sm-commit AT lists.ibiblio.org

Subject: Source Mage code commit list

List archive

[[SM-Commit] ] GIT changes to master grimoire by Pavel Vinogradov (cc8b729168989b366db218b08f904790a694d9a7)

From: Pavel Vinogradov <scm AT sourcemage.org>
To: sm-commit AT lists.ibiblio.org, sm-commit AT lists.sourcemage.org
Subject: [[SM-Commit] ] GIT changes to master grimoire by Pavel Vinogradov (cc8b729168989b366db218b08f904790a694d9a7)
Date: Mon, 16 Sep 2024 19:48:23 +0000

GIT changes to master grimoire by Pavel Vinogradov <public AT sourcemage.org>:

http/firefox/DETAILS
| 2
http/firefox/HISTORY
| 4

http/firefox/patches/0028-bmo-1916038-fix-loading-avif-files-when-using-gcc.patch
| 2675 ++++++++++
http/firefox/patches/0029-bmo-1917964-gcc-15-swgl-fix.patch
| 64
4 files changed, 2744 insertions(+), 1 deletion(-)

New commits:
commit cc8b729168989b366db218b08f904790a694d9a7
Author: Pavel Vinogradov <public AT sourcemage.org>
Commit: Pavel Vinogradov <public AT sourcemage.org>

http/firefox: version 130.0.1

diff --git a/http/firefox/DETAILS b/http/firefox/DETAILS
index c601327..073f400 100755
--- a/http/firefox/DETAILS
+++ b/http/firefox/DETAILS
@@ -1,5 +1,5 @@
SPELL=firefox
- VERSION=130.0
+ VERSION=130.0.1
SECURITY_PATCH=196
SOURCE="${SPELL}-${VERSION}.source.tar.xz"
# Watch: http://releases.mozilla.org/pub/firefox/releases/
/releases/([0-9.]+)/
diff --git a/http/firefox/HISTORY b/http/firefox/HISTORY
index 625f8af..84996fb 100644
--- a/http/firefox/HISTORY
+++ b/http/firefox/HISTORY
@@ -1,3 +1,7 @@
+2024-09-16 Pavel Vinogradov <public AT sourcemage.org>
+ * DETAILS: version 130.0.1
+ * patches/*: updated
+
2024-09-04 Pavel Vinogradov <public AT sourcemage.org>
* DETAILS: version 130.0, SECURITY_PATCH++
* DEPENDS: nss >= 3.103
diff --git
a/http/firefox/patches/0028-bmo-1916038-fix-loading-avif-files-when-using-gcc.patch

b/http/firefox/patches/0028-bmo-1916038-fix-loading-avif-files-when-using-gcc.patch
new file mode 100644
index 0000000..c713d6d
--- /dev/null
+++
b/http/firefox/patches/0028-bmo-1916038-fix-loading-avif-files-when-using-gcc.patch
@@ -0,0 +1,2675 @@
+
+# HG changeset patch
+# User Mike Hommey <mh+mozilla AT glandium.org>
+# Date 1725915543 0
+# Node ID 32a7a66074d14ab077c8c0fbbbd929a72840ed59
+# Parent 43674c61dadccffc2eb1389b15d069f9ee753ff9
+Bug 1916038 - Add volatile for gcc inline to avoid being removed.
r=gfx-reviewers,nical
+
+This extends the fix upstream did in
616bee5420b62a7be09fda0252034e8be85f91b0,
+which was not enough.
+
+Differential Revision: https://phabricator.services.mozilla.com/D221275
+
+diff --git a/media/libyuv/04_add_missing_volatile.patch
b/media/libyuv/04_add_missing_volatile.patch
+new file mode 100644
+--- /dev/null
++++ b/media/libyuv/04_add_missing_volatile.patch
+@@ -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
++
+diff --git a/media/libyuv/libyuv/include/libyuv/macros_msa.h
b/media/libyuv/libyuv/include/libyuv/macros_msa.h
+--- a/media/libyuv/libyuv/include/libyuv/macros_msa.h
++++ b/media/libyuv/libyuv/include/libyuv/macros_msa.h
+@@ -15,28 +15,28 @@
+ #include <msa.h>
+ #include <stdint.h>
+
+ #if (__mips_isa_rev >= 6)
+ #define LW(psrc) \
+ ({ \
+ 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; \
+ })
+
+ #if (__mips == 64)
+ #define LD(psrc) \
+ ({ \
+ 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; \
+ })
+ #else // !(__mips == 64)
+ #define LD(psrc) \
+ ({ \
+ const uint8_t* psrc_ld_m = (const uint8_t*)(psrc); \
+@@ -50,27 +50,27 @@
+ val_m; \
+ })
+ #endif // (__mips == 64)
+
+ #define SW(val, pdst) \
+ ({ \
+ 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)); \
+ })
+
+ #if (__mips == 64)
+ #define SD(val, pdst) \
+ ({ \
+ 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)); \
+ })
+ #else // !(__mips == 64)
+ #define SD(val, pdst) \
+ ({ \
+ uint8_t* pdst_sd_m = (uint8_t*)(pdst); /* NOLINT */ \
+ uint32_t val0_m, val1_m; \
+@@ -81,32 +81,34 @@
+ })
+ #endif // !(__mips == 64)
+ #else // !(__mips_isa_rev >= 6)
+ #define LW(psrc) \
+ ({ \
+ 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) \
+ : [psrc_lw_m] "r"(psrc_lw_m)); \
+ \
+ val_lw_m; \
+ })
+
+ #if (__mips == 64)
+ #define LD(psrc) \
+ ({ \
+ 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) \
+ : [psrc_ld_m] "r"(psrc_ld_m)); \
+ \
+ val_ld_m; \
+ })
+ #else // !(__mips == 64)
+@@ -123,17 +125,17 @@
+ val_m; \
+ })
+ #endif // (__mips == 64)
+
+ #define SW(val, pdst) \
+ ({ \
+ 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)); \
+ })
+
+ #define SD(val, pdst) \
+ ({ \
+ uint8_t* pdst_sd_m = (uint8_t*)(pdst); /* NOLINT */ \
+ uint32_t val0_m, val1_m; \
+diff --git a/media/libyuv/libyuv/source/row_gcc.cc
b/media/libyuv/libyuv/source/row_gcc.cc
+--- a/media/libyuv/libyuv/source/row_gcc.cc
++++ b/media/libyuv/libyuv/source/row_gcc.cc
+@@ -2621,17 +2621,17 @@ void RGBAToUVRow_SSSE3(const uint8_t* sr
+ "lea 0x20(%[dst_ar30]), %[dst_ar30] \n"
+
+ void OMITFP I444ToARGBRow_SSSE3(const uint8_t* y_buf,
+ const uint8_t* u_buf,
+ const uint8_t* v_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"
+
+ LABELALIGN
+ "1: \n"
+ READYUV444
+ YUVTORGB(yuvconstants)
+@@ -2681,17 +2681,17 @@ void OMITFP I444AlphaToARGBRow_SSSE3(con
+ #endif // HAS_I444ALPHATOARGBROW_SSSE3
+
+ void OMITFP I422ToRGB24Row_SSSE3(const uint8_t* y_buf,
+ const uint8_t* u_buf,
+ const uint8_t* v_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"
+ "sub %[u_buf],%[v_buf] \n"
+
+ LABELALIGN
+ "1: \n"
+ READYUV422
+@@ -2717,17 +2717,17 @@ void OMITFP I422ToRGB24Row_SSSE3(const u
+ }
+
+ void OMITFP I444ToRGB24Row_SSSE3(const uint8_t* y_buf,
+ const uint8_t* u_buf,
+ const uint8_t* v_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"
+ "sub %[u_buf],%[v_buf] \n"
+
+ LABELALIGN
+ "1: \n"
+ READYUV444
+@@ -2753,17 +2753,17 @@ void OMITFP I444ToRGB24Row_SSSE3(const u
+ }
+
+ void OMITFP I422ToARGBRow_SSSE3(const uint8_t* y_buf,
+ const uint8_t* u_buf,
+ const uint8_t* v_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"
+
+ LABELALIGN
+ "1: \n"
+ READYUV422
+ YUVTORGB(yuvconstants)
+@@ -2782,17 +2782,17 @@ void OMITFP I422ToARGBRow_SSSE3(const ui
+ }
+
+ void OMITFP I422ToAR30Row_SSSE3(const uint8_t* y_buf,
+ const uint8_t* u_buf,
+ const uint8_t* v_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
+ "psrlw $14,%%xmm5 \n"
+ "psllw $4,%%xmm5 \n" // 2 alpha bits
+ "pxor %%xmm6,%%xmm6 \n" // 0 for min
+ "pcmpeqb %%xmm7,%%xmm7 \n"
+ "psrlw $6,%%xmm7 \n" // 1023 for max
+@@ -2817,17 +2817,17 @@ void OMITFP I422ToAR30Row_SSSE3(const ui
+
+ // 10 bit YUV to ARGB
+ void OMITFP I210ToARGBRow_SSSE3(const uint16_t* y_buf,
+ const uint16_t* u_buf,
+ const uint16_t* v_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"
+
+ LABELALIGN
+ "1: \n"
+ READYUV210
+ YUVTORGB(yuvconstants)
+@@ -2847,17 +2847,17 @@ void OMITFP I210ToARGBRow_SSSE3(const ui
+
+ // 12 bit YUV to ARGB
+ void OMITFP I212ToARGBRow_SSSE3(const uint16_t* y_buf,
+ const uint16_t* u_buf,
+ const uint16_t* v_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"
+
+ LABELALIGN
+ "1: \n"
+ READYUV212
+ YUVTORGB(yuvconstants)
+@@ -2877,17 +2877,17 @@ void OMITFP I212ToARGBRow_SSSE3(const ui
+
+ // 10 bit YUV to AR30
+ void OMITFP I210ToAR30Row_SSSE3(const uint16_t* y_buf,
+ const uint16_t* u_buf,
+ const uint16_t* v_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"
+ "psrlw $14,%%xmm5 \n"
+ "psllw $4,%%xmm5 \n" // 2 alpha bits
+ "pxor %%xmm6,%%xmm6 \n" // 0 for min
+ "pcmpeqb %%xmm7,%%xmm7 \n"
+ "psrlw $6,%%xmm7 \n" // 1023 for max
+@@ -2912,17 +2912,17 @@ void OMITFP I210ToAR30Row_SSSE3(const ui
+
+ // 12 bit YUV to AR30
+ void OMITFP I212ToAR30Row_SSSE3(const uint16_t* y_buf,
+ const uint16_t* u_buf,
+ const uint16_t* v_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"
+ "psrlw $14,%%xmm5 \n"
+ "psllw $4,%%xmm5 \n" // 2 alpha bits
+ "pxor %%xmm6,%%xmm6 \n" // 0 for min
+ "pcmpeqb %%xmm7,%%xmm7 \n"
+ "psrlw $6,%%xmm7 \n" // 1023 for max
+@@ -2947,17 +2947,17 @@ void OMITFP I212ToAR30Row_SSSE3(const ui
+
+ // 10 bit YUV to ARGB
+ void OMITFP I410ToARGBRow_SSSE3(const uint16_t* y_buf,
+ const uint16_t* u_buf,
+ const uint16_t* v_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"
+
+ LABELALIGN
+ "1: \n"
+ READYUV410
+ YUVTORGB(yuvconstants)
+@@ -3040,17 +3040,17 @@ void OMITFP I410AlphaToARGBRow_SSSE3(con
+
+ // 10 bit YUV to AR30
+ void OMITFP I410ToAR30Row_SSSE3(const uint16_t* y_buf,
+ const uint16_t* u_buf,
+ const uint16_t* v_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"
+ "psrlw $14,%%xmm5 \n"
+ "psllw $4,%%xmm5 \n" // 2 alpha bits
+ "pxor %%xmm6,%%xmm6 \n" // 0 for min
+ "pcmpeqb %%xmm7,%%xmm7 \n"
+ "psrlw $6,%%xmm7 \n" // 1023 for max
+@@ -3233,17 +3233,17 @@ void OMITFP P410ToARGBRow_SSSE3(const ui
+ "xmm5");
+ }
+
+ void OMITFP P210ToAR30Row_SSSE3(const uint16_t* y_buf,
+ const uint16_t* uv_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"
+ "psllw $4,%%xmm5 \n" // 2 alpha bits
+ "pxor %%xmm6,%%xmm6 \n" // 0 for min
+ "pcmpeqb %%xmm7,%%xmm7 \n"
+ "psrlw $6,%%xmm7 \n" // 1023 for max
+
+@@ -3264,17 +3264,17 @@ void OMITFP P210ToAR30Row_SSSE3(const ui
+ );
+ }
+
+ void OMITFP P410ToAR30Row_SSSE3(const uint16_t* y_buf,
+ const uint16_t* uv_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"
+ "psllw $4,%%xmm5 \n" // 2 alpha bits
+ "pxor %%xmm6,%%xmm6 \n" // 0 for min
+ "pcmpeqb %%xmm7,%%xmm7 \n"
+ "psrlw $6,%%xmm7 \n" // 1023 for max
+
+@@ -3296,17 +3296,17 @@ void OMITFP P410ToAR30Row_SSSE3(const ui
+ }
+
+ void OMITFP I422ToRGBARow_SSSE3(const uint8_t* y_buf,
+ const uint8_t* u_buf,
+ const uint8_t* v_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"
+
+ LABELALIGN
+ "1: \n"
+ READYUV422
+ YUVTORGB(yuvconstants)
+@@ -3707,17 +3707,17 @@ void OMITFP I422ToRGBARow_SSSE3(const ui
+ // 16 pixels
+ // 16 UV values with 16 Y producing 16 ARGB (64 bytes).
+ void OMITFP I444ToARGBRow_AVX2(const uint8_t* y_buf,
+ const uint8_t* u_buf,
+ const uint8_t* v_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"
+
+ LABELALIGN
+ "1: \n"
+ READYUV444_AVX2
+ YUVTORGB_AVX2(yuvconstants)
+@@ -3741,17 +3741,17 @@ void OMITFP I444ToARGBRow_AVX2(const uin
+ // 16 pixels
+ // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 ARGB (64
bytes).
+ void OMITFP I422ToARGBRow_AVX2(const uint8_t* y_buf,
+ const uint8_t* u_buf,
+ const uint8_t* v_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"
+
+ LABELALIGN
+ "1: \n"
+ READYUV422_AVX2
+ YUVTORGB_AVX2(yuvconstants)
+@@ -3781,17 +3781,17 @@ static const uint64_t kUnpermuteAVX512[8
+ // 16 UV values upsampled to 32 UV, mixed with 32 Y producing 32 ARGB (128
+ // bytes).
+ void OMITFP I422ToARGBRow_AVX512BW(const uint8_t* y_buf,
+ const uint8_t* u_buf,
+ const uint8_t* v_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"
+ "vpbroadcastq %%xmm5,%%zmm5 \n"
+
+ LABELALIGN
+ "1: \n"
+ READYUV422_AVX512BW
+@@ -3820,17 +3820,17 @@ void OMITFP I422ToARGBRow_AVX512BW(const
+ // 16 pixels
+ // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 AR30 (64
bytes).
+ void OMITFP I422ToAR30Row_AVX2(const uint8_t* y_buf,
+ const uint8_t* u_buf,
+ const uint8_t* v_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
+ "vpsrlw $14,%%ymm5,%%ymm5 \n"
+ "vpsllw $4,%%ymm5,%%ymm5 \n" // 2 alpha bits
+ "vpxor %%ymm6,%%ymm6,%%ymm6 \n" // 0 for min
+ "vpcmpeqb %%ymm7,%%ymm7,%%ymm7 \n" // 1023 for max
+ "vpsrlw $6,%%ymm7,%%ymm7 \n"
+@@ -3860,17 +3860,17 @@ void OMITFP I422ToAR30Row_AVX2(const uin
+ // 16 pixels
+ // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 ARGB (64
bytes).
+ void OMITFP I210ToARGBRow_AVX2(const uint16_t* y_buf,
+ const uint16_t* u_buf,
+ const uint16_t* v_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"
+
+ LABELALIGN
+ "1: \n"
+ READYUV210_AVX2
+ YUVTORGB_AVX2(yuvconstants)
+@@ -3895,17 +3895,17 @@ void OMITFP I210ToARGBRow_AVX2(const uin
+ // 16 pixels
+ // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 ARGB (64
bytes).
+ void OMITFP I212ToARGBRow_AVX2(const uint16_t* y_buf,
+ const uint16_t* u_buf,
+ const uint16_t* v_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"
+
+ LABELALIGN
+ "1: \n"
+ READYUV212_AVX2
+ YUVTORGB_AVX2(yuvconstants)
+@@ -3930,17 +3930,17 @@ void OMITFP I212ToARGBRow_AVX2(const uin
+ // 16 pixels
+ // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 AR30 (64
bytes).
+ void OMITFP I210ToAR30Row_AVX2(const uint16_t* y_buf,
+ const uint16_t* u_buf,
+ const uint16_t* v_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
+ "vpsrlw $14,%%ymm5,%%ymm5 \n"
+ "vpsllw $4,%%ymm5,%%ymm5 \n" // 2 alpha bits
+ "vpxor %%ymm6,%%ymm6,%%ymm6 \n" // 0 for min
+ "vpcmpeqb %%ymm7,%%ymm7,%%ymm7 \n" // 1023 for max
+ "vpsrlw $6,%%ymm7,%%ymm7 \n"
+@@ -3970,17 +3970,17 @@ void OMITFP I210ToAR30Row_AVX2(const uin
+ // 16 pixels
+ // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 AR30 (64
bytes).
+ void OMITFP I212ToAR30Row_AVX2(const uint16_t* y_buf,
+ const uint16_t* u_buf,
+ const uint16_t* v_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
+ "vpsrlw $14,%%ymm5,%%ymm5 \n"
+ "vpsllw $4,%%ymm5,%%ymm5 \n" // 2 alpha bits
+ "vpxor %%ymm6,%%ymm6,%%ymm6 \n" // 0 for min
+ "vpcmpeqb %%ymm7,%%ymm7,%%ymm7 \n" // 1023 for max
+ "vpsrlw $6,%%ymm7,%%ymm7 \n"
+@@ -4010,17 +4010,17 @@ void OMITFP I212ToAR30Row_AVX2(const uin
+ // 16 pixels
+ // 16 UV values with 16 Y producing 16 ARGB (64 bytes).
+ void OMITFP I410ToARGBRow_AVX2(const uint16_t* y_buf,
+ const uint16_t* u_buf,
+ const uint16_t* v_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"
+
+ LABELALIGN
+ "1: \n"
+ READYUV410_AVX2
+ YUVTORGB_AVX2(yuvconstants)
+@@ -4115,17 +4115,17 @@ void OMITFP I410AlphaToARGBRow_AVX2(cons
+ // 16 pixels
+ // 16 UV values with 16 Y producing 16 AR30 (64 bytes).
+ void OMITFP I410ToAR30Row_AVX2(const uint16_t* y_buf,
+ const uint16_t* u_buf,
+ const uint16_t* v_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
+ "vpsrlw $14,%%ymm5,%%ymm5 \n"
+ "vpsllw $4,%%ymm5,%%ymm5 \n" // 2 alpha bits
+ "vpxor %%ymm6,%%ymm6,%%ymm6 \n" // 0 for min
+ "vpcmpeqb %%ymm7,%%ymm7,%%ymm7 \n" // 1023 for max
+ "vpsrlw $6,%%ymm7,%%ymm7 \n"
+@@ -4223,17 +4223,17 @@ void OMITFP I422AlphaToARGBRow_AVX2(cons
+ // 16 pixels
+ // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 RGBA (64
bytes).
+ void OMITFP I422ToRGBARow_AVX2(const uint8_t* y_buf,
+ const uint8_t* u_buf,
+ const uint8_t* v_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"
+
+ LABELALIGN
+ "1: \n"
+ READYUV422_AVX2
+ YUVTORGB_AVX2(yuvconstants)
+@@ -4425,17 +4425,17 @@ void OMITFP P410ToARGBRow_AVX2(const uin
+ #if defined(HAS_P210TOAR30ROW_AVX2)
+ // 16 pixels
+ // 16 UV values with 16 Y producing 16 AR30 (64 bytes).
+ void OMITFP P210ToAR30Row_AVX2(const uint16_t* y_buf,
+ const uint16_t* uv_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"
+ "vpsllw $4,%%ymm5,%%ymm5 \n" // 2 alpha bits
+ "vpxor %%ymm6,%%ymm6,%%ymm6 \n" // 0 for min
+ "vpcmpeqb %%ymm7,%%ymm7,%%ymm7 \n" // 1023 for max
+ "vpsrlw $6,%%ymm7,%%ymm7 \n"
+
+@@ -4462,17 +4462,17 @@ void OMITFP P210ToAR30Row_AVX2(const uin
+ #if defined(HAS_P410TOAR30ROW_AVX2)
+ // 16 pixels
+ // 16 UV values with 16 Y producing 16 AR30 (64 bytes).
+ void OMITFP P410ToAR30Row_AVX2(const uint16_t* y_buf,
+ const uint16_t* uv_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"
+ "vpsllw $4,%%ymm5,%%ymm5 \n" // 2 alpha bits
+ "vpxor %%ymm6,%%ymm6,%%ymm6 \n" // 0 for min
+ "vpcmpeqb %%ymm7,%%ymm7,%%ymm7 \n" // 1023 for max
+ "vpsrlw $6,%%ymm7,%%ymm7 \n"
+
+@@ -5676,17 +5676,17 @@ void MergeARGBRow_AVX2(const uint8_t* sr
+ #endif
+
+ #ifdef HAS_MERGEXRGBROW_AVX2
+ void MergeXRGBRow_AVX2(const uint8_t* src_r,
+ const uint8_t* src_g,
+ const uint8_t* src_b,
+ uint8_t* dst_argb,
+ int width) {
+- asm(
++ asm volatile(
+
+ LABELALIGN
+ "1: \n"
+
+ "vmovdqu (%2),%%xmm0 \n" // B
+ "vpcmpeqb %%ymm1,%%ymm1,%%ymm1 \n" // A(255)
+ "vinserti128 $0,(%1),%%ymm1,%%ymm1 \n" // R
+ "vinserti128 $1,(%0),%%ymm0,%%ymm0 \n" // G
+@@ -7376,17 +7376,17 @@ void ARGBAttenuateRow_AVX2(const uint8_t
+ #endif // HAS_ARGBATTENUATEROW_AVX2
+
+ #ifdef HAS_ARGBUNATTENUATEROW_SSE2
+ // Unattenuate 4 pixels at a time.
+ 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"
+ "movdqu (%0),%%xmm0 \n"
+ "movzb 0x03(%0),%3 \n"
+ "punpcklbw %%xmm0,%%xmm0 \n"
+ "movd 0x00(%4,%3,4),%%xmm2 \n"
+ "movzb 0x07(%0),%3 \n"
+@@ -7836,17 +7836,17 @@ void ARGBMultiplyRow_AVX2(const uint8_t*
+ #endif // HAS_ARGBMULTIPLYROW_AVX2
+
+ #ifdef HAS_ARGBADDROW_SSE2
+ // Add 2 rows of ARGB pixels together, 4 pixels at a time.
+ 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"
+ "movdqu (%0),%%xmm0 \n"
+ "lea 0x10(%0),%0 \n"
+ "movdqu (%1),%%xmm1 \n"
+ "lea 0x10(%1),%1 \n"
+ "paddusb %%xmm1,%%xmm0 \n"
+@@ -7864,17 +7864,17 @@ void ARGBAddRow_SSE2(const uint8_t* src_
+ #endif // HAS_ARGBADDROW_SSE2
+
+ #ifdef HAS_ARGBADDROW_AVX2
+ // Add 2 rows of ARGB pixels together, 4 pixels at a time.
+ 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"
+ "vmovdqu (%0),%%ymm0 \n"
+ "lea 0x20(%0),%0 \n"
+ "vpaddusb (%1),%%ymm0,%%ymm0 \n"
+ "lea 0x20(%1),%1 \n"
+ "vmovdqu %%ymm0,(%2) \n"
+@@ -7892,17 +7892,17 @@ void ARGBAddRow_AVX2(const uint8_t* src_
+ #endif // HAS_ARGBADDROW_AVX2
+
+ #ifdef HAS_ARGBSUBTRACTROW_SSE2
+ // Subtract 2 rows of ARGB pixels, 4 pixels at a time.
+ 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"
+ "movdqu (%0),%%xmm0 \n"
+ "lea 0x10(%0),%0 \n"
+ "movdqu (%1),%%xmm1 \n"
+ "lea 0x10(%1),%1 \n"
+ "psubusb %%xmm1,%%xmm0 \n"
+@@ -7920,17 +7920,17 @@ void ARGBSubtractRow_SSE2(const uint8_t*
+ #endif // HAS_ARGBSUBTRACTROW_SSE2
+
+ #ifdef HAS_ARGBSUBTRACTROW_AVX2
+ // Subtract 2 rows of ARGB pixels, 8 pixels at a time.
+ 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"
+ "vmovdqu (%0),%%ymm0 \n"
+ "lea 0x20(%0),%0 \n"
+ "vpsubusb (%1),%%ymm0,%%ymm0 \n"
+ "lea 0x20(%1),%1 \n"
+ "vmovdqu %%ymm0,(%2) \n"
+@@ -9094,17 +9094,17 @@ void HalfFloat1Row_F16C(const uint16_t*
+ #endif // HAS_HALFFLOATROW_F16C
+
+ #ifdef HAS_ARGBCOLORTABLEROW_X86
+ // Tranform ARGB pixels with color table.
+ 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"
+ "movzb (%0),%1 \n"
+ "lea 0x4(%0),%0 \n"
+ "movzb 0x00(%3,%1,4),%1 \n"
+ "mov %b1,-0x4(%0) \n"
+ "movzb -0x3(%0),%1 \n"
+@@ -9127,17 +9127,17 @@ void ARGBColorTableRow_X86(uint8_t* dst_
+ #endif // HAS_ARGBCOLORTABLEROW_X86
+
+ #ifdef HAS_RGBCOLORTABLEROW_X86
+ // Tranform RGB pixels with color table.
+ 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"
+ "movzb (%0),%1 \n"
+ "lea 0x4(%0),%0 \n"
+ "movzb 0x00(%3,%1,4),%1 \n"
+ "mov %b1,-0x4(%0) \n"
+ "movzb -0x3(%0),%1 \n"
+diff --git a/media/libyuv/libyuv/source/row_lsx.cc
b/media/libyuv/libyuv/source/row_lsx.cc
+--- a/media/libyuv/libyuv/source/row_lsx.cc
++++ b/media/libyuv/libyuv/source/row_lsx.cc
+@@ -2800,17 +2800,18 @@ static const struct RgbConstants kRawI60
+ 0x1080,
+ 0};
+
+ // ARGB expects first 3 values to contain RGB and 4th value is ignored.
+ 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
+ "1: \n\t"
+ "vld $vr4, %0, 0 \n\t"
+ "vld $vr5, %0, 16 \n\t"
+ "vld $vr6, %0, 32 \n\t"
+ "vld $vr7, %0, 48 \n\t" // load 16 pixels
of
+@@ -2858,17 +2859,18 @@ void ABGRToYJRow_LSX(const uint8_t* src_
+ }
+
+ // RGBA expects first value to be A and ignored, then 3 values to contain
RGB.
+ // Same code as ARGB, except the LD4
+ 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
+ "1: \n\t"
+ "vld $vr4, %0, 0 \n\t"
+ "vld $vr5, %0, 16 \n\t"
+ "vld $vr6, %0, 32 \n\t"
+ "vld $vr7, %0, 48 \n\t" // load 16 pixels
of
+@@ -2915,17 +2917,18 @@ static void RGBToYMatrixRow_LSX(const ui
+ uint8_t* dst_y,
+ int width,
+ const struct RgbConstants* rgbconstants) {
+ int8_t shuff[64] = {0, 2, 3, 5, 6, 8, 9, 11, 12, 14, 15, 17, 18,
+ 20, 21, 23, 24, 26, 27, 29, 30, 0, 1, 3, 4, 6,
+ 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
+ "vld $vr4, %4, 0 \n\t" // load shuff
+ "vld $vr5, %4, 16 \n\t"
+ "vld $vr6, %4, 32 \n\t"
+ "vld $vr7, %4, 48 \n\t"
+ "1: \n\t"
+diff --git a/media/libyuv/libyuv/source/scale_gcc.cc
b/media/libyuv/libyuv/source/scale_gcc.cc
+--- a/media/libyuv/libyuv/source/scale_gcc.cc
++++ b/media/libyuv/libyuv/source/scale_gcc.cc
+@@ -92,17 +92,17 @@ static const uvec16 kScaleAb2 = {65536 /
+ // Generated using gcc disassembly on Visual C object file:
+ // objdump -D yuvscaler.obj >yuvscaler.txt
+
+ void ScaleRowDown2_SSSE3(const uint8_t* src_ptr,
+ ptrdiff_t src_stride,
+ uint8_t* dst_ptr,
+ int dst_width) {
+ (void)src_stride;
+- asm(
++ asm volatile(
+ // 16 pixel loop.
+ LABELALIGN
+ "1: \n"
+ "movdqu (%0),%%xmm0 \n"
+ "movdqu 0x10(%0),%%xmm1 \n"
+ "lea 0x20(%0),%0 \n"
+ "psrlw $0x8,%%xmm0 \n"
+ "psrlw $0x8,%%xmm1 \n"
+@@ -118,17 +118,17 @@ void ScaleRowDown2_SSSE3(const uint8_t*
+ : "memory", "cc", "xmm0", "xmm1");
+ }
+
+ void ScaleRowDown2Linear_SSSE3(const uint8_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "movdqu (%0),%%xmm0 \n"
+ "movdqu 0x10(%0),%%xmm1 \n"
+@@ -148,17 +148,17 @@ void ScaleRowDown2Linear_SSSE3(const uin
+ :
+ : "memory", "cc", "xmm0", "xmm1", "xmm4", "xmm5");
+ }
+
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "movdqu (%0),%%xmm0 \n"
+ "movdqu 0x10(%0),%%xmm1 \n"
+@@ -214,17 +214,17 @@ void ScaleRowDown2_AVX2(const uint8_t* s
+ : "memory", "cc", "xmm0", "xmm1");
+ }
+
+ void ScaleRowDown2Linear_AVX2(const uint8_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "vmovdqu (%0),%%ymm0 \n"
+ "vmovdqu 0x20(%0),%%ymm1 \n"
+@@ -246,17 +246,17 @@ void ScaleRowDown2Linear_AVX2(const uint
+ :
+ : "memory", "cc", "xmm0", "xmm1", "xmm4", "xmm5");
+ }
+
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "vmovdqu (%0),%%ymm0 \n"
+ "vmovdqu 0x20(%0),%%ymm1 \n"
+@@ -288,17 +288,17 @@ void ScaleRowDown2Box_AVX2(const uint8_t
+ }
+ #endif // HAS_SCALEROWDOWN2_AVX2
+
+ void ScaleRowDown4_SSSE3(const uint8_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "movdqu (%0),%%xmm0 \n"
+ "movdqu 0x10(%0),%%xmm1 \n"
+ "lea 0x20(%0),%0 \n"
+@@ -318,17 +318,17 @@ void ScaleRowDown4_SSSE3(const uint8_t*
+ : "memory", "cc", "xmm0", "xmm1", "xmm5");
+ }
+
+ void ScaleRowDown4Box_SSSE3(const uint8_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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"
+ "psllw $0x3,%%xmm5 \n"
+ "lea 0x00(%4,%4,2),%3 \n"
+
+ LABELALIGN
+ "1: \n"
+@@ -372,17 +372,17 @@ void ScaleRowDown4Box_SSSE3(const uint8_
+ }
+
+ #ifdef HAS_SCALEROWDOWN4_AVX2
+ void ScaleRowDown4_AVX2(const uint8_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "vmovdqu (%0),%%ymm0 \n"
+ "vmovdqu 0x20(%0),%%ymm1 \n"
+ "lea 0x40(%0),%0 \n"
+@@ -404,17 +404,17 @@ void ScaleRowDown4_AVX2(const uint8_t* s
+ :
+ : "memory", "cc", "xmm0", "xmm1", "xmm5");
+ }
+
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "vmovdqu (%0),%%ymm0 \n"
+ "vmovdqu 0x20(%0),%%ymm1 \n"
+@@ -459,17 +459,17 @@ void ScaleRowDown4Box_AVX2(const uint8_t
+ }
+ #endif // HAS_SCALEROWDOWN4_AVX2
+
+ void ScaleRowDown34_SSSE3(const uint8_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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"
+ :
+ : "m"(kShuf0), // %0
+ "m"(kShuf1), // %1
+ "m"(kShuf2) // %2
+ );
+ asm volatile (LABELALIGN
+@@ -494,25 +494,25 @@ void ScaleRowDown34_SSSE3(const uint8_t*
+ :
+ : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5");
+ }
+
+ 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
+ :
+ : "m"(kShuf01), // %0
+ "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
+ :
+ : "m"(kMadd01), // %0
+ "m"(kMadd11), // %1
+ "m"(kRound34) // %2
+ );
+ asm volatile (LABELALIGN
+@@ -556,25 +556,25 @@ void ScaleRowDown34_1_Box_SSSE3(const ui
+ : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5",
"xmm6",
+ "xmm7");
+ }
+
+ 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
+ :
+ : "m"(kShuf01), // %0
+ "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
+ :
+ : "m"(kMadd01), // %0
+ "m"(kMadd11), // %1
+ "m"(kRound34) // %2
+ );
+
+@@ -623,17 +623,17 @@ void ScaleRowDown34_0_Box_SSSE3(const ui
+ "xmm7");
+ }
+
+ void ScaleRowDown38_SSSE3(const uint8_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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
+ "1: \n"
+ "movdqu (%0),%%xmm0 \n"
+ "movdqu 0x10(%0),%%xmm1 \n"
+ "lea 0x20(%0),%0 \n"
+ "pshufb %%xmm4,%%xmm0 \n"
+@@ -652,17 +652,17 @@ void ScaleRowDown38_SSSE3(const uint8_t*
+ "m"(kShuf38b) // %4
+ : "memory", "cc", "xmm0", "xmm1", "xmm4", "xmm5");
+ }
+
+ 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"
+ :
+ : "m"(kShufAb0), // %0
+ "m"(kShufAb1), // %1
+ "m"(kShufAb2), // %2
+ "m"(kScaleAb2) // %3
+@@ -694,17 +694,17 @@ void ScaleRowDown38_2_Box_SSSE3(const ui
+ : "r"((intptr_t)(src_stride)) // %3
+ : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5",
"xmm6");
+ }
+
+ 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"
+ :
+ : "m"(kShufAc), // %0
+ "m"(kShufAc3), // %1
+ "m"(kScaleAc33) // %2
+ );
+@@ -761,17 +761,17 @@ static const uvec8 kLinearShuffleFar = {
+
+ static const uvec8 kLinearMadd31 = {3, 1, 1, 3, 3, 1, 1, 3,
+ 3, 1, 1, 3, 3, 1, 1, 3};
+
+ #ifdef HAS_SCALEROWUP2_LINEAR_SSE2
+ 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
+
+ LABELALIGN
+ "1: \n"
+ "movq (%0),%%xmm1 \n" // 01234567
+ "movq 1(%0),%%xmm2 \n" // 12345678
+@@ -929,17 +929,17 @@ void ScaleRowUp2_Bilinear_SSE2(const uin
+ "xmm7");
+ }
+ #endif
+
+ #ifdef HAS_SCALEROWUP2_LINEAR_12_SSSE3
+ 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
+
+ LABELALIGN
+ "1: \n"
+ "movdqu (%0),%%xmm0 \n" // 01234567 (16)
+ "movdqu 2(%0),%%xmm1 \n" // 12345678 (16)
+@@ -980,17 +980,17 @@ void ScaleRowUp2_Linear_12_SSSE3(const u
+ #endif
+
+ #ifdef HAS_SCALEROWUP2_BILINEAR_12_SSSE3
+ void ScaleRowUp2_Bilinear_12_SSSE3(const uint16_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ // above line
+ "movdqu (%0),%%xmm0 \n" // 01234567 (16)
+@@ -1077,17 +1077,17 @@ void ScaleRowUp2_Bilinear_12_SSSE3(const
+ "xmm7");
+ }
+ #endif
+
+ #ifdef HAS_SCALEROWUP2_LINEAR_16_SSE2
+ 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
+
+ LABELALIGN
+ "1: \n"
+ "movq (%0),%%xmm0 \n" // 0123 (16b)
+ "movq 2(%0),%%xmm1 \n" // 1234 (16b)
+@@ -1129,17 +1129,17 @@ void ScaleRowUp2_Linear_16_SSE2(const ui
+ #endif
+
+ #ifdef HAS_SCALEROWUP2_BILINEAR_16_SSE2
+ void ScaleRowUp2_Bilinear_16_SSE2(const uint16_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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
+
+ LABELALIGN
+ "1: \n"
+ "movq (%0),%%xmm0 \n" // 0011 (16b, 1u1v)
+ "movq 4(%0),%%xmm1 \n" // 1122 (16b, 1u1v)
+@@ -1236,17 +1236,17 @@ void ScaleRowUp2_Bilinear_16_SSE2(const
+ "xmm7");
+ }
+ #endif
+
+ #ifdef HAS_SCALEROWUP2_LINEAR_SSSE3
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "movq (%0),%%xmm0 \n" // 01234567
+ "movq 1(%0),%%xmm1 \n" // 12345678
+@@ -1276,17 +1276,17 @@ void ScaleRowUp2_Linear_SSSE3(const uint
+ #endif
+
+ #ifdef HAS_SCALEROWUP2_BILINEAR_SSSE3
+ void ScaleRowUp2_Bilinear_SSSE3(const uint8_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "movq (%0),%%xmm0 \n" // 01234567
+ "movq 1(%0),%%xmm1 \n" // 12345678
+@@ -1360,17 +1360,17 @@ void ScaleRowUp2_Bilinear_SSSE3(const ui
+ "xmm7");
+ }
+ #endif
+
+ #ifdef HAS_SCALEROWUP2_LINEAR_AVX2
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "vmovdqu (%0),%%xmm0 \n" // 0123456789ABCDEF
+ "vmovdqu 1(%0),%%xmm1 \n" // 123456789ABCDEF0
+@@ -1403,17 +1403,17 @@ void ScaleRowUp2_Linear_AVX2(const uint8
+ #endif
+
+ #ifdef HAS_SCALEROWUP2_BILINEAR_AVX2
+ void ScaleRowUp2_Bilinear_AVX2(const uint8_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "vmovdqu (%0),%%xmm0 \n" // 0123456789ABCDEF
+ "vmovdqu 1(%0),%%xmm1 \n" // 123456789ABCDEF0
+@@ -1484,17 +1484,17 @@ void ScaleRowUp2_Bilinear_AVX2(const uin
+ "xmm7");
+ }
+ #endif
+
+ #ifdef HAS_SCALEROWUP2_LINEAR_12_AVX2
+ 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
+
+ LABELALIGN
+ "1: \n"
+ "vmovdqu (%0),%%ymm0 \n" // 0123456789ABCDEF
(16b)
+ "vmovdqu 2(%0),%%ymm1 \n" // 123456789ABCDEF0
(16b)
+@@ -1535,17 +1535,17 @@ void ScaleRowUp2_Linear_12_AVX2(const ui
+ #endif
+
+ #ifdef HAS_SCALEROWUP2_BILINEAR_12_AVX2
+ void ScaleRowUp2_Bilinear_12_AVX2(const uint16_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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
+
+ LABELALIGN
+ "1: \n"
+
+ "vmovdqu (%0),%%xmm0 \n" // 01234567 (16b)
+@@ -1596,17 +1596,17 @@ void ScaleRowUp2_Bilinear_12_AVX2(const
+ : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5");
+ }
+ #endif
+
+ #ifdef HAS_SCALEROWUP2_LINEAR_16_AVX2
+ 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
+
+ LABELALIGN
+ "1: \n"
+ "vmovdqu (%0),%%xmm0 \n" // 01234567 (16b,
1u1v)
+ "vmovdqu 2(%0),%%xmm1 \n" // 12345678 (16b,
1u1v)
+
+@@ -1645,17 +1645,17 @@ void ScaleRowUp2_Linear_16_AVX2(const ui
+ #endif
+
+ #ifdef HAS_SCALEROWUP2_BILINEAR_16_AVX2
+ void ScaleRowUp2_Bilinear_16_AVX2(const uint16_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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
+
+ LABELALIGN
+ "1: \n"
+
+ "vmovdqu (%0),%%xmm0 \n" // 01234567 (16b,
1u1v)
+ "vmovdqu 2(%0),%%xmm1 \n" // 12345678 (16b,
1u1v)
+@@ -1727,17 +1727,17 @@ void ScaleRowUp2_Bilinear_16_AVX2(const
+ : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5",
"xmm6");
+ }
+ #endif
+
+ // Reads 16xN bytes and produces 16 shorts at a time.
+ 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
+ "1: \n"
+ "movdqu (%0),%%xmm3 \n"
+ "lea 0x10(%0),%0 \n" // src_ptr += 16
+ "movdqu (%1),%%xmm0 \n"
+ "movdqu 0x10(%1),%%xmm1 \n"
+@@ -1758,17 +1758,17 @@ void ScaleAddRow_SSE2(const uint8_t* src
+ : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm5");
+ }
+
+ #ifdef HAS_SCALEADDROW_AVX2
+ // Reads 32 bytes and accumulates to 32 shorts at a time.
+ 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"
+ "vmovdqu (%0),%%ymm3 \n"
+ "lea 0x20(%0),%0 \n" // src_ptr += 32
+ "vpermq $0xd8,%%ymm3,%%ymm3 \n"
+ "vpunpcklbw %%ymm5,%%ymm3,%%ymm2 \n"
+ "vpunpckhbw %%ymm5,%%ymm3,%%ymm3 \n"
+@@ -1799,17 +1799,17 @@ static const uvec16 kFadd40 = {0x4040, 0
+
+ // Bilinear column filtering. SSSE3 version.
+ void ScaleFilterCols_SSSE3(uint8_t* dst_ptr,
+ const uint8_t* src_ptr,
+ int dst_width,
+ 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"
+ "pcmpeqb %%xmm6,%%xmm6 \n"
+ "psrlw $0x9,%%xmm6 \n" // 0x007f007f
+ "pcmpeqb %%xmm7,%%xmm7 \n"
+ "psrlw $15,%%xmm7 \n" // 0x00010001
+
+@@ -2000,17 +2000,17 @@ void ScaleARGBRowDown2Box_SSE2(const uin
+ void ScaleARGBRowDownEven_SSE2(const uint8_t* src_argb,
+ ptrdiff_t src_stride,
+ int src_stepx,
+ uint8_t* dst_argb,
+ int dst_width) {
+ 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
+ "1: \n"
+ "movd (%0),%%xmm0 \n"
+ "movd 0x00(%0,%1,1),%%xmm1 \n"
+ "punpckldq %%xmm1,%%xmm0 \n"
+ "movd 0x00(%0,%1,2),%%xmm2 \n"
+@@ -2036,17 +2036,17 @@ void ScaleARGBRowDownEven_SSE2(const uin
+ void ScaleARGBRowDownEvenBox_SSE2(const uint8_t* src_argb,
+ ptrdiff_t src_stride,
+ int src_stepx,
+ uint8_t* dst_argb,
+ int dst_width) {
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "movq (%0),%%xmm0 \n"
+ "movhps 0x00(%0,%1,1),%%xmm0 \n"
+ "movq 0x00(%0,%1,2),%%xmm1 \n"
+@@ -2078,17 +2078,17 @@ void ScaleARGBRowDownEvenBox_SSE2(const
+ }
+
+ void ScaleARGBCols_SSE2(uint8_t* dst_argb,
+ const uint8_t* src_argb,
+ int dst_width,
+ 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"
+ "paddd %%xmm0,%%xmm2 \n"
+ "paddd %%xmm3,%%xmm3 \n"
+ "pshufd $0x5,%%xmm3,%%xmm0 \n"
+ "paddd %%xmm0,%%xmm2 \n"
+ "paddd %%xmm3,%%xmm3 \n"
+@@ -2186,24 +2186,24 @@ static const uvec8 kShuffleFractions = {
+
+ // Bilinear row filtering combines 4x2 -> 4x1. SSSE3 version
+ void ScaleARGBFilterCols_SSSE3(uint8_t* dst_argb,
+ const uint8_t* src_argb,
+ int dst_width,
+ 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"
+ "pextrw $0x1,%%xmm2,%k3 \n"
+ "sub $0x2,%2 \n"
+ "jl 29f \n"
+ "movdqa %%xmm2,%%xmm0 \n"
+ "paddd %%xmm3,%%xmm0 \n"
+@@ -2255,30 +2255,30 @@ void ScaleARGBFilterCols_SSSE3(uint8_t*
+ "=&r"(x1) // %4
+ : "rm"(x), // %5
+ "rm"(dx) // %6
+ : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5",
"xmm6");
+ }
+
+ // 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"
+ "mov %0, %%eax \n"
+ : "+a"(num) // %0
+ : "c"(div) // %1
+ : "memory", "cc", "edx");
+ return num;
+ }
+
+ // 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"
+ "sbb $0x0,%%edx \n"
+ "sub $0x1,%1 \n"
+ "idiv %1 \n"
+ "mov %0, %%eax \n"
+ : "+a"(num) // %0
+@@ -2299,17 +2299,17 @@ static const uvec8 kShuffleMergeUV = {0u
+ #endif
+
+ #ifdef HAS_SCALEUVROWDOWN2BOX_SSSE3
+
+ 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
+ "movdqa %4,%%xmm1 \n" // split shuffler
+ "movdqa %5,%%xmm3 \n" // merge shuffler
+
+ LABELALIGN
+ "1: \n"
+@@ -2338,17 +2338,17 @@ void ScaleUVRowDown2Box_SSSE3(const uint
+ }
+ #endif // HAS_SCALEUVROWDOWN2BOX_SSSE3
+
+ #ifdef HAS_SCALEUVROWDOWN2BOX_AVX2
+ 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
+ "vbroadcastf128 %4,%%ymm1 \n" // split shuffler
+ "vbroadcastf128 %5,%%ymm3 \n" // merge shuffler
+
+ LABELALIGN
+ "1: \n"
+@@ -2381,17 +2381,17 @@ void ScaleUVRowDown2Box_AVX2(const uint8
+
+ static const uvec8 kUVLinearMadd31 = {3, 1, 3, 1, 1, 3, 1, 3,
+ 3, 1, 3, 1, 1, 3, 1, 3};
+
+ #ifdef HAS_SCALEUVROWUP2_LINEAR_SSSE3
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "movq (%0),%%xmm0 \n" // 00112233 (1u1v)
+ "movq 2(%0),%%xmm1 \n" // 11223344 (1u1v)
+@@ -2421,17 +2421,17 @@ void ScaleUVRowUp2_Linear_SSSE3(const ui
+ #endif
+
+ #ifdef HAS_SCALEUVROWUP2_BILINEAR_SSSE3
+ void ScaleUVRowUp2_Bilinear_SSSE3(const uint8_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "movq (%0),%%xmm0 \n" // 00112233 (1u1v)
+ "movq 2(%0),%%xmm1 \n" // 11223344 (1u1v)
+@@ -2504,17 +2504,17 @@ void ScaleUVRowUp2_Bilinear_SSSE3(const
+ }
+ #endif
+
+ #ifdef HAS_SCALEUVROWUP2_LINEAR_AVX2
+
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "vmovdqu (%0),%%xmm0 \n"
+ "vmovdqu 2(%0),%%xmm1 \n"
+@@ -2546,17 +2546,17 @@ void ScaleUVRowUp2_Linear_AVX2(const uin
+ #endif
+
+ #ifdef HAS_SCALEUVROWUP2_BILINEAR_AVX2
+ void ScaleUVRowUp2_Bilinear_AVX2(const uint8_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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"
+
+ LABELALIGN
+ "1: \n"
+ "vmovdqu (%0),%%xmm0 \n"
+ "vmovdqu 2(%0),%%xmm1 \n"
+@@ -2625,17 +2625,17 @@ void ScaleUVRowUp2_Bilinear_AVX2(const u
+ "xmm7");
+ }
+ #endif
+
+ #ifdef HAS_SCALEUVROWUP2_LINEAR_16_SSE41
+ 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
+
+ LABELALIGN
+ "1: \n"
+ "movq (%0),%%xmm0 \n" // 0011 (16b, 1u1v)
+ "movq 4(%0),%%xmm1 \n" // 1122 (16b, 1u1v)
+@@ -2676,17 +2676,17 @@ void ScaleUVRowUp2_Linear_16_SSE41(const
+ #endif
+
+ #ifdef HAS_SCALEUVROWUP2_BILINEAR_16_SSE41
+ void ScaleUVRowUp2_Bilinear_16_SSE41(const uint16_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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
+
+ LABELALIGN
+ "1: \n"
+ "movq (%0),%%xmm0 \n" // 0011 (16b, 1u1v)
+ "movq 4(%0),%%xmm1 \n" // 1122 (16b, 1u1v)
+@@ -2766,17 +2766,17 @@ void ScaleUVRowUp2_Bilinear_16_SSE41(con
+ "xmm7");
+ }
+ #endif
+
+ #ifdef HAS_SCALEUVROWUP2_LINEAR_16_AVX2
+ 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
+
+ LABELALIGN
+ "1: \n"
+ "vmovdqu (%0),%%xmm0 \n" // 00112233 (16b,
1u1v)
+ "vmovdqu 4(%0),%%xmm1 \n" // 11223344 (16b,
1u1v)
+
+@@ -2814,17 +2814,17 @@ void ScaleUVRowUp2_Linear_16_AVX2(const
+ #endif
+
+ #ifdef HAS_SCALEUVROWUP2_BILINEAR_16_AVX2
+ void ScaleUVRowUp2_Bilinear_16_AVX2(const uint16_t* src_ptr,
+ ptrdiff_t src_stride,
+ 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
+
+ LABELALIGN
+ "1: \n"
+
+ "vmovdqu (%0),%%xmm0 \n" // 00112233 (16b,
1u1v)
+ "vmovdqu 4(%0),%%xmm1 \n" // 11223344 (16b,
1u1v)
+diff --git a/media/libyuv/moz.yaml b/media/libyuv/moz.yaml
+--- a/media/libyuv/moz.yaml
++++ b/media/libyuv/moz.yaml
+@@ -50,8 +50,9 @@ vendoring:
+ exclude:
+ - ".*"
+
+ patches:
+ - 00_fix_build_errors.patch
+ - 01_make_mjpeg_printfs_optional.patch
+ - 02_update_gyp.patch
+ - 03_add_neon64_and_sve_gyp_targets.patch
++ - 04_add_missing_volatile.patch
+
diff --git a/http/firefox/patches/0029-bmo-1917964-gcc-15-swgl-fix.patch
b/http/firefox/patches/0029-bmo-1917964-gcc-15-swgl-fix.patch
new file mode 100644
index 0000000..151289c
--- /dev/null
+++ b/http/firefox/patches/0029-bmo-1917964-gcc-15-swgl-fix.patch
@@ -0,0 +1,64 @@
+
+# HG changeset patch
+# User Sam James <sam AT cmpct.info>
+# Date 1726081531 0
+# Node ID 0249085453e0d5c8378a1a053b94753fd52c956c
+# Parent 3fca72d15659808d8f453b44a13320d4f6ceca86
+Bug 1917964 - Fix swgl build with GCC 15. r=gfx-reviewers,lsalzman
+
+Fix broken specialisations which are exposed by a recent change in GCC trunk.
+
+See https://gcc.gnu.org/PR116666.
+
+Differential Revision: https://phabricator.services.mozilla.com/D221744
+
+diff --git a/gfx/wr/swgl/src/vector_type.h b/gfx/wr/swgl/src/vector_type.h
+--- a/gfx/wr/swgl/src/vector_type.h
++++ b/gfx/wr/swgl/src/vector_type.h
+@@ -235,33 +235,33 @@ struct VectorType {
+ data /= x.data;
+ return *this;
+ }
+ VectorType& operator%=(int x) {
+ data %= x;
+ return *this;
+ }
+
+- VectorType<mask_type, N> operator==(VectorType x) const {
+- return VectorType<mask_type, N>::wrap(data == x.data);
++ VectorType<mask_index, N> operator==(VectorType x) const {
++ return VectorType<mask_index, N>::wrap(data == x.data);
+ }
+- VectorType<mask_type, N> operator!=(VectorType x) const {
+- return VectorType<mask_type, N>::wrap(data != x.data);
++ VectorType<mask_index, N> operator!=(VectorType x) const {
++ return VectorType<mask_index, N>::wrap(data != x.data);
+ }
+- VectorType<mask_type, N> operator<(VectorType x) const {
+- return VectorType<mask_type, N>::wrap(data < x.data);
++ VectorType<mask_index, N> operator<(VectorType x) const {
++ return VectorType<mask_index, N>::wrap(data < x.data);
+ }
+- VectorType<mask_type, N> operator>(VectorType x) const {
+- return VectorType<mask_type, N>::wrap(data > x.data);
++ VectorType<mask_index, N> operator>(VectorType x) const {
++ return VectorType<mask_index, N>::wrap(data > x.data);
+ }
+- VectorType<mask_type, N> operator<=(VectorType x) const {
+- return VectorType<mask_type, N>::wrap(data <= x.data);
++ VectorType<mask_index, N> operator<=(VectorType x) const {
++ return VectorType<mask_index, N>::wrap(data <= x.data);
+ }
+- VectorType<mask_type, N> operator>=(VectorType x) const {
+- return VectorType<mask_type, N>::wrap(data >= x.data);
++ VectorType<mask_index, N> operator>=(VectorType x) const {
++ return VectorType<mask_index, N>::wrap(data >= x.data);
+ }
+
+ VectorType operator!() const { return wrap(!data); }
+ VectorType operator&&(VectorType x) const { return wrap(data & x.data); }
+ VectorType operator||(VectorType x) const { return wrap(data | x.data); }
+
+ VectorType& operator=(VectorType x) {
+ data = x.data;
+

[[SM-Commit] ] GIT changes to master grimoire by Pavel Vinogradov (cc8b729168989b366db218b08f904790a694d9a7), Pavel Vinogradov, 09/16/2024