enhanced the atomic yield patch (SSE2/arm/powerpc)

diff --git a/debian/changelog b/debian/changelog
index 72dbb159db2df1218e55dfe29763bd3d84a24b97..38183bedd3a362bece0e3bf55a2dee8f594ca348 100644 (file)
--- a/debian/changelog
+++ b/debian/changelog
@@ -33,6 +33,10 @@ solvespace (3.0.rc2+repack1-1) UNRELEASED; urgency=medium
   * Do not parse d/changelog (routine-update)
   * Add salsa-ci file (routine-update)
   * debian/upstream/metadata: mention upstream's GitHub repository
+  * debian/patches/03_use_system_threejs.patch: refreshed
+  * debian/patches/10_mimalloc_restrict_cpu_yield.patch: enhanced with better
+    atomic yield for SSE2 (if available) and better/added ops for arm64,
+    armel, and ppc64{el,}/powerpc. Closes: #980169
 
  -- Michael R. Crusoe <crusoe@debian.org>  Sat, 23 Jan 2021 15:40:21 +0100
 

diff --git a/debian/patches/03_use_system_threejs.patch b/debian/patches/03_use_system_threejs.patch
index 46d8b2f39b55f3160e80866bfb8117a5695838e4..d4b206ee911a2e8ba289ced6ff77bab84d6fa1b3 100644 (file)
--- a/debian/patches/03_use_system_threejs.patch
+++ b/debian/patches/03_use_system_threejs.patch
@@ -10,11 +10,9 @@ Date: Tue, 19 Jan 2021 15:05:06 -0600
  src/export.cpp     | 2 +-
  2 files changed, 1 insertion(+), 2 deletions(-)
 
-diff --git a/res/CMakeLists.txt b/res/CMakeLists.txt
-index 5969039..348f812 100644
---- a/res/CMakeLists.txt
-+++ b/res/CMakeLists.txt
-@@ -282,7 +282,6 @@ add_resources(
+--- solvespace.orig/res/CMakeLists.txt
++++ solvespace/res/CMakeLists.txt
+@@ -284,7 +284,6 @@
      shaders/edge.frag
      shaders/edge.vert
      shaders/outline.vert
@@ -22,11 +20,9 @@ index 5969039..348f812 100644
      threejs/hammer-2.0.8.js.gz
      threejs/SolveSpaceControls.js)
  
-diff --git a/src/export.cpp b/src/export.cpp
-index e0c6182..6a670b9 100644
---- a/src/export.cpp
-+++ b/src/export.cpp
-@@ -1064,7 +1064,7 @@ void SolveSpaceUI::ExportMeshAsThreeJsTo(FILE *f, const Platform::Path &filename
+--- solvespace.orig/src/export.cpp
++++ solvespace/src/export.cpp
+@@ -1014,7 +1014,7 @@
  
      if(filename.HasExtension("html")) {
          fprintf(f, htmlbegin,

diff --git a/debian/patches/10_mimalloc_restrict_cpu_yield.patch b/debian/patches/10_mimalloc_restrict_cpu_yield.patch
index aeca59606614260bb9c922dbf3d7f1f3d26fd448..7579cc8960a6c3f221d782eb0324ddefd9134e64 100644 (file)
--- a/debian/patches/10_mimalloc_restrict_cpu_yield.patch
+++ b/debian/patches/10_mimalloc_restrict_cpu_yield.patch
@@ -1,43 +1,61 @@
 Subject: Restricts cpu yield instructions a little.
 From: Ryan Pavlik <ryan.pavlik@collabora.com>
-
-Date: Tue, 19 Jan 2021 13:46:43 -0600
+Last-Updated: 2021-01-23
+Forwarded: https://github.com/microsoft/mimalloc/pull/350
 
 adding clobber for ARM and preventing older 32 bits chips not supporting this instruction.
 
 Backport of https://github.com/microsoft/mimalloc/commit/33a10b48605f8bb419487a03125815ad6ee00a70
 
+Enhanced via https://github.com/mr-c/misc/wiki/pause-techniques-on-many-architectures
+
 ---
  extlib/mimalloc/include/mimalloc-atomic.h | 12 +++++++++---
  1 file changed, 9 insertions(+), 3 deletions(-)
 
-diff --git a/extlib/mimalloc/include/mimalloc-atomic.h b/extlib/mimalloc/include/mimalloc-atomic.h
-index 722b6ad..94010d4 100644
---- a/extlib/mimalloc/include/mimalloc-atomic.h
-+++ b/extlib/mimalloc/include/mimalloc-atomic.h
-@@ -267,16 +267,22 @@ static inline void mi_atomic_maxi64(volatile int64_t* p, int64_t x) {
+--- solvespace.orig/extlib/mimalloc/include/mimalloc-atomic.h
++++ solvespace/extlib/mimalloc/include/mimalloc-atomic.h
+@@ -266,17 +266,37 @@
+   static inline void mi_atomic_yield(void) {
      std::this_thread::yield();
    }
- #elif (defined(__GNUC__) || defined(__clang__)) && \
+-#elif (defined(__GNUC__) || defined(__clang__)) && \
 -      (defined(__x86_64__) || defined(__i386__) || defined(__arm__) || defined(__aarch64__))
-+      (defined(__x86_64__) || defined(__i386__) || (defined(__arm__) && __ARM_ARCH__ >= 7) || defined(__aarch64__))
- #if defined(__x86_64__) || defined(__i386__)
+-#if defined(__x86_64__) || defined(__i386__)
++#elif defined(__SSE2__)  // AMD and Intel
++  #include <emmintrin.h>
++  static inline void mi_atomic_yield(void) {
++    _mm_pause();
++  }
++#elif defined(__x86_64__) || defined(__i386__)
    static inline void mi_atomic_yield(void) {
      asm volatile ("pause" ::: "memory");
    }
 -#elif defined(__arm__) || defined(__aarch64__)
-+#elif (defined(__arm__) && __ARM_ARCH__ >= 7) || defined(__aarch64__)
-   static inline void mi_atomic_yield(void) {
--    asm volatile("yield");
++#elif defined(__aarch64__)
++  static inline void mi_atomic_yield(void) {
++    asm volatile("wfe");
++  }
++#elif defined(__arm__) && __ARM_ARCH__ >= 7
++  static inline void mi_atomic_yield(void) {
 +    __asm__ volatile("yield" ::: "memory");
-   }
- #endif
++  }
++#elif defined(__armel__) || defined(__ARMEL__)
++  static inline void mi_atomic_yield(void) {
++    asm volatile ("nop" ::: "memory");  // default operation - does nothing => Might lead to passive spinning.
++  }
++#elif defined(__powerpc__) || defined(__ppc__) || defined(__PPC__) // PowerPC
++  static inline void mi_atomic_yield(void) {
++     __asm__ __volatile__ ("or 27,27,27" ::: "memory");
++  }
 +#elif defined(__sun)
 +  // Fallback for other archs
 +  #include <synch.h>
-+  static inline void mi_atomic_yield(void) {
+   static inline void mi_atomic_yield(void) {
+-    asm volatile("yield");
 +    smt_pause();
-+  }
+   }
+-#endif
  #elif defined(__wasi__)
    #include <sched.h>
    static inline void mi_atomic_yield(void) {