coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2025-12-06 16:56:55 +08:00
Code Issues Packages Projects Releases Wiki Activity

With an asm() away lint, this cleans up most remaining issues.

Browse Source 
BUG=none
TEST=lint filename
Review URL: https://webrtc-codereview.appspot.com/464001

git-svn-id: http://libyuv.googlecode.com/svn/trunk@230 16f28f9a-4ce2-e073-06de-1de4eb20be90
This commit is contained in:
fbarchard@google.com 2012-03-29 02:19:26 +00:00
parent 427f966645
commit 5b22506b14
19 changed files with 90 additions and 86 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
README.chromium
View File

@ -1,6 +1,6 @@
Name: libyuv Name: libyuv
URL: http://code.google.com/p/libyuv/ URL: http://code.google.com/p/libyuv/
Version: 229 Version: 230
License: BSD License: BSD
License File: LICENSE License File: LICENSE

14
include/libyuv/basic_types.h
View File

@ -31,8 +31,8 @@ typedef __int64 int64;
#define INT64_F "I64" #define INT64_F "I64"
#else // COMPILER_MSVC #else // COMPILER_MSVC
#ifdef __LP64__ #ifdef __LP64__
typedef unsigned long uint64; typedef unsigned long uint64; // NOLINT
typedef long int64; typedef long int64; // NOLINT
#ifndef INT64_C #ifndef INT64_C
#define INT64_C(x) x ## L #define INT64_C(x) x ## L
#endif #endif
@ -41,8 +41,8 @@ typedef long int64;
#endif #endif
#define INT64_F "l" #define INT64_F "l"
#else // __LP64__ #else // __LP64__
typedef unsigned long long uint64; typedef unsigned long long uint64; // NOLINT
typedef long long int64; typedef long long int64; // NOLINT
#ifndef INT64_C #ifndef INT64_C
#define INT64_C(x) x ## LL #define INT64_C(x) x ## LL
#endif #endif
@ -54,8 +54,8 @@ typedef long long int64;
#endif // COMPILER_MSVC #endif // COMPILER_MSVC
typedef unsigned int uint32; typedef unsigned int uint32;
typedef int int32; typedef int int32;
typedef unsigned short uint16; typedef unsigned short uint16; // NOLINT
typedef short int16; typedef short int16; // NOLINT
typedef unsigned char uint8; typedef unsigned char uint8;
typedef signed char int8; typedef signed char int8;
#endif // INT_TYPES_DEFINED #endif // INT_TYPES_DEFINED
@ -70,4 +70,4 @@ typedef signed char int8;
(reinterpret_cast<uint8*>(((reinterpret_cast<uintptr_t>(p) + \ (reinterpret_cast<uint8*>(((reinterpret_cast<uintptr_t>(p) + \
((t)-1)) & ~((t)-1)))) ((t)-1)) & ~((t)-1))))
#endif // INCLUDE_LIBYUV_BASIC_TYPES_H_ #endif // INCLUDE_LIBYUV_BASIC_TYPES_H_

2
include/libyuv/compare.h
View File

@ -62,4 +62,4 @@ double I420Ssim(const uint8* src_y_a, int stride_y_a,
} // namespace libyuv } // namespace libyuv
#endif #endif
#endif // INCLUDE_LIBYUV_COMPARE_H_ #endif // INCLUDE_LIBYUV_COMPARE_H_

2
include/libyuv/convert.h
View File

@ -215,4 +215,4 @@ int ConvertToI420(const uint8* src_frame, size_t src_size,
} // namespace libyuv } // namespace libyuv
#endif #endif
#endif // INCLUDE_LIBYUV_CONVERT_H_ #endif // INCLUDE_LIBYUV_CONVERT_H_

2
include/libyuv/convert_from.h
View File

@ -139,4 +139,4 @@ int ConvertFromI420(const uint8* y, int y_stride,
} // namespace libyuv } // namespace libyuv
#endif #endif
#endif // INCLUDE_LIBYUV_CONVERT_FROM_H_ #endif // INCLUDE_LIBYUV_CONVERT_FROM_H_

12
include/libyuv/format_conversion.h
View File

@ -44,8 +44,8 @@ int BayerRGGBToI420(const uint8* src_bayer, int src_stride_bayer,
int width, int height); int width, int height);
// Temporary API mapper // Temporary API mapper
#define BayerRGBToI420(b,bs,f,y,ys,u,us,v,vs,w,h) \ #define BayerRGBToI420(b, bs, f, y, ys, u, us, v, vs, w, h) \
BayerToI420(b,bs,y,ys,u,us,v,vs,w,h,f) BayerToI420(b, bs, y, ys, u, us, v, vs, w, h, f)
int BayerToI420(const uint8* src_bayer, int src_stride_bayer, int BayerToI420(const uint8* src_bayer, int src_stride_bayer,
uint8* dst_y, int dst_stride_y, uint8* dst_y, int dst_stride_y,
@ -80,8 +80,8 @@ int I420ToBayerRGGB(const uint8* src_y, int src_stride_y,
int width, int height); int width, int height);
// Temporary API mapper // Temporary API mapper
#define I420ToBayerRGB(y,ys,u,us,v,vs,b,bs,f,w,h) \ #define I420ToBayerRGB(y, ys, u, us, v, vs, b, bs, f, w, h) \
I420ToBayer(y,ys,u,us,v,vs,b,bs,w,h,f) I420ToBayer(y, ys, u, us, v, vs, b, bs, w, h, f)
int I420ToBayer(const uint8* src_y, int src_stride_y, int I420ToBayer(const uint8* src_y, int src_stride_y,
const uint8* src_u, int src_stride_u, const uint8* src_u, int src_stride_u,
@ -108,7 +108,7 @@ int BayerRGGBToARGB(const uint8* src_bayer, int src_stride_bayer,
int width, int height); int width, int height);
// Temporary API mapper // Temporary API mapper
#define BayerRGBToARGB(b,bs,f,a,as,w,h) BayerToARGB(b,bs,a,as,w,h,f) #define BayerRGBToARGB(b, bs, f, a, as, w, h) BayerToARGB(b, bs, a, as, w, h, f)
int BayerToARGB(const uint8* src_bayer, int src_stride_bayer, int BayerToARGB(const uint8* src_bayer, int src_stride_bayer,
uint8* dst_argb, int dst_stride_argb, uint8* dst_argb, int dst_stride_argb,
@ -133,7 +133,7 @@ int ARGBToBayerRGGB(const uint8* src_argb, int src_stride_argb,
int width, int height); int width, int height);
// Temporary API mapper // Temporary API mapper
#define ARGBToBayerRGB(a,as,b,bs,f,w,h) ARGBToBayer(b,bs,a,as,w,h,f) #define ARGBToBayerRGB(a, as, b, bs, f, w, h) ARGBToBayer(b, bs, a, as, w, h, f)
int ARGBToBayer(const uint8* src_argb, int src_stride_argb, int ARGBToBayer(const uint8* src_argb, int src_stride_argb,
uint8* dst_bayer, int dst_stride_bayer, uint8* dst_bayer, int dst_stride_bayer,

2
include/libyuv/version.h
View File

@ -11,7 +11,7 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ #ifndef INCLUDE_LIBYUV_VERSION_H_
#define INCLUDE_LIBYUV_VERSION_H_ #define INCLUDE_LIBYUV_VERSION_H_
#define INCLUDE_LIBYUV_VERSION 229 #define INCLUDE_LIBYUV_VERSION 230
#endif // INCLUDE_LIBYUV_VERSION_H_ #endif // INCLUDE_LIBYUV_VERSION_H_

6
source/compare.cc
View File

@ -42,7 +42,7 @@ uint32 HashDjb2(const uint8* src, uint64 count, uint32 seed) {
static uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b, static uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b,
int count) { int count) {
volatile uint32 sse; volatile uint32 sse;
asm volatile( asm volatile (
"vmov.u8 q7, #0 \n" "vmov.u8 q7, #0 \n"
"vmov.u8 q9, #0 \n" "vmov.u8 q9, #0 \n"
"vmov.u8 q8, #0 \n" "vmov.u8 q8, #0 \n"
@ -116,12 +116,12 @@ static uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b,
} }
} }
#elif (defined(__x86_64__) || defined(__i386__)) && !defined(YUV_DISABLE_ASM) #elif defined(__x86_64__) || defined(__i386__) && !defined(YUV_DISABLE_ASM)
#define HAS_SUMSQUAREERROR_SSE2 #define HAS_SUMSQUAREERROR_SSE2
static uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b, static uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b,
int count) { int count) {
uint32 sse; uint32 sse;
asm volatile( asm volatile (
"pxor %%xmm0,%%xmm0 \n" "pxor %%xmm0,%%xmm0 \n"
"pxor %%xmm5,%%xmm5 \n" "pxor %%xmm5,%%xmm5 \n"
"sub %0,%1 \n" "sub %0,%1 \n"

4
source/convert.cc
View File

@ -90,7 +90,7 @@ static void HalfRow_SSE2(const uint8* src_uv, int src_uv_stride,
#define HAS_HALFROW_SSE2 #define HAS_HALFROW_SSE2
static void HalfRow_SSE2(const uint8* src_uv, int src_uv_stride, static void HalfRow_SSE2(const uint8* src_uv, int src_uv_stride,
uint8* dst_uv, int pix) { uint8* dst_uv, int pix) {
asm volatile( asm volatile (
"sub %0,%1 \n" "sub %0,%1 \n"
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa (%0),%%xmm0 \n"
@ -502,7 +502,7 @@ static void SplitYUY2_SSE2(const uint8* src_yuy2,
#define HAS_SPLITYUY2_SSE2 #define HAS_SPLITYUY2_SSE2
static void SplitYUY2_SSE2(const uint8* src_yuy2, uint8* dst_y, static void SplitYUY2_SSE2(const uint8* src_yuy2, uint8* dst_y,
uint8* dst_u, uint8* dst_v, int pix) { uint8* dst_u, uint8* dst_v, int pix) {
asm volatile( asm volatile (
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $0x8,%%xmm5 \n" "psrlw $0x8,%%xmm5 \n"
"1: \n" "1: \n"

4
source/convert_from.cc
View File

@ -289,7 +289,7 @@ static void I42xToYUY2Row_SSE2(const uint8* src_y,
const uint8* src_u, const uint8* src_u,
const uint8* src_v, const uint8* src_v,
uint8* dst_frame, int width) { uint8* dst_frame, int width) {
asm volatile( asm volatile (
"sub %1,%2 \n" "sub %1,%2 \n"
"1: \n" "1: \n"
"movq (%1),%%xmm2 \n" "movq (%1),%%xmm2 \n"
@ -324,7 +324,7 @@ static void I42xToUYVYRow_SSE2(const uint8* src_y,
const uint8* src_u, const uint8* src_u,
const uint8* src_v, const uint8* src_v,
uint8* dst_frame, int width) { uint8* dst_frame, int width) {
asm volatile( asm volatile (
"sub %1,%2 \n" "sub %1,%2 \n"
"1: \n" "1: \n"
"movq (%1),%%xmm2 \n" "movq (%1),%%xmm2 \n"

4
source/cpu_id.cc
View File

@ -24,7 +24,7 @@
// TODO(fbarchard): Use cpuid.h when gcc 4.4 is used on OSX and Linux. // TODO(fbarchard): Use cpuid.h when gcc 4.4 is used on OSX and Linux.
#if (defined(__pic__) || defined(__APPLE__)) && defined(__i386__) #if (defined(__pic__) || defined(__APPLE__)) && defined(__i386__)
static __inline void __cpuid(int cpu_info[4], int info_type) { static __inline void __cpuid(int cpu_info[4], int info_type) {
asm volatile( asm volatile (
"mov %%ebx, %%edi \n" "mov %%ebx, %%edi \n"
"cpuid \n" "cpuid \n"
"xchg %%edi, %%ebx \n" "xchg %%edi, %%ebx \n"
@ -33,7 +33,7 @@ static __inline void __cpuid(int cpu_info[4], int info_type) {
} }
#elif defined(__i386__) || defined(__x86_64__) #elif defined(__i386__) || defined(__x86_64__)
static __inline void __cpuid(int cpu_info[4], int info_type) { static __inline void __cpuid(int cpu_info[4], int info_type) {
asm volatile( asm volatile (
"cpuid \n" "cpuid \n"
: "=a"(cpu_info[0]), "=b"(cpu_info[1]), "=c"(cpu_info[2]), "=d"(cpu_info[3]) : "=a"(cpu_info[0]), "=b"(cpu_info[1]), "=c"(cpu_info[2]), "=d"(cpu_info[3])
: "a"(info_type)); : "a"(info_type));

2
source/format_conversion.cc
View File

@ -53,7 +53,7 @@ static void ARGBToBayerRow_SSSE3(const uint8* src_argb,
#define HAS_ARGBTOBAYERROW_SSSE3 #define HAS_ARGBTOBAYERROW_SSSE3
static void ARGBToBayerRow_SSSE3(const uint8* src_argb, uint8* dst_bayer, static void ARGBToBayerRow_SSSE3(const uint8* src_argb, uint8* dst_bayer,
uint32 selector, int pix) { uint32 selector, int pix) {
asm volatile( asm volatile (
"movd %3,%%xmm5 \n" "movd %3,%%xmm5 \n"
"pshufd $0x0,%%xmm5,%%xmm5 \n" "pshufd $0x0,%%xmm5,%%xmm5 \n"
"1: \n" "1: \n"

6
source/planar_functions.cc
View File

@ -654,7 +654,7 @@ int NV12ToRGB565(const uint8* src_y, int src_stride_y,
#if defined(__ARM_NEON__) && !defined(YUV_DISABLE_ASM) #if defined(__ARM_NEON__) && !defined(YUV_DISABLE_ASM)
#define HAS_SETROW_NEON #define HAS_SETROW_NEON
static void SetRow8_NEON(uint8* dst, uint32 v32, int count) { static void SetRow8_NEON(uint8* dst, uint32 v32, int count) {
asm volatile( asm volatile (
"vdup.u32 q0, %2 \n" // duplicate 4 ints "vdup.u32 q0, %2 \n" // duplicate 4 ints
"1: \n" "1: \n"
"subs %1, %1, #16 \n" // 16 bytes per loop "subs %1, %1, #16 \n" // 16 bytes per loop
@ -723,7 +723,7 @@ static void SetRows32_X86(uint8* dst, uint32 v32, int width,
#define HAS_SETROW_X86 #define HAS_SETROW_X86
static void SetRow8_X86(uint8* dst, uint32 v32, int width) { static void SetRow8_X86(uint8* dst, uint32 v32, int width) {
size_t width_tmp = static_cast<size_t>(width); size_t width_tmp = static_cast<size_t>(width);
asm volatile( asm volatile (
"shr $0x2,%1 \n" "shr $0x2,%1 \n"
"rep stosl \n" "rep stosl \n"
: "+D"(dst), // %0 : "+D"(dst), // %0
@ -737,7 +737,7 @@ static void SetRows32_X86(uint8* dst, uint32 v32, int width,
for (int y = 0; y < height; ++y) { for (int y = 0; y < height; ++y) {
size_t width_tmp = static_cast<size_t>(width); size_t width_tmp = static_cast<size_t>(width);
uint32* d = reinterpret_cast<uint32*>(dst); uint32* d = reinterpret_cast<uint32*>(dst);
asm volatile( asm volatile (
"rep stosl \n" "rep stosl \n"
: "+D"(d), // %0 : "+D"(d), // %0
"+c"(width_tmp) // %1 "+c"(width_tmp) // %1

6
source/rotate.cc
View File

@ -295,7 +295,7 @@ static void TransposeUVWx8_SSE2(const uint8* src, int src_stride,
#define HAS_TRANSPOSE_WX8_SSSE3 #define HAS_TRANSPOSE_WX8_SSSE3
static void TransposeWx8_SSSE3(const uint8* src, int src_stride, static void TransposeWx8_SSSE3(const uint8* src, int src_stride,
uint8* dst, int dst_stride, int width) { uint8* dst, int dst_stride, int width) {
asm volatile( asm volatile (
// Read in the data from the source pointer. // Read in the data from the source pointer.
// First round of bit swap. // First round of bit swap.
"1: \n" "1: \n"
@ -506,7 +506,7 @@ extern "C" void TransposeUVWx8_SSE2(const uint8* src, int src_stride,
#define HAS_TRANSPOSE_WX8_FAST_SSSE3 #define HAS_TRANSPOSE_WX8_FAST_SSSE3
static void TransposeWx8_FAST_SSSE3(const uint8* src, int src_stride, static void TransposeWx8_FAST_SSSE3(const uint8* src, int src_stride,
uint8* dst, int dst_stride, int width) { uint8* dst, int dst_stride, int width) {
asm volatile( asm volatile (
// Read in the data from the source pointer. // Read in the data from the source pointer.
// First round of bit swap. // First round of bit swap.
"1: \n" "1: \n"
@ -646,7 +646,7 @@ static void TransposeUVWx8_SSE2(const uint8* src, int src_stride,
uint8* dst_a, int dst_stride_a, uint8* dst_a, int dst_stride_a,
uint8* dst_b, int dst_stride_b, uint8* dst_b, int dst_stride_b,
int w) { int w) {
asm volatile( asm volatile (
// Read in the data from the source pointer. // Read in the data from the source pointer.
// First round of bit swap. // First round of bit swap.
"1: \n" "1: \n"

4
source/rotate_neon.cc
View File

@ -25,7 +25,7 @@ static const uvec8 vtbl_4x4_transpose =
void TransposeWx8_NEON(const uint8* src, int src_stride, void TransposeWx8_NEON(const uint8* src, int src_stride,
uint8* dst, int dst_stride, uint8* dst, int dst_stride,
int width) { int width) {
asm volatile( asm volatile (
// loops are on blocks of 8. loop will stop when // loops are on blocks of 8. loop will stop when
// counter gets to or below 0. starting the counter // counter gets to or below 0. starting the counter
// at w-8 allow for this // at w-8 allow for this
@ -191,7 +191,7 @@ void TransposeUVWx8_NEON(const uint8* src, int src_stride,
uint8* dst_a, int dst_stride_a, uint8* dst_a, int dst_stride_a,
uint8* dst_b, int dst_stride_b, uint8* dst_b, int dst_stride_b,
int width) { int width) {
asm volatile( asm volatile (
// loops are on blocks of 8. loop will stop when // loops are on blocks of 8. loop will stop when
// counter gets to or below 0. starting the counter // counter gets to or below 0. starting the counter
// at w-8 allow for this // at w-8 allow for this

2
source/row.h
View File

@ -87,7 +87,7 @@ extern "C" {
typedef __declspec(align(16)) int8 vec8[16]; typedef __declspec(align(16)) int8 vec8[16];
typedef __declspec(align(16)) uint8 uvec8[16]; typedef __declspec(align(16)) uint8 uvec8[16];
typedef __declspec(align(16)) int16 vec16[8]; typedef __declspec(align(16)) int16 vec16[8];
#else // __GNUC__ #else // __GNUC__
#define SIMD_ALIGNED(var) var __attribute__((aligned(16))) #define SIMD_ALIGNED(var) var __attribute__((aligned(16)))
typedef int8 __attribute__((vector_size(16))) vec8; typedef int8 __attribute__((vector_size(16))) vec8;
typedef uint8 __attribute__((vector_size(16))) uvec8; typedef uint8 __attribute__((vector_size(16))) uvec8;

28
source/row_common.cc
View File

@ -10,9 +10,10 @@
#include "source/row.h" #include "source/row.h"
#include "libyuv/basic_types.h"
#include <string.h> // For memcpy #include <string.h> // For memcpy
#include "libyuv/basic_types.h"
#ifdef __cplusplus #ifdef __cplusplus
namespace libyuv { namespace libyuv {
extern "C" { extern "C" {
@ -195,7 +196,7 @@ static __inline int RGBToV(uint8 r, uint8 g, uint8 b) {
return ((112 * r - 94 * g - 18 * b + 128) >> 8) + 128; return ((112 * r - 94 * g - 18 * b + 128) >> 8) + 128;
} }
#define MAKEROWY(NAME,R,G,B) \ #define MAKEROWY(NAME, R, G, B) \
void NAME ## ToYRow_C(const uint8* src_argb0, uint8* dst_y, int width) { \ void NAME ## ToYRow_C(const uint8* src_argb0, uint8* dst_y, int width) { \
for (int x = 0; x < width; ++x) { \ for (int x = 0; x < width; ++x) { \
dst_y[0] = RGBToY(src_argb0[R], src_argb0[G], src_argb0[B]); \ dst_y[0] = RGBToY(src_argb0[R], src_argb0[G], src_argb0[B]); \
@ -229,9 +230,9 @@ void NAME ## ToUVRow_C(const uint8* src_rgb0, int src_stride_rgb, \
} \ } \
} }
MAKEROWY(ARGB,2,1,0) MAKEROWY(ARGB, 2, 1, 0)
MAKEROWY(BGRA,1,2,3) MAKEROWY(BGRA, 1, 2, 3)
MAKEROWY(ABGR,0,1,2) MAKEROWY(ABGR, 0, 1, 2)
void I400ToARGBRow_C(const uint8* src_y, uint8* dst_argb, int width) { void I400ToARGBRow_C(const uint8* src_y, uint8* dst_argb, int width) {
// Copy a Y to RGB. // Copy a Y to RGB.
@ -263,11 +264,11 @@ void I400ToARGBRow_C(const uint8* src_y, uint8* dst_argb, int width) {
static __inline uint32 Clip(int32 val) { static __inline uint32 Clip(int32 val) {
if (val < 0) { if (val < 0) {
return (uint32) 0; return static_cast<uint32>(0);
} else if (val > 255){ } else if (val > 255) {
return (uint32) 255; return static_cast<uint32>(255);
} }
return (uint32) val; return static_cast<uint32>(val);
} }
static __inline void YuvPixel(uint8 y, uint8 u, uint8 v, uint8* rgb_buf, static __inline void YuvPixel(uint8 y, uint8 u, uint8 v, uint8* rgb_buf,
@ -469,7 +470,8 @@ void ARGBBlendRow_C(const uint8* src_argb, uint8* dst_argb, int width) {
dst_argb[2] = BLENDER(fr, br, a); dst_argb[2] = BLENDER(fr, br, a);
dst_argb[3] = 255u; dst_argb[3] = 255u;
} else { } else {
*(uint32*)dst_argb = *(uint32*)src_argb; *reinterpret_cast<uint32*>(dst_argb) =
*reinterpret_cast<uint32*>(src_argb);
} }
} }
a = src_argb[4 + 3]; a = src_argb[4 + 3];
@ -486,7 +488,8 @@ void ARGBBlendRow_C(const uint8* src_argb, uint8* dst_argb, int width) {
dst_argb[4 + 2] = BLENDER(fr, br, a); dst_argb[4 + 2] = BLENDER(fr, br, a);
dst_argb[4 + 3] = 255u; dst_argb[4 + 3] = 255u;
} else { } else {
*(uint32*)(dst_argb + 4) = *(uint32*)(src_argb + 4); *reinterpret_cast<uint32*>(dst_argb + 4) =
*reinterpret_cast<uint32*>(src_argb + 4);
} }
} }
src_argb += 8; src_argb += 8;
@ -508,7 +511,8 @@ void ARGBBlendRow_C(const uint8* src_argb, uint8* dst_argb, int width) {
dst_argb[2] = BLENDER(fr, br, a); dst_argb[2] = BLENDER(fr, br, a);
dst_argb[3] = 255u; dst_argb[3] = 255u;
} else { } else {
*(uint32*)dst_argb = *(uint32*)src_argb; *reinterpret_cast<uint32*>(dst_argb) =
*reinterpret_cast<uint32*>(src_argb);
} }
} }
} }

14
source/row_neon.cc
View File

@ -61,7 +61,7 @@ void I420ToARGBRow_NEON(const uint8* y_buf,
const uint8* v_buf, const uint8* v_buf,
uint8* rgb_buf, uint8* rgb_buf,
int width) { int width) {
asm volatile( asm volatile (
"vld1.u8 {d24}, [%5] \n" "vld1.u8 {d24}, [%5] \n"
"vld1.u8 {d25}, [%6] \n" "vld1.u8 {d25}, [%6] \n"
"vmov.u8 d26, #128 \n" "vmov.u8 d26, #128 \n"
@ -93,7 +93,7 @@ void I420ToBGRARow_NEON(const uint8* y_buf,
const uint8* v_buf, const uint8* v_buf,
uint8* rgb_buf, uint8* rgb_buf,
int width) { int width) {
asm volatile( asm volatile (
"vld1.u8 {d24}, [%5] \n" "vld1.u8 {d24}, [%5] \n"
"vld1.u8 {d25}, [%6] \n" "vld1.u8 {d25}, [%6] \n"
"vmov.u8 d26, #128 \n" "vmov.u8 d26, #128 \n"
@ -126,7 +126,7 @@ void I420ToABGRRow_NEON(const uint8* y_buf,
const uint8* v_buf, const uint8* v_buf,
uint8* rgb_buf, uint8* rgb_buf,
int width) { int width) {
asm volatile( asm volatile (
"vld1.u8 {d24}, [%5] \n" "vld1.u8 {d24}, [%5] \n"
"vld1.u8 {d25}, [%6] \n" "vld1.u8 {d25}, [%6] \n"
"vmov.u8 d26, #128 \n" "vmov.u8 d26, #128 \n"
@ -157,7 +157,7 @@ YUVTORGB
// Reads 16 pairs of UV and write even values to dst_u and odd to dst_v // Reads 16 pairs of UV and write even values to dst_u and odd to dst_v
// Alignment requirement: 16 bytes for pointers, and multiple of 16 pixels. // Alignment requirement: 16 bytes for pointers, and multiple of 16 pixels.
void SplitUV_NEON(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width) { void SplitUV_NEON(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width) {
asm volatile( asm volatile (
"1: \n" "1: \n"
"vld2.u8 {q0,q1}, [%0]! \n" // load 16 pairs of UV "vld2.u8 {q0,q1}, [%0]! \n" // load 16 pairs of UV
"subs %3, %3, #16 \n" // 16 processed per loop "subs %3, %3, #16 \n" // 16 processed per loop
@ -177,7 +177,7 @@ void SplitUV_NEON(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width) {
#ifdef HAS_COPYROW_NEON #ifdef HAS_COPYROW_NEON
// Copy multiple of 64 // Copy multiple of 64
void CopyRow_NEON(const uint8* src, uint8* dst, int count) { void CopyRow_NEON(const uint8* src, uint8* dst, int count) {
asm volatile( asm volatile (
"1: \n" "1: \n"
"pld [%0, #0xC0] \n" // preload "pld [%0, #0xC0] \n" // preload
"vldm %0!,{q0,q1,q2,q3} \n" // load 64 "vldm %0!,{q0,q1,q2,q3} \n" // load 64
@ -195,7 +195,7 @@ void CopyRow_NEON(const uint8* src, uint8* dst, int count) {
#ifdef HAS_MIRRORROW_NEON #ifdef HAS_MIRRORROW_NEON
void MirrorRow_NEON(const uint8* src, uint8* dst, int width) { void MirrorRow_NEON(const uint8* src, uint8* dst, int width) {
asm volatile( asm volatile (
// compute where to start writing destination // compute where to start writing destination
"add %1, %2 \n" "add %1, %2 \n"
// work on segments that are multiples of 16 // work on segments that are multiples of 16
@ -270,7 +270,7 @@ void MirrorRow_NEON(const uint8* src, uint8* dst, int width) {
#ifdef HAS_MIRRORROWUV_NEON #ifdef HAS_MIRRORROWUV_NEON
void MirrorRowUV_NEON(const uint8* src, uint8* dst_a, uint8* dst_b, int width) { void MirrorRowUV_NEON(const uint8* src, uint8* dst_a, uint8* dst_b, int width) {
asm volatile( asm volatile (
// compute where to start writing destination // compute where to start writing destination
"add %1, %3 \n" // dst_a + width "add %1, %3 \n" // dst_a + width
"add %2, %3 \n" // dst_b + width "add %2, %3 \n" // dst_b + width

60
source/scale.cc
View File

@ -59,7 +59,7 @@ void SetUseReferenceImpl(bool use) {
#define HAS_SCALEROWDOWN2_NEON #define HAS_SCALEROWDOWN2_NEON
void ScaleRowDown2_NEON(const uint8* src_ptr, int /* src_stride */, void ScaleRowDown2_NEON(const uint8* src_ptr, int /* src_stride */,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
asm volatile( asm volatile (
"1: \n" "1: \n"
"vld2.u8 {q0,q1}, [%0]! \n" // load even pixels into q0, odd into q1 "vld2.u8 {q0,q1}, [%0]! \n" // load even pixels into q0, odd into q1
"vst1.u8 {q0}, [%1]! \n" // store even pixels "vst1.u8 {q0}, [%1]! \n" // store even pixels
@ -75,7 +75,7 @@ void ScaleRowDown2_NEON(const uint8* src_ptr, int /* src_stride */,
void ScaleRowDown2Int_NEON(const uint8* src_ptr, int src_stride, void ScaleRowDown2Int_NEON(const uint8* src_ptr, int src_stride,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
asm volatile( asm volatile (
"add %1, %0 \n" // change the stride to row 2 pointer "add %1, %0 \n" // change the stride to row 2 pointer
"1: \n" "1: \n"
"vld1.u8 {q0,q1}, [%0]! \n" // load row 1 and post increment "vld1.u8 {q0,q1}, [%0]! \n" // load row 1 and post increment
@ -101,7 +101,7 @@ void ScaleRowDown2Int_NEON(const uint8* src_ptr, int src_stride,
#define HAS_SCALEROWDOWN4_NEON #define HAS_SCALEROWDOWN4_NEON
static void ScaleRowDown4_NEON(const uint8* src_ptr, int /* src_stride */, static void ScaleRowDown4_NEON(const uint8* src_ptr, int /* src_stride */,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"1: \n" "1: \n"
"vld2.u8 {d0, d1}, [%0]! \n" "vld2.u8 {d0, d1}, [%0]! \n"
"vtrn.u8 d1, d0 \n" "vtrn.u8 d1, d0 \n"
@ -120,7 +120,7 @@ static void ScaleRowDown4_NEON(const uint8* src_ptr, int /* src_stride */,
static void ScaleRowDown4Int_NEON(const uint8* src_ptr, int src_stride, static void ScaleRowDown4Int_NEON(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"add r4, %0, %3 \n" "add r4, %0, %3 \n"
"add r5, r4, %3 \n" "add r5, r4, %3 \n"
"add %3, r5, %3 \n" "add %3, r5, %3 \n"
@ -159,7 +159,7 @@ static void ScaleRowDown4Int_NEON(const uint8* src_ptr, int src_stride,
// Point samples 32 pixels to 24 pixels. // Point samples 32 pixels to 24 pixels.
static void ScaleRowDown34_NEON(const uint8* src_ptr, int /* src_stride */, static void ScaleRowDown34_NEON(const uint8* src_ptr, int /* src_stride */,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"1: \n" "1: \n"
"vld4.u8 {d0, d1, d2, d3}, [%0]! \n" // src line 0 "vld4.u8 {d0, d1, d2, d3}, [%0]! \n" // src line 0
"vmov d2, d3 \n" // order needs to be d0, d1, d2 "vmov d2, d3 \n" // order needs to be d0, d1, d2
@ -176,7 +176,7 @@ static void ScaleRowDown34_NEON(const uint8* src_ptr, int /* src_stride */,
static void ScaleRowDown34_0_Int_NEON(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_0_Int_NEON(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"vmov.u8 d24, #3 \n" "vmov.u8 d24, #3 \n"
"add %3, %0 \n" "add %3, %0 \n"
"1: \n" "1: \n"
@ -231,7 +231,7 @@ static void ScaleRowDown34_0_Int_NEON(const uint8* src_ptr, int src_stride,
static void ScaleRowDown34_1_Int_NEON(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_1_Int_NEON(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"vmov.u8 d24, #3 \n" "vmov.u8 d24, #3 \n"
"add %3, %0 \n" "add %3, %0 \n"
"1: \n" "1: \n"
@ -283,7 +283,7 @@ const unsigned short mult38_div9[8] __attribute__ ((aligned(16))) =
// 32 -> 12 // 32 -> 12
static void ScaleRowDown38_NEON(const uint8* src_ptr, int, static void ScaleRowDown38_NEON(const uint8* src_ptr, int,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"vld1.u8 {q3}, [%3] \n" "vld1.u8 {q3}, [%3] \n"
"1: \n" "1: \n"
"vld1.u8 {d0, d1, d2, d3}, [%0]! \n" "vld1.u8 {d0, d1, d2, d3}, [%0]! \n"
@ -304,7 +304,7 @@ static void ScaleRowDown38_NEON(const uint8* src_ptr, int,
// 32x3 -> 12x1 // 32x3 -> 12x1
static void ScaleRowDown38_3_Int_NEON(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_3_Int_NEON(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"vld1.u16 {q13}, [%4] \n" "vld1.u16 {q13}, [%4] \n"
"vld1.u8 {q14}, [%5] \n" "vld1.u8 {q14}, [%5] \n"
"vld1.u8 {q15}, [%6] \n" "vld1.u8 {q15}, [%6] \n"
@ -413,7 +413,7 @@ static void ScaleRowDown38_3_Int_NEON(const uint8* src_ptr, int src_stride,
// 32x2 -> 12x1 // 32x2 -> 12x1
static void ScaleRowDown38_2_Int_NEON(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_2_Int_NEON(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"vld1.u16 {q13}, [%4] \n" "vld1.u16 {q13}, [%4] \n"
"vld1.u8 {q14}, [%5] \n" "vld1.u8 {q14}, [%5] \n"
"add %3, %0 \n" "add %3, %0 \n"
@ -508,7 +508,7 @@ static void ScaleRowDown38_2_Int_NEON(const uint8* src_ptr, int src_stride,
static void ScaleFilterRows_NEON(uint8* dst_ptr, static void ScaleFilterRows_NEON(uint8* dst_ptr,
const uint8* src_ptr, int src_stride, const uint8* src_ptr, int src_stride,
int dst_width, int source_y_fraction) { int dst_width, int source_y_fraction) {
asm volatile( asm volatile (
"cmp %4, #0 \n" "cmp %4, #0 \n"
"beq 2f \n" "beq 2f \n"
"add %2, %1 \n" "add %2, %1 \n"
@ -1555,7 +1555,7 @@ static void ScaleFilterCols34_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
#define HAS_SCALEROWDOWN2_SSE2 #define HAS_SCALEROWDOWN2_SSE2
static void ScaleRowDown2_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown2_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $0x8,%%xmm5 \n" "psrlw $0x8,%%xmm5 \n"
"1: \n" "1: \n"
@ -1579,7 +1579,7 @@ static void ScaleRowDown2_SSE2(const uint8* src_ptr, int src_stride,
static void ScaleRowDown2Int_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown2Int_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $0x8,%%xmm5 \n" "psrlw $0x8,%%xmm5 \n"
"1: \n" "1: \n"
@ -1614,7 +1614,7 @@ static void ScaleRowDown2Int_SSE2(const uint8* src_ptr, int src_stride,
#define HAS_SCALEROWDOWN4_SSE2 #define HAS_SCALEROWDOWN4_SSE2
static void ScaleRowDown4_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown4_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrld $0x18,%%xmm5 \n" "psrld $0x18,%%xmm5 \n"
"1: \n" "1: \n"
@ -1640,7 +1640,7 @@ static void ScaleRowDown4_SSE2(const uint8* src_ptr, int src_stride,
static void ScaleRowDown4Int_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown4Int_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
intptr_t temp = 0; intptr_t temp = 0;
asm volatile( asm volatile (
"pcmpeqb %%xmm7,%%xmm7 \n" "pcmpeqb %%xmm7,%%xmm7 \n"
"psrlw $0x8,%%xmm7 \n" "psrlw $0x8,%%xmm7 \n"
"lea (%4,%4,2),%3 \n" "lea (%4,%4,2),%3 \n"
@ -1693,7 +1693,7 @@ static void ScaleRowDown4Int_SSE2(const uint8* src_ptr, int src_stride,
#define HAS_SCALEROWDOWN8_SSE2 #define HAS_SCALEROWDOWN8_SSE2
static void ScaleRowDown8_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown8_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlq $0x38,%%xmm5 \n" "psrlq $0x38,%%xmm5 \n"
"1: \n" "1: \n"
@ -1722,7 +1722,7 @@ static void ScaleAddRows_SSE2(const uint8* src_ptr, int src_stride,
uint16* dst_ptr, int src_width, int src_height) { uint16* dst_ptr, int src_width, int src_height) {
int tmp_height = 0; int tmp_height = 0;
intptr_t tmp_src = 0; intptr_t tmp_src = 0;
asm volatile( asm volatile (
"pxor %%xmm4,%%xmm4 \n" "pxor %%xmm4,%%xmm4 \n"
"sub $0x1,%5 \n" "sub $0x1,%5 \n"
"1: \n" "1: \n"
@ -2263,7 +2263,7 @@ extern "C" void ScaleFilterRows_SSSE3(uint8* dst_ptr,
#elif defined(__x86_64__) #elif defined(__x86_64__)
static void ScaleRowDown8Int_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown8Int_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"lea (%3,%3,2),%%r10 \n" "lea (%3,%3,2),%%r10 \n"
"pxor %%xmm7,%%xmm7 \n" "pxor %%xmm7,%%xmm7 \n"
"1:" "1:"
@ -2322,7 +2322,7 @@ static void ScaleRowDown8Int_SSE2(const uint8* src_ptr, int src_stride,
#define HAS_SCALEROWDOWN34_SSSE3 #define HAS_SCALEROWDOWN34_SSSE3
static void ScaleRowDown34_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"movdqa (%3),%%xmm3 \n" "movdqa (%3),%%xmm3 \n"
"movdqa (%4),%%xmm4 \n" "movdqa (%4),%%xmm4 \n"
"movdqa (%5),%%xmm5 \n" "movdqa (%5),%%xmm5 \n"
@ -2353,7 +2353,7 @@ static void ScaleRowDown34_SSSE3(const uint8* src_ptr, int src_stride,
static void ScaleRowDown34_1_Int_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_1_Int_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"movdqa (%4),%%xmm2 \n" // _shuf01 "movdqa (%4),%%xmm2 \n" // _shuf01
"movdqa (%5),%%xmm3 \n" // _shuf11 "movdqa (%5),%%xmm3 \n" // _shuf11
"movdqa (%6),%%xmm4 \n" // _shuf21 "movdqa (%6),%%xmm4 \n" // _shuf21
@ -2410,7 +2410,7 @@ static void ScaleRowDown34_1_Int_SSSE3(const uint8* src_ptr, int src_stride,
static void ScaleRowDown34_0_Int_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_0_Int_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"movdqa (%4),%%xmm2 \n" // _shuf01 "movdqa (%4),%%xmm2 \n" // _shuf01
"movdqa (%5),%%xmm3 \n" // _shuf11 "movdqa (%5),%%xmm3 \n" // _shuf11
"movdqa (%6),%%xmm4 \n" // _shuf21 "movdqa (%6),%%xmm4 \n" // _shuf21
@ -2471,7 +2471,7 @@ static void ScaleRowDown34_0_Int_SSSE3(const uint8* src_ptr, int src_stride,
#define HAS_SCALEROWDOWN38_SSSE3 #define HAS_SCALEROWDOWN38_SSSE3
static void ScaleRowDown38_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"movdqa (%3),%%xmm4 \n" "movdqa (%3),%%xmm4 \n"
"movdqa (%4),%%xmm5 \n" "movdqa (%4),%%xmm5 \n"
"1:" "1:"
@ -2498,7 +2498,7 @@ static void ScaleRowDown38_SSSE3(const uint8* src_ptr, int src_stride,
static void ScaleRowDown38_3_Int_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_3_Int_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"movdqa (%4),%%xmm4 \n" "movdqa (%4),%%xmm4 \n"
"movdqa (%5),%%xmm5 \n" "movdqa (%5),%%xmm5 \n"
"movdqa (%6),%%xmm6 \n" "movdqa (%6),%%xmm6 \n"
@ -2555,7 +2555,7 @@ static void ScaleRowDown38_3_Int_SSSE3(const uint8* src_ptr, int src_stride,
static void ScaleRowDown38_2_Int_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_2_Int_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile( asm volatile (
"movdqa (%4),%%xmm4 \n" "movdqa (%4),%%xmm4 \n"
"movdqa (%5),%%xmm5 \n" "movdqa (%5),%%xmm5 \n"
"movdqa (%6),%%xmm6 \n" "movdqa (%6),%%xmm6 \n"
@ -2597,7 +2597,7 @@ static void ScaleFilterRows_SSE2(uint8* dst_ptr,
const uint8* src_ptr, int src_stride, const uint8* src_ptr, int src_stride,
int dst_width, int source_y_fraction) { int dst_width, int source_y_fraction) {
if (source_y_fraction == 0) { if (source_y_fraction == 0) {
asm volatile( asm volatile (
"1:" "1:"
"movdqa (%1),%%xmm0 \n" "movdqa (%1),%%xmm0 \n"
"lea 0x10(%1),%1 \n" "lea 0x10(%1),%1 \n"
@ -2615,7 +2615,7 @@ static void ScaleFilterRows_SSE2(uint8* dst_ptr,
); );
return; return;
} else if (source_y_fraction == 128) { } else if (source_y_fraction == 128) {
asm volatile( asm volatile (
"1:" "1:"
"movdqa (%1),%%xmm0 \n" "movdqa (%1),%%xmm0 \n"
"movdqa (%1,%3,1),%%xmm2 \n" "movdqa (%1,%3,1),%%xmm2 \n"
@ -2635,7 +2635,7 @@ static void ScaleFilterRows_SSE2(uint8* dst_ptr,
); );
return; return;
} else { } else {
asm volatile( asm volatile (
"mov %3,%%eax \n" "mov %3,%%eax \n"
"movd %%eax,%%xmm6 \n" "movd %%eax,%%xmm6 \n"
"punpcklwd %%xmm6,%%xmm6 \n" "punpcklwd %%xmm6,%%xmm6 \n"
@ -2688,7 +2688,7 @@ static void ScaleFilterRows_SSSE3(uint8* dst_ptr,
const uint8* src_ptr, int src_stride, const uint8* src_ptr, int src_stride,
int dst_width, int source_y_fraction) { int dst_width, int source_y_fraction) {
if (source_y_fraction <= 1) { if (source_y_fraction <= 1) {
asm volatile( asm volatile (
"1:" "1:"
"movdqa (%1),%%xmm0 \n" "movdqa (%1),%%xmm0 \n"
"lea 0x10(%1),%1 \n" "lea 0x10(%1),%1 \n"
@ -2706,7 +2706,7 @@ static void ScaleFilterRows_SSSE3(uint8* dst_ptr,
); );
return; return;
} else if (source_y_fraction == 128) { } else if (source_y_fraction == 128) {
asm volatile( asm volatile (
"1:" "1:"
"movdqa (%1),%%xmm0 \n" "movdqa (%1),%%xmm0 \n"
"movdqa (%1,%3,1),%%xmm2 \n" "movdqa (%1,%3,1),%%xmm2 \n"
@ -2726,7 +2726,7 @@ static void ScaleFilterRows_SSSE3(uint8* dst_ptr,
); );
return; return;
} else { } else {
asm volatile( asm volatile (
"mov %3,%%eax \n" "mov %3,%%eax \n"
"shr %%eax \n" "shr %%eax \n"
"mov %%al,%%ah \n" "mov %%al,%%ah \n"