llvm-toolchain-6.0.git
5 years agoMerge version 1:6.0.1-11+rpi1 and 1:6.0.1-12 to produce 1:6.0.1-12+rpi1 archive/raspbian/1%6.0.1-12+rpi1 raspbian/1%6.0.1-12+rpi1
Raspbian automatic forward porter [Fri, 31 Jan 2020 19:51:57 +0000 (19:51 +0000)]
Merge version 1:6.0.1-11+rpi1 and 1:6.0.1-12 to produce 1:6.0.1-12+rpi1

5 years agoMerge llvm-toolchain-6.0 (1:6.0.1-12) import into refs/heads/workingbranch
Gianfranco Costamagna [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
Merge llvm-toolchain-6.0 (1:6.0.1-12) import into refs/heads/workingbranch

5 years agollvm-6.0-NVPTX-addrspaces
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
llvm-6.0-NVPTX-addrspaces

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-6.0-NVPTX-addrspaces.patch

5 years ago[PATCH] PerfJITEventListener integration, requires compile flag LLVM_USE_PERF
DokFaust [Mon, 11 Jun 2018 10:59:42 +0000 (12:59 +0200)]
[PATCH] PerfJITEventListener integration, requires compile flag LLVM_USE_PERF

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-D44892-Perf-integration.patch

5 years agollvm-rL327898
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
llvm-rL327898

commit 64c3384f94a1eb3e3510d6f66c3bccdfc9d9050b
Author: Nirav Dave <niravd@google.com>
Date:   Thu Feb 1 16:11:59 2018 +0000

    r327898/dependencies roll up

    This is a squash of 13 commits required in the lead up to r327898,
    which fixes https://github.com/JuliaLang/julia/issues/27603. The squashed
    commits are:

    332d15e981e86b9e058087174bb288ba18a15807
    b659d3fca5d24c25ee73f979edb382f7f24e05e2
    c01d1363ea080170fc5143d72f26eecd9270f03b
    eab8a177a4caef9e42ef1d2aeb4ba15dc788d3f2
    bedb1391781b009ace95f5586e7fae5f03fe0689
    11d041a905f82ac78e7ccf2394773e80b93d147c
    e1ec36c55a0127988f42a3329ca835617b30de09
    b8d2903300c13d8fd151c8e5dc71017269617539
    00884fea345f47ab05174a8f314ecd60d1676d02
    28ab04cec0d9888af9d29946b3a048b8340abe0f
    3dd52e62ea3087efcca63c3772183d9471abc742
    bd3649ff6d6b4d18b3c6de253179d987a120518a
    aea03035b9c633e6d745b6d3fc5b6378699f576c

    Their commit messages follow below:

    [SelectionDAG] Fix UpdateChains handling of TokenFactors

    Summary:
    In Instruction Selection UpdateChains replaces all matched Nodes'
    chain references including interior token factors and deletes them.
    This may allow nodes which depend on these interior nodes but are not
    part of the set of matched nodes to be left with a dangling dependence.
    Avoid this by doing the replacement for matched non-TokenFactor nodes.

    Fixes PR36164.

    Reviewers: jonpa, RKSimon, bogner

    Subscribers: llvm-commits, hiraditya

    Differential Revision: https://reviews.llvm.org/D42754

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323977 91177308-0d34-0410-b5e6-96231b3b80d8

    Regenerate test result for vastart-defs-eflags.ll. NFC.

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323596 91177308-0d34-0410-b5e6-96231b3b80d8

    Regenerate test result for testb-je-fusion.ll. NFC.

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323595 91177308-0d34-0410-b5e6-96231b3b80d8

    [X86] Avoid using high register trick for test instruction

    Summary:
    It seems it's main effect is to create addition copies when values are inr register that do not support this trick, which increase register pressure and makes the code bigger.

    Reviewers: craig.topper, niravd, spatel, hfinkel

    Subscribers: llvm-commits

    Differential Revision: https://reviews.llvm.org/D42646

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323888 91177308-0d34-0410-b5e6-96231b3b80d8

    Add a regression test for problems caused by D42646 . NFC

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323868 91177308-0d34-0410-b5e6-96231b3b80d8

    Add test case for truncated and promotion to test. NFC

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323663 91177308-0d34-0410-b5e6-96231b3b80d8

    [X86] Add test case to ensure testw is generated when optimizing for size. NFC

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323687 91177308-0d34-0410-b5e6-96231b3b80d8

    [X86] Generate testl instruction through truncates.

    Summary:
    This was introduced in D42646 but ended up being reverted because the original implementation was buggy.

    Depends on D42646

    Reviewers: craig.topper, niravd, spatel, hfinkel

    Subscribers: llvm-commits

    Differential Revision: https://reviews.llvm.org/D42741

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323899 91177308-0d34-0410-b5e6-96231b3b80d8

    [X86] Don't look for TEST instruction shrinking opportunities when the root node is a X86ISD::SUB.

    I don't believe we ever create an X86ISD::SUB with a 0 constant which is what the TEST handling needs. The ternary operator at the end of this code shows up as only going one way in the llvm-cov report from the bots.

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@324865 91177308-0d34-0410-b5e6-96231b3b80d8

    [X86] Teach LowerBUILD_VECTOR to recognize pair-wise splats of 32-bit elements and use a 64-bit broadcast

    If we are splatting pairs of 32-bit elements, we can use a 64-bit broadcast to get the job done.

    We could probably could probably do this with other sizes too, for example four 16-bit elements. Or we could broadcast pairs of 16-bit elements using a 32-bit element broadcast. But I've left that as a future improvement.

    I've also restricted this to AVX2 only because we can only broadcast loads under AVX.

    Differential Revision: https://reviews.llvm.org/D42086

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@322730 91177308-0d34-0410-b5e6-96231b3b80d8

    [DAG, X86] Revert r327197 "Revert r327170, r327171, r327172"

    Reland ISel cycle checking improvements after simplifying node id
    invariant traversal and correcting typo.

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@327898 91177308-0d34-0410-b5e6-96231b3b80d8

    [ Modified for cherry-pick: Dropped Hexagon and SystemZ changes"

    [DAG, X86] Fix ISel-time node insertion ids

    As in SystemZ backend, correctly propagate node ids when inserting new
    unselected nodes into the DAG during instruction Seleciton for X86
    target.

    Fixes PR36865.

    Reviewers: jyknight, craig.topper

    Subscribers: hiraditya, llvm-commits

    Differential Revision: https://reviews.llvm.org/D44797

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@328233 91177308-0d34-0410-b5e6-96231b3b80d8

    [DAG] Fix node id invalidation in Instruction Selection.

    Invalidation should be bit negation. Add missing negation.

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@328287 91177308-0d34-0410-b5e6-96231b3b80d8

    Remove failing tests

    This removes tests that are failing due to codegen differences,
    after the latest set of backports. Fixing thse for the backport
    branch does not seem worth it.

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-rL327898.patch

5 years agollvm-rL326967-aligned-load
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
llvm-rL326967-aligned-load

commit b398d8e1fa5a5a914957fa22d0a64db97f6c265e
Author: Craig Topper <craig.topper@intel.com>
Date:   Thu Mar 8 00:21:17 2018 +0000

    [X86] Fix some isel patterns that used aligned vector load instructions with unaligned predicates.

    These patterns weren't checking the alignment of the load, but were using the aligned instructions. This will cause a GP fault if the data isn't aligned.

    I believe these were introduced in r312450.

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@326967 91177308-0d34-0410-b5e6-96231b3b80d8

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-rL326967-aligned-load.patch

5 years agollvm-D50010-VNCoercion-ni
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
llvm-D50010-VNCoercion-ni

commit 8eb2b102a203d83fb713f3bf79acf235dabdd8cd
Author: Keno Fischer <keno@juliacomputing.com>
Date:   Mon Jul 30 16:59:08 2018 -0400

    [VNCoercion] Disallow coercion between different ni addrspaces

    Summary:
    I'm not sure if it would be legal by the IR reference to introduce
    an addrspacecast here, since the IR reference is a bit vague on
    the exact semantics, but at least for our usage of it (and I
    suspect for many other's usage) it is not. For us, addrspacecasts
    between non-integral address spaces carry frontend information that the
    optimizer cannot deduce afterwards in a generic way (though we
    have frontend specific passes in our pipline that do propagate
    these). In any case, I'm sure nobody is using it this way at
    the moment, since it would have introduced inttoptrs, which
    are definitely illegal.

    Fixes PR38375

    Reviewers: sanjoy, reames, dberlin

    Subscribers: llvm-commits

    Differential Revision: https://reviews.llvm.org/D50010

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-D50010-VNCoercion-ni.patch

5 years agollvm-D42262-jumpthreading-not-i1
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
llvm-D42262-jumpthreading-not-i1

commit 6a311a7a804831fea43cfb2f61322adcb407a1af
Author: Keno Fischer <keno@juliacomputing.com>
Date:   Thu Jan 18 15:57:05 2018 -0500

    [JumpThreading] Don't restrict cast-traversal to i1

    Summary:
    In D17663, JumpThreading learned to look trough simple cast instructions,
    but only if the source of those cast instructions was a phi/cmp i1
    (in an effort to limit compile time effects). I think this condition
    is too restrictive. For switches with limited value range, InstCombine
    will readily introduce an extra `trunc` instruction to a smaller
    integer type (e.g. from i8 to i2), leaving us in the somewhat perverse
    situation that jump-threading would work before running instcombine,
    but not after. Since instcombine produces this pattern, I think we
    need to consider it canonical and support it in JumpThreading.
    In general, for limiting recursion, I think the existing restriction
    to phi and cmp nodes should be sufficient to avoid looking through
    unprofitable chains of instructions.

    Reviewers: haicheng, gberry, bmakam, mcrosier

    Subscribers: llvm-commits

    Differential Revision: https://reviews.llvm.org/D42262

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-D42262-jumpthreading-not-i1.patch

5 years ago[PATCH 4/4] Enable support for floating-point division reductions
Yichao Yu [Sat, 10 Jun 2017 12:45:13 +0000 (08:45 -0400)]
[PATCH 4/4] Enable support for floating-point division reductions

Similar to fsub, fdiv can also be vectorized using fmul.

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-D34078-vectorize-fdiv.patch

5 years ago[PATCH] Fix unwind info relocation with large code model on AArch64
Valentin Churavy [Thu, 5 Jul 2018 16:37:50 +0000 (12:37 -0400)]
[PATCH] Fix unwind info relocation with large code model on AArch64

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-D27629-AArch64-large_model_6.0.1.patch

5 years agoclang-arm-default-vfp3-on-armv7a
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
clang-arm-default-vfp3-on-armv7a

Gbp-Pq: Name clang-arm-default-vfp3-on-armv7a.patch

5 years agoD53557-hurd-self-exe-realpath
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
D53557-hurd-self-exe-realpath

[hurd] Make getMainExecutable get the real binary path

On GNU/Hurd, llvm-config is returning bogus value, such as:

$ llvm-config-6.0 --includedir
/usr/include

while it should be:
$ llvm-config-6.0 --includedir
/usr/lib/llvm-6.0/include

This is because getMainExecutable does not get the actual installation
path. On GNU/Hurd, /proc/self/exe is indeed a symlink to the path that
was used to start the program, and not the eventual binary file. Llvm's
getMainExecutable thus needs to run realpath over it to get the actual
place where llvm was installed (/usr/lib/llvm-6.0/bin/llvm-config), and
not /usr/bin/llvm-config-6.0. This will not change the result on Linux,
where /proc/self/exe already points to the eventual file.

Patch by Samuel Thibault!

While making changes here, I reformatted this block a bit to reduce
indentation and match 2 space indent style.

Differential Revision: https://reviews.llvm.org/D53557

Gbp-Pq: Name D53557-hurd-self-exe-realpath.diff

5 years agohurd-tools_llvm-shlib_CMakeLists.txt
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
hurd-tools_llvm-shlib_CMakeLists.txt

===================================================================

Gbp-Pq: Name hurd-tools_llvm-shlib_CMakeLists.txt.diff

5 years agohurd-lib_Support_Unix_Path.inc
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
hurd-lib_Support_Unix_Path.inc

===================================================================

Gbp-Pq: Name hurd-lib_Support_Unix_Path.inc.diff

5 years agoFix missing include and library paths on x32
James Clarke [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
Fix missing include and library paths on x32

Forwarded: https://reviews.llvm.org/D43630
Last-Update: 2018-09-15

Gbp-Pq: Name x32-fix-driver-search-paths.diff

5 years agoFix missing include paths on powerpcspe
John Paul Adrian Glaubitz [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
Fix missing include paths on powerpcspe

Bug-Debian: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=908791
Forwarded: https://reviews.llvm.org/D52066
Last-Update: 2018-09-13

Gbp-Pq: Name powerpcspe-add-missing-include-path.diff

5 years agodisable-sse2-old-x86
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
disable-sse2-old-x86

===================================================================

Gbp-Pq: Name disable-sse2-old-x86.diff

5 years agorL338481-cherry-pick-really-subtle-miscompile
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
rL338481-cherry-pick-really-subtle-miscompile

===================================================================

Gbp-Pq: Name rL338481-cherry-pick-really-subtle-miscompile.diff

5 years agoD51639-optim-issue
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
D51639-optim-issue

===================================================================

Gbp-Pq: Name D51639-optim-issue.diff

5 years agoD51335-alignment-issue
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
D51335-alignment-issue

===================================================================

Gbp-Pq: Name D51335-alignment-issue.diff

5 years agopr38663-pgo-lto-crash
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
pr38663-pgo-lto-crash

===================================================================

Gbp-Pq: Name pr38663-pgo-lto-crash.patch

5 years agoD51108-rust-powerpc
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
D51108-rust-powerpc

===================================================================

Gbp-Pq: Name D51108-rust-powerpc.diff

5 years agoPowerPC-Make-AddrSpaceCast-noop
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
PowerPC-Make-AddrSpaceCast-noop

===================================================================

Gbp-Pq: Name PowerPC-Make-AddrSpaceCast-noop.diff

5 years agollvm-rL323946-LSRTy
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
llvm-rL323946-LSRTy

commit ab60b05a472e8651cbe53c19513b7e62b9ff32df
Author: Mikael Holmen <mikael.holmen@ericsson.com>
Date:   Thu Feb 1 06:38:34 2018 +0000

    [LSR] Don't force bases of foldable formulae to the final type.

    Summary:
    Before emitting code for scaled registers, we prevent
    SCEVExpander from hoisting any scaled addressing mode
    by emitting all the bases first. However, these bases
    are being forced to the final type, resulting in some
    odd code.

    For example, if the type of the base is an integer and
    the final type is a pointer, we will emit an inttoptr
    for the base, a ptrtoint for the scale, and then a
    'reverse' GEP where the GEP pointer is actually the base
    integer and the index is the pointer. It's more intuitive
    to use the pointer as a pointer and the integer as index.

    Patch by: Bevin Hansson

    Reviewers: atrick, qcolombet, sanjoy

    Reviewed By: qcolombet

    Subscribers: llvm-commits

    Differential Revision: https://reviews.llvm.org/D42103

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323946 91177308-0d34-0410-b5e6-96231b3b80d8

Gbp-Pq: Name llvm-rL323946-LSRTy.patch

5 years agollvm-D49832-SCEVPred
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
llvm-D49832-SCEVPred

commit 98592fcc61307968f7df1362771534595a1e1c21
Author: Keno Fischer <keno@juliacomputing.com>
Date:   Wed Jul 25 19:29:02 2018 -0400

    [SCEV] Don't expand Wrap predicate using inttoptr in ni addrspaces

    Summary:
    In non-integral address spaces, we're not allowed to introduce inttoptr/ptrtoint
    intrinsics. Instead, we need to expand any pointer arithmetic as geps on the
    base pointer. Luckily this is a common task for SCEV, so all we have to do here
    is hook up the corresponding helper function and add test case.

    Fixes PR38290

    Reviewers: reames, sanjoy

    Subscribers: javed.absar, llvm-commits

    Differential Revision: https://reviews.llvm.org/D49832

Gbp-Pq: Name llvm-D49832-SCEVPred.patch

5 years agoD42717-JumpThreading-backport-2
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
D42717-JumpThreading-backport-2

===================================================================

Gbp-Pq: Name D42717-JumpThreading-backport-2.diff

5 years agoD40146-JumpThreading-backport-1
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
D40146-JumpThreading-backport-1

===================================================================

Gbp-Pq: Name D40146-JumpThreading-backport-1.diff

5 years agoAdd Ubuntu Cosmic to the distro release list.
Adam Conrad [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
Add Ubuntu Cosmic to the distro release list.

Last-Update: 2018-05-03

Gbp-Pq: Name ubuntu-cosmic-support.patch

5 years agoUse Debian-provided MathJax everywhere.
Nicholas D Steeves [Sun, 11 Feb 2018 02:02:17 +0000 (21:02 -0500)]
Use Debian-provided MathJax everywhere.

Gbp-Pq: Name 0049-Use-Debian-provided-MathJax-everywhere.patch

5 years agoSet html_static_path = ['_static'] everywhere.
Nicholas D Steeves [Sun, 11 Feb 2018 02:00:55 +0000 (21:00 -0500)]
Set html_static_path = ['_static'] everywhere.

Gbp-Pq: Name 0048-Set-html_static_path-_static-everywhere.patch

5 years agoremove-test-freezing
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
remove-test-freezing

===================================================================

Gbp-Pq: Name remove-test-freezing.diff

5 years ago[Sparc] Include __tls_get_addr in symbol table for TLS calls to it
James Clarke [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
[Sparc] Include __tls_get_addr in symbol table for TLS calls to it

Last-Update: 2018-02-14

Global Dynamic and Local Dynamic call relocations only implicitly
reference __tls_get_addr, but it still needs to be in the symbol table
to be bound at link time otherwise it fails to link. For details, see
https://sourceware.org/bugzilla/show_bug.cgi?id=22832.

Gbp-Pq: Name sparc64-add-missing-tls-get-addr.diff

5 years agotest-keep-alive
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
test-keep-alive

===================================================================

Gbp-Pq: Name test-keep-alive.diff

5 years agosilent-llvm-isel-fuzzer
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
silent-llvm-isel-fuzzer

===================================================================

Gbp-Pq: Name silent-llvm-isel-fuzzer.diff

5 years agosilent-test-macho
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
silent-test-macho

===================================================================

Gbp-Pq: Name silent-test-macho.diff

5 years agolldb-disable-swig-error
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
lldb-disable-swig-error

===================================================================

Gbp-Pq: Name lldb-disable-swig-error.diff

5 years agodisable-error-xray
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
disable-error-xray

===================================================================

Gbp-Pq: Name disable-error-xray.diff

5 years agosilent-amd-tet
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
silent-amd-tet

===================================================================

Gbp-Pq: Name silent-amd-tet.diff

5 years agodisable-path-test-failing
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
disable-path-test-failing

===================================================================

Gbp-Pq: Name disable-path-test-failing.diff

5 years agosilent-test-failing-codeverage
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
silent-test-failing-codeverage

===================================================================

Gbp-Pq: Name silent-test-failing-codeverage.diff

5 years agolldb-missing-install
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
lldb-missing-install

===================================================================

Gbp-Pq: Name lldb-missing-install.diff

5 years agolldb-addversion-suffix-to-llvm-server-exec
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
lldb-addversion-suffix-to-llvm-server-exec

lldb-server exec users always /usr/bin/lldb-server. Server is required
for any debugging with lldb which makes it unusable unless default version
package has been installed. Small changes to code and debian/rules allows
a workaround for lldb-server start up.

To use this one needs to add cmake defination during configure. eg
-DDEBIAN_VERSION_SUFFIX=-$(LLVM_VERSION)

Better implementation would be to use /usr/share/llvm-$(VERSION)/bin but
that change seems to require a big change to the path handling code
which could then break something else.

This probably should have upstream bug but I couldn't find any existing report.

Gbp-Pq: Name lldb-addversion-suffix-to-llvm-server-exec.patch

5 years agofix-scan-view-path
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
fix-scan-view-path

===================================================================

Gbp-Pq: Name fix-scan-view-path.diff

5 years agoclang-tidy-run-bin
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
clang-tidy-run-bin

===================================================================

Gbp-Pq: Name clang-tidy-run-bin.diff

5 years agoSilent a test failing on yakkety amd64
Sylvestre [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
Silent a test failing on yakkety amd64

Last-Update: 2016-08-26

/tmp/buildd/llvm-toolchain-snapshot-4.0~svn279801/test/tools/llvm-symbolizer/print_context.c:16:11: error: expected string not found in input
// CHECK: inc
         ^
<stdin>:1:1: note: scanning from here
_fini
^
<stdin>:1:3: note: possible intended match here
_fini
 ^

Gbp-Pq: Name disable-llvm-symbolizer-test.diff

5 years agosilent-gold-utils
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
silent-gold-utils

Gbp-Pq: Name silent-gold-utils.diff

5 years agodisable-source-interleave
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
disable-source-interleave

===================================================================

Gbp-Pq: Name disable-source-interleave.diff

5 years agoLink with -latomic when mips* processor is detected
Gianfranco Costamagna [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
Link with -latomic when mips* processor is detected

Last-Update: 2016-07-27

Gbp-Pq: Name lldb-link-atomic-cmake.patch

5 years ago[PATCH] [clang] cmake: resolve symlinks in ClangConfig.cmake
Peter Wu [Fri, 4 May 2018 13:55:26 +0000 (15:55 +0200)]
[PATCH] [clang] cmake: resolve symlinks in ClangConfig.cmake

Ensure that symlinks such as /usr/lib/cmake/clang-X.Y (pointing to
/usr/lib/llvm-X.Y/lib/cmake/llvm) are resolved. This ensures that
CLANG_INSTALL_PREFIX ends up to be /usr/lib/llvm-X.Y instead of /usr.

Partially addresses PR37128

Gbp-Pq: Name 0001-tools-clang-cmake-resolve-symlinks-in-ClangConfig.cmake.patch

5 years ago[PATCH] [llvm] cmake: resolve symlinks in LLVMConfig.cmake
Peter Wu [Fri, 4 May 2018 13:43:06 +0000 (15:43 +0200)]
[PATCH] [llvm] cmake: resolve symlinks in LLVMConfig.cmake

Ensure that symlinks such as /usr/lib/llvm-X.Y/cmake (pointing to
lib/cmake/llvm) are resolved. This ensures that LLVM_INSTALL_PREFIX
becomes /usr/lib/llvm-X.Y instead of /usr.

Partially addresses PR37128

Gbp-Pq: Name 0001-llvm-cmake-resolve-symlinks-in-LLVMConfig.cmake.patch

5 years agomips-fpxx-enable
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
mips-fpxx-enable

===================================================================

Gbp-Pq: Name mips-fpxx-enable.diff

5 years agoscan-view-fix-path
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
scan-view-fix-path

===================================================================

Gbp-Pq: Name scan-view-fix-path.diff

5 years agoinstall-scan-build-py
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
install-scan-build-py

Gbp-Pq: Name install-scan-build-py.diff

5 years agoclang-analyzer-force-version
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
clang-analyzer-force-version

# Force the version of clang in the analyzer
# This was causing the static analyzer to fail silently if the clang & clang++ are
# not installed

Gbp-Pq: Name clang-analyzer-force-version.diff

5 years agosilent-MCJIIT-tests
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
silent-MCJIIT-tests

Gbp-Pq: Name silent-MCJIIT-tests.diff

5 years agohurd-EIEIO-undef
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
hurd-EIEIO-undef

===================================================================

Gbp-Pq: Name hurd-EIEIO-undef.diff

5 years agolldb-libname
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
lldb-libname

Gbp-Pq: Name lldb-libname.diff

5 years agolldb-soname
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
lldb-soname

Gbp-Pq: Name lldb-soname.diff

5 years agosoname
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
soname

Gbp-Pq: Name 0044-soname.diff

5 years agofix-llvm-config-obj-src-root
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
fix-llvm-config-obj-src-root

Gbp-Pq: Name fix-llvm-config-obj-src-root.patch

5 years agodisable-display-PASS-UNSUPPORTED-XFAIL
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
disable-display-PASS-UNSUPPORTED-XFAIL

===================================================================

Gbp-Pq: Name disable-display-PASS-UNSUPPORTED-XFAIL.diff

5 years agosilent-more-tests
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
silent-more-tests

# Comment the tests for the code coverage (fails otherwise)

Gbp-Pq: Name silent-more-tests.diff

5 years agodo-not-fail-on-unexpected-pass
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
do-not-fail-on-unexpected-pass

===================================================================

Gbp-Pq: Name do-not-fail-on-unexpected-pass.diff

5 years agofix-clang-path-and-build
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
fix-clang-path-and-build

Gbp-Pq: Name fix-clang-path-and-build.diff

5 years agopython-clangpath
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
python-clangpath

Gbp-Pq: Name python-clangpath.diff

5 years agoatomic_library_1
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
atomic_library_1

Gbp-Pq: Name atomic_library_1.diff

5 years agofails on debian unstable amd64
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
fails on debian unstable amd64

Command Output (stderr):
--
/build/llvm-toolchain-snapshot-4.0~svn279916/test/tools/gold/X86/start-lib-common.ll:22:10: error: expected string not found in input
; CHECK: @x = common global i32 0, align 8
         ^
<stdin>:1:1: note: scanning from here
; ModuleID = '/build/llvm-toolchain-snapshot-4.0~svn279916/build-llvm/test/tools/gold/X86/Output/start-lib-common.ll.tmp3.o'
^
<stdin>:4:1: note: possible intended match here
@x = common global i32 0, align 4
^

Gbp-Pq: Name silent-gold-test.diff

5 years agohurd-pathmax
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
hurd-pathmax

===================================================================

Gbp-Pq: Name hurd-pathmax.diff

5 years agounwind-chain-inclusion
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
unwind-chain-inclusion

# Without this patch, the first local include of unwind.h might, with the
# __has_include_next, try to include the one from the system.
# It might be /usr/include/clang/3.4/include/unwind.h
# Because of the #ifndef __CLANG_UNWIND_H, it might never include any declaration
# from the system.

Gbp-Pq: Name unwind-chain-inclusion.diff

5 years agoclang-format-version
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
clang-format-version

Gbp-Pq: Name clang-format-version.diff

5 years agodeclare_clear_cache
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
declare_clear_cache

Gbp-Pq: Name declare_clear_cache.diff

5 years agoscan-build-clang-path
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
scan-build-clang-path

Gbp-Pq: Name scan-build-clang-path.diff

5 years agoDebian-version-info-and-bugreport
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
Debian-version-info-and-bugreport

Gbp-Pq: Name 0003-Debian-version-info-and-bugreport.patch

5 years agoset correct float abi settings for armel and armhf
Peter Michael Green [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
set correct float abi settings for armel and armhf

debian armel supports systems that don't have a fpu so should use a "float abi"
setting of soft by default.

Debian armhf needs a float abi setting of "hard"

Gbp-Pq: Name 26-set-correct-float-abi.diff

5 years agofix_clang_stdint
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
fix_clang_stdint

===================================================================

Gbp-Pq: Name 27-fix_clang_stdint.diff

5 years agostrlcpy_strlcat_warning_removed
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
strlcpy_strlcat_warning_removed

Gbp-Pq: Name 23-strlcpy_strlcat_warning_removed.diff

5 years agoclang_debian_version
LLVM Packaging Team [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
clang_debian_version

===================================================================

Gbp-Pq: Name 19-clang_debian_version.patch

5 years agollvm-toolchain-6.0 (1:6.0.1-12) unstable; urgency=medium
Gianfranco Costamagna [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
llvm-toolchain-6.0 (1:6.0.1-12) unstable; urgency=medium

  [ Andreas Beckmann ]
  * For now, lld doesn't generate shared libs. Removing the files
    (Closes: #857653, LP: #1829677)

  [ Gianfranco Costamagna ]
  * Import the Ubuntu delta
  * Upload to unstable

  [ Matthias Klose ]
  * (Build-)depend on python2 instead of python.
  * Fix the build to use python2 instead of python.
  * Call dh_python2 to rewrite shebangs.

[dgit import unpatched llvm-toolchain-6.0 1:6.0.1-12]

5 years agoImport llvm-toolchain-6.0_6.0.1-12.debian.tar.xz
Gianfranco Costamagna [Mon, 20 Jan 2020 09:26:04 +0000 (09:26 +0000)]
Import llvm-toolchain-6.0_6.0.1-12.debian.tar.xz

[dgit import tarball llvm-toolchain-6.0 1:6.0.1-12 llvm-toolchain-6.0_6.0.1-12.debian.tar.xz]

6 years agoMerge llvm-toolchain-6.0 (1:6.0.1-11+rpi1) import into refs/heads/workingbranch
Raspbian forward porter [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
Merge llvm-toolchain-6.0 (1:6.0.1-11+rpi1) import into refs/heads/workingbranch

6 years agoAutomatically generated patch (1:5.0~+rc2-1+rpi1)
Peter Micahel Green [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
Automatically generated patch (1:5.0~+rc2-1+rpi1)

Last (up to) 3 git changes, FYI:

commit d8bb5e17a1c124bfc65bd4da3d5bad2318b86155
Author: Peter Michael Green <plugwash@raspbian.org>
Date:   Thu Sep 7 07:03:48 2017 +0000

    Remove old cmake patch, dgit will regenerate it.

commit 8479377cb2739eeaa610fbc0370f239e0cabda83
Merge: 706ae0030 46e6d900f
Author: Peter Michael Green <plugwash@raspbian.org>
Date:   Thu Sep 7 07:11:20 2017 +0100

    Merge raspbian changes from llvm-toolchain-4.0

commit 706ae0030c98b01e6594ab4a4aca848e9dbedbd5
Merge: 2b80ff9a0 b60f1e5d8
Author: plugwash <plugwash@thinkpad>
Date:   Thu Sep 7 07:02:37 2017 +0100

    Psuedomerge to tie in 4.0 history.

Gbp-Pq: Name auto-1:5.0~+rc2-1+rpi1-d8bb5e17a1c124bfc65bd4da3d5bad2318b86155-1504768094

6 years agollvm-6.0-NVPTX-addrspaces
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
llvm-6.0-NVPTX-addrspaces

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-6.0-NVPTX-addrspaces.patch

6 years agoPerfJITEventListener integration, requires compile flag LLVM_USE_PERF
DokFaust [Mon, 11 Jun 2018 10:59:42 +0000 (12:59 +0200)]
PerfJITEventListener integration, requires compile flag LLVM_USE_PERF

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-D44892-Perf-integration.patch

6 years agollvm-rL327898
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
llvm-rL327898

commit 64c3384f94a1eb3e3510d6f66c3bccdfc9d9050b
Author: Nirav Dave <niravd@google.com>
Date:   Thu Feb 1 16:11:59 2018 +0000

    r327898/dependencies roll up

    This is a squash of 13 commits required in the lead up to r327898,
    which fixes https://github.com/JuliaLang/julia/issues/27603. The squashed
    commits are:

    332d15e981e86b9e058087174bb288ba18a15807
    b659d3fca5d24c25ee73f979edb382f7f24e05e2
    c01d1363ea080170fc5143d72f26eecd9270f03b
    eab8a177a4caef9e42ef1d2aeb4ba15dc788d3f2
    bedb1391781b009ace95f5586e7fae5f03fe0689
    11d041a905f82ac78e7ccf2394773e80b93d147c
    e1ec36c55a0127988f42a3329ca835617b30de09
    b8d2903300c13d8fd151c8e5dc71017269617539
    00884fea345f47ab05174a8f314ecd60d1676d02
    28ab04cec0d9888af9d29946b3a048b8340abe0f
    3dd52e62ea3087efcca63c3772183d9471abc742
    bd3649ff6d6b4d18b3c6de253179d987a120518a
    aea03035b9c633e6d745b6d3fc5b6378699f576c

    Their commit messages follow below:

    [SelectionDAG] Fix UpdateChains handling of TokenFactors

    Summary:
    In Instruction Selection UpdateChains replaces all matched Nodes'
    chain references including interior token factors and deletes them.
    This may allow nodes which depend on these interior nodes but are not
    part of the set of matched nodes to be left with a dangling dependence.
    Avoid this by doing the replacement for matched non-TokenFactor nodes.

    Fixes PR36164.

    Reviewers: jonpa, RKSimon, bogner

    Subscribers: llvm-commits, hiraditya

    Differential Revision: https://reviews.llvm.org/D42754

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323977 91177308-0d34-0410-b5e6-96231b3b80d8

    Regenerate test result for vastart-defs-eflags.ll. NFC.

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323596 91177308-0d34-0410-b5e6-96231b3b80d8

    Regenerate test result for testb-je-fusion.ll. NFC.

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323595 91177308-0d34-0410-b5e6-96231b3b80d8

    [X86] Avoid using high register trick for test instruction

    Summary:
    It seems it's main effect is to create addition copies when values are inr register that do not support this trick, which increase register pressure and makes the code bigger.

    Reviewers: craig.topper, niravd, spatel, hfinkel

    Subscribers: llvm-commits

    Differential Revision: https://reviews.llvm.org/D42646

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323888 91177308-0d34-0410-b5e6-96231b3b80d8

    Add a regression test for problems caused by D42646 . NFC

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323868 91177308-0d34-0410-b5e6-96231b3b80d8

    Add test case for truncated and promotion to test. NFC

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323663 91177308-0d34-0410-b5e6-96231b3b80d8

    [X86] Add test case to ensure testw is generated when optimizing for size. NFC

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323687 91177308-0d34-0410-b5e6-96231b3b80d8

    [X86] Generate testl instruction through truncates.

    Summary:
    This was introduced in D42646 but ended up being reverted because the original implementation was buggy.

    Depends on D42646

    Reviewers: craig.topper, niravd, spatel, hfinkel

    Subscribers: llvm-commits

    Differential Revision: https://reviews.llvm.org/D42741

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@323899 91177308-0d34-0410-b5e6-96231b3b80d8

    [X86] Don't look for TEST instruction shrinking opportunities when the root node is a X86ISD::SUB.

    I don't believe we ever create an X86ISD::SUB with a 0 constant which is what the TEST handling needs. The ternary operator at the end of this code shows up as only going one way in the llvm-cov report from the bots.

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@324865 91177308-0d34-0410-b5e6-96231b3b80d8

    [X86] Teach LowerBUILD_VECTOR to recognize pair-wise splats of 32-bit elements and use a 64-bit broadcast

    If we are splatting pairs of 32-bit elements, we can use a 64-bit broadcast to get the job done.

    We could probably could probably do this with other sizes too, for example four 16-bit elements. Or we could broadcast pairs of 16-bit elements using a 32-bit element broadcast. But I've left that as a future improvement.

    I've also restricted this to AVX2 only because we can only broadcast loads under AVX.

    Differential Revision: https://reviews.llvm.org/D42086

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@322730 91177308-0d34-0410-b5e6-96231b3b80d8

    [DAG, X86] Revert r327197 "Revert r327170, r327171, r327172"

    Reland ISel cycle checking improvements after simplifying node id
    invariant traversal and correcting typo.

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@327898 91177308-0d34-0410-b5e6-96231b3b80d8

    [ Modified for cherry-pick: Dropped Hexagon and SystemZ changes"

    [DAG, X86] Fix ISel-time node insertion ids

    As in SystemZ backend, correctly propagate node ids when inserting new
    unselected nodes into the DAG during instruction Seleciton for X86
    target.

    Fixes PR36865.

    Reviewers: jyknight, craig.topper

    Subscribers: hiraditya, llvm-commits

    Differential Revision: https://reviews.llvm.org/D44797

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@328233 91177308-0d34-0410-b5e6-96231b3b80d8

    [DAG] Fix node id invalidation in Instruction Selection.

    Invalidation should be bit negation. Add missing negation.

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@328287 91177308-0d34-0410-b5e6-96231b3b80d8

    Remove failing tests

    This removes tests that are failing due to codegen differences,
    after the latest set of backports. Fixing thse for the backport
    branch does not seem worth it.

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-rL327898.patch

6 years agollvm-rL326967-aligned-load
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
llvm-rL326967-aligned-load

commit b398d8e1fa5a5a914957fa22d0a64db97f6c265e
Author: Craig Topper <craig.topper@intel.com>
Date:   Thu Mar 8 00:21:17 2018 +0000

    [X86] Fix some isel patterns that used aligned vector load instructions with unaligned predicates.

    These patterns weren't checking the alignment of the load, but were using the aligned instructions. This will cause a GP fault if the data isn't aligned.

    I believe these were introduced in r312450.

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@326967 91177308-0d34-0410-b5e6-96231b3b80d8

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-rL326967-aligned-load.patch

6 years agollvm-D50010-VNCoercion-ni
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
llvm-D50010-VNCoercion-ni

commit 8eb2b102a203d83fb713f3bf79acf235dabdd8cd
Author: Keno Fischer <keno@juliacomputing.com>
Date:   Mon Jul 30 16:59:08 2018 -0400

    [VNCoercion] Disallow coercion between different ni addrspaces

    Summary:
    I'm not sure if it would be legal by the IR reference to introduce
    an addrspacecast here, since the IR reference is a bit vague on
    the exact semantics, but at least for our usage of it (and I
    suspect for many other's usage) it is not. For us, addrspacecasts
    between non-integral address spaces carry frontend information that the
    optimizer cannot deduce afterwards in a generic way (though we
    have frontend specific passes in our pipline that do propagate
    these). In any case, I'm sure nobody is using it this way at
    the moment, since it would have introduced inttoptrs, which
    are definitely illegal.

    Fixes PR38375

    Reviewers: sanjoy, reames, dberlin

    Subscribers: llvm-commits

    Differential Revision: https://reviews.llvm.org/D50010

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-D50010-VNCoercion-ni.patch

6 years agollvm-D42262-jumpthreading-not-i1
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
llvm-D42262-jumpthreading-not-i1

commit 6a311a7a804831fea43cfb2f61322adcb407a1af
Author: Keno Fischer <keno@juliacomputing.com>
Date:   Thu Jan 18 15:57:05 2018 -0500

    [JumpThreading] Don't restrict cast-traversal to i1

    Summary:
    In D17663, JumpThreading learned to look trough simple cast instructions,
    but only if the source of those cast instructions was a phi/cmp i1
    (in an effort to limit compile time effects). I think this condition
    is too restrictive. For switches with limited value range, InstCombine
    will readily introduce an extra `trunc` instruction to a smaller
    integer type (e.g. from i8 to i2), leaving us in the somewhat perverse
    situation that jump-threading would work before running instcombine,
    but not after. Since instcombine produces this pattern, I think we
    need to consider it canonical and support it in JumpThreading.
    In general, for limiting recursion, I think the existing restriction
    to phi and cmp nodes should be sufficient to avoid looking through
    unprofitable chains of instructions.

    Reviewers: haicheng, gberry, bmakam, mcrosier

    Subscribers: llvm-commits

    Differential Revision: https://reviews.llvm.org/D42262

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-D42262-jumpthreading-not-i1.patch

6 years agoEnable support for floating-point division reductions
Yichao Yu [Sat, 10 Jun 2017 12:45:13 +0000 (08:45 -0400)]
Enable support for floating-point division reductions

Similar to fsub, fdiv can also be vectorized using fmul.

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-D34078-vectorize-fdiv.patch

6 years agoFix unwind info relocation with large code model on AArch64
Valentin Churavy [Thu, 5 Jul 2018 16:37:50 +0000 (12:37 -0400)]
Fix unwind info relocation with large code model on AArch64

Gbp-Pq: Topic julia
Gbp-Pq: Name llvm-D27629-AArch64-large_model_6.0.1.patch

6 years agoclang-arm-default-vfp3-on-armv7a
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
clang-arm-default-vfp3-on-armv7a

Gbp-Pq: Name clang-arm-default-vfp3-on-armv7a.patch

6 years agoD53557-hurd-self-exe-realpath
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
D53557-hurd-self-exe-realpath

[hurd] Make getMainExecutable get the real binary path

On GNU/Hurd, llvm-config is returning bogus value, such as:

$ llvm-config-6.0 --includedir
/usr/include

while it should be:
$ llvm-config-6.0 --includedir
/usr/lib/llvm-6.0/include

This is because getMainExecutable does not get the actual installation
path. On GNU/Hurd, /proc/self/exe is indeed a symlink to the path that
was used to start the program, and not the eventual binary file. Llvm's
getMainExecutable thus needs to run realpath over it to get the actual
place where llvm was installed (/usr/lib/llvm-6.0/bin/llvm-config), and
not /usr/bin/llvm-config-6.0. This will not change the result on Linux,
where /proc/self/exe already points to the eventual file.

Patch by Samuel Thibault!

While making changes here, I reformatted this block a bit to reduce
indentation and match 2 space indent style.

Differential Revision: https://reviews.llvm.org/D53557

Gbp-Pq: Name D53557-hurd-self-exe-realpath.diff

6 years agohurd-tools_llvm-shlib_CMakeLists.txt
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
hurd-tools_llvm-shlib_CMakeLists.txt

===================================================================

Gbp-Pq: Name hurd-tools_llvm-shlib_CMakeLists.txt.diff

6 years agohurd-lib_Support_Unix_Path.inc
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
hurd-lib_Support_Unix_Path.inc

===================================================================

Gbp-Pq: Name hurd-lib_Support_Unix_Path.inc.diff

6 years agox32-fix-driver-search-paths
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
x32-fix-driver-search-paths

Gbp-Pq: Name x32-fix-driver-search-paths.diff

6 years agopowerpcspe-add-missing-include-path
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
powerpcspe-add-missing-include-path

Gbp-Pq: Name powerpcspe-add-missing-include-path.diff

6 years agodisable-sse2-old-x86
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
disable-sse2-old-x86

===================================================================

Gbp-Pq: Name disable-sse2-old-x86.diff

6 years agorL338481-cherry-pick-really-subtle-miscompile
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
rL338481-cherry-pick-really-subtle-miscompile

===================================================================

Gbp-Pq: Name rL338481-cherry-pick-really-subtle-miscompile.diff

6 years agoD51639-optim-issue
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
D51639-optim-issue

===================================================================

Gbp-Pq: Name D51639-optim-issue.diff

6 years agoD51335-alignment-issue
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
D51335-alignment-issue

===================================================================

Gbp-Pq: Name D51335-alignment-issue.diff

6 years agopr38663-pgo-lto-crash
LLVM Packaging Team [Thu, 11 Jul 2019 03:08:05 +0000 (04:08 +0100)]
pr38663-pgo-lto-crash

===================================================================

Gbp-Pq: Name pr38663-pgo-lto-crash.patch