Merge pull request #4250 from facebook/x86_32_avx2

added a CI test for x86 32-bit + avx2 combination
facebook · Jan 19, 2025 · e475dc4 · e475dc4
2 parents a469e7c + d2d7461
commit e475dc4
Show file tree

Hide file tree

Showing 4 changed files with 31 additions and 9 deletions.
diff --git a/.github/workflows/dev-short-tests.yml b/.github/workflows/dev-short-tests.yml
@@ -164,7 +164,17 @@ jobs:
         make clean
         bash tests/libzstd_builds.sh
 
-  gcc-make-tests-32bit:
+  gcc-make-all-avx2:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # tag=v4.1.1
+    - name: Make all, with AVX2
+      run: |
+        sudo apt-get -qqq update
+        make libc6install
+        CFLAGS="-Werror -mavx2" make -j all
+
+  gcc-make-all-32bit:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # tag=v4.1.1
@@ -174,6 +184,17 @@ jobs:
         make libc6install
         CFLAGS="-Werror -m32" make -j all32
 
+  gcc-make-all-32bit-avx2:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # tag=v4.1.1
+    - name: Make all, 32bit + AVX2 mode
+      run: |
+        sudo apt-get -qqq update
+        make libc6install
+        CPPFLAGS="-DSTATIC_BMI2=1" CFLAGS="-Werror -m32 -mavx2 -mbmi2" make -j all32
+
+
   gcc-8-make:
     runs-on: ubuntu-latest
     steps:

diff --git a/lib/common/bitstream.h b/lib/common/bitstream.h
@@ -238,7 +238,7 @@ 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; /* overflow detected */
-    return (bitC->ptr - bitC->startPtr) + (bitC->bitPos > 0);
+    return (size_t)(bitC->ptr - bitC->startPtr) + (bitC->bitPos > 0);
 }
 
 

diff --git a/lib/common/compiler.h b/lib/common/compiler.h
@@ -207,10 +207,10 @@
 #  pragma warning(disable : 4324)        /* disable: C4324: padded structure */
 #endif
 
-/*Like DYNAMIC_BMI2 but for compile time determination of BMI2 support*/
+/* Like DYNAMIC_BMI2 but for compile time determination of BMI2 support */
 #ifndef STATIC_BMI2
 #  if defined(_MSC_VER) && (defined(_M_X64) || defined(_M_I86))
-#    ifdef __AVX2__  //MSVC does not have a BMI2 specific flag, but every CPU that supports AVX2 also supports BMI2
+#    ifdef __AVX2__  /* MSVC does not have a BMI2 specific flag, but every CPU that supports AVX2 also supports BMI2 */
 #       define STATIC_BMI2 1
 #    endif
 #  elif defined(__BMI2__) && defined(__x86_64__) && defined(__GNUC__)
@@ -219,7 +219,7 @@
 #endif
 
 #ifndef STATIC_BMI2
-    #define STATIC_BMI2 0
+#  define STATIC_BMI2 0
 #endif
 
 /* compile time determination of SIMD support */

diff --git a/lib/compress/zstd_compress.c b/lib/compress/zstd_compress.c
@@ -7125,7 +7125,7 @@ size_t ZSTD_compressSequences(ZSTD_CCtx* cctx,
  * @returns > 0 if there is one long length (> 65535),
  * indicating the position, and type.
  */
-size_t convertSequences_noRepcodes(
+static size_t convertSequences_noRepcodes(
     SeqDef* dstSeqs,
     const ZSTD_Sequence* inSeqs,
     size_t nbSequences)
@@ -7277,9 +7277,10 @@ size_t convertSequences_noRepcodes(
 
 #else /* no AVX2 */
 
-static size_t
-convertSequences_noRepcodes(SeqDef* dstSeqs,
-                const ZSTD_Sequence* const inSeqs, size_t nbSequences)
+static size_t convertSequences_noRepcodes(
+    SeqDef* dstSeqs,
+    const ZSTD_Sequence* inSeqs,
+    size_t nbSequences)
 {
     size_t longLen = 0;
     size_t n;