Skip to content

L
libyuv
Commit cf101116 authored by Frank Barchard's avatar Frank Barchard
Browse files
Options

Remove initialize to zero on output variables for inline. 

Inline that uses temporary variables is currently initializing them
to 0 and passing in as output "+r".
This CL replaces the output constraint to "=&r" for most meaning an
output with early write (before inputs).  This allows the initialize
to zero step to be removed, saving 1 instruction.

BUG=libyuv:580
TESTED=local libyuv build on gcc/linux and try bots
R=harryjin@google.com

Review URL: https://codereview.chromium.org/1895743008 .
parent f160ce90
Hide whitespace changes
Inline Side-by-side
Showing with 116 additions and 116 deletions
README.chromium
View file @ cf101116
Name: libyuv
URL: http://code.google.com/p/libyuv/
Version: 1584
Version: 1585
License: BSD
License File: LICENSE
......
include/libyuv/version.h
View file @ cf101116
......@@ -11,6 +11,6 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT
#define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 1584
#define LIBYUV_VERSION 1585
#endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT
source/rotate_neon.cc
View file @ cf101116
......@@ -27,7 +27,7 @@ static uvec8 kVTbl4x4Transpose =
void TransposeWx8_NEON(const uint8* src, int src_stride,
uint8* dst, int dst_stride,
int width) {
const uint8* src_temp = NULL;
const uint8* src_temp;
asm volatile (
// loops are on blocks of 8. loop will stop when
// counter gets to or below 0. starting the counter
......@@ -229,7 +229,7 @@ void TransposeWx8_NEON(const uint8* src, int src_stride,
"4: \n"
: "+r"(src_temp), // %0
: "=&r"(src_temp), // %0
"+r"(src), // %1
"+r"(src_stride), // %2
"+r"(dst), // %3
......@@ -247,7 +247,7 @@ void TransposeUVWx8_NEON(const uint8* src, int src_stride,
uint8* dst_a, int dst_stride_a,
uint8* dst_b, int dst_stride_b,
int width) {
const uint8* src_temp = NULL;
const uint8* src_temp;
asm volatile (
// loops are on blocks of 8. loop will stop when
// counter gets to or below 0. starting the counter
......@@ -512,7 +512,7 @@ void TransposeUVWx8_NEON(const uint8* src, int src_stride,
"4: \n"
: "+r"(src_temp), // %0
: "=&r"(src_temp), // %0
"+r"(src), // %1
"+r"(src_stride), // %2
"+r"(dst_a), // %3
......
source/rotate_neon64.cc
View file @ cf101116
......@@ -26,7 +26,7 @@ static uvec8 kVTbl4x4Transpose =
void TransposeWx8_NEON(const uint8* src, int src_stride,
uint8* dst, int dst_stride, int width) {
const uint8* src_temp = NULL;
const uint8* src_temp;
int64 width64 = (int64) width; // Work around clang 3.4 warning.
asm volatile (
// loops are on blocks of 8. loop will stop when
......@@ -235,7 +235,7 @@ void TransposeWx8_NEON(const uint8* src, int src_stride,
"4: \n"
: "+r"(src_temp), // %0
: "=&r"(src_temp), // %0
"+r"(src), // %1
"+r"(dst), // %2
"+r"(width64) // %3
......@@ -255,7 +255,7 @@ void TransposeUVWx8_NEON(const uint8* src, int src_stride,
uint8* dst_a, int dst_stride_a,
uint8* dst_b, int dst_stride_b,
int width) {
const uint8* src_temp = NULL;
const uint8* src_temp;
int64 width64 = (int64) width; // Work around clang 3.4 warning.
asm volatile (
// loops are on blocks of 8. loop will stop when
......@@ -520,7 +520,7 @@ void TransposeUVWx8_NEON(const uint8* src, int src_stride,
"4: \n"
: "+r"(src_temp), // %0
: "=&r"(src_temp), // %0
"+r"(src), // %1
"+r"(dst_a), // %2
"+r"(dst_b), // %3
......
source/row_gcc.cc
View file @ cf101116
......@@ -1811,7 +1811,7 @@ void OMITFP I411ToARGBRow_SSSE3(const uint8* y_buf,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
int temp = 0;
int temp;
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
......@@ -1823,15 +1823,15 @@ void OMITFP I411ToARGBRow_SSSE3(const uint8* y_buf,
STOREARGB
"subl $0x8,%[width] \n"
"jg 1b \n"
: [y_buf]"+r"(y_buf), // %[y_buf]
[u_buf]"+r"(u_buf), // %[u_buf]
[v_buf]"+r"(v_buf), // %[v_buf]
: [y_buf]"+r"(y_buf), // %[y_buf]
[u_buf]"+r"(u_buf), // %[u_buf]
[v_buf]"+r"(v_buf), // %[v_buf]
[dst_argb]"+r"(dst_argb), // %[dst_argb]
[temp]"+r"(temp), // %[temp]
[temp]"=&r"(temp), // %[temp]
#if defined(__i386__) && defined(__pic__)
[width]"+m"(width) // %[width]
[width]"+m"(width) // %[width]
#else
[width]"+rm"(width) // %[width]
[width]"+rm"(width) // %[width]
#endif
: [yuvconstants]"r"(yuvconstants) // %[yuvconstants]
: "memory", "cc", NACL_R14 YUVTORGB_REGS
......@@ -3732,7 +3732,7 @@ void ARGBAttenuateRow_AVX2(const uint8* src_argb, uint8* dst_argb, int width) {
// Unattenuate 4 pixels at a time.
void ARGBUnattenuateRow_SSE2(const uint8* src_argb, uint8* dst_argb,
int width) {
uintptr_t alpha = 0;
uintptr_t alpha;
asm volatile (
// 4 pixel loop.
LABELALIGN
......@@ -3763,10 +3763,10 @@ void ARGBUnattenuateRow_SSE2(const uint8* src_argb, uint8* dst_argb,
"lea " MEMLEA(0x10,1) ",%1 \n"
"sub $0x4,%2 \n"
"jg 1b \n"
: "+r"(src_argb), // %0
"+r"(dst_argb), // %1
"+r"(width), // %2
"+r"(alpha) // %3
: "+r"(src_argb), // %0
"+r"(dst_argb), // %1
"+r"(width), // %2
"=&r"(alpha) // %3
: "r"(fixed_invtbl8) // %4
: "memory", "cc", NACL_R14
"xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
......@@ -3782,7 +3782,7 @@ static const uvec8 kUnattenShuffleAlpha_AVX2 = {
// Unattenuate 8 pixels at a time.
void ARGBUnattenuateRow_AVX2(const uint8* src_argb, uint8* dst_argb,
int width) {
uintptr_t alpha = 0;
uintptr_t alpha;
asm volatile (
"sub %0,%1 \n"
"vbroadcastf128 %5,%%ymm5 \n"
......@@ -3831,10 +3831,10 @@ void ARGBUnattenuateRow_AVX2(const uint8* src_argb, uint8* dst_argb,
"sub $0x8,%2 \n"
"jg 1b \n"
"vzeroupper \n"
: "+r"(src_argb), // %0
"+r"(dst_argb), // %1
"+r"(width), // %2
"+r"(alpha) // %3
: "+r"(src_argb), // %0
"+r"(dst_argb), // %1
"+r"(width), // %2
"=&r"(alpha) // %3
: "r"(fixed_invtbl8), // %4
"m"(kUnattenShuffleAlpha_AVX2) // %5
: "memory", "cc", NACL_R14
......@@ -4759,7 +4759,7 @@ LIBYUV_API
void ARGBAffineRow_SSE2(const uint8* src_argb, int src_argb_stride,
uint8* dst_argb, const float* src_dudv, int width) {
intptr_t src_argb_stride_temp = src_argb_stride;
intptr_t temp = 0;
intptr_t temp;
asm volatile (
"movq " MEMACCESS(3) ",%%xmm2 \n"
"movq " MEMACCESS2(0x08,3) ",%%xmm7 \n"
......@@ -4831,7 +4831,7 @@ void ARGBAffineRow_SSE2(const uint8* src_argb, int src_argb_stride,
"+r"(dst_argb), // %2
"+r"(src_dudv), // %3
"+rm"(width), // %4
"+r"(temp) // %5
"=&r"(temp) // %5
:
: "memory", "cc", NACL_R14
"xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7"
......@@ -5057,7 +5057,7 @@ void ARGBShuffleRow_AVX2(const uint8* src_argb, uint8* dst_argb,
// For BGRAToARGB, ABGRToARGB, RGBAToARGB, and ARGBToRGBA.
void ARGBShuffleRow_SSE2(const uint8* src_argb, uint8* dst_argb,
const uint8* shuffler, int width) {
uintptr_t pixel_temp = 0u;
uintptr_t pixel_temp;
asm volatile (
"pxor %%xmm5,%%xmm5 \n"
"mov " MEMACCESS(4) ",%k2 \n"
......@@ -5162,11 +5162,11 @@ void ARGBShuffleRow_SSE2(const uint8* src_argb, uint8* dst_argb,
"jg 3012b \n"
"99: \n"
: "+r"(src_argb), // %0
"+r"(dst_argb), // %1
"+d"(pixel_temp), // %2
: "+r"(src_argb), // %0
"+r"(dst_argb), // %1
"=&d"(pixel_temp), // %2
"+r"(width) // %3
: "r"(shuffler) // %4
: "r"(shuffler) // %4
: "memory", "cc", NACL_R14
"xmm0", "xmm1", "xmm5"
);
......@@ -5343,7 +5343,7 @@ void ARGBPolynomialRow_AVX2(const uint8* src_argb,
// Tranform ARGB pixels with color table.
void ARGBColorTableRow_X86(uint8* dst_argb, const uint8* table_argb,
int width) {
uintptr_t pixel_temp = 0u;
uintptr_t pixel_temp;
asm volatile (
// 1 pixel loop.
LABELALIGN
......@@ -5363,10 +5363,10 @@ void ARGBColorTableRow_X86(uint8* dst_argb, const uint8* table_argb,
"mov %b1," MEMACCESS2(-0x1,0) " \n"
"dec %2 \n"
"jg 1b \n"
: "+r"(dst_argb), // %0
"+d"(pixel_temp), // %1
"+r"(width) // %2
: "r"(table_argb) // %3
: "+r"(dst_argb), // %0
"=&d"(pixel_temp), // %1
"+r"(width) // %2
: "r"(table_argb) // %3
: "memory", "cc");
}
#endif // HAS_ARGBCOLORTABLEROW_X86
......@@ -5374,7 +5374,7 @@ void ARGBColorTableRow_X86(uint8* dst_argb, const uint8* table_argb,
#ifdef HAS_RGBCOLORTABLEROW_X86
// Tranform RGB pixels with color table.
void RGBColorTableRow_X86(uint8* dst_argb, const uint8* table_argb, int width) {
uintptr_t pixel_temp = 0u;
uintptr_t pixel_temp;
asm volatile (
// 1 pixel loop.
LABELALIGN
......@@ -5391,10 +5391,10 @@ void RGBColorTableRow_X86(uint8* dst_argb, const uint8* table_argb, int width) {
"mov %b1," MEMACCESS2(-0x2,0) " \n"
"dec %2 \n"
"jg 1b \n"
: "+r"(dst_argb), // %0
"+d"(pixel_temp), // %1
"+r"(width) // %2
: "r"(table_argb) // %3
: "+r"(dst_argb), // %0
"=&d"(pixel_temp), // %1
"+r"(width) // %2
: "r"(table_argb) // %3
: "memory", "cc");
}
#endif // HAS_RGBCOLORTABLEROW_X86
......@@ -5404,8 +5404,8 @@ void RGBColorTableRow_X86(uint8* dst_argb, const uint8* table_argb, int width) {
void ARGBLumaColorTableRow_SSSE3(const uint8* src_argb, uint8* dst_argb,
int width,
const uint8* luma, uint32 lumacoeff) {
uintptr_t pixel_temp = 0u;
uintptr_t table_temp = 0u;
uintptr_t pixel_temp;
uintptr_t table_temp;
asm volatile (
"movd %6,%%xmm3 \n"
"pshufd $0x0,%%xmm3,%%xmm3 \n"
......@@ -5487,13 +5487,13 @@ void ARGBLumaColorTableRow_SSSE3(const uint8* src_argb, uint8* dst_argb,
"lea " MEMLEA(0x10,3) ",%3 \n"
"sub $0x4,%4 \n"
"jg 1b \n"
: "+d"(pixel_temp), // %0
"+a"(table_temp), // %1
"+r"(src_argb), // %2
"+r"(dst_argb), // %3
"+rm"(width) // %4
: "r"(luma), // %5
"rm"(lumacoeff) // %6
: "=&d"(pixel_temp), // %0
"=&a"(table_temp), // %1
"+r"(src_argb), // %2
"+r"(dst_argb), // %3
"+rm"(width) // %4
: "r"(luma), // %5
"rm"(lumacoeff) // %6
: "memory", "cc", "xmm0", "xmm3", "xmm4", "xmm5"
);
}
......
source/row_mips.cc
View file @ cf101116
......@@ -498,8 +498,8 @@ void MirrorRow_DSPR2(const uint8* src, uint8* dst, int width) {
void MirrorUVRow_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
int width) {
int x = 0;
int y = 0;
int x;
int y;
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
......@@ -579,7 +579,7 @@ void MirrorUVRow_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
[dst_u] "+r" (dst_u),
[dst_v] "+r" (dst_v),
[x] "=&r" (x),
[y] "+r" (y)
[y] "=&r" (y)
: [width] "r" (width)
: "t0", "t1", "t2", "t3", "t4",
"t5", "t7", "t8", "t9"
......
source/scale_gcc.cc
View file @ cf101116
......@@ -316,7 +316,7 @@ void ScaleRowDown4_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride,
void ScaleRowDown4Box_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) {
intptr_t stridex3 = 0;
intptr_t stridex3;
asm volatile (
"pcmpeqb %%xmm4,%%xmm4 \n"
"psrlw $0xf,%%xmm4 \n"
......@@ -361,7 +361,7 @@ void ScaleRowDown4Box_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride,
: "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1
"+r"(dst_width), // %2
"+r"(stridex3) // %3
"=&r"(stridex3) // %3
: "r"((intptr_t)(src_stride)) // %4
: "memory", "cc", NACL_R14
"xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
......@@ -824,7 +824,7 @@ void ScaleAddRow_AVX2(const uint8* src_ptr, uint16* dst_ptr, int src_width) {
// Bilinear column filtering. SSSE3 version.
void ScaleFilterCols_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
int dst_width, int x, int dx) {
intptr_t x0 = 0, x1 = 0, temp_pixel = 0;
intptr_t x0, x1, temp_pixel;
asm volatile (
"movd %6,%%xmm2 \n"
"movd %7,%%xmm3 \n"
......@@ -880,14 +880,14 @@ void ScaleFilterCols_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
"movd %%xmm0,%k2 \n"
"mov %b2," MEMACCESS(0) " \n"
"99: \n"
: "+r"(dst_ptr), // %0
"+r"(src_ptr), // %1
"+a"(temp_pixel), // %2
"+r"(x0), // %3
"+r"(x1), // %4
"+rm"(dst_width) // %5
: "rm"(x), // %6
"rm"(dx) // %7
: "+r"(dst_ptr), // %0
"+r"(src_ptr), // %1
"=&a"(temp_pixel), // %2
"=&r"(x0), // %3
"=&r"(x1), // %4
"+rm"(dst_width) // %5
: "rm"(x), // %6
"rm"(dx) // %7
: "memory", "cc", NACL_R14
"xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6"
);
......@@ -998,7 +998,7 @@ void ScaleARGBRowDown2Box_SSE2(const uint8* src_argb,
void ScaleARGBRowDownEven_SSE2(const uint8* src_argb, ptrdiff_t src_stride,
int src_stepx, uint8* dst_argb, int dst_width) {
intptr_t src_stepx_x4 = (intptr_t)(src_stepx);
intptr_t src_stepx_x12 = 0;
intptr_t src_stepx_x12;
asm volatile (
"lea " MEMLEA3(0x00,1,4) ",%1 \n"
"lea " MEMLEA4(0x00,1,1,2) ",%4 \n"
......@@ -1016,11 +1016,11 @@ void ScaleARGBRowDownEven_SSE2(const uint8* src_argb, ptrdiff_t src_stride,
"lea " MEMLEA(0x10,2) ",%2 \n"
"sub $0x4,%3 \n"
"jg 1b \n"
: "+r"(src_argb), // %0
"+r"(src_stepx_x4), // %1
"+r"(dst_argb), // %2
"+r"(dst_width), // %3
"+r"(src_stepx_x12) // %4
: "+r"(src_argb), // %0
"+r"(src_stepx_x4), // %1
"+r"(dst_argb), // %2
"+r"(dst_width), // %3
"=&r"(src_stepx_x12) // %4
:: "memory", "cc", NACL_R14
"xmm0", "xmm1", "xmm2", "xmm3"
);
......@@ -1032,7 +1032,7 @@ void ScaleARGBRowDownEvenBox_SSE2(const uint8* src_argb,
ptrdiff_t src_stride, int src_stepx,
uint8* dst_argb, int dst_width) {
intptr_t src_stepx_x4 = (intptr_t)(src_stepx);
intptr_t src_stepx_x12 = 0;
intptr_t src_stepx_x12;
intptr_t row1 = (intptr_t)(src_stride);
asm volatile (
"lea " MEMLEA3(0x00,1,4) ",%1 \n"
......@@ -1061,12 +1061,12 @@ void ScaleARGBRowDownEvenBox_SSE2(const uint8* src_argb,
"lea " MEMLEA(0x10,2) ",%2 \n"
"sub $0x4,%3 \n"
"jg 1b \n"
: "+r"(src_argb), // %0
"+r"(src_stepx_x4), // %1
"+r"(dst_argb), // %2
"+rm"(dst_width), // %3
"+r"(src_stepx_x12), // %4
"+r"(row1) // %5
: "+r"(src_argb), // %0
"+r"(src_stepx_x4), // %1
"+r"(dst_argb), // %2
"+rm"(dst_width), // %3
"=&r"(src_stepx_x12), // %4
"+r"(row1) // %5
:: "memory", "cc", NACL_R14
"xmm0", "xmm1", "xmm2", "xmm3"
);
......@@ -1074,7 +1074,7 @@ void ScaleARGBRowDownEvenBox_SSE2(const uint8* src_argb,
void ScaleARGBCols_SSE2(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx) {
intptr_t x0 = 0, x1 = 0;
intptr_t x0, x1;
asm volatile (
"movd %5,%%xmm2 \n"
"movd %6,%%xmm3 \n"
......@@ -1127,8 +1127,8 @@ void ScaleARGBCols_SSE2(uint8* dst_argb, const uint8* src_argb,
MEMOPREG(movd,0x00,3,0,4,xmm0) // movd (%3,%0,4),%%xmm0
"movd %%xmm0," MEMACCESS(2) " \n"
"99: \n"
: "+a"(x0), // %0
"+d"(x1), // %1
: "=&a"(x0), // %0
"=&d"(x1), // %1
"+r"(dst_argb), // %2
"+r"(src_argb), // %3
"+r"(dst_width) // %4
......@@ -1179,7 +1179,7 @@ static uvec8 kShuffleFractions = {
// Bilinear row filtering combines 4x2 -> 4x1. SSSE3 version
void ScaleARGBFilterCols_SSSE3(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx) {
intptr_t x0 = 0, x1 = 0;
intptr_t x0, x1;
asm volatile (
"movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm5 \n"
......@@ -1242,8 +1242,8 @@ void ScaleARGBFilterCols_SSSE3(uint8* dst_argb, const uint8* src_argb,
: "+r"(dst_argb), // %0
"+r"(src_argb), // %1
"+rm"(dst_width), // %2
"+r"(x0), // %3
"+r"(x1) // %4
"=&r"(x0), // %3
"=&r"(x1) // %4
: "rm"(x), // %5
"rm"(dx) // %6
: "memory", "cc", NACL_R14
......
source/scale_neon.cc
View file @ cf101116
......@@ -532,7 +532,7 @@ void ScaleRowDown38_2_Box_NEON(const uint8* src_ptr,
void ScaleAddRows_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint16* dst_ptr, int src_width, int src_height) {
const uint8* src_tmp = NULL;
const uint8* src_tmp;
asm volatile (
"1: \n"
"mov %0, %1 \n"
......@@ -552,12 +552,12 @@ void ScaleAddRows_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
"add %1, %1, #16 \n"
"subs %4, %4, #16 \n" // 16 processed per loop
"bgt 1b \n"
: "+r"(src_tmp), // %0
"+r"(src_ptr), // %1
"+r"(dst_ptr), // %2
"+r"(src_stride), // %3
"+r"(src_width), // %4
"+r"(src_height) // %5
: "=&r"(src_tmp), // %0
"+r"(src_ptr), // %1
"+r"(dst_ptr), // %2
"+r"(src_stride), // %3
"+r"(src_width), // %4
"+r"(src_height) // %5
:
: "memory", "cc", "r12", "q0", "q1", "q2", "q3" // Clobber List
);
......@@ -909,7 +909,7 @@ void ScaleARGBRowDownEvenBox_NEON(const uint8* src_argb, ptrdiff_t src_stride,
void ScaleARGBCols_NEON(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx) {
int tmp = 0;
int tmp;
const uint8* src_tmp = src_argb;
asm volatile (
"1: \n"
......@@ -926,13 +926,13 @@ void ScaleARGBCols_NEON(uint8* dst_argb, const uint8* src_argb,
"vst1.32 {q0, q1}, [%0]! \n" // store pixels
"subs %2, %2, #8 \n" // 8 processed per loop
"bgt 1b \n"
: "+r"(dst_argb), // %0
"+r"(src_argb), // %1
"+r"(dst_width), // %2
"+r"(x), // %3
"+r"(dx), // %4
"+r"(tmp), // %5
"+r"(src_tmp) // %6
: "+r"(dst_argb), // %0
"+r"(src_argb), // %1
"+r"(dst_width), // %2
"+r"(x), // %3
"+r"(dx), // %4
"=&r"(tmp), // %5
"+r"(src_tmp) // %6
:
: "memory", "cc", "q0", "q1"
);
......
source/scale_neon64.cc
View file @ cf101116
......@@ -547,7 +547,7 @@ void ScaleRowDown38_2_Box_NEON(const uint8* src_ptr,
void ScaleAddRows_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint16* dst_ptr, int src_width, int src_height) {
const uint8* src_tmp = NULL;
const uint8* src_tmp;
asm volatile (
"1: \n"
"mov %0, %1 \n"
......@@ -567,12 +567,12 @@ void ScaleAddRows_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
"add %1, %1, #16 \n"
"subs %w4, %w4, #16 \n" // 16 processed per loop
"b.gt 1b \n"
: "+r"(src_tmp), // %0
"+r"(src_ptr), // %1
"+r"(dst_ptr), // %2
"+r"(src_stride), // %3
"+r"(src_width), // %4
"+r"(src_height) // %5
: "=&r"(src_tmp), // %0
"+r"(src_ptr), // %1
"+r"(dst_ptr), // %2
"+r"(src_stride), // %3
"+r"(src_width), // %4
"+r"(src_height) // %5
:
: "memory", "cc", "w12", "v0", "v1", "v2", "v3" // Clobber List
);
......@@ -931,7 +931,7 @@ void ScaleARGBCols_NEON(uint8* dst_argb, const uint8* src_argb,
int64 dst_width64 = (int64) dst_width; // Work around ios 64 bit warning.
int64 x64 = (int64) x;
int64 dx64 = (int64) dx;
int64 tmp64 = 0;
int64 tmp64;
asm volatile (
"1: \n"
LOAD1_DATA32_LANE(v0, 0)
......@@ -947,13 +947,13 @@ void ScaleARGBCols_NEON(uint8* dst_argb, const uint8* src_argb,
"st1 {v0.4s, v1.4s}, [%0], #32 \n" // store pixels
"subs %w2, %w2, #8 \n" // 8 processed per loop
"b.gt 1b \n"
: "+r"(dst_argb), // %0
"+r"(src_argb), // %1
"+r"(dst_width64), // %2
"+r"(x64), // %3
"+r"(dx64), // %4
"+r"(tmp64), // %5
"+r"(src_tmp) // %6
: "+r"(dst_argb), // %0
"+r"(src_argb), // %1
"+r"(dst_width64), // %2
"+r"(x64), // %3
"+r"(dx64), // %4
"=&r"(tmp64), // %5
"+r"(src_tmp) // %6
:
: "memory", "cc", "v0", "v1"
);
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment