Address review and fix test

Author: jiepan-intel

site/source/docs/porting/simd.rst

@@ -1220,7 +1220,9 @@ The following table highlights the availability and expected performance of diff
    * - _mm_i64gather_epi64
      - ❌ scalarized
 
-All the 128-bit wide instructions from AVX2 instruction set are listed. Only a small part of the 256-bit AVX2 instruction set are listed, most of the 256-bit wide AVX2 instructions are emulated by two 128-bit wide instructions.
+All the 128-bit wide instructions from AVX2 instruction set are listed.
+Only a small part of the 256-bit AVX2 instruction set are listed, most of the
+256-bit wide AVX2 instructions are emulated by two 128-bit wide instructions.
 
 ====================================================== 
 Compiling SIMD code targeting ARM NEON instruction set

test/sse/test_avx2.cpp

@@ -262,7 +262,8 @@ void test_statisticsa(void) {
   Ret_M256i_M256i(__m256i, _mm256_avg_epu8);
 }
 
-void test_shift(void) {
+// Split test_shift into two functions to reduce memory consumption
+void test_shift1(void) {
   Ret_M256i_Tint(__m256i, _mm256_slli_si256);
   Ret_M256i_Tint(__m256i, _mm256_bslli_epi128);
 
@@ -277,7 +278,9 @@ void test_shift(void) {
   Ret_M256i_Tint(__m256i, _mm256_srai_epi32);
   Ret_M256i_M128i(__m256i, _mm256_sra_epi16);
   Ret_M256i_M128i(__m256i, _mm256_sra_epi32);
+}
 
+void test_shift2(void) {
   Ret_M256i_Tint(__m256i, _mm256_srli_si256);
   Ret_M256i_Tint(__m256i, _mm256_bsrli_epi128);
 
@@ -302,6 +305,11 @@ void test_shift(void) {
   Ret_M256i_M256i(__m256i, _mm256_srlv_epi64);
 }
 
+void test_shift(void) {
+  test_shift1();
+  test_shift2();
+}
+
 int main() {
   assert(numInterestingFloats % 8 == 0);
   assert(numInterestingInts % 8 == 0);

test/test_other.py

@@ -9465,7 +9465,8 @@ def test_standalone_system_headers(self):
         print('header: ' + header)
         # These headers cannot be included in isolation.
         # e.g: error: unknown type name 'EGLDisplay'
-        if header in ['eglext.h', 'SDL_config_macosx.h', 'glext.h', 'gl2ext.h']:
+        # Don't include avxintrin.h and avx2inrin.h directly, include immintrin.h instead
+        if header in ['eglext.h', 'SDL_config_macosx.h', 'glext.h', 'gl2ext.h', 'avxintrin.h', 'avx2intrin.h']:
           continue
         # These headers are C++ only and cannot be included from C code.
         # But we still want to check they can be included on there own without
@@ -9474,12 +9475,14 @@ def test_standalone_system_headers(self):
           'wire.h', 'val.h', 'bind.h',
           'webgpu_cpp.h', 'webgpu_cpp_chained_struct.h', 'webgpu_enum_class_bitmasks.h',
           # Some headers are not yet C compatible
-          'arm_neon.h', 'avxintrin.h', 'immintrin.h',
+          'arm_neon.h', 'immintrin.h',
         ]
         if directory and directory != 'compat':
           header = f'{directory}/{header}'
         inc = f'#include <{header}>\n__attribute__((weak)) int foo;\n'
-        cflags = ['-Werror', '-Wall', '-pedantic', '-mavx', '-msimd128', '-msse3']
+        cflags = ['-Werror', '-Wall', '-pedantic', '-msimd128', '-msse4']
+        if header == 'immintrin.h':
+          cflags.append('-mavx2')
         if cxx_only:
           create_file('a.cxx', inc)
           create_file('b.cxx', inc)