Magic-Mirror/gecko-dev
1
0
Fork 0
mirror of https://github.com/mozilla/gecko-dev.git synced 2024-11-26 22:32:46 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Bug 1916038 - Add volatile for gcc inline to avoid being removed. r=gfx-reviewers,nical

Browse Source 
This extends the fix upstream did in 616bee5420b62a7be09fda0252034e8be85f91b0,
which was not enough.

Differential Revision: https://phabricator.services.mozilla.com/D221275
This commit is contained in:
Mike Hommey 2024-09-09 20:59:03 +00:00
parent 8e3f63734a
commit 01e85f6dea
6 changed files with 976 additions and 95 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

875
media/libyuv/04_add_missing_volatile.patch Normal file
View File

@ -0,0 +1,875 @@
diff --git a/include/libyuv/macros_msa.h b/include/libyuv/macros_msa.h
index 6434a4da0537c..08e8c82927dd0 100644
--- a/include/libyuv/macros_msa.h
+++ b/include/libyuv/macros_msa.h
@@ -20,7 +20,7 @@
({ \
const uint8_t* psrc_lw_m = (const uint8_t*)(psrc); \
uint32_t val_m; \
- asm("lw %[val_m], %[psrc_lw_m] \n" \
+ asm volatile("lw %[val_m], %[psrc_lw_m] \n" \
: [val_m] "=r"(val_m) \
: [psrc_lw_m] "m"(*psrc_lw_m)); \
val_m; \
@@ -31,7 +31,7 @@
({ \
const uint8_t* psrc_ld_m = (const uint8_t*)(psrc); \
uint64_t val_m = 0; \
- asm("ld %[val_m], %[psrc_ld_m] \n" \
+ asm volatile("ld %[val_m], %[psrc_ld_m] \n" \
: [val_m] "=r"(val_m) \
: [psrc_ld_m] "m"(*psrc_ld_m)); \
val_m; \
@@ -55,7 +55,7 @@
({ \
uint8_t* pdst_sw_m = (uint8_t*)(pdst); /* NOLINT */ \
uint32_t val_m = (val); \
- asm("sw %[val_m], %[pdst_sw_m] \n" \
+ asm volatile("sw %[val_m], %[pdst_sw_m] \n" \
: [pdst_sw_m] "=m"(*pdst_sw_m) \
: [val_m] "r"(val_m)); \
})
@@ -65,7 +65,7 @@
({ \
uint8_t* pdst_sd_m = (uint8_t*)(pdst); /* NOLINT */ \
uint64_t val_m = (val); \
- asm("sd %[val_m], %[pdst_sd_m] \n" \
+ asm volatile("sd %[val_m], %[pdst_sd_m] \n" \
: [pdst_sd_m] "=m"(*pdst_sd_m) \
: [val_m] "r"(val_m)); \
})
@@ -86,7 +86,8 @@
uint8_t* psrc_lw_m = (uint8_t*)(psrc); \
uint32_t val_lw_m; \
\
- asm("lwr %[val_lw_m], 0(%[psrc_lw_m]) \n\t" \
+ asm volatile( \
+ "lwr %[val_lw_m], 0(%[psrc_lw_m]) \n\t" \
"lwl %[val_lw_m], 3(%[psrc_lw_m]) \n\t" \
\
: [val_lw_m] "=&r"(val_lw_m) \
@@ -101,7 +102,8 @@
uint8_t* psrc_ld_m = (uint8_t*)(psrc); \
uint64_t val_ld_m = 0; \
\
- asm("ldr %[val_ld_m], 0(%[psrc_ld_m]) \n\t" \
+ asm volatile( \
+ "ldr %[val_ld_m], 0(%[psrc_ld_m]) \n\t" \
"ldl %[val_ld_m], 7(%[psrc_ld_m]) \n\t" \
\
: [val_ld_m] "=&r"(val_ld_m) \
@@ -128,7 +130,7 @@
({ \
uint8_t* pdst_sw_m = (uint8_t*)(pdst); /* NOLINT */ \
uint32_t val_m = (val); \
- asm("usw %[val_m], %[pdst_sw_m] \n" \
+ asm volatile("usw %[val_m], %[pdst_sw_m] \n" \
: [pdst_sw_m] "=m"(*pdst_sw_m) \
: [val_m] "r"(val_m)); \
})
diff --git a/source/row_gcc.cc b/source/row_gcc.cc
index f8f41860ab7c5..6eb3286b053ad 100644
--- a/source/row_gcc.cc
+++ b/source/row_gcc.cc
@@ -2626,7 +2626,7 @@ void OMITFP I444ToARGBRow_SSSE3(const uint8_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@@ -2686,7 +2686,7 @@ void OMITFP I422ToRGB24Row_SSSE3(const uint8_t* y_buf,
uint8_t* dst_rgb24,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"movdqa %[kShuffleMaskARGBToRGB24_0],%%xmm5 \n"
"movdqa %[kShuffleMaskARGBToRGB24],%%xmm6 \n"
@@ -2722,7 +2722,7 @@ void OMITFP I444ToRGB24Row_SSSE3(const uint8_t* y_buf,
uint8_t* dst_rgb24,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"movdqa %[kShuffleMaskARGBToRGB24_0],%%xmm5 \n"
"movdqa %[kShuffleMaskARGBToRGB24],%%xmm6 \n"
@@ -2758,7 +2758,7 @@ void OMITFP I422ToARGBRow_SSSE3(const uint8_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@@ -2787,7 +2787,7 @@ void OMITFP I422ToAR30Row_SSSE3(const uint8_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n" // AR30 constants
@@ -2822,7 +2822,7 @@ void OMITFP I210ToARGBRow_SSSE3(const uint16_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@@ -2852,7 +2852,7 @@ void OMITFP I212ToARGBRow_SSSE3(const uint16_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@@ -2882,7 +2882,7 @@ void OMITFP I210ToAR30Row_SSSE3(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@@ -2917,7 +2917,7 @@ void OMITFP I212ToAR30Row_SSSE3(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@@ -2952,7 +2952,7 @@ void OMITFP I410ToARGBRow_SSSE3(const uint16_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@@ -3045,7 +3045,7 @@ void OMITFP I410ToAR30Row_SSSE3(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@@ -3238,7 +3238,7 @@ void OMITFP P210ToAR30Row_SSSE3(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $14,%%xmm5 \n"
@@ -3269,7 +3269,7 @@ void OMITFP P410ToAR30Row_SSSE3(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $14,%%xmm5 \n"
@@ -3301,7 +3301,7 @@ void OMITFP I422ToRGBARow_SSSE3(const uint8_t* y_buf,
uint8_t* dst_rgba,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@@ -3712,7 +3712,7 @@ void OMITFP I444ToARGBRow_AVX2(const uint8_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
@@ -3746,7 +3746,7 @@ void OMITFP I422ToARGBRow_AVX2(const uint8_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
@@ -3786,7 +3786,7 @@ void OMITFP I422ToARGBRow_AVX512BW(const uint8_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP_AVX512BW(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%xmm5,%%xmm5,%%xmm5 \n"
@@ -3825,7 +3825,7 @@ void OMITFP I422ToAR30Row_AVX2(const uint8_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n" // AR30 constants
@@ -3865,7 +3865,7 @@ void OMITFP I210ToARGBRow_AVX2(const uint16_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
@@ -3900,7 +3900,7 @@ void OMITFP I212ToARGBRow_AVX2(const uint16_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
@@ -3935,7 +3935,7 @@ void OMITFP I210ToAR30Row_AVX2(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n" // AR30 constants
@@ -3975,7 +3975,7 @@ void OMITFP I212ToAR30Row_AVX2(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n" // AR30 constants
@@ -4015,7 +4015,7 @@ void OMITFP I410ToARGBRow_AVX2(const uint16_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
@@ -4120,7 +4120,7 @@ void OMITFP I410ToAR30Row_AVX2(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n" // AR30 constants
@@ -4228,7 +4228,7 @@ void OMITFP I422ToRGBARow_AVX2(const uint8_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
@@ -4430,7 +4430,7 @@ void OMITFP P210ToAR30Row_AVX2(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n" // AR30 constants
"vpsrlw $14,%%ymm5,%%ymm5 \n"
@@ -4467,7 +4467,7 @@ void OMITFP P410ToAR30Row_AVX2(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
- asm (
+ asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n" // AR30 constants
"vpsrlw $14,%%ymm5,%%ymm5 \n"
@@ -5681,7 +5681,7 @@ void MergeXRGBRow_AVX2(const uint8_t* src_r,
const uint8_t* src_b,
uint8_t* dst_argb,
int width) {
- asm(
+ asm volatile(
LABELALIGN
"1: \n"
@@ -7381,7 +7381,7 @@ void ARGBUnattenuateRow_SSE2(const uint8_t* src_argb,
uint8_t* dst_argb,
int width) {
uintptr_t alpha;
- asm(
+ asm volatile(
// 4 pixel loop.
LABELALIGN
"1: \n"
@@ -7841,7 +7841,7 @@ void ARGBAddRow_SSE2(const uint8_t* src_argb,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width) {
- asm(
+ asm volatile(
// 4 pixel loop.
LABELALIGN
"1: \n"
@@ -7869,7 +7869,7 @@ void ARGBAddRow_AVX2(const uint8_t* src_argb,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width) {
- asm(
+ asm volatile(
// 4 pixel loop.
LABELALIGN
"1: \n"
@@ -7897,7 +7897,7 @@ void ARGBSubtractRow_SSE2(const uint8_t* src_argb,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width) {
- asm(
+ asm volatile(
// 4 pixel loop.
LABELALIGN
"1: \n"
@@ -7925,7 +7925,7 @@ void ARGBSubtractRow_AVX2(const uint8_t* src_argb,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width) {
- asm(
+ asm volatile(
// 4 pixel loop.
LABELALIGN
"1: \n"
@@ -9099,7 +9099,7 @@ void ARGBColorTableRow_X86(uint8_t* dst_argb,
const uint8_t* table_argb,
int width) {
uintptr_t pixel_temp;
- asm(
+ asm volatile(
// 1 pixel loop.
LABELALIGN
"1: \n"
@@ -9132,7 +9132,7 @@ void RGBColorTableRow_X86(uint8_t* dst_argb,
const uint8_t* table_argb,
int width) {
uintptr_t pixel_temp;
- asm(
+ asm volatile(
// 1 pixel loop.
LABELALIGN
"1: \n"
diff --git a/source/row_lsx.cc b/source/row_lsx.cc
index 09f206cab93f2..fa088c9e78a94 100644
--- a/source/row_lsx.cc
+++ b/source/row_lsx.cc
@@ -2805,7 +2805,8 @@ static void ARGBToYMatrixRow_LSX(const uint8_t* src_argb,
uint8_t* dst_y,
int width,
const struct RgbConstants* rgbconstants) {
- asm("vldrepl.b $vr0, %3, 0 \n\t" // load rgbconstants
+ asm volatile(
+ "vldrepl.b $vr0, %3, 0 \n\t" // load rgbconstants
"vldrepl.b $vr1, %3, 1 \n\t" // load rgbconstants
"vldrepl.b $vr2, %3, 2 \n\t" // load rgbconstants
"vldrepl.h $vr3, %3, 4 \n\t" // load rgbconstants
@@ -2863,7 +2864,8 @@ static void RGBAToYMatrixRow_LSX(const uint8_t* src_rgba,
uint8_t* dst_y,
int width,
const struct RgbConstants* rgbconstants) {
- asm("vldrepl.b $vr0, %3, 0 \n\t" // load rgbconstants
+ asm volatile(
+ "vldrepl.b $vr0, %3, 0 \n\t" // load rgbconstants
"vldrepl.b $vr1, %3, 1 \n\t" // load rgbconstants
"vldrepl.b $vr2, %3, 2 \n\t" // load rgbconstants
"vldrepl.h $vr3, %3, 4 \n\t" // load rgbconstants
@@ -2920,7 +2922,8 @@ static void RGBToYMatrixRow_LSX(const uint8_t* src_rgba,
7, 9, 10, 12, 13, 15, 1, 0, 4, 0, 7, 0, 10,
0, 13, 0, 16, 0, 19, 0, 22, 0, 25, 0, 28, 0,
31, 0, 2, 0, 5, 0, 8, 0, 11, 0, 14, 0};
- asm("vldrepl.b $vr0, %3, 0 \n\t" // load rgbconstants
+ asm volatile(
+ "vldrepl.b $vr0, %3, 0 \n\t" // load rgbconstants
"vldrepl.b $vr1, %3, 1 \n\t" // load rgbconstants
"vldrepl.b $vr2, %3, 2 \n\t" // load rgbconstants
"vldrepl.h $vr3, %3, 4 \n\t" // load rgbconstants
diff --git a/source/scale_gcc.cc b/source/scale_gcc.cc
index 9dfe64a931808..7556bcb4c1d62 100644
--- a/source/scale_gcc.cc
+++ b/source/scale_gcc.cc
@@ -97,7 +97,7 @@ void ScaleRowDown2_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
- asm(
+ asm volatile(
// 16 pixel loop.
LABELALIGN
"1: \n"
@@ -123,7 +123,7 @@ void ScaleRowDown2Linear_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
- asm("pcmpeqb %%xmm4,%%xmm4 \n"
+ asm volatile("pcmpeqb %%xmm4,%%xmm4 \n"
"psrlw $0xf,%%xmm4 \n"
"packuswb %%xmm4,%%xmm4 \n"
"pxor %%xmm5,%%xmm5 \n"
@@ -153,7 +153,7 @@ void ScaleRowDown2Box_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
- asm("pcmpeqb %%xmm4,%%xmm4 \n"
+ asm volatile("pcmpeqb %%xmm4,%%xmm4 \n"
"psrlw $0xf,%%xmm4 \n"
"packuswb %%xmm4,%%xmm4 \n"
"pxor %%xmm5,%%xmm5 \n"
@@ -219,7 +219,7 @@ void ScaleRowDown2Linear_AVX2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
- asm("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n"
+ asm volatile("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $0xf,%%ymm4,%%ymm4 \n"
"vpackuswb %%ymm4,%%ymm4,%%ymm4 \n"
"vpxor %%ymm5,%%ymm5,%%ymm5 \n"
@@ -251,7 +251,7 @@ void ScaleRowDown2Box_AVX2(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
- asm("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n"
+ asm volatile("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $0xf,%%ymm4,%%ymm4 \n"
"vpackuswb %%ymm4,%%ymm4,%%ymm4 \n"
"vpxor %%ymm5,%%ymm5,%%ymm5 \n"
@@ -293,7 +293,7 @@ void ScaleRowDown4_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
- asm("pcmpeqb %%xmm5,%%xmm5 \n"
+ asm volatile("pcmpeqb %%xmm5,%%xmm5 \n"
"psrld $0x18,%%xmm5 \n"
"pslld $0x10,%%xmm5 \n"
@@ -323,7 +323,7 @@ void ScaleRowDown4Box_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
intptr_t stridex3;
- asm("pcmpeqb %%xmm4,%%xmm4 \n"
+ asm volatile("pcmpeqb %%xmm4,%%xmm4 \n"
"psrlw $0xf,%%xmm4 \n"
"movdqa %%xmm4,%%xmm5 \n"
"packuswb %%xmm4,%%xmm4 \n"
@@ -377,7 +377,7 @@ void ScaleRowDown4_AVX2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
- asm("vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
+ asm volatile("vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
"vpsrld $0x18,%%ymm5,%%ymm5 \n"
"vpslld $0x10,%%ymm5,%%ymm5 \n"
@@ -409,7 +409,7 @@ void ScaleRowDown4Box_AVX2(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
- asm("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n"
+ asm volatile("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $0xf,%%ymm4,%%ymm4 \n"
"vpsllw $0x3,%%ymm4,%%ymm5 \n"
"vpackuswb %%ymm4,%%ymm4,%%ymm4 \n"
@@ -464,7 +464,7 @@ void ScaleRowDown34_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
- asm("movdqa %0,%%xmm3 \n"
+ asm volatile("movdqa %0,%%xmm3 \n"
"movdqa %1,%%xmm4 \n"
"movdqa %2,%%xmm5 \n"
:
@@ -499,7 +499,7 @@ void ScaleRowDown34_1_Box_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
- asm("movdqa %0,%%xmm2 \n" // kShuf01
+ asm volatile("movdqa %0,%%xmm2 \n" // kShuf01
"movdqa %1,%%xmm3 \n" // kShuf11
"movdqa %2,%%xmm4 \n" // kShuf21
:
@@ -507,7 +507,7 @@ void ScaleRowDown34_1_Box_SSSE3(const uint8_t* src_ptr,
"m"(kShuf11), // %1
"m"(kShuf21) // %2
);
- asm("movdqa %0,%%xmm5 \n" // kMadd01
+ asm volatile("movdqa %0,%%xmm5 \n" // kMadd01
"movdqa %1,%%xmm0 \n" // kMadd11
"movdqa %2,%%xmm1 \n" // kRound34
:
@@ -561,7 +561,7 @@ void ScaleRowDown34_0_Box_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
- asm("movdqa %0,%%xmm2 \n" // kShuf01
+ asm volatile("movdqa %0,%%xmm2 \n" // kShuf01
"movdqa %1,%%xmm3 \n" // kShuf11
"movdqa %2,%%xmm4 \n" // kShuf21
:
@@ -569,7 +569,7 @@ void ScaleRowDown34_0_Box_SSSE3(const uint8_t* src_ptr,
"m"(kShuf11), // %1
"m"(kShuf21) // %2
);
- asm("movdqa %0,%%xmm5 \n" // kMadd01
+ asm volatile("movdqa %0,%%xmm5 \n" // kMadd01
"movdqa %1,%%xmm0 \n" // kMadd11
"movdqa %2,%%xmm1 \n" // kRound34
:
@@ -628,7 +628,7 @@ void ScaleRowDown38_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
- asm("movdqa %3,%%xmm4 \n"
+ asm volatile("movdqa %3,%%xmm4 \n"
"movdqa %4,%%xmm5 \n"
LABELALIGN
@@ -657,7 +657,7 @@ void ScaleRowDown38_2_Box_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
- asm("movdqa %0,%%xmm2 \n"
+ asm volatile("movdqa %0,%%xmm2 \n"
"movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm4 \n"
"movdqa %3,%%xmm5 \n"
@@ -699,7 +699,7 @@ void ScaleRowDown38_3_Box_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
- asm("movdqa %0,%%xmm2 \n"
+ asm volatile("movdqa %0,%%xmm2 \n"
"movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm4 \n"
"pxor %%xmm5,%%xmm5 \n"
@@ -766,7 +766,7 @@ static const uvec8 kLinearMadd31 = {3, 1, 1, 3, 3, 1, 1, 3,
void ScaleRowUp2_Linear_SSE2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
- asm("pxor %%xmm0,%%xmm0 \n" // 0
+ asm volatile("pxor %%xmm0,%%xmm0 \n" // 0
"pcmpeqw %%xmm6,%%xmm6 \n"
"psrlw $15,%%xmm6 \n"
"psllw $1,%%xmm6 \n" // all 2
@@ -934,7 +934,7 @@ void ScaleRowUp2_Bilinear_SSE2(const uint8_t* src_ptr,
void ScaleRowUp2_Linear_12_SSSE3(const uint16_t* src_ptr,
uint16_t* dst_ptr,
int dst_width) {
- asm("movdqa %3,%%xmm5 \n"
+ asm volatile("movdqa %3,%%xmm5 \n"
"pcmpeqw %%xmm4,%%xmm4 \n"
"psrlw $15,%%xmm4 \n"
"psllw $1,%%xmm4 \n" // all 2
@@ -985,7 +985,7 @@ void ScaleRowUp2_Bilinear_12_SSSE3(const uint16_t* src_ptr,
uint16_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
- asm("pcmpeqw %%xmm7,%%xmm7 \n"
+ asm volatile("pcmpeqw %%xmm7,%%xmm7 \n"
"psrlw $15,%%xmm7 \n"
"psllw $3,%%xmm7 \n" // all 8
"movdqa %5,%%xmm6 \n"
@@ -1082,7 +1082,7 @@ void ScaleRowUp2_Bilinear_12_SSSE3(const uint16_t* src_ptr,
void ScaleRowUp2_Linear_16_SSE2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
int dst_width) {
- asm("pxor %%xmm5,%%xmm5 \n"
+ asm volatile("pxor %%xmm5,%%xmm5 \n"
"pcmpeqd %%xmm4,%%xmm4 \n"
"psrld $31,%%xmm4 \n"
"pslld $1,%%xmm4 \n" // all 2
@@ -1134,7 +1134,7 @@ void ScaleRowUp2_Bilinear_16_SSE2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
- asm("pxor %%xmm7,%%xmm7 \n"
+ asm volatile("pxor %%xmm7,%%xmm7 \n"
"pcmpeqd %%xmm6,%%xmm6 \n"
"psrld $31,%%xmm6 \n"
"pslld $3,%%xmm6 \n" // all 8
@@ -1241,7 +1241,7 @@ void ScaleRowUp2_Bilinear_16_SSE2(const uint16_t* src_ptr,
void ScaleRowUp2_Linear_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
- asm("pcmpeqw %%xmm4,%%xmm4 \n"
+ asm volatile("pcmpeqw %%xmm4,%%xmm4 \n"
"psrlw $15,%%xmm4 \n"
"psllw $1,%%xmm4 \n" // all 2
"movdqa %3,%%xmm3 \n"
@@ -1281,7 +1281,7 @@ void ScaleRowUp2_Bilinear_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
- asm("pcmpeqw %%xmm6,%%xmm6 \n"
+ asm volatile("pcmpeqw %%xmm6,%%xmm6 \n"
"psrlw $15,%%xmm6 \n"
"psllw $3,%%xmm6 \n" // all 8
"movdqa %5,%%xmm7 \n"
@@ -1365,7 +1365,7 @@ void ScaleRowUp2_Bilinear_SSSE3(const uint8_t* src_ptr,
void ScaleRowUp2_Linear_AVX2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
- asm("vpcmpeqw %%ymm4,%%ymm4,%%ymm4 \n"
+ asm volatile("vpcmpeqw %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $15,%%ymm4,%%ymm4 \n"
"vpsllw $1,%%ymm4,%%ymm4 \n" // all 2
"vbroadcastf128 %3,%%ymm3 \n"
@@ -1408,7 +1408,7 @@ void ScaleRowUp2_Bilinear_AVX2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
- asm("vpcmpeqw %%ymm6,%%ymm6,%%ymm6 \n"
+ asm volatile("vpcmpeqw %%ymm6,%%ymm6,%%ymm6 \n"
"vpsrlw $15,%%ymm6,%%ymm6 \n"
"vpsllw $3,%%ymm6,%%ymm6 \n" // all 8
"vbroadcastf128 %5,%%ymm7 \n"
@@ -1489,7 +1489,7 @@ void ScaleRowUp2_Bilinear_AVX2(const uint8_t* src_ptr,
void ScaleRowUp2_Linear_12_AVX2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
int dst_width) {
- asm("vbroadcastf128 %3,%%ymm5 \n"
+ asm volatile("vbroadcastf128 %3,%%ymm5 \n"
"vpcmpeqw %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $15,%%ymm4,%%ymm4 \n"
"vpsllw $1,%%ymm4,%%ymm4 \n" // all 2
@@ -1540,7 +1540,7 @@ void ScaleRowUp2_Bilinear_12_AVX2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
- asm("vbroadcastf128 %5,%%ymm5 \n"
+ asm volatile("vbroadcastf128 %5,%%ymm5 \n"
"vpcmpeqw %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $15,%%ymm4,%%ymm4 \n"
"vpsllw $3,%%ymm4,%%ymm4 \n" // all 8
@@ -1601,7 +1601,7 @@ void ScaleRowUp2_Bilinear_12_AVX2(const uint16_t* src_ptr,
void ScaleRowUp2_Linear_16_AVX2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
int dst_width) {
- asm("vpcmpeqd %%ymm4,%%ymm4,%%ymm4 \n"
+ asm volatile("vpcmpeqd %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrld $31,%%ymm4,%%ymm4 \n"
"vpslld $1,%%ymm4,%%ymm4 \n" // all 2
@@ -1650,7 +1650,7 @@ void ScaleRowUp2_Bilinear_16_AVX2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
- asm("vpcmpeqd %%ymm6,%%ymm6,%%ymm6 \n"
+ asm volatile("vpcmpeqd %%ymm6,%%ymm6,%%ymm6 \n"
"vpsrld $31,%%ymm6,%%ymm6 \n"
"vpslld $3,%%ymm6,%%ymm6 \n" // all 8
@@ -1732,7 +1732,7 @@ void ScaleRowUp2_Bilinear_16_AVX2(const uint16_t* src_ptr,
void ScaleAddRow_SSE2(const uint8_t* src_ptr,
uint16_t* dst_ptr,
int src_width) {
- asm("pxor %%xmm5,%%xmm5 \n"
+ asm volatile("pxor %%xmm5,%%xmm5 \n"
// 16 pixel loop.
LABELALIGN
@@ -1763,7 +1763,7 @@ void ScaleAddRow_SSE2(const uint8_t* src_ptr,
void ScaleAddRow_AVX2(const uint8_t* src_ptr,
uint16_t* dst_ptr,
int src_width) {
- asm("vpxor %%ymm5,%%ymm5,%%ymm5 \n"
+ asm volatile("vpxor %%ymm5,%%ymm5,%%ymm5 \n"
LABELALIGN
"1: \n"
@@ -1804,7 +1804,7 @@ void ScaleFilterCols_SSSE3(uint8_t* dst_ptr,
int x,
int dx) {
intptr_t x0, x1, temp_pixel;
- asm("movd %6,%%xmm2 \n"
+ asm volatile("movd %6,%%xmm2 \n"
"movd %7,%%xmm3 \n"
"movl $0x04040000,%k2 \n"
"movd %k2,%%xmm5 \n"
@@ -2005,7 +2005,7 @@ void ScaleARGBRowDownEven_SSE2(const uint8_t* src_argb,
intptr_t src_stepx_x4 = (intptr_t)(src_stepx);
intptr_t src_stepx_x12;
(void)src_stride;
- asm("lea 0x00(,%1,4),%1 \n"
+ asm volatile("lea 0x00(,%1,4),%1 \n"
"lea 0x00(%1,%1,2),%4 \n"
LABELALIGN
@@ -2041,7 +2041,7 @@ void ScaleARGBRowDownEvenBox_SSE2(const uint8_t* src_argb,
intptr_t src_stepx_x4 = (intptr_t)(src_stepx);
intptr_t src_stepx_x12;
intptr_t row1 = (intptr_t)(src_stride);
- asm("lea 0x00(,%1,4),%1 \n"
+ asm volatile("lea 0x00(,%1,4),%1 \n"
"lea 0x00(%1,%1,2),%4 \n"
"lea 0x00(%0,%5,1),%5 \n"
@@ -2083,7 +2083,7 @@ void ScaleARGBCols_SSE2(uint8_t* dst_argb,
int x,
int dx) {
intptr_t x0, x1;
- asm("movd %5,%%xmm2 \n"
+ asm volatile("movd %5,%%xmm2 \n"
"movd %6,%%xmm3 \n"
"pshufd $0x0,%%xmm2,%%xmm2 \n"
"pshufd $0x11,%%xmm3,%%xmm0 \n"
@@ -2191,14 +2191,14 @@ void ScaleARGBFilterCols_SSSE3(uint8_t* dst_argb,
int x,
int dx) {
intptr_t x0, x1;
- asm("movdqa %0,%%xmm4 \n"
+ asm volatile("movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm5 \n"
:
: "m"(kShuffleColARGB), // %0
"m"(kShuffleFractions) // %1
);
- asm("movd %5,%%xmm2 \n"
+ asm volatile("movd %5,%%xmm2 \n"
"movd %6,%%xmm3 \n"
"pcmpeqb %%xmm6,%%xmm6 \n"
"psrlw $0x9,%%xmm6 \n"
@@ -2260,7 +2260,7 @@ void ScaleARGBFilterCols_SSSE3(uint8_t* dst_argb,
// Divide num by div and return as 16.16 fixed point result.
int FixedDiv_X86(int num, int div) {
- asm("cdq \n"
+ asm volatile("cdq \n"
"shld $0x10,%%eax,%%edx \n"
"shl $0x10,%%eax \n"
"idiv %1 \n"
@@ -2273,7 +2273,7 @@ int FixedDiv_X86(int num, int div) {
// Divide num - 1 by div - 1 and return as 16.16 fixed point result.
int FixedDiv1_X86(int num, int div) {
- asm("cdq \n"
+ asm volatile("cdq \n"
"shld $0x10,%%eax,%%edx \n"
"shl $0x10,%%eax \n"
"sub $0x10001,%%eax \n"
@@ -2304,7 +2304,7 @@ void ScaleUVRowDown2Box_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
- asm("pcmpeqb %%xmm4,%%xmm4 \n" // 01010101
+ asm volatile("pcmpeqb %%xmm4,%%xmm4 \n" // 01010101
"psrlw $0xf,%%xmm4 \n"
"packuswb %%xmm4,%%xmm4 \n"
"pxor %%xmm5, %%xmm5 \n" // zero
@@ -2343,7 +2343,7 @@ void ScaleUVRowDown2Box_AVX2(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
- asm("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n" // 01010101
+ asm volatile("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n" // 01010101
"vpsrlw $0xf,%%ymm4,%%ymm4 \n"
"vpackuswb %%ymm4,%%ymm4,%%ymm4 \n"
"vpxor %%ymm5,%%ymm5,%%ymm5 \n" // zero
@@ -2386,7 +2386,7 @@ static const uvec8 kUVLinearMadd31 = {3, 1, 3, 1, 1, 3, 1, 3,
void ScaleUVRowUp2_Linear_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
- asm("pcmpeqw %%xmm4,%%xmm4 \n"
+ asm volatile("pcmpeqw %%xmm4,%%xmm4 \n"
"psrlw $15,%%xmm4 \n"
"psllw $1,%%xmm4 \n" // all 2
"movdqa %3,%%xmm3 \n"
@@ -2426,7 +2426,7 @@ void ScaleUVRowUp2_Bilinear_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
- asm("pcmpeqw %%xmm6,%%xmm6 \n"
+ asm volatile("pcmpeqw %%xmm6,%%xmm6 \n"
"psrlw $15,%%xmm6 \n"
"psllw $3,%%xmm6 \n" // all 8
"movdqa %5,%%xmm7 \n"
@@ -2509,7 +2509,7 @@ void ScaleUVRowUp2_Bilinear_SSSE3(const uint8_t* src_ptr,
void ScaleUVRowUp2_Linear_AVX2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
- asm("vpcmpeqw %%ymm4,%%ymm4,%%ymm4 \n"
+ asm volatile("vpcmpeqw %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $15,%%ymm4,%%ymm4 \n"
"vpsllw $1,%%ymm4,%%ymm4 \n" // all 2
"vbroadcastf128 %3,%%ymm3 \n"
@@ -2551,7 +2551,7 @@ void ScaleUVRowUp2_Bilinear_AVX2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
- asm("vpcmpeqw %%ymm6,%%ymm6,%%ymm6 \n"
+ asm volatile("vpcmpeqw %%ymm6,%%ymm6,%%ymm6 \n"
"vpsrlw $15,%%ymm6,%%ymm6 \n"
"vpsllw $3,%%ymm6,%%ymm6 \n" // all 8
"vbroadcastf128 %5,%%ymm7 \n"
@@ -2630,7 +2630,7 @@ void ScaleUVRowUp2_Bilinear_AVX2(const uint8_t* src_ptr,
void ScaleUVRowUp2_Linear_16_SSE41(const uint16_t* src_ptr,
uint16_t* dst_ptr,
int dst_width) {
- asm("pxor %%xmm5,%%xmm5 \n"
+ asm volatile("pxor %%xmm5,%%xmm5 \n"
"pcmpeqd %%xmm4,%%xmm4 \n"
"psrld $31,%%xmm4 \n"
"pslld $1,%%xmm4 \n" // all 2
@@ -2681,7 +2681,7 @@ void ScaleUVRowUp2_Bilinear_16_SSE41(const uint16_t* src_ptr,
uint16_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
- asm("pxor %%xmm7,%%xmm7 \n"
+ asm volatile("pxor %%xmm7,%%xmm7 \n"
"pcmpeqd %%xmm6,%%xmm6 \n"
"psrld $31,%%xmm6 \n"
"pslld $3,%%xmm6 \n" // all 8
@@ -2771,7 +2771,7 @@ void ScaleUVRowUp2_Bilinear_16_SSE41(const uint16_t* src_ptr,
void ScaleUVRowUp2_Linear_16_AVX2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
int dst_width) {
- asm("vpcmpeqd %%ymm4,%%ymm4,%%ymm4 \n"
+ asm volatile("vpcmpeqd %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrld $31,%%ymm4,%%ymm4 \n"
"vpslld $1,%%ymm4,%%ymm4 \n" // all 2
@@ -2819,7 +2819,7 @@ void ScaleUVRowUp2_Bilinear_16_AVX2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
- asm("vpcmpeqd %%ymm6,%%ymm6,%%ymm6 \n"
+ asm volatile("vpcmpeqd %%ymm6,%%ymm6,%%ymm6 \n"
"vpsrld $31,%%ymm6,%%ymm6 \n"
"vpslld $3,%%ymm6,%%ymm6 \n" // all 8

16
media/libyuv/libyuv/include/libyuv/macros_msa.h
View File

@ -20,7 +20,7 @@
({ \
const uint8_t* psrc_lw_m = (const uint8_t*)(psrc); \
uint32_t val_m; \
asm("lw %[val_m], %[psrc_lw_m] \n" \
asm volatile("lw %[val_m], %[psrc_lw_m] \n" \
: [val_m] "=r"(val_m) \
: [psrc_lw_m] "m"(*psrc_lw_m)); \
val_m; \
@ -31,7 +31,7 @@
({ \
const uint8_t* psrc_ld_m = (const uint8_t*)(psrc); \
uint64_t val_m = 0; \
asm("ld %[val_m], %[psrc_ld_m] \n" \
asm volatile("ld %[val_m], %[psrc_ld_m] \n" \
: [val_m] "=r"(val_m) \
: [psrc_ld_m] "m"(*psrc_ld_m)); \
val_m; \
@ -55,7 +55,7 @@
({ \
uint8_t* pdst_sw_m = (uint8_t*)(pdst); /* NOLINT */ \
uint32_t val_m = (val); \
asm("sw %[val_m], %[pdst_sw_m] \n" \
asm volatile("sw %[val_m], %[pdst_sw_m] \n" \
: [pdst_sw_m] "=m"(*pdst_sw_m) \
: [val_m] "r"(val_m)); \
})
@ -65,7 +65,7 @@
({ \
uint8_t* pdst_sd_m = (uint8_t*)(pdst); /* NOLINT */ \
uint64_t val_m = (val); \
asm("sd %[val_m], %[pdst_sd_m] \n" \
asm volatile("sd %[val_m], %[pdst_sd_m] \n" \
: [pdst_sd_m] "=m"(*pdst_sd_m) \
: [val_m] "r"(val_m)); \
})
@ -86,7 +86,8 @@
uint8_t* psrc_lw_m = (uint8_t*)(psrc); \
uint32_t val_lw_m; \
\
asm("lwr %[val_lw_m], 0(%[psrc_lw_m]) \n\t" \
asm volatile( \
"lwr %[val_lw_m], 0(%[psrc_lw_m]) \n\t" \
"lwl %[val_lw_m], 3(%[psrc_lw_m]) \n\t" \
\
: [val_lw_m] "=&r"(val_lw_m) \
@ -101,7 +102,8 @@
uint8_t* psrc_ld_m = (uint8_t*)(psrc); \
uint64_t val_ld_m = 0; \
\
asm("ldr %[val_ld_m], 0(%[psrc_ld_m]) \n\t" \
asm volatile( \
"ldr %[val_ld_m], 0(%[psrc_ld_m]) \n\t" \
"ldl %[val_ld_m], 7(%[psrc_ld_m]) \n\t" \
\
: [val_ld_m] "=&r"(val_ld_m) \
@ -128,7 +130,7 @@
({ \
uint8_t* pdst_sw_m = (uint8_t*)(pdst); /* NOLINT */ \
uint32_t val_m = (val); \
asm("usw %[val_m], %[pdst_sw_m] \n" \
asm volatile("usw %[val_m], %[pdst_sw_m] \n" \
: [pdst_sw_m] "=m"(*pdst_sw_m) \
: [val_m] "r"(val_m)); \
})

70
media/libyuv/libyuv/source/row_gcc.cc
View File

@ -2626,7 +2626,7 @@ void OMITFP I444ToARGBRow_SSSE3(const uint8_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@ -2686,7 +2686,7 @@ void OMITFP I422ToRGB24Row_SSSE3(const uint8_t* y_buf,
uint8_t* dst_rgb24,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"movdqa %[kShuffleMaskARGBToRGB24_0],%%xmm5 \n"
"movdqa %[kShuffleMaskARGBToRGB24],%%xmm6 \n"
@ -2722,7 +2722,7 @@ void OMITFP I444ToRGB24Row_SSSE3(const uint8_t* y_buf,
uint8_t* dst_rgb24,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"movdqa %[kShuffleMaskARGBToRGB24_0],%%xmm5 \n"
"movdqa %[kShuffleMaskARGBToRGB24],%%xmm6 \n"
@ -2758,7 +2758,7 @@ void OMITFP I422ToARGBRow_SSSE3(const uint8_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@ -2787,7 +2787,7 @@ void OMITFP I422ToAR30Row_SSSE3(const uint8_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n" // AR30 constants
@ -2822,7 +2822,7 @@ void OMITFP I210ToARGBRow_SSSE3(const uint16_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@ -2852,7 +2852,7 @@ void OMITFP I212ToARGBRow_SSSE3(const uint16_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@ -2882,7 +2882,7 @@ void OMITFP I210ToAR30Row_SSSE3(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@ -2917,7 +2917,7 @@ void OMITFP I212ToAR30Row_SSSE3(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@ -2952,7 +2952,7 @@ void OMITFP I410ToARGBRow_SSSE3(const uint16_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@ -3045,7 +3045,7 @@ void OMITFP I410ToAR30Row_SSSE3(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@ -3238,7 +3238,7 @@ void OMITFP P210ToAR30Row_SSSE3(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $14,%%xmm5 \n"
@ -3269,7 +3269,7 @@ void OMITFP P410ToAR30Row_SSSE3(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $14,%%xmm5 \n"
@ -3301,7 +3301,7 @@ void OMITFP I422ToRGBARow_SSSE3(const uint8_t* y_buf,
uint8_t* dst_rgba,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
@ -3712,7 +3712,7 @@ void OMITFP I444ToARGBRow_AVX2(const uint8_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
@ -3746,7 +3746,7 @@ void OMITFP I422ToARGBRow_AVX2(const uint8_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
@ -3786,7 +3786,7 @@ void OMITFP I422ToARGBRow_AVX512BW(const uint8_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP_AVX512BW(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%xmm5,%%xmm5,%%xmm5 \n"
@ -3825,7 +3825,7 @@ void OMITFP I422ToAR30Row_AVX2(const uint8_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n" // AR30 constants
@ -3865,7 +3865,7 @@ void OMITFP I210ToARGBRow_AVX2(const uint16_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
@ -3900,7 +3900,7 @@ void OMITFP I212ToARGBRow_AVX2(const uint16_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
@ -3935,7 +3935,7 @@ void OMITFP I210ToAR30Row_AVX2(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n" // AR30 constants
@ -3975,7 +3975,7 @@ void OMITFP I212ToAR30Row_AVX2(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n" // AR30 constants
@ -4015,7 +4015,7 @@ void OMITFP I410ToARGBRow_AVX2(const uint16_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
@ -4120,7 +4120,7 @@ void OMITFP I410ToAR30Row_AVX2(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n" // AR30 constants
@ -4228,7 +4228,7 @@ void OMITFP I422ToRGBARow_AVX2(const uint8_t* y_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
@ -4430,7 +4430,7 @@ void OMITFP P210ToAR30Row_AVX2(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n" // AR30 constants
"vpsrlw $14,%%ymm5,%%ymm5 \n"
@ -4467,7 +4467,7 @@ void OMITFP P410ToAR30Row_AVX2(const uint16_t* y_buf,
uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants,
int width) {
asm (
asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n" // AR30 constants
"vpsrlw $14,%%ymm5,%%ymm5 \n"
@ -5681,7 +5681,7 @@ void MergeXRGBRow_AVX2(const uint8_t* src_r,
const uint8_t* src_b,
uint8_t* dst_argb,
int width) {
asm(
asm volatile(
LABELALIGN
"1: \n"
@ -7381,7 +7381,7 @@ void ARGBUnattenuateRow_SSE2(const uint8_t* src_argb,
uint8_t* dst_argb,
int width) {
uintptr_t alpha;
asm(
asm volatile(
// 4 pixel loop.
LABELALIGN
"1: \n"
@ -7841,7 +7841,7 @@ void ARGBAddRow_SSE2(const uint8_t* src_argb,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width) {
asm(
asm volatile(
// 4 pixel loop.
LABELALIGN
"1: \n"
@ -7869,7 +7869,7 @@ void ARGBAddRow_AVX2(const uint8_t* src_argb,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width) {
asm(
asm volatile(
// 4 pixel loop.
LABELALIGN
"1: \n"
@ -7897,7 +7897,7 @@ void ARGBSubtractRow_SSE2(const uint8_t* src_argb,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width) {
asm(
asm volatile(
// 4 pixel loop.
LABELALIGN
"1: \n"
@ -7925,7 +7925,7 @@ void ARGBSubtractRow_AVX2(const uint8_t* src_argb,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width) {
asm(
asm volatile(
// 4 pixel loop.
LABELALIGN
"1: \n"
@ -9099,7 +9099,7 @@ void ARGBColorTableRow_X86(uint8_t* dst_argb,
const uint8_t* table_argb,
int width) {
uintptr_t pixel_temp;
asm(
asm volatile(
// 1 pixel loop.
LABELALIGN
"1: \n"
@ -9132,7 +9132,7 @@ void RGBColorTableRow_X86(uint8_t* dst_argb,
const uint8_t* table_argb,
int width) {
uintptr_t pixel_temp;
asm(
asm volatile(
// 1 pixel loop.
LABELALIGN
"1: \n"

9
media/libyuv/libyuv/source/row_lsx.cc
View File

@ -2805,7 +2805,8 @@ static void ARGBToYMatrixRow_LSX(const uint8_t* src_argb,
uint8_t* dst_y,
int width,
const struct RgbConstants* rgbconstants) {
asm("vldrepl.b $vr0, %3, 0 \n\t" // load rgbconstants
asm volatile(
"vldrepl.b $vr0, %3, 0 \n\t" // load rgbconstants
"vldrepl.b $vr1, %3, 1 \n\t" // load rgbconstants
"vldrepl.b $vr2, %3, 2 \n\t" // load rgbconstants
"vldrepl.h $vr3, %3, 4 \n\t" // load rgbconstants
@ -2863,7 +2864,8 @@ static void RGBAToYMatrixRow_LSX(const uint8_t* src_rgba,
uint8_t* dst_y,
int width,
const struct RgbConstants* rgbconstants) {
asm("vldrepl.b $vr0, %3, 0 \n\t" // load rgbconstants
asm volatile(
"vldrepl.b $vr0, %3, 0 \n\t" // load rgbconstants
"vldrepl.b $vr1, %3, 1 \n\t" // load rgbconstants
"vldrepl.b $vr2, %3, 2 \n\t" // load rgbconstants
"vldrepl.h $vr3, %3, 4 \n\t" // load rgbconstants
@ -2920,7 +2922,8 @@ static void RGBToYMatrixRow_LSX(const uint8_t* src_rgba,
7, 9, 10, 12, 13, 15, 1, 0, 4, 0, 7, 0, 10,
0, 13, 0, 16, 0, 19, 0, 22, 0, 25, 0, 28, 0,
31, 0, 2, 0, 5, 0, 8, 0, 11, 0, 14, 0};
asm("vldrepl.b $vr0, %3, 0 \n\t" // load rgbconstants
asm volatile(
"vldrepl.b $vr0, %3, 0 \n\t" // load rgbconstants
"vldrepl.b $vr1, %3, 1 \n\t" // load rgbconstants
"vldrepl.b $vr2, %3, 2 \n\t" // load rgbconstants
"vldrepl.h $vr3, %3, 4 \n\t" // load rgbconstants

100
media/libyuv/libyuv/source/scale_gcc.cc
View File

@ -97,7 +97,7 @@ void ScaleRowDown2_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
asm(
asm volatile(
// 16 pixel loop.
LABELALIGN
"1: \n"
@ -123,7 +123,7 @@ void ScaleRowDown2Linear_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
asm("pcmpeqb %%xmm4,%%xmm4 \n"
asm volatile("pcmpeqb %%xmm4,%%xmm4 \n"
"psrlw $0xf,%%xmm4 \n"
"packuswb %%xmm4,%%xmm4 \n"
"pxor %%xmm5,%%xmm5 \n"
@ -153,7 +153,7 @@ void ScaleRowDown2Box_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
asm("pcmpeqb %%xmm4,%%xmm4 \n"
asm volatile("pcmpeqb %%xmm4,%%xmm4 \n"
"psrlw $0xf,%%xmm4 \n"
"packuswb %%xmm4,%%xmm4 \n"
"pxor %%xmm5,%%xmm5 \n"
@ -219,7 +219,7 @@ void ScaleRowDown2Linear_AVX2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
asm("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n"
asm volatile("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $0xf,%%ymm4,%%ymm4 \n"
"vpackuswb %%ymm4,%%ymm4,%%ymm4 \n"
"vpxor %%ymm5,%%ymm5,%%ymm5 \n"
@ -251,7 +251,7 @@ void ScaleRowDown2Box_AVX2(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
asm("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n"
asm volatile("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $0xf,%%ymm4,%%ymm4 \n"
"vpackuswb %%ymm4,%%ymm4,%%ymm4 \n"
"vpxor %%ymm5,%%ymm5,%%ymm5 \n"
@ -293,7 +293,7 @@ void ScaleRowDown4_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
asm("pcmpeqb %%xmm5,%%xmm5 \n"
asm volatile("pcmpeqb %%xmm5,%%xmm5 \n"
"psrld $0x18,%%xmm5 \n"
"pslld $0x10,%%xmm5 \n"
@ -323,7 +323,7 @@ void ScaleRowDown4Box_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
intptr_t stridex3;
asm("pcmpeqb %%xmm4,%%xmm4 \n"
asm volatile("pcmpeqb %%xmm4,%%xmm4 \n"
"psrlw $0xf,%%xmm4 \n"
"movdqa %%xmm4,%%xmm5 \n"
"packuswb %%xmm4,%%xmm4 \n"
@ -377,7 +377,7 @@ void ScaleRowDown4_AVX2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
asm("vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
asm volatile("vpcmpeqb %%ymm5,%%ymm5,%%ymm5 \n"
"vpsrld $0x18,%%ymm5,%%ymm5 \n"
"vpslld $0x10,%%ymm5,%%ymm5 \n"
@ -409,7 +409,7 @@ void ScaleRowDown4Box_AVX2(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
asm("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n"
asm volatile("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $0xf,%%ymm4,%%ymm4 \n"
"vpsllw $0x3,%%ymm4,%%ymm5 \n"
"vpackuswb %%ymm4,%%ymm4,%%ymm4 \n"
@ -464,7 +464,7 @@ void ScaleRowDown34_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
asm("movdqa %0,%%xmm3 \n"
asm volatile("movdqa %0,%%xmm3 \n"
"movdqa %1,%%xmm4 \n"
"movdqa %2,%%xmm5 \n"
:
@ -499,7 +499,7 @@ void ScaleRowDown34_1_Box_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
asm("movdqa %0,%%xmm2 \n" // kShuf01
asm volatile("movdqa %0,%%xmm2 \n" // kShuf01
"movdqa %1,%%xmm3 \n" // kShuf11
"movdqa %2,%%xmm4 \n" // kShuf21
:
@ -507,7 +507,7 @@ void ScaleRowDown34_1_Box_SSSE3(const uint8_t* src_ptr,
"m"(kShuf11), // %1
"m"(kShuf21) // %2
);
asm("movdqa %0,%%xmm5 \n" // kMadd01
asm volatile("movdqa %0,%%xmm5 \n" // kMadd01
"movdqa %1,%%xmm0 \n" // kMadd11
"movdqa %2,%%xmm1 \n" // kRound34
:
@ -561,7 +561,7 @@ void ScaleRowDown34_0_Box_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
asm("movdqa %0,%%xmm2 \n" // kShuf01
asm volatile("movdqa %0,%%xmm2 \n" // kShuf01
"movdqa %1,%%xmm3 \n" // kShuf11
"movdqa %2,%%xmm4 \n" // kShuf21
:
@ -569,7 +569,7 @@ void ScaleRowDown34_0_Box_SSSE3(const uint8_t* src_ptr,
"m"(kShuf11), // %1
"m"(kShuf21) // %2
);
asm("movdqa %0,%%xmm5 \n" // kMadd01
asm volatile("movdqa %0,%%xmm5 \n" // kMadd01
"movdqa %1,%%xmm0 \n" // kMadd11
"movdqa %2,%%xmm1 \n" // kRound34
:
@ -628,7 +628,7 @@ void ScaleRowDown38_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
(void)src_stride;
asm("movdqa %3,%%xmm4 \n"
asm volatile("movdqa %3,%%xmm4 \n"
"movdqa %4,%%xmm5 \n"
LABELALIGN
@ -657,7 +657,7 @@ void ScaleRowDown38_2_Box_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
asm("movdqa %0,%%xmm2 \n"
asm volatile("movdqa %0,%%xmm2 \n"
"movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm4 \n"
"movdqa %3,%%xmm5 \n"
@ -699,7 +699,7 @@ void ScaleRowDown38_3_Box_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
asm("movdqa %0,%%xmm2 \n"
asm volatile("movdqa %0,%%xmm2 \n"
"movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm4 \n"
"pxor %%xmm5,%%xmm5 \n"
@ -766,7 +766,7 @@ static const uvec8 kLinearMadd31 = {3, 1, 1, 3, 3, 1, 1, 3,
void ScaleRowUp2_Linear_SSE2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
asm("pxor %%xmm0,%%xmm0 \n" // 0
asm volatile("pxor %%xmm0,%%xmm0 \n" // 0
"pcmpeqw %%xmm6,%%xmm6 \n"
"psrlw $15,%%xmm6 \n"
"psllw $1,%%xmm6 \n" // all 2
@ -934,7 +934,7 @@ void ScaleRowUp2_Bilinear_SSE2(const uint8_t* src_ptr,
void ScaleRowUp2_Linear_12_SSSE3(const uint16_t* src_ptr,
uint16_t* dst_ptr,
int dst_width) {
asm("movdqa %3,%%xmm5 \n"
asm volatile("movdqa %3,%%xmm5 \n"
"pcmpeqw %%xmm4,%%xmm4 \n"
"psrlw $15,%%xmm4 \n"
"psllw $1,%%xmm4 \n" // all 2
@ -985,7 +985,7 @@ void ScaleRowUp2_Bilinear_12_SSSE3(const uint16_t* src_ptr,
uint16_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
asm("pcmpeqw %%xmm7,%%xmm7 \n"
asm volatile("pcmpeqw %%xmm7,%%xmm7 \n"
"psrlw $15,%%xmm7 \n"
"psllw $3,%%xmm7 \n" // all 8
"movdqa %5,%%xmm6 \n"
@ -1082,7 +1082,7 @@ void ScaleRowUp2_Bilinear_12_SSSE3(const uint16_t* src_ptr,
void ScaleRowUp2_Linear_16_SSE2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
int dst_width) {
asm("pxor %%xmm5,%%xmm5 \n"
asm volatile("pxor %%xmm5,%%xmm5 \n"
"pcmpeqd %%xmm4,%%xmm4 \n"
"psrld $31,%%xmm4 \n"
"pslld $1,%%xmm4 \n" // all 2
@ -1134,7 +1134,7 @@ void ScaleRowUp2_Bilinear_16_SSE2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
asm("pxor %%xmm7,%%xmm7 \n"
asm volatile("pxor %%xmm7,%%xmm7 \n"
"pcmpeqd %%xmm6,%%xmm6 \n"
"psrld $31,%%xmm6 \n"
"pslld $3,%%xmm6 \n" // all 8
@ -1241,7 +1241,7 @@ void ScaleRowUp2_Bilinear_16_SSE2(const uint16_t* src_ptr,
void ScaleRowUp2_Linear_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
asm("pcmpeqw %%xmm4,%%xmm4 \n"
asm volatile("pcmpeqw %%xmm4,%%xmm4 \n"
"psrlw $15,%%xmm4 \n"
"psllw $1,%%xmm4 \n" // all 2
"movdqa %3,%%xmm3 \n"
@ -1281,7 +1281,7 @@ void ScaleRowUp2_Bilinear_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
asm("pcmpeqw %%xmm6,%%xmm6 \n"
asm volatile("pcmpeqw %%xmm6,%%xmm6 \n"
"psrlw $15,%%xmm6 \n"
"psllw $3,%%xmm6 \n" // all 8
"movdqa %5,%%xmm7 \n"
@ -1365,7 +1365,7 @@ void ScaleRowUp2_Bilinear_SSSE3(const uint8_t* src_ptr,
void ScaleRowUp2_Linear_AVX2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
asm("vpcmpeqw %%ymm4,%%ymm4,%%ymm4 \n"
asm volatile("vpcmpeqw %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $15,%%ymm4,%%ymm4 \n"
"vpsllw $1,%%ymm4,%%ymm4 \n" // all 2
"vbroadcastf128 %3,%%ymm3 \n"
@ -1408,7 +1408,7 @@ void ScaleRowUp2_Bilinear_AVX2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
asm("vpcmpeqw %%ymm6,%%ymm6,%%ymm6 \n"
asm volatile("vpcmpeqw %%ymm6,%%ymm6,%%ymm6 \n"
"vpsrlw $15,%%ymm6,%%ymm6 \n"
"vpsllw $3,%%ymm6,%%ymm6 \n" // all 8
"vbroadcastf128 %5,%%ymm7 \n"
@ -1489,7 +1489,7 @@ void ScaleRowUp2_Bilinear_AVX2(const uint8_t* src_ptr,
void ScaleRowUp2_Linear_12_AVX2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
int dst_width) {
asm("vbroadcastf128 %3,%%ymm5 \n"
asm volatile("vbroadcastf128 %3,%%ymm5 \n"
"vpcmpeqw %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $15,%%ymm4,%%ymm4 \n"
"vpsllw $1,%%ymm4,%%ymm4 \n" // all 2
@ -1540,7 +1540,7 @@ void ScaleRowUp2_Bilinear_12_AVX2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
asm("vbroadcastf128 %5,%%ymm5 \n"
asm volatile("vbroadcastf128 %5,%%ymm5 \n"
"vpcmpeqw %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $15,%%ymm4,%%ymm4 \n"
"vpsllw $3,%%ymm4,%%ymm4 \n" // all 8
@ -1601,7 +1601,7 @@ void ScaleRowUp2_Bilinear_12_AVX2(const uint16_t* src_ptr,
void ScaleRowUp2_Linear_16_AVX2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
int dst_width) {
asm("vpcmpeqd %%ymm4,%%ymm4,%%ymm4 \n"
asm volatile("vpcmpeqd %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrld $31,%%ymm4,%%ymm4 \n"
"vpslld $1,%%ymm4,%%ymm4 \n" // all 2
@ -1650,7 +1650,7 @@ void ScaleRowUp2_Bilinear_16_AVX2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
asm("vpcmpeqd %%ymm6,%%ymm6,%%ymm6 \n"
asm volatile("vpcmpeqd %%ymm6,%%ymm6,%%ymm6 \n"
"vpsrld $31,%%ymm6,%%ymm6 \n"
"vpslld $3,%%ymm6,%%ymm6 \n" // all 8
@ -1732,7 +1732,7 @@ void ScaleRowUp2_Bilinear_16_AVX2(const uint16_t* src_ptr,
void ScaleAddRow_SSE2(const uint8_t* src_ptr,
uint16_t* dst_ptr,
int src_width) {
asm("pxor %%xmm5,%%xmm5 \n"
asm volatile("pxor %%xmm5,%%xmm5 \n"
// 16 pixel loop.
LABELALIGN
@ -1763,7 +1763,7 @@ void ScaleAddRow_SSE2(const uint8_t* src_ptr,
void ScaleAddRow_AVX2(const uint8_t* src_ptr,
uint16_t* dst_ptr,
int src_width) {
asm("vpxor %%ymm5,%%ymm5,%%ymm5 \n"
asm volatile("vpxor %%ymm5,%%ymm5,%%ymm5 \n"
LABELALIGN
"1: \n"
@ -1804,7 +1804,7 @@ void ScaleFilterCols_SSSE3(uint8_t* dst_ptr,
int x,
int dx) {
intptr_t x0, x1, temp_pixel;
asm("movd %6,%%xmm2 \n"
asm volatile("movd %6,%%xmm2 \n"
"movd %7,%%xmm3 \n"
"movl $0x04040000,%k2 \n"
"movd %k2,%%xmm5 \n"
@ -2005,7 +2005,7 @@ void ScaleARGBRowDownEven_SSE2(const uint8_t* src_argb,
intptr_t src_stepx_x4 = (intptr_t)(src_stepx);
intptr_t src_stepx_x12;
(void)src_stride;
asm("lea 0x00(,%1,4),%1 \n"
asm volatile("lea 0x00(,%1,4),%1 \n"
"lea 0x00(%1,%1,2),%4 \n"
LABELALIGN
@ -2041,7 +2041,7 @@ void ScaleARGBRowDownEvenBox_SSE2(const uint8_t* src_argb,
intptr_t src_stepx_x4 = (intptr_t)(src_stepx);
intptr_t src_stepx_x12;
intptr_t row1 = (intptr_t)(src_stride);
asm("lea 0x00(,%1,4),%1 \n"
asm volatile("lea 0x00(,%1,4),%1 \n"
"lea 0x00(%1,%1,2),%4 \n"
"lea 0x00(%0,%5,1),%5 \n"
@ -2083,7 +2083,7 @@ void ScaleARGBCols_SSE2(uint8_t* dst_argb,
int x,
int dx) {
intptr_t x0, x1;
asm("movd %5,%%xmm2 \n"
asm volatile("movd %5,%%xmm2 \n"
"movd %6,%%xmm3 \n"
"pshufd $0x0,%%xmm2,%%xmm2 \n"
"pshufd $0x11,%%xmm3,%%xmm0 \n"
@ -2191,14 +2191,14 @@ void ScaleARGBFilterCols_SSSE3(uint8_t* dst_argb,
int x,
int dx) {
intptr_t x0, x1;
asm("movdqa %0,%%xmm4 \n"
asm volatile("movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm5 \n"
:
: "m"(kShuffleColARGB), // %0
"m"(kShuffleFractions) // %1
);
asm("movd %5,%%xmm2 \n"
asm volatile("movd %5,%%xmm2 \n"
"movd %6,%%xmm3 \n"
"pcmpeqb %%xmm6,%%xmm6 \n"
"psrlw $0x9,%%xmm6 \n"
@ -2260,7 +2260,7 @@ void ScaleARGBFilterCols_SSSE3(uint8_t* dst_argb,
// Divide num by div and return as 16.16 fixed point result.
int FixedDiv_X86(int num, int div) {
asm("cdq \n"
asm volatile("cdq \n"
"shld $0x10,%%eax,%%edx \n"
"shl $0x10,%%eax \n"
"idiv %1 \n"
@ -2273,7 +2273,7 @@ int FixedDiv_X86(int num, int div) {
// Divide num - 1 by div - 1 and return as 16.16 fixed point result.
int FixedDiv1_X86(int num, int div) {
asm("cdq \n"
asm volatile("cdq \n"
"shld $0x10,%%eax,%%edx \n"
"shl $0x10,%%eax \n"
"sub $0x10001,%%eax \n"
@ -2304,7 +2304,7 @@ void ScaleUVRowDown2Box_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
asm("pcmpeqb %%xmm4,%%xmm4 \n" // 01010101
asm volatile("pcmpeqb %%xmm4,%%xmm4 \n" // 01010101
"psrlw $0xf,%%xmm4 \n"
"packuswb %%xmm4,%%xmm4 \n"
"pxor %%xmm5, %%xmm5 \n" // zero
@ -2343,7 +2343,7 @@ void ScaleUVRowDown2Box_AVX2(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
asm("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n" // 01010101
asm volatile("vpcmpeqb %%ymm4,%%ymm4,%%ymm4 \n" // 01010101
"vpsrlw $0xf,%%ymm4,%%ymm4 \n"
"vpackuswb %%ymm4,%%ymm4,%%ymm4 \n"
"vpxor %%ymm5,%%ymm5,%%ymm5 \n" // zero
@ -2386,7 +2386,7 @@ static const uvec8 kUVLinearMadd31 = {3, 1, 3, 1, 1, 3, 1, 3,
void ScaleUVRowUp2_Linear_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
asm("pcmpeqw %%xmm4,%%xmm4 \n"
asm volatile("pcmpeqw %%xmm4,%%xmm4 \n"
"psrlw $15,%%xmm4 \n"
"psllw $1,%%xmm4 \n" // all 2
"movdqa %3,%%xmm3 \n"
@ -2426,7 +2426,7 @@ void ScaleUVRowUp2_Bilinear_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
asm("pcmpeqw %%xmm6,%%xmm6 \n"
asm volatile("pcmpeqw %%xmm6,%%xmm6 \n"
"psrlw $15,%%xmm6 \n"
"psllw $3,%%xmm6 \n" // all 8
"movdqa %5,%%xmm7 \n"
@ -2509,7 +2509,7 @@ void ScaleUVRowUp2_Bilinear_SSSE3(const uint8_t* src_ptr,
void ScaleUVRowUp2_Linear_AVX2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int dst_width) {
asm("vpcmpeqw %%ymm4,%%ymm4,%%ymm4 \n"
asm volatile("vpcmpeqw %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrlw $15,%%ymm4,%%ymm4 \n"
"vpsllw $1,%%ymm4,%%ymm4 \n" // all 2
"vbroadcastf128 %3,%%ymm3 \n"
@ -2551,7 +2551,7 @@ void ScaleUVRowUp2_Bilinear_AVX2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
asm("vpcmpeqw %%ymm6,%%ymm6,%%ymm6 \n"
asm volatile("vpcmpeqw %%ymm6,%%ymm6,%%ymm6 \n"
"vpsrlw $15,%%ymm6,%%ymm6 \n"
"vpsllw $3,%%ymm6,%%ymm6 \n" // all 8
"vbroadcastf128 %5,%%ymm7 \n"
@ -2630,7 +2630,7 @@ void ScaleUVRowUp2_Bilinear_AVX2(const uint8_t* src_ptr,
void ScaleUVRowUp2_Linear_16_SSE41(const uint16_t* src_ptr,
uint16_t* dst_ptr,
int dst_width) {
asm("pxor %%xmm5,%%xmm5 \n"
asm volatile("pxor %%xmm5,%%xmm5 \n"
"pcmpeqd %%xmm4,%%xmm4 \n"
"psrld $31,%%xmm4 \n"
"pslld $1,%%xmm4 \n" // all 2
@ -2681,7 +2681,7 @@ void ScaleUVRowUp2_Bilinear_16_SSE41(const uint16_t* src_ptr,
uint16_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
asm("pxor %%xmm7,%%xmm7 \n"
asm volatile("pxor %%xmm7,%%xmm7 \n"
"pcmpeqd %%xmm6,%%xmm6 \n"
"psrld $31,%%xmm6 \n"
"pslld $3,%%xmm6 \n" // all 8
@ -2771,7 +2771,7 @@ void ScaleUVRowUp2_Bilinear_16_SSE41(const uint16_t* src_ptr,
void ScaleUVRowUp2_Linear_16_AVX2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
int dst_width) {
asm("vpcmpeqd %%ymm4,%%ymm4,%%ymm4 \n"
asm volatile("vpcmpeqd %%ymm4,%%ymm4,%%ymm4 \n"
"vpsrld $31,%%ymm4,%%ymm4 \n"
"vpslld $1,%%ymm4,%%ymm4 \n" // all 2
@ -2819,7 +2819,7 @@ void ScaleUVRowUp2_Bilinear_16_AVX2(const uint16_t* src_ptr,
uint16_t* dst_ptr,
ptrdiff_t dst_stride,
int dst_width) {
asm("vpcmpeqd %%ymm6,%%ymm6,%%ymm6 \n"
asm volatile("vpcmpeqd %%ymm6,%%ymm6,%%ymm6 \n"
"vpsrld $31,%%ymm6,%%ymm6 \n"
"vpslld $3,%%ymm6,%%ymm6 \n" // all 8

1
media/libyuv/moz.yaml
View File

@ -55,3 +55,4 @@ vendoring:
- 01_make_mjpeg_printfs_optional.patch
- 02_update_gyp.patch
- 03_add_neon64_and_sve_gyp_targets.patch
- 04_add_missing_volatile.patch