Attachment 7477 Details for Bug 34722 – Debian's patch to disable sse2

[patch] Debian's patch to disable sse2

disable-sse2 (text/plain), 12.75 KB, created by mikhailnov on 2018-04-01 00:51:53 MSK

(hide)

Description:

Filename:

MIME Type:

Creator: mikhailnov

Created: 2018-04-01 00:51:53 MSK

Size: 12.75 KB

patch

obsolete

>description: don't set sse2 compiler flags on i386
>author: Michael Gilbert <mgilbert@debian.org>
>debian-bug: http://bugs.debian.org/750361
>
>--- a/cc/raster/texture_compressor.cc
>+++ b/cc/raster/texture_compressor.cc
>@@ -8,7 +8,7 @@
> #include "base/memory/ptr_util.h"
> #include "cc/raster/texture_compressor_etc1.h"
> 
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
> #include "base/cpu.h"
> #include "cc/raster/texture_compressor_etc1_sse.h"
> #endif
>@@ -18,7 +18,7 @@ namespace cc {
> std::unique_ptr<TextureCompressor> TextureCompressor::Create(Format format) {
>   switch (format) {
>     case kFormatETC1: {
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>       base::CPU cpu;
>       if (cpu.has_sse2()) {
>         return base::WrapUnique(new TextureCompressorETC1SSE());
>--- a/media/base/sinc_resampler.cc
>+++ b/media/base/sinc_resampler.cc
>@@ -81,7 +81,7 @@
> #include "base/numerics/math_constants.h"
> #include "build/build_config.h"
> 
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(__x86_64__)
> #include <xmmintrin.h>
> #define CONVOLVE_FUNC Convolve_SSE
> #elif defined(ARCH_CPU_ARM_FAMILY) && defined(USE_NEON)
>@@ -326,7 +326,7 @@ float SincResampler::Convolve_C(const fl
>       kernel_interpolation_factor * sum2);
> }
> 
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(__x86_64__)
> float SincResampler::Convolve_SSE(const float* input_ptr, const float* k1,
>                                   const float* k2,
>                                   double kernel_interpolation_factor) {
>--- a/media/base/sinc_resampler.h
>+++ b/media/base/sinc_resampler.h
>@@ -98,7 +98,7 @@ class MEDIA_EXPORT SincResampler {
>   // ARM, NEON support is chosen at compile time based on compilation flags.
>   static float Convolve_C(const float* input_ptr, const float* k1,
>                           const float* k2, double kernel_interpolation_factor);
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(__x86_64__)
>   static float Convolve_SSE(const float* input_ptr, const float* k1,
>                             const float* k2,
>                             double kernel_interpolation_factor);
>--- a/media/base/sinc_resampler_perftest.cc
>+++ b/media/base/sinc_resampler_perftest.cc
>@@ -22,7 +22,7 @@ static const double kKernelInterpolation
> static void DoNothing(int frames, float* destination) {}
> 
> // Define platform independent function name for Convolve* tests.
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(__x86_64__)
> #define CONVOLVE_FUNC Convolve_SSE
> #elif defined(ARCH_CPU_ARM_FAMILY) && defined(USE_NEON)
> #define CONVOLVE_FUNC Convolve_NEON
>--- a/media/base/sinc_resampler_unittest.cc
>+++ b/media/base/sinc_resampler_unittest.cc
>@@ -150,7 +150,7 @@ TEST(SincResamplerTest, DISABLED_SetRati
> 
> 
> // Define platform independent function name for Convolve* tests.
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(__x86_64__)
> #define CONVOLVE_FUNC Convolve_SSE
> #elif defined(ARCH_CPU_ARM_FAMILY) && defined(USE_NEON)
> #define CONVOLVE_FUNC Convolve_NEON
>--- a/media/base/vector_math.cc
>+++ b/media/base/vector_math.cc
>@@ -11,7 +11,7 @@
> #include "build/build_config.h"
> 
> // NaCl does not allow intrinsics.
>-#if defined(ARCH_CPU_X86_FAMILY) && !defined(OS_NACL)
>+#if defined(__x86_64__) && !defined(OS_NACL)
> #include <xmmintrin.h>
> // Don't use custom SSE versions where the auto-vectorized C version performs
> // better, which is anywhere clang is used.
>@@ -91,7 +91,7 @@ std::pair<float, float> EWMAAndMaxPower_
>   return result;
> }
> 
>-#if defined(ARCH_CPU_X86_FAMILY) && !defined(OS_NACL)
>+#if defined(__x86_64__) && !defined(OS_NACL)
> void FMUL_SSE(const float src[], float scale, int len, float dest[]) {
>   const int rem = len % 4;
>   const int last_index = len - rem;
>--- a/media/base/vector_math_perftest.cc
>+++ b/media/base/vector_math_perftest.cc
>@@ -83,7 +83,7 @@ class VectorMathPerfTest : public testin
> };
> 
> // Define platform dependent function names for SIMD optimized methods.
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(__x86_64__)
> #define FMAC_FUNC FMAC_SSE
> #define FMUL_FUNC FMUL_SSE
> #define EWMAAndMaxPower_FUNC EWMAAndMaxPower_SSE
>--- a/media/base/vector_math_testing.h
>+++ b/media/base/vector_math_testing.h
>@@ -28,7 +28,7 @@ MEDIA_SHMEM_EXPORT std::pair<float, floa
>     int len,
>     float smoothing_factor);
> 
>-#if defined(ARCH_CPU_X86_FAMILY) && !defined(OS_NACL)
>+#if defined(__x86_64__) && !defined(OS_NACL)
> MEDIA_SHMEM_EXPORT void FMAC_SSE(const float src[],
>                                  float scale,
>                                  int len,
>--- a/media/base/vector_math_unittest.cc
>+++ b/media/base/vector_math_unittest.cc
>@@ -73,7 +73,7 @@ TEST_F(VectorMathTest, FMAC) {
>     VerifyOutput(kResult);
>   }
> 
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(__x86_64__)
>   {
>     SCOPED_TRACE("FMAC_SSE");
>     FillTestVectors(kInputFillValue, kOutputFillValue);
>@@ -114,7 +114,7 @@ TEST_F(VectorMathTest, FMUL) {
>     VerifyOutput(kResult);
>   }
> 
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(__x86_64__)
>   {
>     SCOPED_TRACE("FMUL_SSE");
>     FillTestVectors(kInputFillValue, kOutputFillValue);
>@@ -222,7 +222,7 @@ class EWMATestScenario {
>       EXPECT_NEAR(expected_max_, result.second, 0.0000001f);
>     }
> 
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(__x86_64__)
>     {
>       SCOPED_TRACE("EWMAAndMaxPower_SSE");
>       const std::pair<float, float>& result = vector_math::EWMAAndMaxPower_SSE(
>--- a/skia/ext/convolver.h
>+++ b/skia/ext/convolver.h
>@@ -17,7 +17,6 @@
> // We can build SSE2 optimized versions for all x86 CPUs
> // except when building for the IOS emulator.
> #if defined(ARCH_CPU_X86_FAMILY) && !defined(OS_IOS)
>-#define SIMD_SSE2 1
> #define SIMD_PADDING 8  // 8 * int16_t
> #endif
> 
>--- a/third_party/WebKit/Source/platform/audio/DirectConvolver.cpp
>+++ b/third_party/WebKit/Source/platform/audio/DirectConvolver.cpp
>@@ -35,7 +35,7 @@
> #include <Accelerate/Accelerate.h>
> #endif
> 
>-#if defined(ARCH_CPU_X86_FAMILY) && !defined(OS_MACOSX)
>+#if defined(ARCH_CPU_X86_64) && !defined(OS_MACOSX)
> #include <emmintrin.h>
> #endif
> 
>@@ -83,7 +83,7 @@ void DirectConvolver::Process(AudioFloat
> #endif  // ARCH_CPU_X86
> #else
>   size_t i = 0;
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>   // Convolution using SSE2. Currently only do this if both |kernelSize| and
>   // |framesToProcess| are multiples of 4. If not, use the straightforward loop
>   // below.
>@@ -397,7 +397,7 @@ void DirectConvolver::Process(AudioFloat
>       }
>       dest_p[i++] = sum;
>     }
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>   }
> #endif
> #endif  // OS_MACOSX
>--- a/third_party/WebKit/Source/platform/audio/SincResampler.cpp
>+++ b/third_party/WebKit/Source/platform/audio/SincResampler.cpp
>@@ -32,7 +32,7 @@
> #include "platform/audio/AudioBus.h"
> #include "platform/wtf/MathExtras.h"
> 
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
> #include <emmintrin.h>
> #endif
> 
>@@ -277,7 +277,7 @@ void SincResampler::Process(AudioSourceP
>       {
>         float input;
> 
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>         // If the sourceP address is not 16-byte aligned, the first several
>         // frames (at most three) should be processed seperately.
>         while ((reinterpret_cast<uintptr_t>(input_p) & 0x0F) && n) {
>--- a/third_party/WebKit/Source/platform/audio/VectorMath.cpp
>+++ b/third_party/WebKit/Source/platform/audio/VectorMath.cpp
>@@ -37,7 +37,7 @@
> #include "platform/audio/cpu/arm/VectorMathNEON.h"
> #elif HAVE_MIPS_MSA_INTRINSICS
> #include "platform/audio/cpu/mips/VectorMathMSA.h"
>-#elif defined(ARCH_CPU_X86_FAMILY)
>+#elif defined(ARCH_CPU_X86_64)
> #include "platform/audio/cpu/x86/VectorMathX86.h"
> #else
> #include "platform/audio/VectorMathScalar.h"
>@@ -54,7 +54,7 @@ namespace Impl = Mac;
> namespace Impl = NEON;
> #elif HAVE_MIPS_MSA_INTRINSICS
> namespace Impl = MSA;
>-#elif defined(ARCH_CPU_X86_FAMILY)
>+#elif defined(ARCH_CPU_X86_64)
> namespace Impl = X86;
> #else
> namespace Impl = Scalar;
>--- a/third_party/WebKit/Source/platform/graphics/cpu/x86/WebGLImageConversionSSE.h
>+++ b/third_party/WebKit/Source/platform/graphics/cpu/x86/WebGLImageConversionSSE.h
>@@ -7,7 +7,7 @@
> 
> #include "build/build_config.h"
> 
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
> #include <emmintrin.h>
> 
> namespace blink {
>@@ -195,6 +195,6 @@ ALWAYS_INLINE void PackOneRowOfRGBA8Litt
> }  // namespace SIMD
> }  // namespace blink
> 
>-#endif  // ARCH_CPU_X86_FAMILY
>+#endif  // ARCH_CPU_X86_64
> 
> #endif  // WebGLImageConversionSSE_h
>--- a/third_party/WebKit/Source/platform/graphics/gpu/WebGLImageConversion.cpp
>+++ b/third_party/WebKit/Source/platform/graphics/gpu/WebGLImageConversion.cpp
>@@ -444,7 +444,7 @@ void Unpack<WebGLImageConversion::kDataF
>   const uint32_t* source32 = reinterpret_cast_ptr<const uint32_t*>(source);
>   uint32_t* destination32 = reinterpret_cast_ptr<uint32_t*>(destination);
> 
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>   SIMD::UnpackOneRowOfBGRA8LittleToRGBA8(source32, destination32,
>                                          pixels_per_row);
> #endif
>@@ -472,7 +472,7 @@ void Unpack<WebGLImageConversion::kDataF
>     const uint16_t* source,
>     uint8_t* destination,
>     unsigned pixels_per_row) {
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>   SIMD::UnpackOneRowOfRGBA5551LittleToRGBA8(source, destination,
>                                             pixels_per_row);
> #endif
>@@ -502,7 +502,7 @@ void Unpack<WebGLImageConversion::kDataF
>     const uint16_t* source,
>     uint8_t* destination,
>     unsigned pixels_per_row) {
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>   SIMD::UnpackOneRowOfRGBA4444LittleToRGBA8(source, destination,
>                                             pixels_per_row);
> #endif
>@@ -718,7 +718,7 @@ void Pack<WebGLImageConversion::kDataFor
>           uint8_t>(const uint8_t* source,
>                    uint8_t* destination,
>                    unsigned pixels_per_row) {
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>   SIMD::PackOneRowOfRGBA8LittleToR8(source, destination, pixels_per_row);
> #endif
> #if HAVE_MIPS_MSA_INTRINSICS
>@@ -775,7 +775,7 @@ void Pack<WebGLImageConversion::kDataFor
>           uint8_t>(const uint8_t* source,
>                    uint8_t* destination,
>                    unsigned pixels_per_row) {
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>   SIMD::PackOneRowOfRGBA8LittleToRA8(source, destination, pixels_per_row);
> #endif
> #if HAVE_MIPS_MSA_INTRINSICS
>@@ -887,7 +887,7 @@ void Pack<WebGLImageConversion::kDataFor
>           uint8_t>(const uint8_t* source,
>                    uint8_t* destination,
>                    unsigned pixels_per_row) {
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>   SIMD::PackOneRowOfRGBA8LittleToRGBA8(source, destination, pixels_per_row);
> #endif
> #if HAVE_MIPS_MSA_INTRINSICS
>--- a/third_party/WebKit/Source/modules/webaudio/AudioParamTimeline.cpp
>+++ b/third_party/WebKit/Source/modules/webaudio/AudioParamTimeline.cpp
>@@ -39,7 +39,7 @@
> #include "platform/wtf/CPU.h"
> #include "platform/wtf/MathExtras.h"
> 
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
> #include <emmintrin.h>
> #endif
> 
>@@ -1333,7 +1333,7 @@ std::tuple<size_t, float, unsigned> Audi
>     size_t current_frame,
>     float value,
>     unsigned write_index) {
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>   auto number_of_values = current_state.number_of_values;
> #endif
>   auto fill_to_frame = current_state.fill_to_frame;
>@@ -1346,7 +1346,7 @@ std::tuple<size_t, float, unsigned> Audi
>   double delta_time = time2 - time1;
>   float k = delta_time > 0 ? 1 / delta_time : 0;
>   const float value_delta = value2 - value1;
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>   if (fill_to_frame > write_index) {
>     // Minimize in-loop operations. Calculate starting value and increment.
>     // Next step: value += inc.
>@@ -1474,7 +1474,7 @@ std::tuple<size_t, float, unsigned> Audi
>     size_t current_frame,
>     float value,
>     unsigned write_index) {
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>   auto number_of_values = current_state.number_of_values;
> #endif
>   auto fill_to_frame = current_state.fill_to_frame;
>@@ -1524,7 +1524,7 @@ std::tuple<size_t, float, unsigned> Audi
>     for (; write_index < fill_to_frame; ++write_index)
>       values[write_index] = target;
>   } else {
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>     if (fill_to_frame > write_index) {
>       // Resolve recursion by expanding constants to achieve a 4-step
>       // loop unrolling.
>@@ -1658,7 +1658,7 @@ std::tuple<size_t, float, unsigned> Audi
>   // Oversampled curve data can be provided if sharp discontinuities are
>   // desired.
>   unsigned k = 0;
>-#if defined(ARCH_CPU_X86_FAMILY)
>+#if defined(ARCH_CPU_X86_64)
>   if (fill_to_frame > write_index) {
>     const __m128 v_curve_virtual_index = _mm_set_ps1(curve_virtual_index);
>     const __m128 v_curve_points_per_frame = _mm_set_ps1(curve_points_per_frame);

Actions: View | Diff

Attachments on bug 34722: 7457 | 7474 | 7475 | 7477 | 7480 | 7481