize);
+    if (HUF_isError(iSize)) return iSize;
+
+    /* check result */
+    if (tableLog > memLog) return ERROR(tableLog_tooLarge);   /* DTable can't fit code depth */
+
+    /* find maxWeight */
+    for (maxW = tableLog; rankStats[maxW]==0; maxW--)
+        {if (!maxW) return ERROR(GENERIC); }  /* necessarily finds a solution before maxW==0 */
+
+    /* Get start index of each weight */
+    {
+        U32 w, nextRankStart = 0;
+        for (w=1; w<=maxW; w++)
+        {
+            U32 current = nextRankStart;
+            nextRankStart += rankStats[w];
+            rankStart[w] = current;
+        }
+        rankStart[0] = nextRankStart;   /* put all 0w symbols at the end of sorted list*/
+        sizeOfSort = nextRankStart;
+    }
+
+    /* sort symbols by weight */
+    {
+        U32 s;
+        for (s=0; s> consumed;
+            }
+        }
+    }
+
+    HUF_fillDTableX4(dt, memLog,
+                   sortedSymbol, sizeOfSort,
+                   rankStart0, rankVal, maxW,
+                   tableLog+1);
+
+    return iSize;
+}
+
+
+static U32 HUF_decodeSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DEltX4* dt, const U32 dtLog)
+{
+    const size_t val = BIT_lookBitsFast(DStream, dtLog);   /* note : dtLog >= 1 */
+    memcpy(op, dt+val, 2);
+    BIT_skipBits(DStream, dt[val].nbBits);
+    return dt[val].length;
+}
+
+static U32 HUF_decodeLastSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DEltX4* dt, const U32 dtLog)
+{
+    const size_t val = BIT_lookBitsFast(DStream, dtLog);   /* note : dtLog >= 1 */
+    memcpy(op, dt+val, 1);
+    if (dt[val].length==1) BIT_skipBits(DStream, dt[val].nbBits);
+    else
+    {
+        if (DStream->bitsConsumed < (sizeof(DStream->bitContainer)*8))
+        {
+            BIT_skipBits(DStream, dt[val].nbBits);
+            if (DStream->bitsConsumed > (sizeof(DStream->bitContainer)*8))
+                DStream->bitsConsumed = (sizeof(DStream->bitContainer)*8);   /* ugly hack; works only because it's the last symbol. Note : can't easily extract nbBits from just this symbol */
+        }
+    }
+    return 1;
+}
+
+
+#define HUF_DECODE_SYMBOLX4_0(ptr, DStreamPtr) \
+    ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
+
+#define HUF_DECODE_SYMBOLX4_1(ptr, DStreamPtr) \
+    if (MEM_64bits() || (HUF_MAX_TABLELOG<=12)) \
+        ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
+
+#define HUF_DECODE_SYMBOLX4_2(ptr, DStreamPtr) \
+    if (MEM_64bits()) \
+        ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
+
+static inline size_t HUF_decodeStreamX4(BYTE* p, BIT_DStream_t* bitDPtr, BYTE* const pEnd, const HUF_DEltX4* const dt, const U32 dtLog)
+{
+    BYTE* const pStart = p;
+
+    /* up to 8 symbols at a time */
+    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) && (p < pEnd-7))
+    {
+        HUF_DECODE_SYMBOLX4_2(p, bitDPtr);
+        HUF_DECODE_SYMBOLX4_1(p, bitDPtr);
+        HUF_DECODE_SYMBOLX4_2(p, bitDPtr);
+        HUF_DECODE_SYMBOLX4_0(p, bitDPtr);
+    }
+
+    /* closer to the end */
+    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) && (p <= pEnd-2))
+        HUF_DECODE_SYMBOLX4_0(p, bitDPtr);
+
+    while (p <= pEnd-2)
+        HUF_DECODE_SYMBOLX4_0(p, bitDPtr);   /* no need to reload : reached the end of DStream */
+
+    if (p < pEnd)
+        p += HUF_decodeLastSymbolX4(p, bitDPtr, dt, dtLog);
+
+    return p-pStart;
+}
+
+
+
+static size_t HUF_decompress4X4_usingDTable(
+          void* dst,  size_t dstSize,
+    const void* cSrc, size_t cSrcSize,
+    const U32* DTable)
+{
+    if (cSrcSize < 10) return ERROR(corruption_detected);   /* strict minimum : jump table + 1 byte per stream */
+
+    {
+        const BYTE* const istart = (const BYTE*) cSrc;
+        BYTE* const ostart = (BYTE*) dst;
+        BYTE* const oend = ostart + dstSize;
+
+        const void* ptr = DTable;
+        const HUF_DEltX4* const dt = ((const HUF_DEltX4*)ptr) +1;
+        const U32 dtLog = DTable[0];
+        size_t errorCode;
+
+        /* Init */
+        BIT_DStream_t bitD1;
+        BIT_DStream_t bitD2;
+        BIT_DStream_t bitD3;
+        BIT_DStream_t bitD4;
+        const size_t length1 = MEM_readLE16(istart);
+        const size_t length2 = MEM_readLE16(istart+2);
+        const size_t length3 = MEM_readLE16(istart+4);
+        size_t length4;
+        const BYTE* const istart1 = istart + 6;  /* jumpTable */
+        const BYTE* const istart2 = istart1 + length1;
+        const BYTE* const istart3 = istart2 + length2;
+        const BYTE* const istart4 = istart3 + length3;
+        const size_t segmentSize = (dstSize+3) / 4;
+        BYTE* const opStart2 = ostart + segmentSize;
+        BYTE* const opStart3 = opStart2 + segmentSize;
+        BYTE* const opStart4 = opStart3 + segmentSize;
+        BYTE* op1 = ostart;
+        BYTE* op2 = opStart2;
+        BYTE* op3 = opStart3;
+        BYTE* op4 = opStart4;
+        U32 endSignal;
+
+        length4 = cSrcSize - (length1 + length2 + length3 + 6);
+        if (length4 > cSrcSize) return ERROR(corruption_detected);   /* overflow */
+        errorCode = BIT_initDStream(&bitD1, istart1, length1);
+        if (HUF_isError(errorCode)) return errorCode;
+        errorCode = BIT_initDStream(&bitD2, istart2, length2);
+        if (HUF_isError(errorCode)) return errorCode;
+        errorCode = BIT_initDStream(&bitD3, istart3, length3);
+        if (HUF_isError(errorCode)) return errorCode;
+        errorCode = BIT_initDStream(&bitD4, istart4, length4);
+        if (HUF_isError(errorCode)) return errorCode;
+
+        /* 16-32 symbols per loop (4-8 symbols per stream) */
+        endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
+        for ( ; (endSignal==BIT_DStream_unfinished) && (op4<(oend-7)) ; )
+        {
+            HUF_DECODE_SYMBOLX4_2(op1, &bitD1);
+            HUF_DECODE_SYMBOLX4_2(op2, &bitD2);
+            HUF_DECODE_SYMBOLX4_2(op3, &bitD3);
+            HUF_DECODE_SYMBOLX4_2(op4, &bitD4);
+            HUF_DECODE_SYMBOLX4_1(op1, &bitD1);
+            HUF_DECODE_SYMBOLX4_1(op2, &bitD2);
+            HUF_DECODE_SYMBOLX4_1(op3, &bitD3);
+            HUF_DECODE_SYMBOLX4_1(op4, &bitD4);
+            HUF_DECODE_SYMBOLX4_2(op1, &bitD1);
+            HUF_DECODE_SYMBOLX4_2(op2, &bitD2);
+            HUF_DECODE_SYMBOLX4_2(op3, &bitD3);
+            HUF_DECODE_SYMBOLX4_2(op4, &bitD4);
+            HUF_DECODE_SYMBOLX4_0(op1, &bitD1);
+            HUF_DECODE_SYMBOLX4_0(op2, &bitD2);
+            HUF_DECODE_SYMBOLX4_0(op3, &bitD3);
+            HUF_DECODE_SYMBOLX4_0(op4, &bitD4);
+
+            endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
+        }
+
+        /* check corruption */
+        if (op1 > opStart2) return ERROR(corruption_detected);
+        if (op2 > opStart3) return ERROR(corruption_detected);
+        if (op3 > opStart4) return ERROR(corruption_detected);
+        /* note : op4 supposed already verified within main loop */
+
+        /* finish bitStreams one by one */
+        HUF_decodeStreamX4(op1, &bitD1, opStart2, dt, dtLog);
+        HUF_decodeStreamX4(op2, &bitD2, opStart3, dt, dtLog);
+        HUF_decodeStreamX4(op3, &bitD3, opStart4, dt, dtLog);
+        HUF_decodeStreamX4(op4, &bitD4, oend,     dt, dtLog);
+
+        /* check */
+        endSignal = BIT_endOfDStream(&bitD1) & BIT_endOfDStream(&bitD2) & BIT_endOfDStream(&bitD3) & BIT_endOfDStream(&bitD4);
+        if (!endSignal) return ERROR(corruption_detected);
+
+        /* decoded size */
+        return dstSize;
+    }
+}
+
+
+static size_t HUF_decompress4X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    HUF_CREATE_STATIC_DTABLEX4(DTable, HUF_MAX_TABLELOG);
+    const BYTE* ip = (const BYTE*) cSrc;
+
+    size_t hSize = HUF_readDTableX4 (DTable, cSrc, cSrcSize);
+    if (HUF_isError(hSize)) return hSize;
+    if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
+    ip += hSize;
+    cSrcSize -= hSize;
+
+    return HUF_decompress4X4_usingDTable (dst, dstSize, ip, cSrcSize, DTable);
+}
+
+
+/**********************************/
+/* quad-symbol decoding           */
+/**********************************/
+typedef struct { BYTE nbBits; BYTE nbBytes; } HUF_DDescX6;
+typedef union { BYTE byte[4]; U32 sequence; } HUF_DSeqX6;
+
+/* recursive, up to level 3; may benefit from