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

yasm ALIGN uppercase

Browse Source 
BUG=none
TEST=untested
R=tpsiaki@google.com

Review URL: https://webrtc-codereview.appspot.com/4769005

git-svn-id: http://libyuv.googlecode.com/svn/trunk@885 16f28f9a-4ce2-e073-06de-1de4eb20be90
This commit is contained in:
fbarchard@google.com 2013-12-03 00:51:14 +00:00
parent 545a51c1d3
commit 04f40278df
10 changed files with 1678 additions and 1678 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
URL: http://code.google.com/p/libyuv/
Version: 884
Version: 885
License: BSD
License File: LICENSE

336
include/libyuv/convert_from_argb.h
View File

@ -1,168 +1,168 @@
/*
* Copyright 2012 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef INCLUDE_LIBYUV_CONVERT_FROM_ARGB_H_ // NOLINT
#define INCLUDE_LIBYUV_CONVERT_FROM_ARGB_H_
#include "libyuv/basic_types.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// Copy ARGB to ARGB.
#define ARGBToARGB ARGBCopy
LIBYUV_API
int ARGBCopy(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Convert ARGB To BGRA. (alias)
#define ARGBToBGRA BGRAToARGB
LIBYUV_API
int BGRAToARGB(const uint8* src_frame, int src_stride_frame,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Convert ARGB To ABGR. (alias)
#define ARGBToABGR ABGRToARGB
LIBYUV_API
int ABGRToARGB(const uint8* src_frame, int src_stride_frame,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Convert ARGB To RGBA.
LIBYUV_API
int ARGBToRGBA(const uint8* src_frame, int src_stride_frame,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Convert ARGB To RGB24.
LIBYUV_API
int ARGBToRGB24(const uint8* src_argb, int src_stride_argb,
uint8* dst_rgb24, int dst_stride_rgb24,
int width, int height);
// Convert ARGB To RAW.
LIBYUV_API
int ARGBToRAW(const uint8* src_argb, int src_stride_argb,
uint8* dst_rgb, int dst_stride_rgb,
int width, int height);
// Convert ARGB To RGB565.
LIBYUV_API
int ARGBToRGB565(const uint8* src_argb, int src_stride_argb,
uint8* dst_rgb565, int dst_stride_rgb565,
int width, int height);
// Convert ARGB To ARGB1555.
LIBYUV_API
int ARGBToARGB1555(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb1555, int dst_stride_argb1555,
int width, int height);
// Convert ARGB To ARGB4444.
LIBYUV_API
int ARGBToARGB4444(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb4444, int dst_stride_argb4444,
int width, int height);
// Convert ARGB To I444.
LIBYUV_API
int ARGBToI444(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int width, int height);
// Convert ARGB To I422.
LIBYUV_API
int ARGBToI422(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int width, int height);
// Convert ARGB To I420. (also in convert.h)
LIBYUV_API
int ARGBToI420(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int width, int height);
// Convert ARGB to J420. (JPeg full range I420).
LIBYUV_API
int ARGBToJ420(const uint8* src_argb, int src_stride_argb,
uint8* dst_yj, int dst_stride_yj,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int width, int height);
// Convert ARGB To I411.
LIBYUV_API
int ARGBToI411(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int width, int height);
// Convert ARGB to J400. (JPeg full range).
LIBYUV_API
int ARGBToJ400(const uint8* src_argb, int src_stride_argb,
uint8* dst_yj, int dst_stride_yj,
int width, int height);
// Convert ARGB to I400.
LIBYUV_API
int ARGBToI400(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
int width, int height);
// Convert ARGB To NV12.
LIBYUV_API
int ARGBToNV12(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_uv, int dst_stride_uv,
int width, int height);
// Convert ARGB To NV21.
LIBYUV_API
int ARGBToNV21(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_vu, int dst_stride_vu,
int width, int height);
// Convert ARGB To NV21.
LIBYUV_API
int ARGBToNV21(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_vu, int dst_stride_vu,
int width, int height);
// Convert ARGB To YUY2.
LIBYUV_API
int ARGBToYUY2(const uint8* src_argb, int src_stride_argb,
uint8* dst_yuy2, int dst_stride_yuy2,
int width, int height);
// Convert ARGB To UYVY.
LIBYUV_API
int ARGBToUYVY(const uint8* src_argb, int src_stride_argb,
uint8* dst_uyvy, int dst_stride_uyvy,
int width, int height);
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
#endif // INCLUDE_LIBYUV_CONVERT_FROM_ARGB_H_ NOLINT
/*
* Copyright 2012 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef INCLUDE_LIBYUV_CONVERT_FROM_ARGB_H_ // NOLINT
#define INCLUDE_LIBYUV_CONVERT_FROM_ARGB_H_
#include "libyuv/basic_types.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// Copy ARGB to ARGB.
#define ARGBToARGB ARGBCopy
LIBYUV_API
int ARGBCopy(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Convert ARGB To BGRA. (alias)
#define ARGBToBGRA BGRAToARGB
LIBYUV_API
int BGRAToARGB(const uint8* src_frame, int src_stride_frame,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Convert ARGB To ABGR. (alias)
#define ARGBToABGR ABGRToARGB
LIBYUV_API
int ABGRToARGB(const uint8* src_frame, int src_stride_frame,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Convert ARGB To RGBA.
LIBYUV_API
int ARGBToRGBA(const uint8* src_frame, int src_stride_frame,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Convert ARGB To RGB24.
LIBYUV_API
int ARGBToRGB24(const uint8* src_argb, int src_stride_argb,
uint8* dst_rgb24, int dst_stride_rgb24,
int width, int height);
// Convert ARGB To RAW.
LIBYUV_API
int ARGBToRAW(const uint8* src_argb, int src_stride_argb,
uint8* dst_rgb, int dst_stride_rgb,
int width, int height);
// Convert ARGB To RGB565.
LIBYUV_API
int ARGBToRGB565(const uint8* src_argb, int src_stride_argb,
uint8* dst_rgb565, int dst_stride_rgb565,
int width, int height);
// Convert ARGB To ARGB1555.
LIBYUV_API
int ARGBToARGB1555(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb1555, int dst_stride_argb1555,
int width, int height);
// Convert ARGB To ARGB4444.
LIBYUV_API
int ARGBToARGB4444(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb4444, int dst_stride_argb4444,
int width, int height);
// Convert ARGB To I444.
LIBYUV_API
int ARGBToI444(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int width, int height);
// Convert ARGB To I422.
LIBYUV_API
int ARGBToI422(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int width, int height);
// Convert ARGB To I420. (also in convert.h)
LIBYUV_API
int ARGBToI420(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int width, int height);
// Convert ARGB to J420. (JPeg full range I420).
LIBYUV_API
int ARGBToJ420(const uint8* src_argb, int src_stride_argb,
uint8* dst_yj, int dst_stride_yj,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int width, int height);
// Convert ARGB To I411.
LIBYUV_API
int ARGBToI411(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int width, int height);
// Convert ARGB to J400. (JPeg full range).
LIBYUV_API
int ARGBToJ400(const uint8* src_argb, int src_stride_argb,
uint8* dst_yj, int dst_stride_yj,
int width, int height);
// Convert ARGB to I400.
LIBYUV_API
int ARGBToI400(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
int width, int height);
// Convert ARGB To NV12.
LIBYUV_API
int ARGBToNV12(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_uv, int dst_stride_uv,
int width, int height);
// Convert ARGB To NV21.
LIBYUV_API
int ARGBToNV21(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_vu, int dst_stride_vu,
int width, int height);
// Convert ARGB To NV21.
LIBYUV_API
int ARGBToNV21(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
uint8* dst_vu, int dst_stride_vu,
int width, int height);
// Convert ARGB To YUY2.
LIBYUV_API
int ARGBToYUY2(const uint8* src_argb, int src_stride_argb,
uint8* dst_yuy2, int dst_stride_yuy2,
int width, int height);
// Convert ARGB To UYVY.
LIBYUV_API
int ARGBToUYVY(const uint8* src_argb, int src_stride_argb,
uint8* dst_uyvy, int dst_stride_uyvy,
int width, int height);
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
#endif // INCLUDE_LIBYUV_CONVERT_FROM_ARGB_H_ NOLINT

2
include/libyuv/version.h
View File

@ -11,6 +11,6 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT
#define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 884
#define LIBYUV_VERSION 885
#endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT

80
source/compare_common.cc
View File

@ -1,40 +1,40 @@
/*
* Copyright 2012 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "libyuv/basic_types.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
uint32 SumSquareError_C(const uint8* src_a, const uint8* src_b, int count) {
uint32 sse = 0u;
for (int i = 0; i < count; ++i) {
int diff = src_a[i] - src_b[i];
sse += static_cast<uint32>(diff * diff);
}
return sse;
}
// hash seed of 5381 recommended.
// Internal C version of HashDjb2 with int sized count for efficiency.
uint32 HashDjb2_C(const uint8* src, int count, uint32 seed) {
uint32 hash = seed;
for (int i = 0; i < count; ++i) {
hash += (hash << 5) + src[i];
}
return hash;
}
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
/*
* Copyright 2012 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "libyuv/basic_types.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
uint32 SumSquareError_C(const uint8* src_a, const uint8* src_b, int count) {
uint32 sse = 0u;
for (int i = 0; i < count; ++i) {
int diff = src_a[i] - src_b[i];
sse += static_cast<uint32>(diff * diff);
}
return sse;
}
// hash seed of 5381 recommended.
// Internal C version of HashDjb2 with int sized count for efficiency.
uint32 HashDjb2_C(const uint8* src, int count, uint32 seed) {
uint32 hash = seed;
for (int i = 0; i < count; ++i) {
hash += (hash << 5) + src[i];
}
return hash;
}
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif

122
source/compare_neon.cc
View File

@ -1,61 +1,61 @@
/*
* Copyright 2012 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "libyuv/basic_types.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
#if !defined(LIBYUV_DISABLE_NEON) && defined(__ARM_NEON__)
uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b, int count) {
volatile uint32 sse;
asm volatile (
"vmov.u8 q8, #0 \n"
"vmov.u8 q10, #0 \n"
"vmov.u8 q9, #0 \n"
"vmov.u8 q11, #0 \n"
".p2align 2 \n"
"1: \n"
"vld1.8 {q0}, [%0]! \n"
"vld1.8 {q1}, [%1]! \n"
"subs %2, %2, #16 \n"
"vsubl.u8 q2, d0, d2 \n"
"vsubl.u8 q3, d1, d3 \n"
"vmlal.s16 q8, d4, d4 \n"
"vmlal.s16 q9, d6, d6 \n"
"vmlal.s16 q10, d5, d5 \n"
"vmlal.s16 q11, d7, d7 \n"
"bgt 1b \n"
"vadd.u32 q8, q8, q9 \n"
"vadd.u32 q10, q10, q11 \n"
"vadd.u32 q11, q8, q10 \n"
"vpaddl.u32 q1, q11 \n"
"vadd.u64 d0, d2, d3 \n"
"vmov.32 %3, d0[0] \n"
: "+r"(src_a),
"+r"(src_b),
"+r"(count),
"=r"(sse)
:
: "memory", "cc", "q0", "q1", "q2", "q3", "q8", "q9", "q10", "q11");
return sse;
}
#endif // __ARM_NEON__
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
/*
* Copyright 2012 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "libyuv/basic_types.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
#if !defined(LIBYUV_DISABLE_NEON) && defined(__ARM_NEON__)
uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b, int count) {
volatile uint32 sse;
asm volatile (
"vmov.u8 q8, #0 \n"
"vmov.u8 q10, #0 \n"
"vmov.u8 q9, #0 \n"
"vmov.u8 q11, #0 \n"
".p2align 2 \n"
"1: \n"
"vld1.8 {q0}, [%0]! \n"
"vld1.8 {q1}, [%1]! \n"
"subs %2, %2, #16 \n"
"vsubl.u8 q2, d0, d2 \n"
"vsubl.u8 q3, d1, d3 \n"
"vmlal.s16 q8, d4, d4 \n"
"vmlal.s16 q9, d6, d6 \n"
"vmlal.s16 q10, d5, d5 \n"
"vmlal.s16 q11, d7, d7 \n"
"bgt 1b \n"
"vadd.u32 q8, q8, q9 \n"
"vadd.u32 q10, q10, q11 \n"
"vadd.u32 q11, q8, q10 \n"
"vpaddl.u32 q1, q11 \n"
"vadd.u64 d0, d2, d3 \n"
"vmov.32 %3, d0[0] \n"
: "+r"(src_a),
"+r"(src_b),
"+r"(count),
"=r"(sse)
:
: "memory", "cc", "q0", "q1", "q2", "q3", "q8", "q9", "q10", "q11");
return sse;
}
#endif // __ARM_NEON__
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif

332
source/compare_posix.cc
View File

@ -1,166 +1,166 @@
/*
* Copyright 2012 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "libyuv/basic_types.h"
#include "libyuv/row.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
#if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__))
#if defined(__native_client__) && defined(__x86_64__)
#define MEMACCESS(base) "%%nacl:(%%r15,%q" #base ")"
#define MEMLEA(offset, base) #offset "(%q" #base ")"
#else
#define MEMACCESS(base) "(%" #base ")"
#define MEMLEA(offset, base) #offset "(%" #base ")"
#endif
uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b, int count) {
uint32 sse;
asm volatile ( // NOLINT
"pxor %%xmm0,%%xmm0 \n"
"pxor %%xmm5,%%xmm5 \n"
".p2align 2 \n"
"1: \n"
"movdqa " MEMACCESS(0) ",%%xmm1 \n"
"lea " MEMLEA(0x10, 0) ",%0 \n"
"movdqa " MEMACCESS(1) ",%%xmm2 \n"
"lea " MEMLEA(0x10, 1) ",%1 \n"
"sub $0x10,%2 \n"
"movdqa %%xmm1,%%xmm3 \n"
"psubusb %%xmm2,%%xmm1 \n"
"psubusb %%xmm3,%%xmm2 \n"
"por %%xmm2,%%xmm1 \n"
"movdqa %%xmm1,%%xmm2 \n"
"punpcklbw %%xmm5,%%xmm1 \n"
"punpckhbw %%xmm5,%%xmm2 \n"
"pmaddwd %%xmm1,%%xmm1 \n"
"pmaddwd %%xmm2,%%xmm2 \n"
"paddd %%xmm1,%%xmm0 \n"
"paddd %%xmm2,%%xmm0 \n"
"jg 1b \n"
"pshufd $0xee,%%xmm0,%%xmm1 \n"
"paddd %%xmm1,%%xmm0 \n"
"pshufd $0x1,%%xmm0,%%xmm1 \n"
"paddd %%xmm1,%%xmm0 \n"
"movd %%xmm0,%3 \n"
: "+r"(src_a), // %0
"+r"(src_b), // %1
"+r"(count), // %2
"=g"(sse) // %3
:
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
#endif
); // NOLINT
return sse;
}
#endif // defined(__x86_64__) || defined(__i386__)
#if !defined(LIBYUV_DISABLE_X86) && \
(defined(__x86_64__) || (defined(__i386__) && !defined(__pic__)))
#define HAS_HASHDJB2_SSE41
static uvec32 kHash16x33 = { 0x92d9e201, 0, 0, 0 }; // 33 ^ 16
static uvec32 kHashMul0 = {
0x0c3525e1, // 33 ^ 15
0xa3476dc1, // 33 ^ 14
0x3b4039a1, // 33 ^ 13
0x4f5f0981, // 33 ^ 12
};
static uvec32 kHashMul1 = {
0x30f35d61, // 33 ^ 11
0x855cb541, // 33 ^ 10
0x040a9121, // 33 ^ 9
0x747c7101, // 33 ^ 8
};
static uvec32 kHashMul2 = {
0xec41d4e1, // 33 ^ 7
0x4cfa3cc1, // 33 ^ 6
0x025528a1, // 33 ^ 5
0x00121881, // 33 ^ 4
};
static uvec32 kHashMul3 = {
0x00008c61, // 33 ^ 3
0x00000441, // 33 ^ 2
0x00000021, // 33 ^ 1
0x00000001, // 33 ^ 0
};
uint32 HashDjb2_SSE41(const uint8* src, int count, uint32 seed) {
uint32 hash;
asm volatile ( // NOLINT
"movd %2,%%xmm0 \n"
"pxor %%xmm7,%%xmm7 \n"
"movdqa %4,%%xmm6 \n"
".p2align 2 \n"
"1: \n"
"movdqu " MEMACCESS(0) ",%%xmm1 \n"
"lea " MEMLEA(0x10, 0) ",%0 \n"
"pmulld %%xmm6,%%xmm0 \n"
"movdqa %5,%%xmm5 \n"
"movdqa %%xmm1,%%xmm2 \n"
"punpcklbw %%xmm7,%%xmm2 \n"
"movdqa %%xmm2,%%xmm3 \n"
"punpcklwd %%xmm7,%%xmm3 \n"
"pmulld %%xmm5,%%xmm3 \n"
"movdqa %6,%%xmm5 \n"
"movdqa %%xmm2,%%xmm4 \n"
"punpckhwd %%xmm7,%%xmm4 \n"
"pmulld %%xmm5,%%xmm4 \n"
"movdqa %7,%%xmm5 \n"
"punpckhbw %%xmm7,%%xmm1 \n"
"movdqa %%xmm1,%%xmm2 \n"
"punpcklwd %%xmm7,%%xmm2 \n"
"pmulld %%xmm5,%%xmm2 \n"
"movdqa %8,%%xmm5 \n"
"punpckhwd %%xmm7,%%xmm1 \n"
"pmulld %%xmm5,%%xmm1 \n"
"paddd %%xmm4,%%xmm3 \n"
"paddd %%xmm2,%%xmm1 \n"
"sub $0x10,%1 \n"
"paddd %%xmm3,%%xmm1 \n"
"pshufd $0xe,%%xmm1,%%xmm2 \n"
"paddd %%xmm2,%%xmm1 \n"
"pshufd $0x1,%%xmm1,%%xmm2 \n"
"paddd %%xmm2,%%xmm1 \n"
"paddd %%xmm1,%%xmm0 \n"
"jg 1b \n"
"movd %%xmm0,%3 \n"
: "+r"(src), // %0
"+r"(count), // %1
"+rm"(seed), // %2
"=g"(hash) // %3
: "m"(kHash16x33), // %4
"m"(kHashMul0), // %5
"m"(kHashMul1), // %6
"m"(kHashMul2), // %7
"m"(kHashMul3) // %8
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7"
#endif
); // NOLINT
return hash;
}
#endif // defined(__x86_64__) || (defined(__i386__) && !defined(__pic__)))
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
/*
* Copyright 2012 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "libyuv/basic_types.h"
#include "libyuv/row.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
#if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__))
#if defined(__native_client__) && defined(__x86_64__)
#define MEMACCESS(base) "%%nacl:(%%r15,%q" #base ")"
#define MEMLEA(offset, base) #offset "(%q" #base ")"
#else
#define MEMACCESS(base) "(%" #base ")"
#define MEMLEA(offset, base) #offset "(%" #base ")"
#endif
uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b, int count) {
uint32 sse;
asm volatile ( // NOLINT
"pxor %%xmm0,%%xmm0 \n"
"pxor %%xmm5,%%xmm5 \n"
".p2align 2 \n"
"1: \n"
"movdqa " MEMACCESS(0) ",%%xmm1 \n"
"lea " MEMLEA(0x10, 0) ",%0 \n"
"movdqa " MEMACCESS(1) ",%%xmm2 \n"
"lea " MEMLEA(0x10, 1) ",%1 \n"
"sub $0x10,%2 \n"
"movdqa %%xmm1,%%xmm3 \n"
"psubusb %%xmm2,%%xmm1 \n"
"psubusb %%xmm3,%%xmm2 \n"
"por %%xmm2,%%xmm1 \n"
"movdqa %%xmm1,%%xmm2 \n"
"punpcklbw %%xmm5,%%xmm1 \n"
"punpckhbw %%xmm5,%%xmm2 \n"
"pmaddwd %%xmm1,%%xmm1 \n"
"pmaddwd %%xmm2,%%xmm2 \n"
"paddd %%xmm1,%%xmm0 \n"
"paddd %%xmm2,%%xmm0 \n"
"jg 1b \n"
"pshufd $0xee,%%xmm0,%%xmm1 \n"
"paddd %%xmm1,%%xmm0 \n"
"pshufd $0x1,%%xmm0,%%xmm1 \n"
"paddd %%xmm1,%%xmm0 \n"
"movd %%xmm0,%3 \n"
: "+r"(src_a), // %0
"+r"(src_b), // %1
"+r"(count), // %2
"=g"(sse) // %3
:
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
#endif
); // NOLINT
return sse;
}
#endif // defined(__x86_64__) || defined(__i386__)
#if !defined(LIBYUV_DISABLE_X86) && \
(defined(__x86_64__) || (defined(__i386__) && !defined(__pic__)))
#define HAS_HASHDJB2_SSE41
static uvec32 kHash16x33 = { 0x92d9e201, 0, 0, 0 }; // 33 ^ 16
static uvec32 kHashMul0 = {
0x0c3525e1, // 33 ^ 15
0xa3476dc1, // 33 ^ 14
0x3b4039a1, // 33 ^ 13
0x4f5f0981, // 33 ^ 12
};
static uvec32 kHashMul1 = {
0x30f35d61, // 33 ^ 11
0x855cb541, // 33 ^ 10
0x040a9121, // 33 ^ 9
0x747c7101, // 33 ^ 8
};
static uvec32 kHashMul2 = {
0xec41d4e1, // 33 ^ 7
0x4cfa3cc1, // 33 ^ 6
0x025528a1, // 33 ^ 5
0x00121881, // 33 ^ 4
};
static uvec32 kHashMul3 = {
0x00008c61, // 33 ^ 3
0x00000441, // 33 ^ 2
0x00000021, // 33 ^ 1
0x00000001, // 33 ^ 0
};
uint32 HashDjb2_SSE41(const uint8* src, int count, uint32 seed) {
uint32 hash;
asm volatile ( // NOLINT
"movd %2,%%xmm0 \n"
"pxor %%xmm7,%%xmm7 \n"
"movdqa %4,%%xmm6 \n"
".p2align 2 \n"
"1: \n"
"movdqu " MEMACCESS(0) ",%%xmm1 \n"
"lea " MEMLEA(0x10, 0) ",%0 \n"
"pmulld %%xmm6,%%xmm0 \n"
"movdqa %5,%%xmm5 \n"
"movdqa %%xmm1,%%xmm2 \n"
"punpcklbw %%xmm7,%%xmm2 \n"
"movdqa %%xmm2,%%xmm3 \n"
"punpcklwd %%xmm7,%%xmm3 \n"
"pmulld %%xmm5,%%xmm3 \n"
"movdqa %6,%%xmm5 \n"
"movdqa %%xmm2,%%xmm4 \n"
"punpckhwd %%xmm7,%%xmm4 \n"
"pmulld %%xmm5,%%xmm4 \n"
"movdqa %7,%%xmm5 \n"
"punpckhbw %%xmm7,%%xmm1 \n"
"movdqa %%xmm1,%%xmm2 \n"
"punpcklwd %%xmm7,%%xmm2 \n"
"pmulld %%xmm5,%%xmm2 \n"
"movdqa %8,%%xmm5 \n"
"punpckhwd %%xmm7,%%xmm1 \n"
"pmulld %%xmm5,%%xmm1 \n"
"paddd %%xmm4,%%xmm3 \n"
"paddd %%xmm2,%%xmm1 \n"
"sub $0x10,%1 \n"
"paddd %%xmm3,%%xmm1 \n"
"pshufd $0xe,%%xmm1,%%xmm2 \n"
"paddd %%xmm2,%%xmm1 \n"
"pshufd $0x1,%%xmm1,%%xmm2 \n"
"paddd %%xmm2,%%xmm1 \n"
"paddd %%xmm1,%%xmm0 \n"
"jg 1b \n"
"movd %%xmm0,%3 \n"
: "+r"(src), // %0
"+r"(count), // %1
"+rm"(seed), // %2
"=g"(hash) // %3
: "m"(kHash16x33), // %4
"m"(kHashMul0), // %5
"m"(kHashMul1), // %6
"m"(kHashMul2), // %7
"m"(kHashMul3) // %8
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7"
#endif
); // NOLINT
return hash;
}
#endif // defined(__x86_64__) || (defined(__i386__) && !defined(__pic__)))
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif

464
source/compare_win.cc
View File

@ -1,232 +1,232 @@
/*
* Copyright 2012 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "libyuv/basic_types.h"
#include "libyuv/row.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
#if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && defined(_MSC_VER)
__declspec(naked) __declspec(align(16))
uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b, int count) {
__asm {
mov eax, [esp + 4] // src_a
mov edx, [esp + 8] // src_b
mov ecx, [esp + 12] // count
pxor xmm0, xmm0
pxor xmm5, xmm5
align 4
wloop:
movdqa xmm1, [eax]
lea eax, [eax + 16]
movdqa xmm2, [edx]
lea edx, [edx + 16]
sub ecx, 16
movdqa xmm3, xmm1 // abs trick
psubusb xmm1, xmm2
psubusb xmm2, xmm3
por xmm1, xmm2
movdqa xmm2, xmm1
punpcklbw xmm1, xmm5
punpckhbw xmm2, xmm5
pmaddwd xmm1, xmm1
pmaddwd xmm2, xmm2
paddd xmm0, xmm1
paddd xmm0, xmm2
jg wloop
pshufd xmm1, xmm0, 0xee
paddd xmm0, xmm1
pshufd xmm1, xmm0, 0x01
paddd xmm0, xmm1
movd eax, xmm0
ret
}
}
// Visual C 2012 required for AVX2.
#if _MSC_VER >= 1700
// C4752: found Intel(R) Advanced Vector Extensions; consider using /arch:AVX.
#pragma warning(disable: 4752)
__declspec(naked) __declspec(align(16))
uint32 SumSquareError_AVX2(const uint8* src_a, const uint8* src_b, int count) {
__asm {
mov eax, [esp + 4] // src_a
mov edx, [esp + 8] // src_b
mov ecx, [esp + 12] // count
vpxor ymm0, ymm0, ymm0 // sum
vpxor ymm5, ymm5, ymm5 // constant 0 for unpck
sub edx, eax
align 4
wloop:
vmovdqu ymm1, [eax]
vmovdqu ymm2, [eax + edx]
lea eax, [eax + 32]
sub ecx, 32
vpsubusb ymm3, ymm1, ymm2 // abs difference trick
vpsubusb ymm2, ymm2, ymm1
vpor ymm1, ymm2, ymm3
vpunpcklbw ymm2, ymm1, ymm5 // u16. mutates order.
vpunpckhbw ymm1, ymm1, ymm5
vpmaddwd ymm2, ymm2, ymm2 // square + hadd to u32.
vpmaddwd ymm1, ymm1, ymm1
vpaddd ymm0, ymm0, ymm1
vpaddd ymm0, ymm0, ymm2
jg wloop
vpshufd ymm1, ymm0, 0xee // 3, 2 + 1, 0 both lanes.
vpaddd ymm0, ymm0, ymm1
vpshufd ymm1, ymm0, 0x01 // 1 + 0 both lanes.
vpaddd ymm0, ymm0, ymm1
vpermq ymm1, ymm0, 0x02 // high + low lane.
vpaddd ymm0, ymm0, ymm1
vmovd eax, xmm0
vzeroupper
ret
}
}
#endif // _MSC_VER >= 1700
#define HAS_HASHDJB2_SSE41
static uvec32 kHash16x33 = { 0x92d9e201, 0, 0, 0 }; // 33 ^ 16
static uvec32 kHashMul0 = {
0x0c3525e1, // 33 ^ 15
0xa3476dc1, // 33 ^ 14
0x3b4039a1, // 33 ^ 13
0x4f5f0981, // 33 ^ 12
};
static uvec32 kHashMul1 = {
0x30f35d61, // 33 ^ 11
0x855cb541, // 33 ^ 10
0x040a9121, // 33 ^ 9
0x747c7101, // 33 ^ 8
};
static uvec32 kHashMul2 = {
0xec41d4e1, // 33 ^ 7
0x4cfa3cc1, // 33 ^ 6
0x025528a1, // 33 ^ 5
0x00121881, // 33 ^ 4
};
static uvec32 kHashMul3 = {
0x00008c61, // 33 ^ 3
0x00000441, // 33 ^ 2
0x00000021, // 33 ^ 1
0x00000001, // 33 ^ 0
};
// 27: 66 0F 38 40 C6 pmulld xmm0,xmm6
// 44: 66 0F 38 40 DD pmulld xmm3,xmm5
// 59: 66 0F 38 40 E5 pmulld xmm4,xmm5
// 72: 66 0F 38 40 D5 pmulld xmm2,xmm5
// 83: 66 0F 38 40 CD pmulld xmm1,xmm5
#define pmulld(reg) _asm _emit 0x66 _asm _emit 0x0F _asm _emit 0x38 \
_asm _emit 0x40 _asm _emit reg
__declspec(naked) __declspec(align(16))
uint32 HashDjb2_SSE41(const uint8* src, int count, uint32 seed) {
__asm {
mov eax, [esp + 4] // src
mov ecx, [esp + 8] // count
movd xmm0, [esp + 12] // seed
pxor xmm7, xmm7 // constant 0 for unpck
movdqa xmm6, kHash16x33
align 4
wloop:
movdqu xmm1, [eax] // src[0-15]
lea eax, [eax + 16]
pmulld(0xc6) // pmulld xmm0,xmm6 hash *= 33 ^ 16
movdqa xmm5, kHashMul0
movdqa xmm2, xmm1
punpcklbw xmm2, xmm7 // src[0-7]
movdqa xmm3, xmm2
punpcklwd xmm3, xmm7 // src[0-3]
pmulld(0xdd) // pmulld xmm3, xmm5
movdqa xmm5, kHashMul1
movdqa xmm4, xmm2
punpckhwd xmm4, xmm7 // src[4-7]
pmulld(0xe5) // pmulld xmm4, xmm5
movdqa xmm5, kHashMul2
punpckhbw xmm1, xmm7 // src[8-15]
movdqa xmm2, xmm1
punpcklwd xmm2, xmm7 // src[8-11]
pmulld(0xd5) // pmulld xmm2, xmm5
movdqa xmm5, kHashMul3
punpckhwd xmm1, xmm7 // src[12-15]
pmulld(0xcd) // pmulld xmm1, xmm5
paddd xmm3, xmm4 // add 16 results
paddd xmm1, xmm2
sub ecx, 16
paddd xmm1, xmm3
pshufd xmm2, xmm1, 0x0e // upper 2 dwords
paddd xmm1, xmm2
pshufd xmm2, xmm1, 0x01
paddd xmm1, xmm2
paddd xmm0, xmm1
jg wloop
movd eax, xmm0 // return hash
ret
}
}
// Visual C 2012 required for AVX2.
#if _MSC_VER >= 1700
__declspec(naked) __declspec(align(16))
uint32 HashDjb2_AVX2(const uint8* src, int count, uint32 seed) {
__asm {
mov eax, [esp + 4] // src
mov ecx, [esp + 8] // count
movd xmm0, [esp + 12] // seed
movdqa xmm6, kHash16x33
align 4
wloop:
vpmovzxbd xmm3, dword ptr [eax] // src[0-3]
pmulld xmm0, xmm6 // hash *= 33 ^ 16
vpmovzxbd xmm4, dword ptr [eax + 4] // src[4-7]
pmulld xmm3, kHashMul0
vpmovzxbd xmm2, dword ptr [eax + 8] // src[8-11]
pmulld xmm4, kHashMul1
vpmovzxbd xmm1, dword ptr [eax + 12] // src[12-15]
pmulld xmm2, kHashMul2
lea eax, [eax + 16]
pmulld xmm1, kHashMul3
paddd xmm3, xmm4 // add 16 results
paddd xmm1, xmm2
sub ecx, 16
paddd xmm1, xmm3
pshufd xmm2, xmm1, 0x0e // upper 2 dwords
paddd xmm1, xmm2
pshufd xmm2, xmm1, 0x01
paddd xmm1, xmm2
paddd xmm0, xmm1
jg wloop
movd eax, xmm0 // return hash
ret
}
}
#endif // _MSC_VER >= 1700
#endif // !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && defined(_MSC_VER)
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
/*
* Copyright 2012 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "libyuv/basic_types.h"
#include "libyuv/row.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
#if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && defined(_MSC_VER)
__declspec(naked) __declspec(align(16))
uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b, int count) {
__asm {
mov eax, [esp + 4] // src_a
mov edx, [esp + 8] // src_b
mov ecx, [esp + 12] // count
pxor xmm0, xmm0
pxor xmm5, xmm5
align 4
wloop:
movdqa xmm1, [eax]
lea eax, [eax + 16]
movdqa xmm2, [edx]
lea edx, [edx + 16]
sub ecx, 16
movdqa xmm3, xmm1 // abs trick
psubusb xmm1, xmm2
psubusb xmm2, xmm3
por xmm1, xmm2
movdqa xmm2, xmm1
punpcklbw xmm1, xmm5
punpckhbw xmm2, xmm5
pmaddwd xmm1, xmm1
pmaddwd xmm2, xmm2
paddd xmm0, xmm1
paddd xmm0, xmm2
jg wloop
pshufd xmm1, xmm0, 0xee
paddd xmm0, xmm1
pshufd xmm1, xmm0, 0x01
paddd xmm0, xmm1
movd eax, xmm0
ret
}
}
// Visual C 2012 required for AVX2.
#if _MSC_VER >= 1700
// C4752: found Intel(R) Advanced Vector Extensions; consider using /arch:AVX.
#pragma warning(disable: 4752)
__declspec(naked) __declspec(align(16))
uint32 SumSquareError_AVX2(const uint8* src_a, const uint8* src_b, int count) {
__asm {
mov eax, [esp + 4] // src_a
mov edx, [esp + 8] // src_b
mov ecx, [esp + 12] // count
vpxor ymm0, ymm0, ymm0 // sum
vpxor ymm5, ymm5, ymm5 // constant 0 for unpck
sub edx, eax
align 4
wloop:
vmovdqu ymm1, [eax]
vmovdqu ymm2, [eax + edx]
lea eax, [eax + 32]
sub ecx, 32
vpsubusb ymm3, ymm1, ymm2 // abs difference trick
vpsubusb ymm2, ymm2, ymm1
vpor ymm1, ymm2, ymm3
vpunpcklbw ymm2, ymm1, ymm5 // u16. mutates order.
vpunpckhbw ymm1, ymm1, ymm5
vpmaddwd ymm2, ymm2, ymm2 // square + hadd to u32.
vpmaddwd ymm1, ymm1, ymm1
vpaddd ymm0, ymm0, ymm1
vpaddd ymm0, ymm0, ymm2
jg wloop
vpshufd ymm1, ymm0, 0xee // 3, 2 + 1, 0 both lanes.
vpaddd ymm0, ymm0, ymm1
vpshufd ymm1, ymm0, 0x01 // 1 + 0 both lanes.
vpaddd ymm0, ymm0, ymm1
vpermq ymm1, ymm0, 0x02 // high + low lane.
vpaddd ymm0, ymm0, ymm1
vmovd eax, xmm0
vzeroupper
ret
}
}
#endif // _MSC_VER >= 1700
#define HAS_HASHDJB2_SSE41
static uvec32 kHash16x33 = { 0x92d9e201, 0, 0, 0 }; // 33 ^ 16
static uvec32 kHashMul0 = {
0x0c3525e1, // 33 ^ 15
0xa3476dc1, // 33 ^ 14
0x3b4039a1, // 33 ^ 13
0x4f5f0981, // 33 ^ 12
};
static uvec32 kHashMul1 = {
0x30f35d61, // 33 ^ 11
0x855cb541, // 33 ^ 10
0x040a9121, // 33 ^ 9
0x747c7101, // 33 ^ 8
};
static uvec32 kHashMul2 = {
0xec41d4e1, // 33 ^ 7
0x4cfa3cc1, // 33 ^ 6
0x025528a1, // 33 ^ 5
0x00121881, // 33 ^ 4
};
static uvec32 kHashMul3 = {
0x00008c61, // 33 ^ 3
0x00000441, // 33 ^ 2
0x00000021, // 33 ^ 1
0x00000001, // 33 ^ 0
};
// 27: 66 0F 38 40 C6 pmulld xmm0,xmm6
// 44: 66 0F 38 40 DD pmulld xmm3,xmm5
// 59: 66 0F 38 40 E5 pmulld xmm4,xmm5
// 72: 66 0F 38 40 D5 pmulld xmm2,xmm5
// 83: 66 0F 38 40 CD pmulld xmm1,xmm5
#define pmulld(reg) _asm _emit 0x66 _asm _emit 0x0F _asm _emit 0x38 \
_asm _emit 0x40 _asm _emit reg
__declspec(naked) __declspec(align(16))
uint32 HashDjb2_SSE41(const uint8* src, int count, uint32 seed) {
__asm {
mov eax, [esp + 4] // src
mov ecx, [esp + 8] // count
movd xmm0, [esp + 12] // seed
pxor xmm7, xmm7 // constant 0 for unpck
movdqa xmm6, kHash16x33
align 4
wloop:
movdqu xmm1, [eax] // src[0-15]
lea eax, [eax + 16]
pmulld(0xc6) // pmulld xmm0,xmm6 hash *= 33 ^ 16
movdqa xmm5, kHashMul0
movdqa xmm2, xmm1
punpcklbw xmm2, xmm7 // src[0-7]
movdqa xmm3, xmm2
punpcklwd xmm3, xmm7 // src[0-3]
pmulld(0xdd) // pmulld xmm3, xmm5
movdqa xmm5, kHashMul1
movdqa xmm4, xmm2
punpckhwd xmm4, xmm7 // src[4-7]
pmulld(0xe5) // pmulld xmm4, xmm5
movdqa xmm5, kHashMul2
punpckhbw xmm1, xmm7 // src[8-15]
movdqa xmm2, xmm1
punpcklwd xmm2, xmm7 // src[8-11]
pmulld(0xd5) // pmulld xmm2, xmm5
movdqa xmm5, kHashMul3
punpckhwd xmm1, xmm7 // src[12-15]
pmulld(0xcd) // pmulld xmm1, xmm5
paddd xmm3, xmm4 // add 16 results
paddd xmm1, xmm2
sub ecx, 16
paddd xmm1, xmm3
pshufd xmm2, xmm1, 0x0e // upper 2 dwords
paddd xmm1, xmm2
pshufd xmm2, xmm1, 0x01
paddd xmm1, xmm2
paddd xmm0, xmm1
jg wloop
movd eax, xmm0 // return hash
ret
}
}
// Visual C 2012 required for AVX2.
#if _MSC_VER >= 1700
__declspec(naked) __declspec(align(16))
uint32 HashDjb2_AVX2(const uint8* src, int count, uint32 seed) {
__asm {
mov eax, [esp + 4] // src
mov ecx, [esp + 8] // count
movd xmm0, [esp + 12] // seed
movdqa xmm6, kHash16x33
align 4
wloop:
vpmovzxbd xmm3, dword ptr [eax] // src[0-3]
pmulld xmm0, xmm6 // hash *= 33 ^ 16
vpmovzxbd xmm4, dword ptr [eax + 4] // src[4-7]
pmulld xmm3, kHashMul0
vpmovzxbd xmm2, dword ptr [eax + 8] // src[8-11]
pmulld xmm4, kHashMul1
vpmovzxbd xmm1, dword ptr [eax + 12] // src[12-15]
pmulld xmm2, kHashMul2
lea eax, [eax + 16]
pmulld xmm1, kHashMul3
paddd xmm3, xmm4 // add 16 results
paddd xmm1, xmm2
sub ecx, 16
paddd xmm1, xmm3
pshufd xmm2, xmm1, 0x0e // upper 2 dwords
paddd xmm1, xmm2
pshufd xmm2, xmm1, 0x01
paddd xmm1, xmm2
paddd xmm0, xmm1
jg wloop
movd eax, xmm0 // return hash
ret
}
}
#endif // _MSC_VER >= 1700
#endif // !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && defined(_MSC_VER)
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif

22
source/convert_from_argb.cc
View File

@ -49,17 +49,17 @@ int ARGBToI444(const uint8* src_argb, int src_stride_argb,
ARGBToYRow_C;
void (*ARGBToUV444Row)(const uint8* src_argb, uint8* dst_u, uint8* dst_v,
int pix) = ARGBToUV444Row_C;
#if defined(HAS_ARGBTOUV444ROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
ARGBToUV444Row = ARGBToUV444Row_Any_SSSE3;
if (IS_ALIGNED(width, 16)) {
ARGBToUV444Row = ARGBToUV444Row_Unaligned_SSSE3;
if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
ARGBToUV444Row = ARGBToUV444Row_SSSE3;
}
}
}
#endif
#if defined(HAS_ARGBTOUV444ROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
ARGBToUV444Row = ARGBToUV444Row_Any_SSSE3;
if (IS_ALIGNED(width, 16)) {
ARGBToUV444Row = ARGBToUV444Row_Unaligned_SSSE3;
if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
ARGBToUV444Row = ARGBToUV444Row_SSSE3;
}
}
}
#endif
#if defined(HAS_ARGBTOYROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
ARGBToYRow = ARGBToYRow_Any_SSSE3;

6
source/row_x86.asm
View File

@ -28,7 +28,7 @@ cglobal %1ToYRow%3, 3, 3, 3, src_yuy2, dst_y, pix
psrlw m2, m2, 8
%endif
align 4
ALIGN 4
.convertloop:
mov%2 m0, [src_yuy2q]
mov%2 m1, [src_yuy2q + mmsize]
@ -74,7 +74,7 @@ cglobal SplitUVRow%2, 4, 4, 5, src_uv, dst_u, dst_v, pix
psrlw m4, m4, 8
sub dst_vq, dst_uq
align 4
ALIGN 4
.convertloop:
mov%1 m0, [src_uvq]
mov%1 m1, [src_uvq + mmsize]
@ -113,7 +113,7 @@ SplitUVRow a,
cglobal MergeUVRow_%2, 4, 4, 3, src_u, src_v, dst_uv, pix
sub src_vq, src_uq
align 4
ALIGN 4
.convertloop:
mov%1 m0, [src_uq]
mov%1 m1, [src_vq]

1990
unit_test/convert_test.cc
View File

File diff suppressed because it is too large Load Diff