generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Updates libyuv to version 1005

Browse Source 
Also adds compile check and a libyuv configure flag

Change-Id: Ib9f0f4a71c4083e6f0aea7b5a5d175531ef0f66b
This commit is contained in:
Deb Mukherjee 2014-05-16 18:52:01 -07:00
parent ef750d8472
commit 47031c0a54
28 changed files with 32514 additions and 4290 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
configure vendored
View File

@ -52,6 +52,7 @@ Advanced options:
${toggle_multi_res_encoding} enable multiple-resolution encoding
${toggle_temporal_denoising} enable temporal denoising and disable the spatial denoiser
${toggle_webm_io} enable input from and output to WebM container
${toggle_libyuv} enable libyuv
Codecs:
Codecs can be selectively enabled or disabled individually, or by family:
@ -315,6 +316,7 @@ CONFIG_LIST="
os_support
unit_tests
webm_io
libyuv
decode_perf_tests
multi_res_encoding
temporal_denoising
@ -368,6 +370,7 @@ CMDLINE_SELECT="
postproc_visualizer
unit_tests
webm_io
libyuv
decode_perf_tests
multi_res_encoding
temporal_denoising
@ -709,9 +712,11 @@ process_toolchain() {
*-vs*)
soft_enable unit_tests
soft_enable webm_io
soft_enable libyuv
;;
*-android-*)
soft_enable webm_io
soft_enable libyuv
# GTestLog must be modified to use Android logging utilities.
;;
*-darwin-*)
@ -727,6 +732,9 @@ int z;
EOF
check_cxx "$@" <<EOF && soft_enable webm_io
int z;
EOF
check_cxx "$@" <<EOF && soft_enable libyuv
int z;
EOF
;;
*)
@ -735,6 +743,9 @@ int z;
EOF
check_cxx "$@" <<EOF && soft_enable webm_io
int z;
EOF
check_cxx "$@" <<EOF && soft_enable libyuv
int z;
EOF
;;
esac

30
examples.mk
View File

@ -10,10 +10,24 @@
LIBYUV_SRCS += third_party/libyuv/include/libyuv/basic_types.h \
third_party/libyuv/include/libyuv/cpu_id.h \
third_party/libyuv/include/libyuv/planar_functions.h \
third_party/libyuv/include/libyuv/row.h \
third_party/libyuv/include/libyuv/scale.h \
third_party/libyuv/source/row.h \
third_party/libyuv/source/scale.c \
third_party/libyuv/source/cpu_id.c
third_party/libyuv/include/libyuv/scale_row.h \
third_party/libyuv/source/cpu_id.cc \
third_party/libyuv/source/planar_functions.cc \
third_party/libyuv/source/row_any.cc \
third_party/libyuv/source/row_common.cc \
third_party/libyuv/source/row_mips.cc \
third_party/libyuv/source/row_neon.cc \
third_party/libyuv/source/row_posix.cc \
third_party/libyuv/source/row_win.cc \
third_party/libyuv/source/scale.cc \
third_party/libyuv/source/scale_common.cc \
third_party/libyuv/source/scale_mips.cc \
third_party/libyuv/source/scale_neon.cc \
third_party/libyuv/source/scale_posix.cc \
third_party/libyuv/source/scale_win.cc
LIBWEBM_MUXER_SRCS += third_party/libwebm/mkvmuxer.cpp \
third_party/libwebm/mkvmuxerutil.cpp \
@ -42,7 +56,9 @@ vpxdec.SRCS += args.c args.h
vpxdec.SRCS += ivfdec.c ivfdec.h
vpxdec.SRCS += tools_common.c tools_common.h
vpxdec.SRCS += y4menc.c y4menc.h
vpxdec.SRCS += $(LIBYUV_SRCS)
ifeq ($(CONFIG_LIBYUV),yes)
vpxdec.SRCS += $(LIBYUV_SRCS)
endif
ifeq ($(CONFIG_WEBM_IO),yes)
vpxdec.SRCS += $(LIBWEBM_PARSER_SRCS)
vpxdec.SRCS += webmdec.cc webmdec.h
@ -60,7 +76,9 @@ vpxenc.SRCS += vpx_ports/mem_ops.h
vpxenc.SRCS += vpx_ports/mem_ops_aligned.h
vpxenc.SRCS += vpx_ports/vpx_timer.h
vpxenc.SRCS += vpxstats.c vpxstats.h
vpxenc.SRCS += $(LIBYUV_SRCS)
ifeq ($(CONFIG_LIBYUV),yes)
vpxenc.SRCS += $(LIBYUV_SRCS)
endif
ifeq ($(CONFIG_WEBM_IO),yes)
vpxenc.SRCS += $(LIBWEBM_MUXER_SRCS)
vpxenc.SRCS += webmenc.cc webmenc.h
@ -160,11 +178,13 @@ vp8cx_set_ref.DESCRIPTION = VP8 set encoder reference frame
ifeq ($(CONFIG_MULTI_RES_ENCODING),yes)
ifeq ($(CONFIG_LIBYUV),yes)
EXAMPLES-$(CONFIG_VP8_DECODER) += vp8_multi_resolution_encoder.c
vp8_multi_resolution_encoder.SRCS += $(LIBYUV_SRCS)
vp8_multi_resolution_encoder.GUID = 04f8738e-63c8-423b-90fa-7c2703a374de
vp8_multi_resolution_encoder.DESCRIPTION = VP8 Multiple-resolution Encoding
endif
endif
# Handle extra library flags depending on codec configuration

6
third_party/libyuv/README.libvpx vendored
View File

@ -1,6 +1,6 @@
Name: libyuv
URL: http://code.google.com/p/libyuv/
Version: 102
Version: 1005
License: BSD
License File: LICENSE
@ -13,5 +13,5 @@ which down-samples the original input video (f.g. 1280x720) a number of times
in order to encode multiple resolution bit streams.
Local Modifications:
Modified the original scaler code from C++ to C to fit in our current build
system. This is a temporal solution, and will be improved later.
Modified the original scaler code minimally with include file changes to fit
in our current build system.

77
third_party/libyuv/include/libyuv/basic_types.h vendored
View File

@ -1,22 +1,25 @@
/*
* Copyright (c) 2011 The LibYuv project authors. All Rights Reserved.
* Copyright 2011 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
* 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_BASIC_TYPES_H_
#ifndef INCLUDE_LIBYUV_BASIC_TYPES_H_ // NOLINT
#define INCLUDE_LIBYUV_BASIC_TYPES_H_
#include <stddef.h> // for NULL, size_t
#if !(defined(_MSC_VER) && (_MSC_VER < 1600))
#if defined(__ANDROID__) || (defined(_MSC_VER) && (_MSC_VER < 1600))
#include <sys/types.h> // for uintptr_t on x86
#else
#include <stdint.h> // for uintptr_t
#endif
#ifndef GG_LONGLONG
#ifndef INT_TYPES_DEFINED
#define INT_TYPES_DEFINED
#ifdef COMPILER_MSVC
@ -30,9 +33,9 @@ typedef __int64 int64;
#endif
#define INT64_F "I64"
#else // COMPILER_MSVC
#ifdef __LP64__
typedef unsigned long uint64;
typedef long int64;
#if defined(__LP64__) && !defined(__OpenBSD__) && !defined(__APPLE__)
typedef unsigned long uint64; // NOLINT
typedef long int64; // NOLINT
#ifndef INT64_C
#define INT64_C(x) x ## L
#endif
@ -40,9 +43,9 @@ typedef long int64;
#define UINT64_C(x) x ## UL
#endif
#define INT64_F "l"
#else // __LP64__
typedef unsigned long long uint64;
typedef long long int64;
#else // defined(__LP64__) && !defined(__OpenBSD__) && !defined(__APPLE__)
typedef unsigned long long uint64; // NOLINT
typedef long long int64; // NOLINT
#ifndef INT64_C
#define INT64_C(x) x ## LL
#endif
@ -54,20 +57,62 @@ typedef long long int64;
#endif // COMPILER_MSVC
typedef unsigned int uint32;
typedef int int32;
typedef unsigned short uint16;
typedef short int16;
typedef unsigned short uint16; // NOLINT
typedef short int16; // NOLINT
typedef unsigned char uint8;
typedef char int8;
typedef signed char int8;
#endif // INT_TYPES_DEFINED
#endif // GG_LONGLONG
// Detect compiler is for x86 or x64.
#if defined(__x86_64__) || defined(_M_X64) || \
defined(__i386__) || defined(_M_IX86)
#define CPU_X86 1
#endif
// Detect compiler is for ARM.
#if defined(__arm__) || defined(_M_ARM)
#define CPU_ARM 1
#endif
#ifndef ALIGNP
#ifdef __cplusplus
#define ALIGNP(p, t) \
((uint8*)((((uintptr_t)(p) + \
((t)-1)) & ~((t)-1))))
(reinterpret_cast<uint8*>(((reinterpret_cast<uintptr_t>(p) + \
((t) - 1)) & ~((t) - 1))))
#else
#define ALIGNP(p, t) \
((uint8*)((((uintptr_t)(p) + ((t) - 1)) & ~((t) - 1)))) /* NOLINT */
#endif
#endif
#endif // INCLUDE_LIBYUV_BASIC_TYPES_H_
#if !defined(LIBYUV_API)
#if defined(_WIN32) || defined(__CYGWIN__)
#if defined(LIBYUV_BUILDING_SHARED_LIBRARY)
#define LIBYUV_API __declspec(dllexport)
#elif defined(LIBYUV_USING_SHARED_LIBRARY)
#define LIBYUV_API __declspec(dllimport)
#else
#define LIBYUV_API
#endif // LIBYUV_BUILDING_SHARED_LIBRARY
#elif defined(__GNUC__) && (__GNUC__ >= 4) && !defined(__APPLE__) && \
(defined(LIBYUV_BUILDING_SHARED_LIBRARY) || \
defined(LIBYUV_USING_SHARED_LIBRARY))
#define LIBYUV_API __attribute__ ((visibility ("default")))
#else
#define LIBYUV_API
#endif // __GNUC__
#endif // LIBYUV_API
#define LIBYUV_BOOL int
#define LIBYUV_FALSE 0
#define LIBYUV_TRUE 1
// Visual C x86 or GCC little endian.
#if defined(__x86_64__) || defined(_M_X64) || \
defined(__i386__) || defined(_M_IX86) || \
defined(__arm__) || defined(_M_ARM) || \
(defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
#define LIBYUV_LITTLE_ENDIAN
#endif
#endif // INCLUDE_LIBYUV_BASIC_TYPES_H_ NOLINT

66
third_party/libyuv/include/libyuv/cpu_id.h vendored
View File

@ -1,49 +1,81 @@
/*
* Copyright (c) 2011 The LibYuv project authors. All Rights Reserved.
* Copyright 2011 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
* 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_CPU_ID_H_
#ifndef INCLUDE_LIBYUV_CPU_ID_H_ // NOLINT
#define INCLUDE_LIBYUV_CPU_ID_H_
#include "basic_types.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// These flags are only valid on x86 processors
static const int kCpuHasSSE2 = 1;
static const int kCpuHasSSSE3 = 2;
// TODO(fbarchard): Consider overlapping bits for different architectures.
// Internal flag to indicate cpuid requires initialization.
#define kCpuInit 0x1
// These flags are only valid on ARM processors
static const int kCpuHasNEON = 4;
// These flags are only valid on ARM processors.
static const int kCpuHasARM = 0x2;
static const int kCpuHasNEON = 0x4;
// 0x8 reserved for future ARM flag.
// Internal flag to indicate cpuid is initialized.
static const int kCpuInitialized = 8;
// These flags are only valid on x86 processors.
static const int kCpuHasX86 = 0x10;
static const int kCpuHasSSE2 = 0x20;
static const int kCpuHasSSSE3 = 0x40;
static const int kCpuHasSSE41 = 0x80;
static const int kCpuHasSSE42 = 0x100;
static const int kCpuHasAVX = 0x200;
static const int kCpuHasAVX2 = 0x400;
static const int kCpuHasERMS = 0x800;
static const int kCpuHasFMA3 = 0x1000;
// 0x2000, 0x4000, 0x8000 reserved for future X86 flags.
// These flags are only valid on MIPS processors.
static const int kCpuHasMIPS = 0x10000;
static const int kCpuHasMIPS_DSP = 0x20000;
static const int kCpuHasMIPS_DSPR2 = 0x40000;
// Internal function used to auto-init.
LIBYUV_API
int InitCpuFlags(void);
// Internal function for parsing /proc/cpuinfo.
LIBYUV_API
int ArmCpuCaps(const char* cpuinfo_name);
// Detect CPU has SSE2 etc.
// test_flag parameter should be one of kCpuHas constants above
// Test_flag parameter should be one of kCpuHas constants above.
// returns non-zero if instruction set is detected
static __inline int TestCpuFlag(int test_flag) {
extern int cpu_info_;
extern int InitCpuFlags();
return (cpu_info_ ? cpu_info_ : InitCpuFlags()) & test_flag;
LIBYUV_API extern int cpu_info_;
return (cpu_info_ == kCpuInit ? InitCpuFlags() : cpu_info_) & test_flag;
}
// For testing, allow CPU flags to be disabled.
// ie MaskCpuFlags(~kCpuHasSSSE3) to disable SSSE3.
// -1 to enable all cpu specific optimizations.
// 0 to disable all cpu specific optimizations.
// MaskCpuFlags(-1) to enable all cpu specific optimizations.
// MaskCpuFlags(0) to disable all cpu specific optimizations.
LIBYUV_API
void MaskCpuFlags(int enable_flags);
// Low level cpuid for X86. Returns zeros on other CPUs.
// eax is the info type that you want.
// ecx is typically the cpu number, and should normally be zero.
LIBYUV_API
void CpuId(uint32 eax, uint32 ecx, uint32* cpu_info);
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
#endif // INCLUDE_LIBYUV_CPU_ID_H_
#endif // INCLUDE_LIBYUV_CPU_ID_H_ NOLINT

439
third_party/libyuv/include/libyuv/planar_functions.h vendored Normal file
View File

@ -0,0 +1,439 @@
/*
* Copyright 2011 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_PLANAR_FUNCTIONS_H_ // NOLINT
#define INCLUDE_LIBYUV_PLANAR_FUNCTIONS_H_
#include "basic_types.h"
// TODO(fbarchard): Remove the following headers includes.
// #include "convert.h"
// #include "convert_argb.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// Copy a plane of data.
LIBYUV_API
void CopyPlane(const uint8* src_y, int src_stride_y,
uint8* dst_y, int dst_stride_y,
int width, int height);
LIBYUV_API
void CopyPlane_16(const uint16* src_y, int src_stride_y,
uint16* dst_y, int dst_stride_y,
int width, int height);
// Set a plane of data to a 32 bit value.
LIBYUV_API
void SetPlane(uint8* dst_y, int dst_stride_y,
int width, int height,
uint32 value);
// Copy I400. Supports inverting.
LIBYUV_API
int I400ToI400(const uint8* src_y, int src_stride_y,
uint8* dst_y, int dst_stride_y,
int width, int height);
// Copy I422 to I422.
#define I422ToI422 I422Copy
LIBYUV_API
int I422Copy(const uint8* src_y, int src_stride_y,
const uint8* src_u, int src_stride_u,
const uint8* src_v, int src_stride_v,
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);
// Copy I444 to I444.
#define I444ToI444 I444Copy
LIBYUV_API
int I444Copy(const uint8* src_y, int src_stride_y,
const uint8* src_u, int src_stride_u,
const uint8* src_v, int src_stride_v,
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 YUY2 to I422.
LIBYUV_API
int YUY2ToI422(const uint8* src_yuy2, int src_stride_yuy2,
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 UYVY to I422.
LIBYUV_API
int UYVYToI422(const uint8* src_uyvy, int src_stride_uyvy,
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 I420 to I400. (calls CopyPlane ignoring u/v).
LIBYUV_API
int I420ToI400(const uint8* src_y, int src_stride_y,
const uint8* src_u, int src_stride_u,
const uint8* src_v, int src_stride_v,
uint8* dst_y, int dst_stride_y,
int width, int height);
// Alias
#define I420ToI420Mirror I420Mirror
// I420 mirror.
LIBYUV_API
int I420Mirror(const uint8* src_y, int src_stride_y,
const uint8* src_u, int src_stride_u,
const uint8* src_v, int src_stride_v,
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);
// Alias
#define I400ToI400Mirror I400Mirror
// I400 mirror. A single plane is mirrored horizontally.
// Pass negative height to achieve 180 degree rotation.
LIBYUV_API
int I400Mirror(const uint8* src_y, int src_stride_y,
uint8* dst_y, int dst_stride_y,
int width, int height);
// Alias
#define ARGBToARGBMirror ARGBMirror
// ARGB mirror.
LIBYUV_API
int ARGBMirror(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Convert NV12 to RGB565.
LIBYUV_API
int NV12ToRGB565(const uint8* src_y, int src_stride_y,
const uint8* src_uv, int src_stride_uv,
uint8* dst_rgb565, int dst_stride_rgb565,
int width, int height);
// Convert NV21 to RGB565.
LIBYUV_API
int NV21ToRGB565(const uint8* src_y, int src_stride_y,
const uint8* src_uv, int src_stride_uv,
uint8* dst_rgb565, int dst_stride_rgb565,
int width, int height);
// I422ToARGB is in convert_argb.h
// Convert I422 to BGRA.
LIBYUV_API
int I422ToBGRA(const uint8* src_y, int src_stride_y,
const uint8* src_u, int src_stride_u,
const uint8* src_v, int src_stride_v,
uint8* dst_bgra, int dst_stride_bgra,
int width, int height);
// Convert I422 to ABGR.
LIBYUV_API
int I422ToABGR(const uint8* src_y, int src_stride_y,
const uint8* src_u, int src_stride_u,
const uint8* src_v, int src_stride_v,
uint8* dst_abgr, int dst_stride_abgr,
int width, int height);
// Convert I422 to RGBA.
LIBYUV_API
int I422ToRGBA(const uint8* src_y, int src_stride_y,
const uint8* src_u, int src_stride_u,
const uint8* src_v, int src_stride_v,
uint8* dst_rgba, int dst_stride_rgba,
int width, int height);
// Draw a rectangle into I420.
LIBYUV_API
int I420Rect(uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int x, int y, int width, int height,
int value_y, int value_u, int value_v);
// Draw a rectangle into ARGB.
LIBYUV_API
int ARGBRect(uint8* dst_argb, int dst_stride_argb,
int x, int y, int width, int height, uint32 value);
// Convert ARGB to gray scale ARGB.
LIBYUV_API
int ARGBGrayTo(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Make a rectangle of ARGB gray scale.
LIBYUV_API
int ARGBGray(uint8* dst_argb, int dst_stride_argb,
int x, int y, int width, int height);
// Make a rectangle of ARGB Sepia tone.
LIBYUV_API
int ARGBSepia(uint8* dst_argb, int dst_stride_argb,
int x, int y, int width, int height);
// Apply a matrix rotation to each ARGB pixel.
// matrix_argb is 4 signed ARGB values. -128 to 127 representing -2 to 2.
// The first 4 coefficients apply to B, G, R, A and produce B of the output.
// The next 4 coefficients apply to B, G, R, A and produce G of the output.
// The next 4 coefficients apply to B, G, R, A and produce R of the output.
// The last 4 coefficients apply to B, G, R, A and produce A of the output.
LIBYUV_API
int ARGBColorMatrix(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
const int8* matrix_argb,
int width, int height);
// Deprecated. Use ARGBColorMatrix instead.
// Apply a matrix rotation to each ARGB pixel.
// matrix_argb is 3 signed ARGB values. -128 to 127 representing -1 to 1.
// The first 4 coefficients apply to B, G, R, A and produce B of the output.
// The next 4 coefficients apply to B, G, R, A and produce G of the output.
// The last 4 coefficients apply to B, G, R, A and produce R of the output.
LIBYUV_API
int RGBColorMatrix(uint8* dst_argb, int dst_stride_argb,
const int8* matrix_rgb,
int x, int y, int width, int height);
// Apply a color table each ARGB pixel.
// Table contains 256 ARGB values.
LIBYUV_API
int ARGBColorTable(uint8* dst_argb, int dst_stride_argb,
const uint8* table_argb,
int x, int y, int width, int height);
// Apply a color table each ARGB pixel but preserve destination alpha.
// Table contains 256 ARGB values.
LIBYUV_API
int RGBColorTable(uint8* dst_argb, int dst_stride_argb,
const uint8* table_argb,
int x, int y, int width, int height);
// Apply a luma/color table each ARGB pixel but preserve destination alpha.
// Table contains 32768 values indexed by [Y][C] where 7 it 7 bit luma from
// RGB (YJ style) and C is an 8 bit color component (R, G or B).
LIBYUV_API
int ARGBLumaColorTable(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
const uint8* luma_rgb_table,
int width, int height);
// Apply a 3 term polynomial to ARGB values.
// poly points to a 4x4 matrix. The first row is constants. The 2nd row is
// coefficients for b, g, r and a. The 3rd row is coefficients for b squared,
// g squared, r squared and a squared. The 4rd row is coefficients for b to
// the 3, g to the 3, r to the 3 and a to the 3. The values are summed and
// result clamped to 0 to 255.
// A polynomial approximation can be dirived using software such as 'R'.
LIBYUV_API
int ARGBPolynomial(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
const float* poly,
int width, int height);
// Quantize a rectangle of ARGB. Alpha unaffected.
// scale is a 16 bit fractional fixed point scaler between 0 and 65535.
// interval_size should be a value between 1 and 255.
// interval_offset should be a value between 0 and 255.
LIBYUV_API
int ARGBQuantize(uint8* dst_argb, int dst_stride_argb,
int scale, int interval_size, int interval_offset,
int x, int y, int width, int height);
// Copy ARGB to ARGB.
LIBYUV_API
int ARGBCopy(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Copy ARGB to ARGB.
LIBYUV_API
int ARGBCopyAlpha(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Copy ARGB to ARGB.
LIBYUV_API
int ARGBCopyYToAlpha(const uint8* src_y, int src_stride_y,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
typedef void (*ARGBBlendRow)(const uint8* src_argb0, const uint8* src_argb1,
uint8* dst_argb, int width);
// Get function to Alpha Blend ARGB pixels and store to destination.
LIBYUV_API
ARGBBlendRow GetARGBBlend();
// Alpha Blend ARGB images and store to destination.
// Alpha of destination is set to 255.
LIBYUV_API
int ARGBBlend(const uint8* src_argb0, int src_stride_argb0,
const uint8* src_argb1, int src_stride_argb1,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Multiply ARGB image by ARGB image. Shifted down by 8. Saturates to 255.
LIBYUV_API
int ARGBMultiply(const uint8* src_argb0, int src_stride_argb0,
const uint8* src_argb1, int src_stride_argb1,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Add ARGB image with ARGB image. Saturates to 255.
LIBYUV_API
int ARGBAdd(const uint8* src_argb0, int src_stride_argb0,
const uint8* src_argb1, int src_stride_argb1,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Subtract ARGB image (argb1) from ARGB image (argb0). Saturates to 0.
LIBYUV_API
int ARGBSubtract(const uint8* src_argb0, int src_stride_argb0,
const uint8* src_argb1, int src_stride_argb1,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Convert I422 to YUY2.
LIBYUV_API
int I422ToYUY2(const uint8* src_y, int src_stride_y,
const uint8* src_u, int src_stride_u,
const uint8* src_v, int src_stride_v,
uint8* dst_frame, int dst_stride_frame,
int width, int height);
// Convert I422 to UYVY.
LIBYUV_API
int I422ToUYVY(const uint8* src_y, int src_stride_y,
const uint8* src_u, int src_stride_u,
const uint8* src_v, int src_stride_v,
uint8* dst_frame, int dst_stride_frame,
int width, int height);
// Convert unattentuated ARGB to preattenuated ARGB.
LIBYUV_API
int ARGBAttenuate(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Convert preattentuated ARGB to unattenuated ARGB.
LIBYUV_API
int ARGBUnattenuate(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Convert MJPG to ARGB.
LIBYUV_API
int MJPGToARGB(const uint8* sample, size_t sample_size,
uint8* argb, int argb_stride,
int w, int h, int dw, int dh);
// Internal function - do not call directly.
// Computes table of cumulative sum for image where the value is the sum
// of all values above and to the left of the entry. Used by ARGBBlur.
LIBYUV_API
int ARGBComputeCumulativeSum(const uint8* src_argb, int src_stride_argb,
int32* dst_cumsum, int dst_stride32_cumsum,
int width, int height);
// Blur ARGB image.
// dst_cumsum table of width * (height + 1) * 16 bytes aligned to
// 16 byte boundary.
// dst_stride32_cumsum is number of ints in a row (width * 4).
// radius is number of pixels around the center. e.g. 1 = 3x3. 2=5x5.
// Blur is optimized for radius of 5 (11x11) or less.
LIBYUV_API
int ARGBBlur(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
int32* dst_cumsum, int dst_stride32_cumsum,
int width, int height, int radius);
// Multiply ARGB image by ARGB value.
LIBYUV_API
int ARGBShade(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
int width, int height, uint32 value);
// Interpolate between two ARGB images using specified amount of interpolation
// (0 to 255) and store to destination.
// 'interpolation' is specified as 8 bit fraction where 0 means 100% src_argb0
// and 255 means 1% src_argb0 and 99% src_argb1.
// Internally uses ARGBScale bilinear filtering.
// Caveat: This function will write up to 16 bytes beyond the end of dst_argb.
LIBYUV_API
int ARGBInterpolate(const uint8* src_argb0, int src_stride_argb0,
const uint8* src_argb1, int src_stride_argb1,
uint8* dst_argb, int dst_stride_argb,
int width, int height, int interpolation);
#if defined(__pnacl__) || defined(__CLR_VER) || defined(COVERAGE_ENABLED) || \
defined(TARGET_IPHONE_SIMULATOR)
#define LIBYUV_DISABLE_X86
#endif
// Row functions for copying a pixels from a source with a slope to a row
// of destination. Useful for scaling, rotation, mirror, texture mapping.
LIBYUV_API
void ARGBAffineRow_C(const uint8* src_argb, int src_argb_stride,
uint8* dst_argb, const float* uv_dudv, int width);
// The following are available on all x86 platforms:
#if !defined(LIBYUV_DISABLE_X86) && \
(defined(_M_IX86) || defined(__x86_64__) || defined(__i386__))
LIBYUV_API
void ARGBAffineRow_SSE2(const uint8* src_argb, int src_argb_stride,
uint8* dst_argb, const float* uv_dudv, int width);
#define HAS_ARGBAFFINEROW_SSE2
#endif // LIBYUV_DISABLE_X86
// Shuffle ARGB channel order. e.g. BGRA to ARGB.
// shuffler is 16 bytes and must be aligned.
LIBYUV_API
int ARGBShuffle(const uint8* src_bgra, int src_stride_bgra,
uint8* dst_argb, int dst_stride_argb,
const uint8* shuffler, int width, int height);
// Sobel ARGB effect with planar output.
LIBYUV_API
int ARGBSobelToPlane(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
int width, int height);
// Sobel ARGB effect.
LIBYUV_API
int ARGBSobel(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
// Sobel ARGB effect w/ Sobel X, Sobel, Sobel Y in ARGB.
LIBYUV_API
int ARGBSobelXY(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb,
int width, int height);
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
#endif // INCLUDE_LIBYUV_PLANAR_FUNCTIONS_H_ NOLINT

1704
third_party/libyuv/include/libyuv/row.h vendored Normal file
View File

File diff suppressed because it is too large Load Diff

70
third_party/libyuv/include/libyuv/scale.h vendored
View File

@ -1,30 +1,45 @@
/*
* Copyright (c) 2011 The LibYuv project authors. All Rights Reserved.
* Copyright 2011 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
* 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_SCALE_H_
#ifndef INCLUDE_LIBYUV_SCALE_H_ // NOLINT
#define INCLUDE_LIBYUV_SCALE_H_
#include "third_party/libyuv/include/libyuv/basic_types.h"
#include "basic_types.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// Supported filtering
typedef enum {
kFilterNone = 0, // Point sample; Fastest
kFilterBilinear = 1, // Faster than box, but lower quality scaling down.
kFilterBox = 2 // Highest quality
// Supported filtering.
typedef enum FilterMode {
kFilterNone = 0, // Point sample; Fastest.
kFilterLinear = 1, // Filter horizontally only.
kFilterBilinear = 2, // Faster than box, but lower quality scaling down.
kFilterBox = 3 // Highest quality.
} FilterModeEnum;
// Scale a YUV plane.
LIBYUV_API
void ScalePlane(const uint8* src, int src_stride,
int src_width, int src_height,
uint8* dst, int dst_stride,
int dst_width, int dst_height,
enum FilterMode filtering);
void ScalePlane_16(const uint16* src, int src_stride,
int src_width, int src_height,
uint16* dst, int dst_stride,
int dst_width, int dst_height,
enum FilterMode filtering);
// Scales a YUV 4:2:0 image from the src width and height to the
// dst width and height.
// If filtering is kFilterNone, a simple nearest-neighbor algorithm is
@ -35,6 +50,7 @@ typedef enum {
// quality image, at further expense of speed.
// Returns 0 if successful.
LIBYUV_API
int I420Scale(const uint8* src_y, int src_stride_y,
const uint8* src_u, int src_stride_u,
const uint8* src_v, int src_stride_v,
@ -43,28 +59,44 @@ int I420Scale(const uint8* src_y, int src_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int dst_width, int dst_height,
FilterModeEnum filtering);
enum FilterMode filtering);
// Legacy API. Deprecated
LIBYUV_API
int I420Scale_16(const uint16* src_y, int src_stride_y,
const uint16* src_u, int src_stride_u,
const uint16* src_v, int src_stride_v,
int src_width, int src_height,
uint16* dst_y, int dst_stride_y,
uint16* dst_u, int dst_stride_u,
uint16* dst_v, int dst_stride_v,
int dst_width, int dst_height,
enum FilterMode filtering);
#ifdef __cplusplus
// Legacy API. Deprecated.
LIBYUV_API
int Scale(const uint8* src_y, const uint8* src_u, const uint8* src_v,
int src_stride_y, int src_stride_u, int src_stride_v,
int src_width, int src_height,
uint8* dst_y, uint8* dst_u, uint8* dst_v,
int dst_stride_y, int dst_stride_u, int dst_stride_v,
int dst_width, int dst_height,
int interpolate);
LIBYUV_BOOL interpolate);
// Legacy API. Deprecated
int ScaleOffset(const uint8* src, int src_width, int src_height,
uint8* dst, int dst_width, int dst_height, int dst_yoffset,
int interpolate);
// Legacy API. Deprecated.
LIBYUV_API
int ScaleOffset(const uint8* src_i420, int src_width, int src_height,
uint8* dst_i420, int dst_width, int dst_height, int dst_yoffset,
LIBYUV_BOOL interpolate);
// For testing, allow disabling of optimizations.
void SetUseReferenceImpl(int use);
// For testing, allow disabling of specialized scalers.
LIBYUV_API
void SetUseReferenceImpl(LIBYUV_BOOL use);
#endif // __cplusplus
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
#endif // INCLUDE_LIBYUV_SCALE_H_
#endif // INCLUDE_LIBYUV_SCALE_H_ NOLINT

341
third_party/libyuv/include/libyuv/scale_row.h vendored Normal file
View File

@ -0,0 +1,341 @@
/*
* Copyright 2013 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_SCALE_ROW_H_ // NOLINT
#define INCLUDE_LIBYUV_SCALE_ROW_H_
#include "basic_types.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
#if defined(__pnacl__) || defined(__CLR_VER) || defined(COVERAGE_ENABLED) || \
defined(TARGET_IPHONE_SIMULATOR)
#define LIBYUV_DISABLE_X86
#endif
// The following are available on all x86 platforms:
#if !defined(LIBYUV_DISABLE_X86) && \
(defined(_M_IX86) || defined(__x86_64__) || defined(__i386__))
#define HAS_SCALEROWDOWN2_SSE2
#define HAS_SCALEROWDOWN4_SSE2
#define HAS_SCALEROWDOWN34_SSSE3
#define HAS_SCALEROWDOWN38_SSSE3
#define HAS_SCALEADDROWS_SSE2
#define HAS_SCALEFILTERCOLS_SSSE3
#define HAS_SCALECOLSUP2_SSE2
#define HAS_SCALEARGBROWDOWN2_SSE2
#define HAS_SCALEARGBROWDOWNEVEN_SSE2
#define HAS_SCALEARGBCOLS_SSE2
#define HAS_SCALEARGBFILTERCOLS_SSSE3
#define HAS_SCALEARGBCOLSUP2_SSE2
#define HAS_FIXEDDIV_X86
#define HAS_FIXEDDIV1_X86
#endif
// The following are available on Neon platforms:
#if !defined(LIBYUV_DISABLE_NEON) && !defined(__native_client__) && \
(defined(__ARM_NEON__) || defined(LIBYUV_NEON))
#define HAS_SCALEROWDOWN2_NEON
#define HAS_SCALEROWDOWN4_NEON
#define HAS_SCALEROWDOWN34_NEON
#define HAS_SCALEROWDOWN38_NEON
#define HAS_SCALEARGBROWDOWNEVEN_NEON
#define HAS_SCALEARGBROWDOWN2_NEON
#endif
// The following are available on Mips platforms:
#if !defined(LIBYUV_DISABLE_MIPS) && !defined(__native_client__) && \
defined(__mips__) && defined(__mips_dsp) && (__mips_dsp_rev >= 2)
#define HAS_SCALEROWDOWN2_MIPS_DSPR2
#define HAS_SCALEROWDOWN4_MIPS_DSPR2
#define HAS_SCALEROWDOWN34_MIPS_DSPR2
#define HAS_SCALEROWDOWN38_MIPS_DSPR2
#endif
// Scale ARGB vertically with bilinear interpolation.
void ScalePlaneVertical(int src_height,
int dst_width, int dst_height,
int src_stride, int dst_stride,
const uint8* src_argb, uint8* dst_argb,
int x, int y, int dy,
int bpp, enum FilterMode filtering);
void ScalePlaneVertical_16(int src_height,
int dst_width, int dst_height,
int src_stride, int dst_stride,
const uint16* src_argb, uint16* dst_argb,
int x, int y, int dy,
int wpp, enum FilterMode filtering);
// Simplify the filtering based on scale factors.
enum FilterMode ScaleFilterReduce(int src_width, int src_height,
int dst_width, int dst_height,
enum FilterMode filtering);
// Divide num by div and return as 16.16 fixed point result.
int FixedDiv_C(int num, int div);
int FixedDiv_X86(int num, int div);
// Divide num - 1 by div - 1 and return as 16.16 fixed point result.
int FixedDiv1_C(int num, int div);
int FixedDiv1_X86(int num, int div);
#ifdef HAS_FIXEDDIV_X86
#define FixedDiv FixedDiv_X86
#define FixedDiv1 FixedDiv1_X86
#else
#define FixedDiv FixedDiv_C
#define FixedDiv1 FixedDiv1_C
#endif
// Compute slope values for stepping.
void ScaleSlope(int src_width, int src_height,
int dst_width, int dst_height,
enum FilterMode filtering,
int* x, int* y, int* dx, int* dy);
void ScaleRowDown2_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown2_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
uint16* dst, int dst_width);
void ScaleRowDown2Linear_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown2Linear_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
uint16* dst, int dst_width);
void ScaleRowDown2Box_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown2Box_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
uint16* dst, int dst_width);
void ScaleRowDown4_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown4_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
uint16* dst, int dst_width);
void ScaleRowDown4Box_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown4Box_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
uint16* dst, int dst_width);
void ScaleRowDown34_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown34_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
uint16* dst, int dst_width);
void ScaleRowDown34_0_Box_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* d, int dst_width);
void ScaleRowDown34_0_Box_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
uint16* d, int dst_width);
void ScaleRowDown34_1_Box_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* d, int dst_width);
void ScaleRowDown34_1_Box_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
uint16* d, int dst_width);
void ScaleCols_C(uint8* dst_ptr, const uint8* src_ptr,
int dst_width, int x, int dx);
void ScaleCols_16_C(uint16* dst_ptr, const uint16* src_ptr,
int dst_width, int x, int dx);
void ScaleColsUp2_C(uint8* dst_ptr, const uint8* src_ptr,
int dst_width, int, int);
void ScaleColsUp2_16_C(uint16* dst_ptr, const uint16* src_ptr,
int dst_width, int, int);
void ScaleFilterCols_C(uint8* dst_ptr, const uint8* src_ptr,
int dst_width, int x, int dx);
void ScaleFilterCols_16_C(uint16* dst_ptr, const uint16* src_ptr,
int dst_width, int x, int dx);
void ScaleFilterCols64_C(uint8* dst_ptr, const uint8* src_ptr,
int dst_width, int x, int dx);
void ScaleFilterCols64_16_C(uint16* dst_ptr, const uint16* src_ptr,
int dst_width, int x, int dx);
void ScaleRowDown38_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown38_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
uint16* dst, int dst_width);
void ScaleRowDown38_3_Box_C(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown38_3_Box_16_C(const uint16* src_ptr,
ptrdiff_t src_stride,
uint16* dst_ptr, int dst_width);
void ScaleRowDown38_2_Box_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown38_2_Box_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
uint16* dst_ptr, int dst_width);
void ScaleAddRows_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint16* dst_ptr, int src_width, int src_height);
void ScaleAddRows_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
uint32* dst_ptr, int src_width, int src_height);
void ScaleARGBRowDown2_C(const uint8* src_argb,
ptrdiff_t src_stride,
uint8* dst_argb, int dst_width);
void ScaleARGBRowDown2Linear_C(const uint8* src_argb,
ptrdiff_t src_stride,
uint8* dst_argb, int dst_width);
void ScaleARGBRowDown2Box_C(const uint8* src_argb, ptrdiff_t src_stride,
uint8* dst_argb, int dst_width);
void ScaleARGBRowDownEven_C(const uint8* src_argb, ptrdiff_t src_stride,
int src_stepx,
uint8* dst_argb, int dst_width);
void ScaleARGBRowDownEvenBox_C(const uint8* src_argb,
ptrdiff_t src_stride,
int src_stepx,
uint8* dst_argb, int dst_width);
void ScaleARGBCols_C(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx);
void ScaleARGBCols64_C(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx);
void ScaleARGBColsUp2_C(uint8* dst_argb, const uint8* src_argb,
int dst_width, int, int);
void ScaleARGBFilterCols_C(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx);
void ScaleARGBFilterCols64_C(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx);
void ScaleRowDown2_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown2Linear_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown2Box_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown2_Unaligned_SSE2(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown2Linear_Unaligned_SSE2(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown2Box_Unaligned_SSE2(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown4_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown4Box_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown34_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown34_1_Box_SSSE3(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown34_0_Box_SSSE3(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown38_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown38_3_Box_SSSE3(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown38_2_Box_SSSE3(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleAddRows_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint16* dst_ptr, int src_width,
int src_height);
void ScaleFilterCols_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
int dst_width, int x, int dx);
void ScaleColsUp2_SSE2(uint8* dst_ptr, const uint8* src_ptr,
int dst_width, int x, int dx);
void ScaleARGBRowDown2_SSE2(const uint8* src_argb,
ptrdiff_t src_stride,
uint8* dst_argb, int dst_width);
void ScaleARGBRowDown2Linear_SSE2(const uint8* src_argb,
ptrdiff_t src_stride,
uint8* dst_argb, int dst_width);
void ScaleARGBRowDown2Box_SSE2(const uint8* src_argb,
ptrdiff_t src_stride,
uint8* dst_argb, int dst_width);
void ScaleARGBRowDownEven_SSE2(const uint8* src_argb, ptrdiff_t src_stride,
int src_stepx,
uint8* dst_argb, int dst_width);
void ScaleARGBRowDownEvenBox_SSE2(const uint8* src_argb,
ptrdiff_t src_stride,
int src_stepx,
uint8* dst_argb, int dst_width);
void ScaleARGBCols_SSE2(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx);
void ScaleARGBFilterCols_SSSE3(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx);
void ScaleARGBColsUp2_SSE2(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx);
// Row functions.
void ScaleARGBRowDownEven_NEON(const uint8* src_argb, int src_stride,
int src_stepx,
uint8* dst_argb, int dst_width);
void ScaleARGBRowDownEvenBox_NEON(const uint8* src_argb, int src_stride,
int src_stepx,
uint8* dst_argb, int dst_width);
void ScaleARGBRowDown2_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleARGBRowDown2Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
// ScaleRowDown2Box also used by planar functions
// NEON downscalers with interpolation.
// Note - not static due to reuse in convert for 444 to 420.
void ScaleRowDown2_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown2Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown4_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown4Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
// Down scale from 4 to 3 pixels. Use the neon multilane read/write
// to load up the every 4th pixel into a 4 different registers.
// Point samples 32 pixels to 24 pixels.
void ScaleRowDown34_NEON(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown34_0_Box_NEON(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown34_1_Box_NEON(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
// 32 -> 12
void ScaleRowDown38_NEON(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
// 32x3 -> 12x1
void ScaleRowDown38_3_Box_NEON(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
// 32x2 -> 12x1
void ScaleRowDown38_2_Box_NEON(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown2_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown2Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown4_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown4Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown34_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown34_0_Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* d, int dst_width);
void ScaleRowDown34_1_Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* d, int dst_width);
void ScaleRowDown38_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width);
void ScaleRowDown38_2_Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
void ScaleRowDown38_3_Box_MIPS_DSPR2(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width);
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
#endif // INCLUDE_LIBYUV_SCALE_ROW_H_ NOLINT

81
third_party/libyuv/source/cpu_id.c vendored
View File

@ -1,81 +0,0 @@
/*
* Copyright (c) 2011 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 "third_party/libyuv/include/libyuv/cpu_id.h"
#ifdef _MSC_VER
#include <intrin.h>
#endif
#ifdef __ANDROID__
#include <cpu-features.h>
#endif
#include "third_party/libyuv/include/libyuv/basic_types.h" // for CPU_X86
// TODO(fbarchard): Use cpuid.h when gcc 4.4 is used on OSX and Linux.
#if (defined(__pic__) || defined(__APPLE__)) && defined(__i386__)
static inline void __cpuid(int cpu_info[4], int info_type) {
asm volatile (
"mov %%ebx, %%edi \n"
"cpuid \n"
"xchg %%edi, %%ebx \n"
: "=a"(cpu_info[0]), "=D"(cpu_info[1]), "=c"(cpu_info[2]), "=d"(cpu_info[3])
: "a"(info_type)
);
}
#elif defined(__i386__) || defined(__x86_64__)
static inline void __cpuid(int cpu_info[4], int info_type) {
asm volatile (
"cpuid \n"
: "=a"(cpu_info[0]), "=b"(cpu_info[1]), "=c"(cpu_info[2]), "=d"(cpu_info[3])
: "a"(info_type)
);
}
#endif
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// CPU detect function for SIMD instruction sets.
int cpu_info_ = 0;
int InitCpuFlags() {
#ifdef CPU_X86
int cpu_info[4];
__cpuid(cpu_info, 1);
cpu_info_ = (cpu_info[3] & 0x04000000 ? kCpuHasSSE2 : 0) |
(cpu_info[2] & 0x00000200 ? kCpuHasSSSE3 : 0) |
kCpuInitialized;
#elif defined(__ANDROID__) && defined(__ARM_NEON__)
uint64_t features = android_getCpuFeatures();
cpu_info_ = ((features & ANDROID_CPU_ARM_FEATURE_NEON) ? kCpuHasNEON : 0) |
kCpuInitialized;
#elif defined(__ARM_NEON__)
// gcc -mfpu=neon defines __ARM_NEON__
// Enable Neon if you want support for Neon and Arm, and use MaskCpuFlags
// to disable Neon on devices that do not have it.
cpu_info_ = kCpuHasNEON | kCpuInitialized;
#else
cpu_info_ = kCpuInitialized;
#endif
return cpu_info_;
}
void MaskCpuFlags(int enable_flags) {
InitCpuFlags();
cpu_info_ = (cpu_info_ & enable_flags) | kCpuInitialized;
}
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif

283
third_party/libyuv/source/cpu_id.cc vendored Normal file
View File

@ -0,0 +1,283 @@
/*
* Copyright 2011 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 "third_party/libyuv/include/libyuv/cpu_id.h"
#ifdef _MSC_VER
#include <intrin.h> // For __cpuidex()
#endif
#if !defined(__pnacl__) && !defined(__CLR_VER) && \
!defined(__native_client__) && defined(_M_X64) && \
defined(_MSC_VER) && (_MSC_FULL_VER >= 160040219)
#include <immintrin.h> // For _xgetbv()
#endif
#if !defined(__native_client__)
#include <stdlib.h> // For getenv()
#endif
// For ArmCpuCaps() but unittested on all platforms
#include <stdio.h>
#include <string.h>
#include "third_party/libyuv/include/libyuv/basic_types.h" // For CPU_X86
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// For functions that use the stack and have runtime checks for overflow,
// use SAFEBUFFERS to avoid additional check.
#if defined(_MSC_VER) && (_MSC_FULL_VER >= 160040219)
#define SAFEBUFFERS __declspec(safebuffers)
#else
#define SAFEBUFFERS
#endif
// Low level cpuid for X86. Returns zeros on other CPUs.
#if !defined(__pnacl__) && !defined(__CLR_VER) && \
(defined(_M_IX86) || defined(_M_X64) || \
defined(__i386__) || defined(__x86_64__))
LIBYUV_API
void CpuId(uint32 info_eax, uint32 info_ecx, uint32* cpu_info) {
#if defined(_MSC_VER)
#if (_MSC_FULL_VER >= 160040219)
__cpuidex((int*)(cpu_info), info_eax, info_ecx);
#elif defined(_M_IX86)
__asm {
mov eax, info_eax
mov ecx, info_ecx
mov edi, cpu_info
cpuid
mov [edi], eax
mov [edi + 4], ebx
mov [edi + 8], ecx
mov [edi + 12], edx
}
#else
if (info_ecx == 0) {
__cpuid((int*)(cpu_info), info_eax);
} else {
cpu_info[3] = cpu_info[2] = cpu_info[1] = cpu_info[0] = 0;
}
#endif
#else // defined(_MSC_VER)
uint32 info_ebx, info_edx;
asm volatile ( // NOLINT
#if defined( __i386__) && defined(__PIC__)
// Preserve ebx for fpic 32 bit.
"mov %%ebx, %%edi \n"
"cpuid \n"
"xchg %%edi, %%ebx \n"
: "=D" (info_ebx),
#else
"cpuid \n"
: "=b" (info_ebx),
#endif // defined( __i386__) && defined(__PIC__)
"+a" (info_eax), "+c" (info_ecx), "=d" (info_edx));
cpu_info[0] = info_eax;
cpu_info[1] = info_ebx;
cpu_info[2] = info_ecx;
cpu_info[3] = info_edx;
#endif // defined(_MSC_VER)
}
#if !defined(__native_client__)
#define HAS_XGETBV
// X86 CPUs have xgetbv to detect OS saves high parts of ymm registers.
int TestOsSaveYmm() {
uint32 xcr0 = 0u;
#if defined(_MSC_VER) && (_MSC_FULL_VER >= 160040219)
xcr0 = (uint32)(_xgetbv(0)); // VS2010 SP1 required.
#elif defined(_M_IX86)
__asm {
xor ecx, ecx // xcr 0
_asm _emit 0x0f _asm _emit 0x01 _asm _emit 0xd0 // For VS2010 and earlier.
mov xcr0, eax
}
#elif defined(__i386__) || defined(__x86_64__)
asm(".byte 0x0f, 0x01, 0xd0" : "=a" (xcr0) : "c" (0) : "%edx");
#endif // defined(_MSC_VER)
return((xcr0 & 6) == 6); // Is ymm saved?
}
#endif // !defined(__native_client__)
#else
LIBYUV_API
void CpuId(uint32 eax, uint32 ecx, uint32* cpu_info) {
cpu_info[0] = cpu_info[1] = cpu_info[2] = cpu_info[3] = 0;
}
#endif
// based on libvpx arm_cpudetect.c
// For Arm, but public to allow testing on any CPU
LIBYUV_API SAFEBUFFERS
int ArmCpuCaps(const char* cpuinfo_name) {
char cpuinfo_line[512];
FILE* f = fopen(cpuinfo_name, "r");
if (!f) {
// Assume Neon if /proc/cpuinfo is unavailable.
// This will occur for Chrome sandbox for Pepper or Render process.
return kCpuHasNEON;
}
while (fgets(cpuinfo_line, sizeof(cpuinfo_line) - 1, f)) {
if (memcmp(cpuinfo_line, "Features", 8) == 0) {
char* p = strstr(cpuinfo_line, " neon");
if (p && (p[5] == ' ' || p[5] == '\n')) {
fclose(f);
return kCpuHasNEON;
}
}
}
fclose(f);
return 0;
}
#if defined(__mips__) && defined(__linux__)
static int MipsCpuCaps(const char* search_string) {
char cpuinfo_line[512];
const char* file_name = "/proc/cpuinfo";
FILE* f = fopen(file_name, "r");
if (!f) {
// Assume DSP if /proc/cpuinfo is unavailable.
// This will occur for Chrome sandbox for Pepper or Render process.
return kCpuHasMIPS_DSP;
}
while (fgets(cpuinfo_line, sizeof(cpuinfo_line) - 1, f) != NULL) {
if (strstr(cpuinfo_line, search_string) != NULL) {
fclose(f);
return kCpuHasMIPS_DSP;
}
}
fclose(f);
return 0;
}
#endif
// CPU detect function for SIMD instruction sets.
LIBYUV_API
int cpu_info_ = kCpuInit; // cpu_info is not initialized yet.
// Test environment variable for disabling CPU features. Any non-zero value
// to disable. Zero ignored to make it easy to set the variable on/off.
#if !defined(__native_client__) && !defined(_M_ARM)
static LIBYUV_BOOL TestEnv(const char* name) {
const char* var = getenv(name);
if (var) {
if (var[0] != '0') {
return LIBYUV_TRUE;
}
}
return LIBYUV_FALSE;
}
#else // nacl does not support getenv().
static LIBYUV_BOOL TestEnv(const char*) {
return LIBYUV_FALSE;
}
#endif
LIBYUV_API SAFEBUFFERS
int InitCpuFlags(void) {
#if !defined(__pnacl__) && !defined(__CLR_VER) && defined(CPU_X86)
uint32 cpu_info1[4] = { 0, 0, 0, 0 };
uint32 cpu_info7[4] = { 0, 0, 0, 0 };
CpuId(1, 0, cpu_info1);
CpuId(7, 0, cpu_info7);
cpu_info_ = ((cpu_info1[3] & 0x04000000) ? kCpuHasSSE2 : 0) |
((cpu_info1[2] & 0x00000200) ? kCpuHasSSSE3 : 0) |
((cpu_info1[2] & 0x00080000) ? kCpuHasSSE41 : 0) |
((cpu_info1[2] & 0x00100000) ? kCpuHasSSE42 : 0) |
((cpu_info7[1] & 0x00000200) ? kCpuHasERMS : 0) |
((cpu_info1[2] & 0x00001000) ? kCpuHasFMA3 : 0) |
kCpuHasX86;
#ifdef HAS_XGETBV
if ((cpu_info1[2] & 0x18000000) == 0x18000000 && // AVX and OSSave
TestOsSaveYmm()) { // Saves YMM.
cpu_info_ |= ((cpu_info7[1] & 0x00000020) ? kCpuHasAVX2 : 0) |
kCpuHasAVX;
}
#endif
// Environment variable overrides for testing.
if (TestEnv("LIBYUV_DISABLE_X86")) {
cpu_info_ &= ~kCpuHasX86;
}
if (TestEnv("LIBYUV_DISABLE_SSE2")) {
cpu_info_ &= ~kCpuHasSSE2;
}
if (TestEnv("LIBYUV_DISABLE_SSSE3")) {
cpu_info_ &= ~kCpuHasSSSE3;
}
if (TestEnv("LIBYUV_DISABLE_SSE41")) {
cpu_info_ &= ~kCpuHasSSE41;
}
if (TestEnv("LIBYUV_DISABLE_SSE42")) {
cpu_info_ &= ~kCpuHasSSE42;
}
if (TestEnv("LIBYUV_DISABLE_AVX")) {
cpu_info_ &= ~kCpuHasAVX;
}
if (TestEnv("LIBYUV_DISABLE_AVX2")) {
cpu_info_ &= ~kCpuHasAVX2;
}
if (TestEnv("LIBYUV_DISABLE_ERMS")) {
cpu_info_ &= ~kCpuHasERMS;
}
if (TestEnv("LIBYUV_DISABLE_FMA3")) {
cpu_info_ &= ~kCpuHasFMA3;
}
#elif defined(__mips__) && defined(__linux__)
// Linux mips parse text file for dsp detect.
cpu_info_ = MipsCpuCaps("dsp"); // set kCpuHasMIPS_DSP.
#if defined(__mips_dspr2)
cpu_info_ |= kCpuHasMIPS_DSPR2;
#endif
cpu_info_ |= kCpuHasMIPS;
if (getenv("LIBYUV_DISABLE_MIPS")) {
cpu_info_ &= ~kCpuHasMIPS;
}
if (getenv("LIBYUV_DISABLE_MIPS_DSP")) {
cpu_info_ &= ~kCpuHasMIPS_DSP;
}
if (getenv("LIBYUV_DISABLE_MIPS_DSPR2")) {
cpu_info_ &= ~kCpuHasMIPS_DSPR2;
}
#elif defined(__arm__)
// gcc -mfpu=neon defines __ARM_NEON__
// __ARM_NEON__ generates code that requires Neon. NaCL also requires Neon.
// For Linux, /proc/cpuinfo can be tested but without that assume Neon.
#if defined(__ARM_NEON__) || defined(__native_client__) || !defined(__linux__)
cpu_info_ = kCpuHasNEON;
#else
// Linux arm parse text file for neon detect.
cpu_info_ = ArmCpuCaps("/proc/cpuinfo");
#endif
cpu_info_ |= kCpuHasARM;
if (TestEnv("LIBYUV_DISABLE_NEON")) {
cpu_info_ &= ~kCpuHasNEON;
}
#endif // __arm__
if (TestEnv("LIBYUV_DISABLE_ASM")) {
cpu_info_ = 0;
}
return cpu_info_;
}
LIBYUV_API
void MaskCpuFlags(int enable_flags) {
cpu_info_ = InitCpuFlags() & enable_flags;
}
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif

2287
third_party/libyuv/source/planar_functions.cc vendored Normal file
View File

File diff suppressed because it is too large Load Diff

264
third_party/libyuv/source/row.h vendored
View File

@ -1,264 +0,0 @@
/*
* Copyright (c) 2011 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 LIBYUV_SOURCE_ROW_H_
#define LIBYUV_SOURCE_ROW_H_
#include "third_party/libyuv/include/libyuv/basic_types.h"
#define kMaxStride (2048 * 4)
#define IS_ALIGNED(p, a) (!((uintptr_t)(p) & ((a) - 1)))
#if defined(COVERAGE_ENABLED) || defined(TARGET_IPHONE_SIMULATOR)
#define YUV_DISABLE_ASM
#endif
#if defined(__ARM_NEON__) && !defined(YUV_DISABLE_ASM)
#define HAS_FASTCONVERTYUVTOARGBROW_NEON
void FastConvertYUVToARGBRow_NEON(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
#define HAS_FASTCONVERTYUVTOBGRAROW_NEON
void FastConvertYUVToBGRARow_NEON(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
#define HAS_FASTCONVERTYUVTOABGRROW_NEON
void FastConvertYUVToABGRRow_NEON(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
#endif
// The following are available on all x86 platforms
#if (defined(_M_IX86) || defined(__x86_64__) || defined(__i386__)) && \
!defined(YUV_DISABLE_ASM)
#define HAS_ABGRTOARGBROW_SSSE3
#define HAS_BGRATOARGBROW_SSSE3
#define HAS_BG24TOARGBROW_SSSE3
#define HAS_RAWTOARGBROW_SSSE3
#define HAS_RGB24TOYROW_SSSE3
#define HAS_RAWTOYROW_SSSE3
#define HAS_RGB24TOUVROW_SSSE3
#define HAS_RAWTOUVROW_SSSE3
#define HAS_ARGBTOYROW_SSSE3
#define HAS_BGRATOYROW_SSSE3
#define HAS_ABGRTOYROW_SSSE3
#define HAS_ARGBTOUVROW_SSSE3
#define HAS_BGRATOUVROW_SSSE3
#define HAS_ABGRTOUVROW_SSSE3
#define HAS_I400TOARGBROW_SSE2
#define HAS_FASTCONVERTYTOARGBROW_SSE2
#define HAS_FASTCONVERTYUVTOARGBROW_SSSE3
#define HAS_FASTCONVERTYUVTOBGRAROW_SSSE3
#define HAS_FASTCONVERTYUVTOABGRROW_SSSE3
#define HAS_FASTCONVERTYUV444TOARGBROW_SSSE3
#define HAS_REVERSE_ROW_SSSE3
#endif
// The following are available on Neon platforms
#if defined(__ARM_NEON__) && !defined(YUV_DISABLE_ASM)
#define HAS_REVERSE_ROW_NEON
#endif
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
#ifdef HAS_ARGBTOYROW_SSSE3
void ARGBToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix);
void BGRAToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix);
void ABGRToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix);
void ARGBToUVRow_SSSE3(const uint8* src_argb0, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int width);
void BGRAToUVRow_SSSE3(const uint8* src_argb0, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int width);
void ABGRToUVRow_SSSE3(const uint8* src_argb0, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int width);
#endif
#if defined(HAS_BG24TOARGBROW_SSSE3) && defined(HAS_ARGBTOYROW_SSSE3)
#define HASRGB24TOYROW_SSSE3
#endif
#ifdef HASRGB24TOYROW_SSSE3
void RGB24ToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix);
void RAWToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix);
void RGB24ToUVRow_SSSE3(const uint8* src_argb0, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int width);
void RAWToUVRow_SSSE3(const uint8* src_argb0, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int width);
#endif
#ifdef HAS_REVERSE_ROW_SSSE3
void ReverseRow_SSSE3(const uint8* src, uint8* dst, int width);
#endif
#ifdef HAS_REVERSE_ROW_NEON
void ReverseRow_NEON(const uint8* src, uint8* dst, int width);
#endif
void ReverseRow_C(const uint8* src, uint8* dst, int width);
void ARGBToYRow_C(const uint8* src_argb, uint8* dst_y, int pix);
void BGRAToYRow_C(const uint8* src_argb, uint8* dst_y, int pix);
void ABGRToYRow_C(const uint8* src_argb, uint8* dst_y, int pix);
void RGB24ToYRow_C(const uint8* src_argb, uint8* dst_y, int pix);
void RAWToYRow_C(const uint8* src_argb, uint8* dst_y, int pix);
void ARGBToUVRow_C(const uint8* src_argb0, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int width);
void BGRAToUVRow_C(const uint8* src_argb0, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int width);
void ABGRToUVRow_C(const uint8* src_argb0, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int width);
void RGB24ToUVRow_C(const uint8* src_argb0, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int width);
void RAWToUVRow_C(const uint8* src_argb0, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int width);
#ifdef HAS_BG24TOARGBROW_SSSE3
void ABGRToARGBRow_SSSE3(const uint8* src_abgr, uint8* dst_argb, int pix);
void BGRAToARGBRow_SSSE3(const uint8* src_bgra, uint8* dst_argb, int pix);
void BG24ToARGBRow_SSSE3(const uint8* src_bg24, uint8* dst_argb, int pix);
void RAWToARGBRow_SSSE3(const uint8* src_bg24, uint8* dst_argb, int pix);
#endif
void ABGRToARGBRow_C(const uint8* src_abgr, uint8* dst_argb, int pix);
void BGRAToARGBRow_C(const uint8* src_bgra, uint8* dst_argb, int pix);
void BG24ToARGBRow_C(const uint8* src_bg24, uint8* dst_argb, int pix);
void RAWToARGBRow_C(const uint8* src_bg24, uint8* dst_argb, int pix);
#ifdef HAS_I400TOARGBROW_SSE2
void I400ToARGBRow_SSE2(const uint8* src_y, uint8* dst_argb, int pix);
#endif
void I400ToARGBRow_C(const uint8* src_y, uint8* dst_argb, int pix);
#if defined(_MSC_VER)
#define SIMD_ALIGNED(var) __declspec(align(16)) var
typedef __declspec(align(16)) signed char vec8[16];
typedef __declspec(align(16)) unsigned char uvec8[16];
typedef __declspec(align(16)) signed short vec16[8];
#else // __GNUC__
#define SIMD_ALIGNED(var) var __attribute__((aligned(16)))
typedef signed char __attribute__((vector_size(16))) vec8;
typedef unsigned char __attribute__((vector_size(16))) uvec8;
typedef signed short __attribute__((vector_size(16))) vec16;
#endif
//extern "C"
SIMD_ALIGNED(const int16 kCoefficientsRgbY[768][4]);
//extern "C"
SIMD_ALIGNED(const int16 kCoefficientsBgraY[768][4]);
//extern "C"
SIMD_ALIGNED(const int16 kCoefficientsAbgrY[768][4]);
void FastConvertYUVToARGBRow_C(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
void FastConvertYUVToBGRARow_C(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
void FastConvertYUVToABGRRow_C(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
void FastConvertYUV444ToARGBRow_C(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
void FastConvertYToARGBRow_C(const uint8* y_buf,
uint8* rgb_buf,
int width);
#ifdef HAS_FASTCONVERTYUVTOARGBROW_SSE2
void FastConvertYUVToARGBRow_SSE2(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
void FastConvertYUVToARGBRow4_SSE2(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
void FastConvertYUVToBGRARow_SSE2(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
void FastConvertYUVToABGRRow_SSE2(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
void FastConvertYUV444ToARGBRow_SSE2(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
void FastConvertYToARGBRow_SSE2(const uint8* y_buf,
uint8* rgb_buf,
int width);
#endif
#ifdef HAS_FASTCONVERTYUVTOARGBROW_SSSE3
void FastConvertYUVToARGBRow_SSSE3(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
void FastConvertYUVToBGRARow_SSSE3(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
void FastConvertYUVToABGRRow_SSSE3(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
void FastConvertYUV444ToARGBRow_SSSE3(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width);
#endif
#ifdef HAS_FASTCONVERTYTOARGBROW_SSE2
void FastConvertYToARGBRow_SSE2(const uint8* y_buf,
uint8* rgb_buf,
int width);
#endif
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
#endif // LIBYUV_SOURCE_ROW_H_

542
third_party/libyuv/source/row_any.cc vendored Normal file
View File

@ -0,0 +1,542 @@
/*
* 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 "third_party/libyuv/include/libyuv/row.h"
#include "third_party/libyuv/include/libyuv/basic_types.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// TODO(fbarchard): Consider 'any' functions handling any quantity of pixels.
// TODO(fbarchard): Consider 'any' functions handling odd alignment.
// YUV to RGB does multiple of 8 with SIMD and remainder with C.
#define YANY(NAMEANY, I420TORGB_SIMD, I420TORGB_C, UV_SHIFT, BPP, MASK) \
void NAMEANY(const uint8* y_buf, \
const uint8* u_buf, \
const uint8* v_buf, \
uint8* rgb_buf, \
int width) { \
int n = width & ~MASK; \
I420TORGB_SIMD(y_buf, u_buf, v_buf, rgb_buf, n); \
I420TORGB_C(y_buf + n, \
u_buf + (n >> UV_SHIFT), \
v_buf + (n >> UV_SHIFT), \
rgb_buf + n * BPP, width & MASK); \
}
#ifdef HAS_I422TOARGBROW_SSSE3
YANY(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_Unaligned_SSSE3, I444ToARGBRow_C,
0, 4, 7)
YANY(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_Unaligned_SSSE3, I422ToARGBRow_C,
1, 4, 7)
YANY(I411ToARGBRow_Any_SSSE3, I411ToARGBRow_Unaligned_SSSE3, I411ToARGBRow_C,
2, 4, 7)
YANY(I422ToBGRARow_Any_SSSE3, I422ToBGRARow_Unaligned_SSSE3, I422ToBGRARow_C,
1, 4, 7)
YANY(I422ToABGRRow_Any_SSSE3, I422ToABGRRow_Unaligned_SSSE3, I422ToABGRRow_C,
1, 4, 7)
YANY(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_Unaligned_SSSE3, I422ToRGBARow_C,
1, 4, 7)
// I422ToRGB565Row_SSSE3 is unaligned.
YANY(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, I422ToARGB4444Row_C,
1, 2, 7)
YANY(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, I422ToARGB1555Row_C,
1, 2, 7)
YANY(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, I422ToRGB565Row_C,
1, 2, 7)
// I422ToRGB24Row_SSSE3 is unaligned.
YANY(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, I422ToRGB24Row_C, 1, 3, 7)
YANY(I422ToRAWRow_Any_SSSE3, I422ToRAWRow_SSSE3, I422ToRAWRow_C, 1, 3, 7)
YANY(I422ToYUY2Row_Any_SSE2, I422ToYUY2Row_SSE2, I422ToYUY2Row_C, 1, 2, 15)
YANY(I422ToUYVYRow_Any_SSE2, I422ToUYVYRow_SSE2, I422ToUYVYRow_C, 1, 2, 15)
#endif // HAS_I422TOARGBROW_SSSE3
#ifdef HAS_I422TOARGBROW_AVX2
YANY(I422ToARGBRow_Any_AVX2, I422ToARGBRow_AVX2, I422ToARGBRow_C, 1, 4, 15)
#endif // HAS_I422TOARGBROW_AVX2
#ifdef HAS_I422TOARGBROW_NEON
YANY(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, I444ToARGBRow_C, 0, 4, 7)
YANY(I422ToARGBRow_Any_NEON, I422ToARGBRow_NEON, I422ToARGBRow_C, 1, 4, 7)
YANY(I411ToARGBRow_Any_NEON, I411ToARGBRow_NEON, I411ToARGBRow_C, 2, 4, 7)
YANY(I422ToBGRARow_Any_NEON, I422ToBGRARow_NEON, I422ToBGRARow_C, 1, 4, 7)
YANY(I422ToABGRRow_Any_NEON, I422ToABGRRow_NEON, I422ToABGRRow_C, 1, 4, 7)
YANY(I422ToRGBARow_Any_NEON, I422ToRGBARow_NEON, I422ToRGBARow_C, 1, 4, 7)
YANY(I422ToRGB24Row_Any_NEON, I422ToRGB24Row_NEON, I422ToRGB24Row_C, 1, 3, 7)
YANY(I422ToRAWRow_Any_NEON, I422ToRAWRow_NEON, I422ToRAWRow_C, 1, 3, 7)
YANY(I422ToARGB4444Row_Any_NEON, I422ToARGB4444Row_NEON, I422ToARGB4444Row_C,
1, 2, 7)
YANY(I422ToARGB1555Row_Any_NEON, I422ToARGB1555Row_NEON, I422ToARGB1555Row_C,
1, 2, 7)
YANY(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, I422ToRGB565Row_C, 1, 2, 7)
YANY(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, I422ToYUY2Row_C, 1, 2, 15)
YANY(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, I422ToUYVYRow_C, 1, 2, 15)
#endif // HAS_I422TOARGBROW_NEON
#undef YANY
// Wrappers to handle odd width
#define NV2NY(NAMEANY, NV12TORGB_SIMD, NV12TORGB_C, UV_SHIFT, BPP) \
void NAMEANY(const uint8* y_buf, \
const uint8* uv_buf, \
uint8* rgb_buf, \
int width) { \
int n = width & ~7; \
NV12TORGB_SIMD(y_buf, uv_buf, rgb_buf, n); \
NV12TORGB_C(y_buf + n, \
uv_buf + (n >> UV_SHIFT), \
rgb_buf + n * BPP, width & 7); \
}
#ifdef HAS_NV12TOARGBROW_SSSE3
NV2NY(NV12ToARGBRow_Any_SSSE3, NV12ToARGBRow_Unaligned_SSSE3, NV12ToARGBRow_C,
0, 4)
NV2NY(NV21ToARGBRow_Any_SSSE3, NV21ToARGBRow_Unaligned_SSSE3, NV21ToARGBRow_C,
0, 4)
#endif // HAS_NV12TOARGBROW_SSSE3
#ifdef HAS_NV12TOARGBROW_NEON
NV2NY(NV12ToARGBRow_Any_NEON, NV12ToARGBRow_NEON, NV12ToARGBRow_C, 0, 4)
NV2NY(NV21ToARGBRow_Any_NEON, NV21ToARGBRow_NEON, NV21ToARGBRow_C, 0, 4)
#endif // HAS_NV12TOARGBROW_NEON
#ifdef HAS_NV12TORGB565ROW_SSSE3
NV2NY(NV12ToRGB565Row_Any_SSSE3, NV12ToRGB565Row_SSSE3, NV12ToRGB565Row_C,
0, 2)
NV2NY(NV21ToRGB565Row_Any_SSSE3, NV21ToRGB565Row_SSSE3, NV21ToRGB565Row_C,
0, 2)
#endif // HAS_NV12TORGB565ROW_SSSE3
#ifdef HAS_NV12TORGB565ROW_NEON
NV2NY(NV12ToRGB565Row_Any_NEON, NV12ToRGB565Row_NEON, NV12ToRGB565Row_C, 0, 2)
NV2NY(NV21ToRGB565Row_Any_NEON, NV21ToRGB565Row_NEON, NV21ToRGB565Row_C, 0, 2)
#endif // HAS_NV12TORGB565ROW_NEON
#undef NVANY
#define RGBANY(NAMEANY, ARGBTORGB_SIMD, ARGBTORGB_C, MASK, SBPP, BPP) \
void NAMEANY(const uint8* src, \
uint8* dst, \
int width) { \
int n = width & ~MASK; \
ARGBTORGB_SIMD(src, dst, n); \
ARGBTORGB_C(src + n * SBPP, dst + n * BPP, width & MASK); \
}
#if defined(HAS_ARGBTORGB24ROW_SSSE3)
RGBANY(ARGBToRGB24Row_Any_SSSE3, ARGBToRGB24Row_SSSE3, ARGBToRGB24Row_C,
15, 4, 3)
RGBANY(ARGBToRAWRow_Any_SSSE3, ARGBToRAWRow_SSSE3, ARGBToRAWRow_C,
15, 4, 3)
RGBANY(ARGBToRGB565Row_Any_SSE2, ARGBToRGB565Row_SSE2, ARGBToRGB565Row_C,
3, 4, 2)
RGBANY(ARGBToARGB1555Row_Any_SSE2, ARGBToARGB1555Row_SSE2, ARGBToARGB1555Row_C,
3, 4, 2)
RGBANY(ARGBToARGB4444Row_Any_SSE2, ARGBToARGB4444Row_SSE2, ARGBToARGB4444Row_C,
3, 4, 2)
#endif
#if defined(HAS_I400TOARGBROW_SSE2)
RGBANY(I400ToARGBRow_Any_SSE2, I400ToARGBRow_Unaligned_SSE2, I400ToARGBRow_C,
7, 1, 4)
#endif
#if defined(HAS_YTOARGBROW_SSE2)
RGBANY(YToARGBRow_Any_SSE2, YToARGBRow_SSE2, YToARGBRow_C,
7, 1, 4)
RGBANY(YUY2ToARGBRow_Any_SSSE3, YUY2ToARGBRow_Unaligned_SSSE3, YUY2ToARGBRow_C,
15, 2, 4)
RGBANY(UYVYToARGBRow_Any_SSSE3, UYVYToARGBRow_Unaligned_SSSE3, UYVYToARGBRow_C,
15, 2, 4)
// These require alignment on ARGB, so C is used for remainder.
RGBANY(RGB24ToARGBRow_Any_SSSE3, RGB24ToARGBRow_SSSE3, RGB24ToARGBRow_C,
15, 3, 4)
RGBANY(RAWToARGBRow_Any_SSSE3, RAWToARGBRow_SSSE3, RAWToARGBRow_C,
15, 3, 4)
RGBANY(RGB565ToARGBRow_Any_SSE2, RGB565ToARGBRow_SSE2, RGB565ToARGBRow_C,
7, 2, 4)
RGBANY(ARGB1555ToARGBRow_Any_SSE2, ARGB1555ToARGBRow_SSE2, ARGB1555ToARGBRow_C,
7, 2, 4)
RGBANY(ARGB4444ToARGBRow_Any_SSE2, ARGB4444ToARGBRow_SSE2, ARGB4444ToARGBRow_C,
7, 2, 4)
#endif
#if defined(HAS_ARGBTORGB24ROW_NEON)
RGBANY(ARGBToRGB24Row_Any_NEON, ARGBToRGB24Row_NEON, ARGBToRGB24Row_C, 7, 4, 3)
RGBANY(ARGBToRAWRow_Any_NEON, ARGBToRAWRow_NEON, ARGBToRAWRow_C, 7, 4, 3)
RGBANY(ARGBToRGB565Row_Any_NEON, ARGBToRGB565Row_NEON, ARGBToRGB565Row_C,
7, 4, 2)
RGBANY(ARGBToARGB1555Row_Any_NEON, ARGBToARGB1555Row_NEON, ARGBToARGB1555Row_C,
7, 4, 2)
RGBANY(ARGBToARGB4444Row_Any_NEON, ARGBToARGB4444Row_NEON, ARGBToARGB4444Row_C,
7, 4, 2)
RGBANY(I400ToARGBRow_Any_NEON, I400ToARGBRow_NEON, I400ToARGBRow_C,
7, 1, 4)
RGBANY(YToARGBRow_Any_NEON, YToARGBRow_NEON, YToARGBRow_C,
7, 1, 4)
RGBANY(YUY2ToARGBRow_Any_NEON, YUY2ToARGBRow_NEON, YUY2ToARGBRow_C,
7, 2, 4)
RGBANY(UYVYToARGBRow_Any_NEON, UYVYToARGBRow_NEON, UYVYToARGBRow_C,
7, 2, 4)
#endif
#undef RGBANY
// ARGB to Bayer does multiple of 4 pixels, SSSE3 aligned src, unaligned dst.
#define BAYERANY(NAMEANY, ARGBTORGB_SIMD, ARGBTORGB_C, MASK, SBPP, BPP) \
void NAMEANY(const uint8* src, \
uint8* dst, uint32 selector, \
int width) { \
int n = width & ~MASK; \
ARGBTORGB_SIMD(src, dst, selector, n); \
ARGBTORGB_C(src + n * SBPP, dst + n * BPP, selector, width & MASK); \
}
#if defined(HAS_ARGBTOBAYERROW_SSSE3)
BAYERANY(ARGBToBayerRow_Any_SSSE3, ARGBToBayerRow_SSSE3, ARGBToBayerRow_C,
7, 4, 1)
#endif
#if defined(HAS_ARGBTOBAYERROW_NEON)
BAYERANY(ARGBToBayerRow_Any_NEON, ARGBToBayerRow_NEON, ARGBToBayerRow_C,
7, 4, 1)
#endif
#if defined(HAS_ARGBTOBAYERGGROW_SSE2)
BAYERANY(ARGBToBayerGGRow_Any_SSE2, ARGBToBayerGGRow_SSE2, ARGBToBayerGGRow_C,
7, 4, 1)
#endif
#if defined(HAS_ARGBTOBAYERGGROW_NEON)
BAYERANY(ARGBToBayerGGRow_Any_NEON, ARGBToBayerGGRow_NEON, ARGBToBayerGGRow_C,
7, 4, 1)
#endif
#undef BAYERANY
// RGB/YUV to Y does multiple of 16 with SIMD and last 16 with SIMD.
#define YANY(NAMEANY, ARGBTOY_SIMD, SBPP, BPP, NUM) \
void NAMEANY(const uint8* src_argb, uint8* dst_y, int width) { \
ARGBTOY_SIMD(src_argb, dst_y, width - NUM); \
ARGBTOY_SIMD(src_argb + (width - NUM) * SBPP, \
dst_y + (width - NUM) * BPP, NUM); \
}
#ifdef HAS_ARGBTOYROW_AVX2
YANY(ARGBToYRow_Any_AVX2, ARGBToYRow_AVX2, 4, 1, 32)
YANY(ARGBToYJRow_Any_AVX2, ARGBToYJRow_AVX2, 4, 1, 32)
YANY(YUY2ToYRow_Any_AVX2, YUY2ToYRow_AVX2, 2, 1, 32)
YANY(UYVYToYRow_Any_AVX2, UYVYToYRow_AVX2, 2, 1, 32)
#endif
#ifdef HAS_ARGBTOYROW_SSSE3
YANY(ARGBToYRow_Any_SSSE3, ARGBToYRow_Unaligned_SSSE3, 4, 1, 16)
#endif
#ifdef HAS_BGRATOYROW_SSSE3
YANY(BGRAToYRow_Any_SSSE3, BGRAToYRow_Unaligned_SSSE3, 4, 1, 16)
YANY(ABGRToYRow_Any_SSSE3, ABGRToYRow_Unaligned_SSSE3, 4, 1, 16)
YANY(RGBAToYRow_Any_SSSE3, RGBAToYRow_Unaligned_SSSE3, 4, 1, 16)
YANY(YUY2ToYRow_Any_SSE2, YUY2ToYRow_Unaligned_SSE2, 2, 1, 16)
YANY(UYVYToYRow_Any_SSE2, UYVYToYRow_Unaligned_SSE2, 2, 1, 16)
#endif
#ifdef HAS_ARGBTOYJROW_SSSE3
YANY(ARGBToYJRow_Any_SSSE3, ARGBToYJRow_Unaligned_SSSE3, 4, 1, 16)
#endif
#ifdef HAS_ARGBTOYROW_NEON
YANY(ARGBToYRow_Any_NEON, ARGBToYRow_NEON, 4, 1, 8)
YANY(ARGBToYJRow_Any_NEON, ARGBToYJRow_NEON, 4, 1, 8)
YANY(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, 4, 1, 8)
YANY(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, 4, 1, 8)
YANY(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, 4, 1, 8)
YANY(RGB24ToYRow_Any_NEON, RGB24ToYRow_NEON, 3, 1, 8)
YANY(RAWToYRow_Any_NEON, RAWToYRow_NEON, 3, 1, 8)
YANY(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, 2, 1, 8)
YANY(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 2, 1, 8)
YANY(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, 2, 1, 8)
YANY(YUY2ToYRow_Any_NEON, YUY2ToYRow_NEON, 2, 1, 16)
YANY(UYVYToYRow_Any_NEON, UYVYToYRow_NEON, 2, 1, 16)
YANY(RGB24ToARGBRow_Any_NEON, RGB24ToARGBRow_NEON, 3, 4, 8)
YANY(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, 3, 4, 8)
YANY(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, 2, 4, 8)
YANY(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, 2, 4, 8)
YANY(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, 2, 4, 8)
#endif
#undef YANY
#define YANY(NAMEANY, ARGBTOY_SIMD, ARGBTOY_C, SBPP, BPP, MASK) \
void NAMEANY(const uint8* src_argb, uint8* dst_y, int width) { \
int n = width & ~MASK; \
ARGBTOY_SIMD(src_argb, dst_y, n); \
ARGBTOY_C(src_argb + n * SBPP, \
dst_y + n * BPP, width & MASK); \
}
// Attenuate is destructive so last16 method can not be used due to overlap.
#ifdef HAS_ARGBATTENUATEROW_SSSE3
YANY(ARGBAttenuateRow_Any_SSSE3, ARGBAttenuateRow_SSSE3, ARGBAttenuateRow_C,
4, 4, 3)
#endif
#ifdef HAS_ARGBATTENUATEROW_SSE2
YANY(ARGBAttenuateRow_Any_SSE2, ARGBAttenuateRow_SSE2, ARGBAttenuateRow_C,
4, 4, 3)
#endif
#ifdef HAS_ARGBUNATTENUATEROW_SSE2
YANY(ARGBUnattenuateRow_Any_SSE2, ARGBUnattenuateRow_SSE2, ARGBUnattenuateRow_C,
4, 4, 3)
#endif
#ifdef HAS_ARGBATTENUATEROW_AVX2
YANY(ARGBAttenuateRow_Any_AVX2, ARGBAttenuateRow_AVX2, ARGBAttenuateRow_C,
4, 4, 7)
#endif
#ifdef HAS_ARGBUNATTENUATEROW_AVX2
YANY(ARGBUnattenuateRow_Any_AVX2, ARGBUnattenuateRow_AVX2, ARGBUnattenuateRow_C,
4, 4, 7)
#endif
#ifdef HAS_ARGBATTENUATEROW_NEON
YANY(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, ARGBAttenuateRow_C,
4, 4, 7)
#endif
#undef YANY
// RGB/YUV to UV does multiple of 16 with SIMD and remainder with C.
#define UVANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, BPP, MASK) \
void NAMEANY(const uint8* src_argb, int src_stride_argb, \
uint8* dst_u, uint8* dst_v, int width) { \
int n = width & ~MASK; \
ANYTOUV_SIMD(src_argb, src_stride_argb, dst_u, dst_v, n); \
ANYTOUV_C(src_argb + n * BPP, src_stride_argb, \
dst_u + (n >> 1), \
dst_v + (n >> 1), \
width & MASK); \
}
#ifdef HAS_ARGBTOUVROW_AVX2
UVANY(ARGBToUVRow_Any_AVX2, ARGBToUVRow_AVX2, ARGBToUVRow_C, 4, 31)
UVANY(YUY2ToUVRow_Any_AVX2, YUY2ToUVRow_AVX2, YUY2ToUVRow_C, 2, 31)
UVANY(UYVYToUVRow_Any_AVX2, UYVYToUVRow_AVX2, UYVYToUVRow_C, 2, 31)
#endif
#ifdef HAS_ARGBTOUVROW_SSSE3
UVANY(ARGBToUVRow_Any_SSSE3, ARGBToUVRow_Unaligned_SSSE3, ARGBToUVRow_C, 4, 15)
UVANY(ARGBToUVJRow_Any_SSSE3, ARGBToUVJRow_Unaligned_SSSE3, ARGBToUVJRow_C,
4, 15)
UVANY(BGRAToUVRow_Any_SSSE3, BGRAToUVRow_Unaligned_SSSE3, BGRAToUVRow_C, 4, 15)
UVANY(ABGRToUVRow_Any_SSSE3, ABGRToUVRow_Unaligned_SSSE3, ABGRToUVRow_C, 4, 15)
UVANY(RGBAToUVRow_Any_SSSE3, RGBAToUVRow_Unaligned_SSSE3, RGBAToUVRow_C, 4, 15)
UVANY(YUY2ToUVRow_Any_SSE2, YUY2ToUVRow_Unaligned_SSE2, YUY2ToUVRow_C, 2, 15)
UVANY(UYVYToUVRow_Any_SSE2, UYVYToUVRow_Unaligned_SSE2, UYVYToUVRow_C, 2, 15)
#endif
#ifdef HAS_ARGBTOUVROW_NEON
UVANY(ARGBToUVRow_Any_NEON, ARGBToUVRow_NEON, ARGBToUVRow_C, 4, 15)
UVANY(ARGBToUVJRow_Any_NEON, ARGBToUVJRow_NEON, ARGBToUVJRow_C, 4, 15)
UVANY(BGRAToUVRow_Any_NEON, BGRAToUVRow_NEON, BGRAToUVRow_C, 4, 15)
UVANY(ABGRToUVRow_Any_NEON, ABGRToUVRow_NEON, ABGRToUVRow_C, 4, 15)
UVANY(RGBAToUVRow_Any_NEON, RGBAToUVRow_NEON, RGBAToUVRow_C, 4, 15)
UVANY(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, RGB24ToUVRow_C, 3, 15)
UVANY(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, RAWToUVRow_C, 3, 15)
UVANY(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, RGB565ToUVRow_C, 2, 15)
UVANY(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, ARGB1555ToUVRow_C, 2, 15)
UVANY(ARGB4444ToUVRow_Any_NEON, ARGB4444ToUVRow_NEON, ARGB4444ToUVRow_C, 2, 15)
UVANY(YUY2ToUVRow_Any_NEON, YUY2ToUVRow_NEON, YUY2ToUVRow_C, 2, 15)
UVANY(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, UYVYToUVRow_C, 2, 15)
#endif
#undef UVANY
#define UV422ANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, BPP, MASK, SHIFT) \
void NAMEANY(const uint8* src_uv, \
uint8* dst_u, uint8* dst_v, int width) { \
int n = width & ~MASK; \
ANYTOUV_SIMD(src_uv, dst_u, dst_v, n); \
ANYTOUV_C(src_uv + n * BPP, \
dst_u + (n >> SHIFT), \
dst_v + (n >> SHIFT), \
width & MASK); \
}
#ifdef HAS_ARGBTOUV444ROW_SSSE3
UV422ANY(ARGBToUV444Row_Any_SSSE3, ARGBToUV444Row_Unaligned_SSSE3,
ARGBToUV444Row_C, 4, 15, 0)
#endif
#ifdef HAS_YUY2TOUV422ROW_AVX2
UV422ANY(YUY2ToUV422Row_Any_AVX2, YUY2ToUV422Row_AVX2,
YUY2ToUV422Row_C, 2, 31, 1)
UV422ANY(UYVYToUV422Row_Any_AVX2, UYVYToUV422Row_AVX2,
UYVYToUV422Row_C, 2, 31, 1)
#endif
#ifdef HAS_ARGBTOUVROW_SSSE3
UV422ANY(ARGBToUV422Row_Any_SSSE3, ARGBToUV422Row_Unaligned_SSSE3,
ARGBToUV422Row_C, 4, 15, 1)
UV422ANY(YUY2ToUV422Row_Any_SSE2, YUY2ToUV422Row_Unaligned_SSE2,
YUY2ToUV422Row_C, 2, 15, 1)
UV422ANY(UYVYToUV422Row_Any_SSE2, UYVYToUV422Row_Unaligned_SSE2,
UYVYToUV422Row_C, 2, 15, 1)
#endif
#ifdef HAS_YUY2TOUV422ROW_NEON
UV422ANY(ARGBToUV444Row_Any_NEON, ARGBToUV444Row_NEON,
ARGBToUV444Row_C, 4, 7, 0)
UV422ANY(ARGBToUV422Row_Any_NEON, ARGBToUV422Row_NEON,
ARGBToUV422Row_C, 4, 15, 1)
UV422ANY(ARGBToUV411Row_Any_NEON, ARGBToUV411Row_NEON,
ARGBToUV411Row_C, 4, 31, 2)
UV422ANY(YUY2ToUV422Row_Any_NEON, YUY2ToUV422Row_NEON,
YUY2ToUV422Row_C, 2, 15, 1)
UV422ANY(UYVYToUV422Row_Any_NEON, UYVYToUV422Row_NEON,
UYVYToUV422Row_C, 2, 15, 1)
#endif
#undef UV422ANY
#define SPLITUVROWANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, MASK) \
void NAMEANY(const uint8* src_uv, \
uint8* dst_u, uint8* dst_v, int width) { \
int n = width & ~MASK; \
ANYTOUV_SIMD(src_uv, dst_u, dst_v, n); \
ANYTOUV_C(src_uv + n * 2, \
dst_u + n, \
dst_v + n, \
width & MASK); \
}
#ifdef HAS_SPLITUVROW_SSE2
SPLITUVROWANY(SplitUVRow_Any_SSE2, SplitUVRow_Unaligned_SSE2, SplitUVRow_C, 15)
#endif
#ifdef HAS_SPLITUVROW_AVX2
SPLITUVROWANY(SplitUVRow_Any_AVX2, SplitUVRow_AVX2, SplitUVRow_C, 31)
#endif
#ifdef HAS_SPLITUVROW_NEON
SPLITUVROWANY(SplitUVRow_Any_NEON, SplitUVRow_NEON, SplitUVRow_C, 15)
#endif
#ifdef HAS_SPLITUVROW_MIPS_DSPR2
SPLITUVROWANY(SplitUVRow_Any_MIPS_DSPR2, SplitUVRow_Unaligned_MIPS_DSPR2,
SplitUVRow_C, 15)
#endif
#undef SPLITUVROWANY
#define MERGEUVROW_ANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, MASK) \
void NAMEANY(const uint8* src_u, const uint8* src_v, \
uint8* dst_uv, int width) { \
int n = width & ~MASK; \
ANYTOUV_SIMD(src_u, src_v, dst_uv, n); \
ANYTOUV_C(src_u + n, \
src_v + n, \
dst_uv + n * 2, \
width & MASK); \
}
#ifdef HAS_MERGEUVROW_SSE2
MERGEUVROW_ANY(MergeUVRow_Any_SSE2, MergeUVRow_Unaligned_SSE2, MergeUVRow_C, 15)
#endif
#ifdef HAS_MERGEUVROW_AVX2
MERGEUVROW_ANY(MergeUVRow_Any_AVX2, MergeUVRow_AVX2, MergeUVRow_C, 31)
#endif
#ifdef HAS_MERGEUVROW_NEON
MERGEUVROW_ANY(MergeUVRow_Any_NEON, MergeUVRow_NEON, MergeUVRow_C, 15)
#endif
#undef MERGEUVROW_ANY
#define MATHROW_ANY(NAMEANY, ARGBMATH_SIMD, ARGBMATH_C, MASK) \
void NAMEANY(const uint8* src_argb0, const uint8* src_argb1, \
uint8* dst_argb, int width) { \
int n = width & ~MASK; \
ARGBMATH_SIMD(src_argb0, src_argb1, dst_argb, n); \
ARGBMATH_C(src_argb0 + n * 4, \
src_argb1 + n * 4, \
dst_argb + n * 4, \
width & MASK); \
}
#ifdef HAS_ARGBMULTIPLYROW_SSE2
MATHROW_ANY(ARGBMultiplyRow_Any_SSE2, ARGBMultiplyRow_SSE2, ARGBMultiplyRow_C,
3)
#endif
#ifdef HAS_ARGBADDROW_SSE2
MATHROW_ANY(ARGBAddRow_Any_SSE2, ARGBAddRow_SSE2, ARGBAddRow_C, 3)
#endif
#ifdef HAS_ARGBSUBTRACTROW_SSE2
MATHROW_ANY(ARGBSubtractRow_Any_SSE2, ARGBSubtractRow_SSE2, ARGBSubtractRow_C,
3)
#endif
#ifdef HAS_ARGBMULTIPLYROW_AVX2
MATHROW_ANY(ARGBMultiplyRow_Any_AVX2, ARGBMultiplyRow_AVX2, ARGBMultiplyRow_C,
7)
#endif
#ifdef HAS_ARGBADDROW_AVX2
MATHROW_ANY(ARGBAddRow_Any_AVX2, ARGBAddRow_AVX2, ARGBAddRow_C, 7)
#endif
#ifdef HAS_ARGBSUBTRACTROW_AVX2
MATHROW_ANY(ARGBSubtractRow_Any_AVX2, ARGBSubtractRow_AVX2, ARGBSubtractRow_C,
7)
#endif
#ifdef HAS_ARGBMULTIPLYROW_NEON
MATHROW_ANY(ARGBMultiplyRow_Any_NEON, ARGBMultiplyRow_NEON, ARGBMultiplyRow_C,
7)
#endif
#ifdef HAS_ARGBADDROW_NEON
MATHROW_ANY(ARGBAddRow_Any_NEON, ARGBAddRow_NEON, ARGBAddRow_C, 7)
#endif
#ifdef HAS_ARGBSUBTRACTROW_NEON
MATHROW_ANY(ARGBSubtractRow_Any_NEON, ARGBSubtractRow_NEON, ARGBSubtractRow_C,
7)
#endif
#undef MATHROW_ANY
// Shuffle may want to work in place, so last16 method can not be used.
#define YANY(NAMEANY, ARGBTOY_SIMD, ARGBTOY_C, SBPP, BPP, MASK) \
void NAMEANY(const uint8* src_argb, uint8* dst_argb, \
const uint8* shuffler, int width) { \
int n = width & ~MASK; \
ARGBTOY_SIMD(src_argb, dst_argb, shuffler, n); \
ARGBTOY_C(src_argb + n * SBPP, \
dst_argb + n * BPP, shuffler, width & MASK); \
}
#ifdef HAS_ARGBSHUFFLEROW_SSE2
YANY(ARGBShuffleRow_Any_SSE2, ARGBShuffleRow_SSE2,
ARGBShuffleRow_C, 4, 4, 3)
#endif
#ifdef HAS_ARGBSHUFFLEROW_SSSE3
YANY(ARGBShuffleRow_Any_SSSE3, ARGBShuffleRow_Unaligned_SSSE3,
ARGBShuffleRow_C, 4, 4, 7)
#endif
#ifdef HAS_ARGBSHUFFLEROW_AVX2
YANY(ARGBShuffleRow_Any_AVX2, ARGBShuffleRow_AVX2,
ARGBShuffleRow_C, 4, 4, 15)
#endif
#ifdef HAS_ARGBSHUFFLEROW_NEON
YANY(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON,
ARGBShuffleRow_C, 4, 4, 3)
#endif
#undef YANY
// Interpolate may want to work in place, so last16 method can not be used.
#define NANY(NAMEANY, TERP_SIMD, TERP_C, SBPP, BPP, MASK) \
void NAMEANY(uint8* dst_ptr, const uint8* src_ptr, \
ptrdiff_t src_stride_ptr, int width, \
int source_y_fraction) { \
int n = width & ~MASK; \
TERP_SIMD(dst_ptr, src_ptr, src_stride_ptr, \
n, source_y_fraction); \
TERP_C(dst_ptr + n * BPP, \
src_ptr + n * SBPP, src_stride_ptr, \
width & MASK, source_y_fraction); \
}
#ifdef HAS_INTERPOLATEROW_AVX2
NANY(InterpolateRow_Any_AVX2, InterpolateRow_AVX2,
InterpolateRow_C, 1, 1, 32)
#endif
#ifdef HAS_INTERPOLATEROW_SSSE3
NANY(InterpolateRow_Any_SSSE3, InterpolateRow_Unaligned_SSSE3,
InterpolateRow_C, 1, 1, 15)
#endif
#ifdef HAS_INTERPOLATEROW_SSE2
NANY(InterpolateRow_Any_SSE2, InterpolateRow_Unaligned_SSE2,
InterpolateRow_C, 1, 1, 15)
#endif
#ifdef HAS_INTERPOLATEROW_NEON
NANY(InterpolateRow_Any_NEON, InterpolateRow_NEON,
InterpolateRow_C, 1, 1, 15)
#endif
#ifdef HAS_INTERPOLATEROW_MIPS_DSPR2
NANY(InterpolateRow_Any_MIPS_DSPR2, InterpolateRow_MIPS_DSPR2,
InterpolateRow_C, 1, 1, 3)
#endif
#undef NANY
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif

2286
third_party/libyuv/source/row_common.cc vendored Normal file
View File

File diff suppressed because it is too large Load Diff

991
third_party/libyuv/source/row_mips.cc vendored Normal file
View File

@ -0,0 +1,991 @@
/*
* Copyright (c) 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 "third_party/libyuv/include/libyuv/row.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// The following are available on Mips platforms:
#if !defined(LIBYUV_DISABLE_MIPS) && defined(__mips__)
#ifdef HAS_COPYROW_MIPS
void CopyRow_MIPS(const uint8* src, uint8* dst, int count) {
__asm__ __volatile__ (
".set noreorder \n"
".set noat \n"
"slti $at, %[count], 8 \n"
"bne $at ,$zero, $last8 \n"
"xor $t8, %[src], %[dst] \n"
"andi $t8, $t8, 0x3 \n"
"bne $t8, $zero, unaligned \n"
"negu $a3, %[dst] \n"
// make dst/src aligned
"andi $a3, $a3, 0x3 \n"
"beq $a3, $zero, $chk16w \n"
// word-aligned now count is the remining bytes count
"subu %[count], %[count], $a3 \n"
"lwr $t8, 0(%[src]) \n"
"addu %[src], %[src], $a3 \n"
"swr $t8, 0(%[dst]) \n"
"addu %[dst], %[dst], $a3 \n"
// Now the dst/src are mutually word-aligned with word-aligned addresses
"$chk16w: \n"
"andi $t8, %[count], 0x3f \n" // whole 64-B chunks?
// t8 is the byte count after 64-byte chunks
"beq %[count], $t8, chk8w \n"
// There will be at most 1 32-byte chunk after it
"subu $a3, %[count], $t8 \n" // the reminder
// Here a3 counts bytes in 16w chunks
"addu $a3, %[dst], $a3 \n"
// Now a3 is the final dst after 64-byte chunks
"addu $t0, %[dst], %[count] \n"
// t0 is the "past the end" address
// When in the loop we exercise "pref 30,x(a1)", the a1+x should not be past
// the "t0-32" address
// This means: for x=128 the last "safe" a1 address is "t0-160"
// Alternatively, for x=64 the last "safe" a1 address is "t0-96"
// we will use "pref 30,128(a1)", so "t0-160" is the limit
"subu $t9, $t0, 160 \n"
// t9 is the "last safe pref 30,128(a1)" address
"pref 0, 0(%[src]) \n" // first line of src
"pref 0, 32(%[src]) \n" // second line of src
"pref 0, 64(%[src]) \n"
"pref 30, 32(%[dst]) \n"
// In case the a1 > t9 don't use "pref 30" at all
"sgtu $v1, %[dst], $t9 \n"
"bgtz $v1, $loop16w \n"
"nop \n"
// otherwise, start with using pref30
"pref 30, 64(%[dst]) \n"
"$loop16w: \n"
"pref 0, 96(%[src]) \n"
"lw $t0, 0(%[src]) \n"
"bgtz $v1, $skip_pref30_96 \n" // skip
"lw $t1, 4(%[src]) \n"
"pref 30, 96(%[dst]) \n" // continue
"$skip_pref30_96: \n"
"lw $t2, 8(%[src]) \n"
"lw $t3, 12(%[src]) \n"
"lw $t4, 16(%[src]) \n"
"lw $t5, 20(%[src]) \n"
"lw $t6, 24(%[src]) \n"
"lw $t7, 28(%[src]) \n"
"pref 0, 128(%[src]) \n"
// bring the next lines of src, addr 128
"sw $t0, 0(%[dst]) \n"
"sw $t1, 4(%[dst]) \n"
"sw $t2, 8(%[dst]) \n"
"sw $t3, 12(%[dst]) \n"
"sw $t4, 16(%[dst]) \n"
"sw $t5, 20(%[dst]) \n"
"sw $t6, 24(%[dst]) \n"
"sw $t7, 28(%[dst]) \n"
"lw $t0, 32(%[src]) \n"
"bgtz $v1, $skip_pref30_128 \n" // skip pref 30,128(a1)
"lw $t1, 36(%[src]) \n"
"pref 30, 128(%[dst]) \n" // set dest, addr 128
"$skip_pref30_128: \n"
"lw $t2, 40(%[src]) \n"
"lw $t3, 44(%[src]) \n"
"lw $t4, 48(%[src]) \n"
"lw $t5, 52(%[src]) \n"
"lw $t6, 56(%[src]) \n"
"lw $t7, 60(%[src]) \n"
"pref 0, 160(%[src]) \n"
// bring the next lines of src, addr 160
"sw $t0, 32(%[dst]) \n"
"sw $t1, 36(%[dst]) \n"
"sw $t2, 40(%[dst]) \n"
"sw $t3, 44(%[dst]) \n"
"sw $t4, 48(%[dst]) \n"
"sw $t5, 52(%[dst]) \n"
"sw $t6, 56(%[dst]) \n"
"sw $t7, 60(%[dst]) \n"
"addiu %[dst], %[dst], 64 \n" // adding 64 to dest
"sgtu $v1, %[dst], $t9 \n"
"bne %[dst], $a3, $loop16w \n"
" addiu %[src], %[src], 64 \n" // adding 64 to src
"move %[count], $t8 \n"
// Here we have src and dest word-aligned but less than 64-bytes to go
"chk8w: \n"
"pref 0, 0x0(%[src]) \n"
"andi $t8, %[count], 0x1f \n" // 32-byte chunk?
// the t8 is the reminder count past 32-bytes
"beq %[count], $t8, chk1w \n"
// count=t8,no 32-byte chunk
" nop \n"
"lw $t0, 0(%[src]) \n"
"lw $t1, 4(%[src]) \n"
"lw $t2, 8(%[src]) \n"
"lw $t3, 12(%[src]) \n"
"lw $t4, 16(%[src]) \n"
"lw $t5, 20(%[src]) \n"
"lw $t6, 24(%[src]) \n"
"lw $t7, 28(%[src]) \n"
"addiu %[src], %[src], 32 \n"
"sw $t0, 0(%[dst]) \n"
"sw $t1, 4(%[dst]) \n"
"sw $t2, 8(%[dst]) \n"
"sw $t3, 12(%[dst]) \n"
"sw $t4, 16(%[dst]) \n"
"sw $t5, 20(%[dst]) \n"
"sw $t6, 24(%[dst]) \n"
"sw $t7, 28(%[dst]) \n"
"addiu %[dst], %[dst], 32 \n"
"chk1w: \n"
"andi %[count], $t8, 0x3 \n"
// now count is the reminder past 1w chunks
"beq %[count], $t8, $last8 \n"
" subu $a3, $t8, %[count] \n"
// a3 is count of bytes in 1w chunks
"addu $a3, %[dst], $a3 \n"
// now a3 is the dst address past the 1w chunks
// copying in words (4-byte chunks)
"$wordCopy_loop: \n"
"lw $t3, 0(%[src]) \n"
// the first t3 may be equal t0 ... optimize?
"addiu %[src], %[src],4 \n"
"addiu %[dst], %[dst],4 \n"
"bne %[dst], $a3,$wordCopy_loop \n"
" sw $t3, -4(%[dst]) \n"
// For the last (<8) bytes
"$last8: \n"
"blez %[count], leave \n"
" addu $a3, %[dst], %[count] \n" // a3 -last dst address
"$last8loop: \n"
"lb $v1, 0(%[src]) \n"
"addiu %[src], %[src], 1 \n"
"addiu %[dst], %[dst], 1 \n"
"bne %[dst], $a3, $last8loop \n"
" sb $v1, -1(%[dst]) \n"
"leave: \n"
" j $ra \n"
" nop \n"
//
// UNALIGNED case
//
"unaligned: \n"
// got here with a3="negu a1"
"andi $a3, $a3, 0x3 \n" // a1 is word aligned?
"beqz $a3, $ua_chk16w \n"
" subu %[count], %[count], $a3 \n"
// bytes left after initial a3 bytes
"lwr $v1, 0(%[src]) \n"
"lwl $v1, 3(%[src]) \n"
"addu %[src], %[src], $a3 \n" // a3 may be 1, 2 or 3
"swr $v1, 0(%[dst]) \n"
"addu %[dst], %[dst], $a3 \n"
// below the dst will be word aligned (NOTE1)
"$ua_chk16w: \n"
"andi $t8, %[count], 0x3f \n" // whole 64-B chunks?
// t8 is the byte count after 64-byte chunks
"beq %[count], $t8, ua_chk8w \n"
// if a2==t8, no 64-byte chunks
// There will be at most 1 32-byte chunk after it
"subu $a3, %[count], $t8 \n" // the reminder
// Here a3 counts bytes in 16w chunks
"addu $a3, %[dst], $a3 \n"
// Now a3 is the final dst after 64-byte chunks
"addu $t0, %[dst], %[count] \n" // t0 "past the end"
"subu $t9, $t0, 160 \n"
// t9 is the "last safe pref 30,128(a1)" address
"pref 0, 0(%[src]) \n" // first line of src
"pref 0, 32(%[src]) \n" // second line addr 32
"pref 0, 64(%[src]) \n"
"pref 30, 32(%[dst]) \n"
// safe, as we have at least 64 bytes ahead
// In case the a1 > t9 don't use "pref 30" at all
"sgtu $v1, %[dst], $t9 \n"
"bgtz $v1, $ua_loop16w \n"
// skip "pref 30,64(a1)" for too short arrays
" nop \n"
// otherwise, start with using pref30
"pref 30, 64(%[dst]) \n"
"$ua_loop16w: \n"
"pref 0, 96(%[src]) \n"
"lwr $t0, 0(%[src]) \n"
"lwl $t0, 3(%[src]) \n"
"lwr $t1, 4(%[src]) \n"
"bgtz $v1, $ua_skip_pref30_96 \n"
" lwl $t1, 7(%[src]) \n"
"pref 30, 96(%[dst]) \n"
// continue setting up the dest, addr 96
"$ua_skip_pref30_96: \n"
"lwr $t2, 8(%[src]) \n"
"lwl $t2, 11(%[src]) \n"
"lwr $t3, 12(%[src]) \n"
"lwl $t3, 15(%[src]) \n"
"lwr $t4, 16(%[src]) \n"
"lwl $t4, 19(%[src]) \n"
"lwr $t5, 20(%[src]) \n"
"lwl $t5, 23(%[src]) \n"
"lwr $t6, 24(%[src]) \n"
"lwl $t6, 27(%[src]) \n"
"lwr $t7, 28(%[src]) \n"
"lwl $t7, 31(%[src]) \n"
"pref 0, 128(%[src]) \n"
// bring the next lines of src, addr 128
"sw $t0, 0(%[dst]) \n"
"sw $t1, 4(%[dst]) \n"
"sw $t2, 8(%[dst]) \n"
"sw $t3, 12(%[dst]) \n"
"sw $t4, 16(%[dst]) \n"
"sw $t5, 20(%[dst]) \n"
"sw $t6, 24(%[dst]) \n"
"sw $t7, 28(%[dst]) \n"
"lwr $t0, 32(%[src]) \n"
"lwl $t0, 35(%[src]) \n"
"lwr $t1, 36(%[src]) \n"
"bgtz $v1, ua_skip_pref30_128 \n"
" lwl $t1, 39(%[src]) \n"
"pref 30, 128(%[dst]) \n"
// continue setting up the dest, addr 128
"ua_skip_pref30_128: \n"
"lwr $t2, 40(%[src]) \n"
"lwl $t2, 43(%[src]) \n"
"lwr $t3, 44(%[src]) \n"
"lwl $t3, 47(%[src]) \n"
"lwr $t4, 48(%[src]) \n"
"lwl $t4, 51(%[src]) \n"
"lwr $t5, 52(%[src]) \n"
"lwl $t5, 55(%[src]) \n"
"lwr $t6, 56(%[src]) \n"
"lwl $t6, 59(%[src]) \n"
"lwr $t7, 60(%[src]) \n"
"lwl $t7, 63(%[src]) \n"
"pref 0, 160(%[src]) \n"
// bring the next lines of src, addr 160
"sw $t0, 32(%[dst]) \n"
"sw $t1, 36(%[dst]) \n"
"sw $t2, 40(%[dst]) \n"
"sw $t3, 44(%[dst]) \n"
"sw $t4, 48(%[dst]) \n"
"sw $t5, 52(%[dst]) \n"
"sw $t6, 56(%[dst]) \n"
"sw $t7, 60(%[dst]) \n"
"addiu %[dst],%[dst],64 \n" // adding 64 to dest
"sgtu $v1,%[dst],$t9 \n"
"bne %[dst],$a3,$ua_loop16w \n"
" addiu %[src],%[src],64 \n" // adding 64 to src
"move %[count],$t8 \n"
// Here we have src and dest word-aligned but less than 64-bytes to go
"ua_chk8w: \n"
"pref 0, 0x0(%[src]) \n"
"andi $t8, %[count], 0x1f \n" // 32-byte chunk?
// the t8 is the reminder count
"beq %[count], $t8, $ua_chk1w \n"
// when count==t8, no 32-byte chunk
"lwr $t0, 0(%[src]) \n"
"lwl $t0, 3(%[src]) \n"
"lwr $t1, 4(%[src]) \n"
"lwl $t1, 7(%[src]) \n"
"lwr $t2, 8(%[src]) \n"
"lwl $t2, 11(%[src]) \n"
"lwr $t3, 12(%[src]) \n"
"lwl $t3, 15(%[src]) \n"
"lwr $t4, 16(%[src]) \n"
"lwl $t4, 19(%[src]) \n"
"lwr $t5, 20(%[src]) \n"
"lwl $t5, 23(%[src]) \n"
"lwr $t6, 24(%[src]) \n"
"lwl $t6, 27(%[src]) \n"
"lwr $t7, 28(%[src]) \n"
"lwl $t7, 31(%[src]) \n"
"addiu %[src], %[src], 32 \n"
"sw $t0, 0(%[dst]) \n"
"sw $t1, 4(%[dst]) \n"
"sw $t2, 8(%[dst]) \n"
"sw $t3, 12(%[dst]) \n"
"sw $t4, 16(%[dst]) \n"
"sw $t5, 20(%[dst]) \n"
"sw $t6, 24(%[dst]) \n"
"sw $t7, 28(%[dst]) \n"
"addiu %[dst], %[dst], 32 \n"
"$ua_chk1w: \n"
"andi %[count], $t8, 0x3 \n"
// now count is the reminder past 1w chunks
"beq %[count], $t8, ua_smallCopy \n"
"subu $a3, $t8, %[count] \n"
// a3 is count of bytes in 1w chunks
"addu $a3, %[dst], $a3 \n"
// now a3 is the dst address past the 1w chunks
// copying in words (4-byte chunks)
"$ua_wordCopy_loop: \n"
"lwr $v1, 0(%[src]) \n"
"lwl $v1, 3(%[src]) \n"
"addiu %[src], %[src], 4 \n"
"addiu %[dst], %[dst], 4 \n"
// note: dst=a1 is word aligned here, see NOTE1
"bne %[dst], $a3, $ua_wordCopy_loop \n"
" sw $v1,-4(%[dst]) \n"
// Now less than 4 bytes (value in count) left to copy
"ua_smallCopy: \n"
"beqz %[count], leave \n"
" addu $a3, %[dst], %[count] \n" // a3 = last dst address
"$ua_smallCopy_loop: \n"
"lb $v1, 0(%[src]) \n"
"addiu %[src], %[src], 1 \n"
"addiu %[dst], %[dst], 1 \n"
"bne %[dst],$a3,$ua_smallCopy_loop \n"
" sb $v1, -1(%[dst]) \n"
"j $ra \n"
" nop \n"
".set at \n"
".set reorder \n"
: [dst] "+r" (dst), [src] "+r" (src)
: [count] "r" (count)
: "t0", "t1", "t2", "t3", "t4", "t5", "t6", "t7",
"t8", "t9", "a3", "v1", "at"
);
}
#endif // HAS_COPYROW_MIPS
// MIPS DSPR2 functions
#if !defined(LIBYUV_DISABLE_MIPS) && defined(__mips_dsp) && \
(__mips_dsp_rev >= 2)
void SplitUVRow_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
int width) {
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"srl $t4, %[width], 4 \n" // multiplies of 16
"blez $t4, 2f \n"
" andi %[width], %[width], 0xf \n" // residual
".p2align 2 \n"
"1: \n"
"addiu $t4, $t4, -1 \n"
"lw $t0, 0(%[src_uv]) \n" // V1 | U1 | V0 | U0
"lw $t1, 4(%[src_uv]) \n" // V3 | U3 | V2 | U2
"lw $t2, 8(%[src_uv]) \n" // V5 | U5 | V4 | U4
"lw $t3, 12(%[src_uv]) \n" // V7 | U7 | V6 | U6
"lw $t5, 16(%[src_uv]) \n" // V9 | U9 | V8 | U8
"lw $t6, 20(%[src_uv]) \n" // V11 | U11 | V10 | U10
"lw $t7, 24(%[src_uv]) \n" // V13 | U13 | V12 | U12
"lw $t8, 28(%[src_uv]) \n" // V15 | U15 | V14 | U14
"addiu %[src_uv], %[src_uv], 32 \n"
"precrq.qb.ph $t9, $t1, $t0 \n" // V3 | V2 | V1 | V0
"precr.qb.ph $t0, $t1, $t0 \n" // U3 | U2 | U1 | U0
"precrq.qb.ph $t1, $t3, $t2 \n" // V7 | V6 | V5 | V4
"precr.qb.ph $t2, $t3, $t2 \n" // U7 | U6 | U5 | U4
"precrq.qb.ph $t3, $t6, $t5 \n" // V11 | V10 | V9 | V8
"precr.qb.ph $t5, $t6, $t5 \n" // U11 | U10 | U9 | U8
"precrq.qb.ph $t6, $t8, $t7 \n" // V15 | V14 | V13 | V12
"precr.qb.ph $t7, $t8, $t7 \n" // U15 | U14 | U13 | U12
"sw $t9, 0(%[dst_v]) \n"
"sw $t0, 0(%[dst_u]) \n"
"sw $t1, 4(%[dst_v]) \n"
"sw $t2, 4(%[dst_u]) \n"
"sw $t3, 8(%[dst_v]) \n"
"sw $t5, 8(%[dst_u]) \n"
"sw $t6, 12(%[dst_v]) \n"
"sw $t7, 12(%[dst_u]) \n"
"addiu %[dst_v], %[dst_v], 16 \n"
"bgtz $t4, 1b \n"
" addiu %[dst_u], %[dst_u], 16 \n"
"beqz %[width], 3f \n"
" nop \n"
"2: \n"
"lbu $t0, 0(%[src_uv]) \n"
"lbu $t1, 1(%[src_uv]) \n"
"addiu %[src_uv], %[src_uv], 2 \n"
"addiu %[width], %[width], -1 \n"
"sb $t0, 0(%[dst_u]) \n"
"sb $t1, 0(%[dst_v]) \n"
"addiu %[dst_u], %[dst_u], 1 \n"
"bgtz %[width], 2b \n"
" addiu %[dst_v], %[dst_v], 1 \n"
"3: \n"
".set pop \n"
: [src_uv] "+r" (src_uv),
[width] "+r" (width),
[dst_u] "+r" (dst_u),
[dst_v] "+r" (dst_v)
:
: "t0", "t1", "t2", "t3",
"t4", "t5", "t6", "t7", "t8", "t9"
);
}
void SplitUVRow_Unaligned_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u,
uint8* dst_v, int width) {
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"srl $t4, %[width], 4 \n" // multiplies of 16
"blez $t4, 2f \n"
" andi %[width], %[width], 0xf \n" // residual
".p2align 2 \n"
"1: \n"
"addiu $t4, $t4, -1 \n"
"lwr $t0, 0(%[src_uv]) \n"
"lwl $t0, 3(%[src_uv]) \n" // V1 | U1 | V0 | U0
"lwr $t1, 4(%[src_uv]) \n"
"lwl $t1, 7(%[src_uv]) \n" // V3 | U3 | V2 | U2
"lwr $t2, 8(%[src_uv]) \n"
"lwl $t2, 11(%[src_uv]) \n" // V5 | U5 | V4 | U4
"lwr $t3, 12(%[src_uv]) \n"
"lwl $t3, 15(%[src_uv]) \n" // V7 | U7 | V6 | U6
"lwr $t5, 16(%[src_uv]) \n"
"lwl $t5, 19(%[src_uv]) \n" // V9 | U9 | V8 | U8
"lwr $t6, 20(%[src_uv]) \n"
"lwl $t6, 23(%[src_uv]) \n" // V11 | U11 | V10 | U10
"lwr $t7, 24(%[src_uv]) \n"
"lwl $t7, 27(%[src_uv]) \n" // V13 | U13 | V12 | U12
"lwr $t8, 28(%[src_uv]) \n"
"lwl $t8, 31(%[src_uv]) \n" // V15 | U15 | V14 | U14
"precrq.qb.ph $t9, $t1, $t0 \n" // V3 | V2 | V1 | V0
"precr.qb.ph $t0, $t1, $t0 \n" // U3 | U2 | U1 | U0
"precrq.qb.ph $t1, $t3, $t2 \n" // V7 | V6 | V5 | V4
"precr.qb.ph $t2, $t3, $t2 \n" // U7 | U6 | U5 | U4
"precrq.qb.ph $t3, $t6, $t5 \n" // V11 | V10 | V9 | V8
"precr.qb.ph $t5, $t6, $t5 \n" // U11 | U10 | U9 | U8
"precrq.qb.ph $t6, $t8, $t7 \n" // V15 | V14 | V13 | V12
"precr.qb.ph $t7, $t8, $t7 \n" // U15 | U14 | U13 | U12
"addiu %[src_uv], %[src_uv], 32 \n"
"swr $t9, 0(%[dst_v]) \n"
"swl $t9, 3(%[dst_v]) \n"
"swr $t0, 0(%[dst_u]) \n"
"swl $t0, 3(%[dst_u]) \n"
"swr $t1, 4(%[dst_v]) \n"
"swl $t1, 7(%[dst_v]) \n"
"swr $t2, 4(%[dst_u]) \n"
"swl $t2, 7(%[dst_u]) \n"
"swr $t3, 8(%[dst_v]) \n"
"swl $t3, 11(%[dst_v]) \n"
"swr $t5, 8(%[dst_u]) \n"
"swl $t5, 11(%[dst_u]) \n"
"swr $t6, 12(%[dst_v]) \n"
"swl $t6, 15(%[dst_v]) \n"
"swr $t7, 12(%[dst_u]) \n"
"swl $t7, 15(%[dst_u]) \n"
"addiu %[dst_u], %[dst_u], 16 \n"
"bgtz $t4, 1b \n"
" addiu %[dst_v], %[dst_v], 16 \n"
"beqz %[width], 3f \n"
" nop \n"
"2: \n"
"lbu $t0, 0(%[src_uv]) \n"
"lbu $t1, 1(%[src_uv]) \n"
"addiu %[src_uv], %[src_uv], 2 \n"
"addiu %[width], %[width], -1 \n"
"sb $t0, 0(%[dst_u]) \n"
"sb $t1, 0(%[dst_v]) \n"
"addiu %[dst_u], %[dst_u], 1 \n"
"bgtz %[width], 2b \n"
" addiu %[dst_v], %[dst_v], 1 \n"
"3: \n"
".set pop \n"
: [src_uv] "+r" (src_uv),
[width] "+r" (width),
[dst_u] "+r" (dst_u),
[dst_v] "+r" (dst_v)
:
: "t0", "t1", "t2", "t3",
"t4", "t5", "t6", "t7", "t8", "t9"
);
}
void MirrorRow_MIPS_DSPR2(const uint8* src, uint8* dst, int width) {
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"srl $t4, %[width], 4 \n" // multiplies of 16
"andi $t5, %[width], 0xf \n"
"blez $t4, 2f \n"
" addu %[src], %[src], %[width] \n" // src += width
".p2align 2 \n"
"1: \n"
"lw $t0, -16(%[src]) \n" // |3|2|1|0|
"lw $t1, -12(%[src]) \n" // |7|6|5|4|
"lw $t2, -8(%[src]) \n" // |11|10|9|8|
"lw $t3, -4(%[src]) \n" // |15|14|13|12|
"wsbh $t0, $t0 \n" // |2|3|0|1|
"wsbh $t1, $t1 \n" // |6|7|4|5|
"wsbh $t2, $t2 \n" // |10|11|8|9|
"wsbh $t3, $t3 \n" // |14|15|12|13|
"rotr $t0, $t0, 16 \n" // |0|1|2|3|
"rotr $t1, $t1, 16 \n" // |4|5|6|7|
"rotr $t2, $t2, 16 \n" // |8|9|10|11|
"rotr $t3, $t3, 16 \n" // |12|13|14|15|
"addiu %[src], %[src], -16 \n"
"addiu $t4, $t4, -1 \n"
"sw $t3, 0(%[dst]) \n" // |15|14|13|12|
"sw $t2, 4(%[dst]) \n" // |11|10|9|8|
"sw $t1, 8(%[dst]) \n" // |7|6|5|4|
"sw $t0, 12(%[dst]) \n" // |3|2|1|0|
"bgtz $t4, 1b \n"
" addiu %[dst], %[dst], 16 \n"
"beqz $t5, 3f \n"
" nop \n"
"2: \n"
"lbu $t0, -1(%[src]) \n"
"addiu $t5, $t5, -1 \n"
"addiu %[src], %[src], -1 \n"
"sb $t0, 0(%[dst]) \n"
"bgez $t5, 2b \n"
" addiu %[dst], %[dst], 1 \n"
"3: \n"
".set pop \n"
: [src] "+r" (src), [dst] "+r" (dst)
: [width] "r" (width)
: "t0", "t1", "t2", "t3", "t4", "t5"
);
}
void MirrorUVRow_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
int width) {
int x = 0;
int y = 0;
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"addu $t4, %[width], %[width] \n"
"srl %[x], %[width], 4 \n"
"andi %[y], %[width], 0xf \n"
"blez %[x], 2f \n"
" addu %[src_uv], %[src_uv], $t4 \n"
".p2align 2 \n"
"1: \n"
"lw $t0, -32(%[src_uv]) \n" // |3|2|1|0|
"lw $t1, -28(%[src_uv]) \n" // |7|6|5|4|
"lw $t2, -24(%[src_uv]) \n" // |11|10|9|8|
"lw $t3, -20(%[src_uv]) \n" // |15|14|13|12|
"lw $t4, -16(%[src_uv]) \n" // |19|18|17|16|
"lw $t6, -12(%[src_uv]) \n" // |23|22|21|20|
"lw $t7, -8(%[src_uv]) \n" // |27|26|25|24|
"lw $t8, -4(%[src_uv]) \n" // |31|30|29|28|
"rotr $t0, $t0, 16 \n" // |1|0|3|2|
"rotr $t1, $t1, 16 \n" // |5|4|7|6|
"rotr $t2, $t2, 16 \n" // |9|8|11|10|
"rotr $t3, $t3, 16 \n" // |13|12|15|14|
"rotr $t4, $t4, 16 \n" // |17|16|19|18|
"rotr $t6, $t6, 16 \n" // |21|20|23|22|
"rotr $t7, $t7, 16 \n" // |25|24|27|26|
"rotr $t8, $t8, 16 \n" // |29|28|31|30|
"precr.qb.ph $t9, $t0, $t1 \n" // |0|2|4|6|
"precrq.qb.ph $t5, $t0, $t1 \n" // |1|3|5|7|
"precr.qb.ph $t0, $t2, $t3 \n" // |8|10|12|14|
"precrq.qb.ph $t1, $t2, $t3 \n" // |9|11|13|15|
"precr.qb.ph $t2, $t4, $t6 \n" // |16|18|20|22|
"precrq.qb.ph $t3, $t4, $t6 \n" // |17|19|21|23|
"precr.qb.ph $t4, $t7, $t8 \n" // |24|26|28|30|
"precrq.qb.ph $t6, $t7, $t8 \n" // |25|27|29|31|
"addiu %[src_uv], %[src_uv], -32 \n"
"addiu %[x], %[x], -1 \n"
"swr $t4, 0(%[dst_u]) \n"
"swl $t4, 3(%[dst_u]) \n" // |30|28|26|24|
"swr $t6, 0(%[dst_v]) \n"
"swl $t6, 3(%[dst_v]) \n" // |31|29|27|25|
"swr $t2, 4(%[dst_u]) \n"
"swl $t2, 7(%[dst_u]) \n" // |22|20|18|16|
"swr $t3, 4(%[dst_v]) \n"
"swl $t3, 7(%[dst_v]) \n" // |23|21|19|17|
"swr $t0, 8(%[dst_u]) \n"
"swl $t0, 11(%[dst_u]) \n" // |14|12|10|8|
"swr $t1, 8(%[dst_v]) \n"
"swl $t1, 11(%[dst_v]) \n" // |15|13|11|9|
"swr $t9, 12(%[dst_u]) \n"
"swl $t9, 15(%[dst_u]) \n" // |6|4|2|0|
"swr $t5, 12(%[dst_v]) \n"
"swl $t5, 15(%[dst_v]) \n" // |7|5|3|1|
"addiu %[dst_v], %[dst_v], 16 \n"
"bgtz %[x], 1b \n"
" addiu %[dst_u], %[dst_u], 16 \n"
"beqz %[y], 3f \n"
" nop \n"
"b 2f \n"
" nop \n"
"2: \n"
"lbu $t0, -2(%[src_uv]) \n"
"lbu $t1, -1(%[src_uv]) \n"
"addiu %[src_uv], %[src_uv], -2 \n"
"addiu %[y], %[y], -1 \n"
"sb $t0, 0(%[dst_u]) \n"
"sb $t1, 0(%[dst_v]) \n"
"addiu %[dst_u], %[dst_u], 1 \n"
"bgtz %[y], 2b \n"
" addiu %[dst_v], %[dst_v], 1 \n"
"3: \n"
".set pop \n"
: [src_uv] "+r" (src_uv),
[dst_u] "+r" (dst_u),
[dst_v] "+r" (dst_v),
[x] "=&r" (x),
[y] "+r" (y)
: [width] "r" (width)
: "t0", "t1", "t2", "t3", "t4",
"t5", "t7", "t8", "t9"
);
}
// Convert (4 Y and 2 VU) I422 and arrange RGB values into
// t5 = | 0 | B0 | 0 | b0 |
// t4 = | 0 | B1 | 0 | b1 |
// t9 = | 0 | G0 | 0 | g0 |
// t8 = | 0 | G1 | 0 | g1 |
// t2 = | 0 | R0 | 0 | r0 |
// t1 = | 0 | R1 | 0 | r1 |
#define I422ToTransientMipsRGB \
"lw $t0, 0(%[y_buf]) \n" \
"lhu $t1, 0(%[u_buf]) \n" \
"lhu $t2, 0(%[v_buf]) \n" \
"preceu.ph.qbr $t1, $t1 \n" \
"preceu.ph.qbr $t2, $t2 \n" \
"preceu.ph.qbra $t3, $t0 \n" \
"preceu.ph.qbla $t0, $t0 \n" \
"subu.ph $t1, $t1, $s5 \n" \
"subu.ph $t2, $t2, $s5 \n" \
"subu.ph $t3, $t3, $s4 \n" \
"subu.ph $t0, $t0, $s4 \n" \
"mul.ph $t3, $t3, $s0 \n" \
"mul.ph $t0, $t0, $s0 \n" \
"shll.ph $t4, $t1, 0x7 \n" \
"subu.ph $t4, $t4, $t1 \n" \
"mul.ph $t6, $t1, $s1 \n" \
"mul.ph $t1, $t2, $s2 \n" \
"addq_s.ph $t5, $t4, $t3 \n" \
"addq_s.ph $t4, $t4, $t0 \n" \
"shra.ph $t5, $t5, 6 \n" \
"shra.ph $t4, $t4, 6 \n" \
"addiu %[u_buf], 2 \n" \
"addiu %[v_buf], 2 \n" \
"addu.ph $t6, $t6, $t1 \n" \
"mul.ph $t1, $t2, $s3 \n" \
"addu.ph $t9, $t6, $t3 \n" \
"addu.ph $t8, $t6, $t0 \n" \
"shra.ph $t9, $t9, 6 \n" \
"shra.ph $t8, $t8, 6 \n" \
"addu.ph $t2, $t1, $t3 \n" \
"addu.ph $t1, $t1, $t0 \n" \
"shra.ph $t2, $t2, 6 \n" \
"shra.ph $t1, $t1, 6 \n" \
"subu.ph $t5, $t5, $s5 \n" \
"subu.ph $t4, $t4, $s5 \n" \
"subu.ph $t9, $t9, $s5 \n" \
"subu.ph $t8, $t8, $s5 \n" \
"subu.ph $t2, $t2, $s5 \n" \
"subu.ph $t1, $t1, $s5 \n" \
"shll_s.ph $t5, $t5, 8 \n" \
"shll_s.ph $t4, $t4, 8 \n" \
"shll_s.ph $t9, $t9, 8 \n" \
"shll_s.ph $t8, $t8, 8 \n" \
"shll_s.ph $t2, $t2, 8 \n" \
"shll_s.ph $t1, $t1, 8 \n" \
"shra.ph $t5, $t5, 8 \n" \
"shra.ph $t4, $t4, 8 \n" \
"shra.ph $t9, $t9, 8 \n" \
"shra.ph $t8, $t8, 8 \n" \
"shra.ph $t2, $t2, 8 \n" \
"shra.ph $t1, $t1, 8 \n" \
"addu.ph $t5, $t5, $s5 \n" \
"addu.ph $t4, $t4, $s5 \n" \
"addu.ph $t9, $t9, $s5 \n" \
"addu.ph $t8, $t8, $s5 \n" \
"addu.ph $t2, $t2, $s5 \n" \
"addu.ph $t1, $t1, $s5 \n"
void I422ToARGBRow_MIPS_DSPR2(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width) {
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"beqz %[width], 2f \n"
" repl.ph $s0, 74 \n" // |YG|YG| = |74|74|
"repl.ph $s1, -25 \n" // |UG|UG| = |-25|-25|
"repl.ph $s2, -52 \n" // |VG|VG| = |-52|-52|
"repl.ph $s3, 102 \n" // |VR|VR| = |102|102|
"repl.ph $s4, 16 \n" // |0|16|0|16|
"repl.ph $s5, 128 \n" // |128|128| // clipping
"lui $s6, 0xff00 \n"
"ori $s6, 0xff00 \n" // |ff|00|ff|00|ff|
".p2align 2 \n"
"1: \n"
I422ToTransientMipsRGB
// Arranging into argb format
"precr.qb.ph $t4, $t8, $t4 \n" // |G1|g1|B1|b1|
"precr.qb.ph $t5, $t9, $t5 \n" // |G0|g0|B0|b0|
"addiu %[width], -4 \n"
"precrq.qb.ph $t8, $t4, $t5 \n" // |G1|B1|G0|B0|
"precr.qb.ph $t9, $t4, $t5 \n" // |g1|b1|g0|b0|
"precr.qb.ph $t2, $t1, $t2 \n" // |R1|r1|R0|r0|
"addiu %[y_buf], 4 \n"
"preceu.ph.qbla $t1, $t2 \n" // |0 |R1|0 |R0|
"preceu.ph.qbra $t2, $t2 \n" // |0 |r1|0 |r0|
"or $t1, $t1, $s6 \n" // |ff|R1|ff|R0|
"or $t2, $t2, $s6 \n" // |ff|r1|ff|r0|
"precrq.ph.w $t0, $t2, $t9 \n" // |ff|r1|g1|b1|
"precrq.ph.w $t3, $t1, $t8 \n" // |ff|R1|G1|B1|
"sll $t9, $t9, 16 \n"
"sll $t8, $t8, 16 \n"
"packrl.ph $t2, $t2, $t9 \n" // |ff|r0|g0|b0|
"packrl.ph $t1, $t1, $t8 \n" // |ff|R0|G0|B0|
// Store results.
"sw $t2, 0(%[rgb_buf]) \n"
"sw $t0, 4(%[rgb_buf]) \n"
"sw $t1, 8(%[rgb_buf]) \n"
"sw $t3, 12(%[rgb_buf]) \n"
"bnez %[width], 1b \n"
" addiu %[rgb_buf], 16 \n"
"2: \n"
".set pop \n"
:[y_buf] "+r" (y_buf),
[u_buf] "+r" (u_buf),
[v_buf] "+r" (v_buf),
[width] "+r" (width),
[rgb_buf] "+r" (rgb_buf)
:
: "t0", "t1", "t2", "t3", "t4", "t5",
"t6", "t7", "t8", "t9",
"s0", "s1", "s2", "s3",
"s4", "s5", "s6"
);
}
void I422ToABGRRow_MIPS_DSPR2(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width) {
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"beqz %[width], 2f \n"
" repl.ph $s0, 74 \n" // |YG|YG| = |74|74|
"repl.ph $s1, -25 \n" // |UG|UG| = |-25|-25|
"repl.ph $s2, -52 \n" // |VG|VG| = |-52|-52|
"repl.ph $s3, 102 \n" // |VR|VR| = |102|102|
"repl.ph $s4, 16 \n" // |0|16|0|16|
"repl.ph $s5, 128 \n" // |128|128|
"lui $s6, 0xff00 \n"
"ori $s6, 0xff00 \n" // |ff|00|ff|00|
".p2align 2 \n"
"1: \n"
I422ToTransientMipsRGB
// Arranging into abgr format
"precr.qb.ph $t0, $t8, $t1 \n" // |G1|g1|R1|r1|
"precr.qb.ph $t3, $t9, $t2 \n" // |G0|g0|R0|r0|
"precrq.qb.ph $t8, $t0, $t3 \n" // |G1|R1|G0|R0|
"precr.qb.ph $t9, $t0, $t3 \n" // |g1|r1|g0|r0|
"precr.qb.ph $t2, $t4, $t5 \n" // |B1|b1|B0|b0|
"addiu %[width], -4 \n"
"addiu %[y_buf], 4 \n"
"preceu.ph.qbla $t1, $t2 \n" // |0 |B1|0 |B0|
"preceu.ph.qbra $t2, $t2 \n" // |0 |b1|0 |b0|
"or $t1, $t1, $s6 \n" // |ff|B1|ff|B0|
"or $t2, $t2, $s6 \n" // |ff|b1|ff|b0|
"precrq.ph.w $t0, $t2, $t9 \n" // |ff|b1|g1|r1|
"precrq.ph.w $t3, $t1, $t8 \n" // |ff|B1|G1|R1|
"sll $t9, $t9, 16 \n"
"sll $t8, $t8, 16 \n"
"packrl.ph $t2, $t2, $t9 \n" // |ff|b0|g0|r0|
"packrl.ph $t1, $t1, $t8 \n" // |ff|B0|G0|R0|
// Store results.
"sw $t2, 0(%[rgb_buf]) \n"
"sw $t0, 4(%[rgb_buf]) \n"
"sw $t1, 8(%[rgb_buf]) \n"
"sw $t3, 12(%[rgb_buf]) \n"
"bnez %[width], 1b \n"
" addiu %[rgb_buf], 16 \n"
"2: \n"
".set pop \n"
:[y_buf] "+r" (y_buf),
[u_buf] "+r" (u_buf),
[v_buf] "+r" (v_buf),
[width] "+r" (width),
[rgb_buf] "+r" (rgb_buf)
:
: "t0", "t1", "t2", "t3", "t4", "t5",
"t6", "t7", "t8", "t9",
"s0", "s1", "s2", "s3",
"s4", "s5", "s6"
);
}
void I422ToBGRARow_MIPS_DSPR2(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width) {
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"beqz %[width], 2f \n"
" repl.ph $s0, 74 \n" // |YG|YG| = |74 |74 |
"repl.ph $s1, -25 \n" // |UG|UG| = |-25|-25|
"repl.ph $s2, -52 \n" // |VG|VG| = |-52|-52|
"repl.ph $s3, 102 \n" // |VR|VR| = |102|102|
"repl.ph $s4, 16 \n" // |0|16|0|16|
"repl.ph $s5, 128 \n" // |128|128|
"lui $s6, 0xff \n"
"ori $s6, 0xff \n" // |00|ff|00|ff|
".p2align 2 \n"
"1: \n"
I422ToTransientMipsRGB
// Arranging into bgra format
"precr.qb.ph $t4, $t4, $t8 \n" // |B1|b1|G1|g1|
"precr.qb.ph $t5, $t5, $t9 \n" // |B0|b0|G0|g0|
"precrq.qb.ph $t8, $t4, $t5 \n" // |B1|G1|B0|G0|
"precr.qb.ph $t9, $t4, $t5 \n" // |b1|g1|b0|g0|
"precr.qb.ph $t2, $t1, $t2 \n" // |R1|r1|R0|r0|
"addiu %[width], -4 \n"
"addiu %[y_buf], 4 \n"
"preceu.ph.qbla $t1, $t2 \n" // |0 |R1|0 |R0|
"preceu.ph.qbra $t2, $t2 \n" // |0 |r1|0 |r0|
"sll $t1, $t1, 8 \n" // |R1|0 |R0|0 |
"sll $t2, $t2, 8 \n" // |r1|0 |r0|0 |
"or $t1, $t1, $s6 \n" // |R1|ff|R0|ff|
"or $t2, $t2, $s6 \n" // |r1|ff|r0|ff|
"precrq.ph.w $t0, $t9, $t2 \n" // |b1|g1|r1|ff|
"precrq.ph.w $t3, $t8, $t1 \n" // |B1|G1|R1|ff|
"sll $t1, $t1, 16 \n"
"sll $t2, $t2, 16 \n"
"packrl.ph $t2, $t9, $t2 \n" // |b0|g0|r0|ff|
"packrl.ph $t1, $t8, $t1 \n" // |B0|G0|R0|ff|
// Store results.
"sw $t2, 0(%[rgb_buf]) \n"
"sw $t0, 4(%[rgb_buf]) \n"
"sw $t1, 8(%[rgb_buf]) \n"
"sw $t3, 12(%[rgb_buf]) \n"
"bnez %[width], 1b \n"
" addiu %[rgb_buf], 16 \n"
"2: \n"
".set pop \n"
:[y_buf] "+r" (y_buf),
[u_buf] "+r" (u_buf),
[v_buf] "+r" (v_buf),
[width] "+r" (width),
[rgb_buf] "+r" (rgb_buf)
:
: "t0", "t1", "t2", "t3", "t4", "t5",
"t6", "t7", "t8", "t9",
"s0", "s1", "s2", "s3",
"s4", "s5", "s6"
);
}
// Bilinear filter 8x2 -> 8x1
void InterpolateRows_MIPS_DSPR2(uint8* dst_ptr, const uint8* src_ptr,
ptrdiff_t src_stride, int dst_width,
int source_y_fraction) {
int y0_fraction = 256 - source_y_fraction;
const uint8* src_ptr1 = src_ptr + src_stride;
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"replv.ph $t0, %[y0_fraction] \n"
"replv.ph $t1, %[source_y_fraction] \n"
".p2align 2 \n"
"1: \n"
"lw $t2, 0(%[src_ptr]) \n"
"lw $t3, 0(%[src_ptr1]) \n"
"lw $t4, 4(%[src_ptr]) \n"
"lw $t5, 4(%[src_ptr1]) \n"
"muleu_s.ph.qbl $t6, $t2, $t0 \n"
"muleu_s.ph.qbr $t7, $t2, $t0 \n"
"muleu_s.ph.qbl $t8, $t3, $t1 \n"
"muleu_s.ph.qbr $t9, $t3, $t1 \n"
"muleu_s.ph.qbl $t2, $t4, $t0 \n"
"muleu_s.ph.qbr $t3, $t4, $t0 \n"
"muleu_s.ph.qbl $t4, $t5, $t1 \n"
"muleu_s.ph.qbr $t5, $t5, $t1 \n"
"addq.ph $t6, $t6, $t8 \n"
"addq.ph $t7, $t7, $t9 \n"
"addq.ph $t2, $t2, $t4 \n"
"addq.ph $t3, $t3, $t5 \n"
"shra.ph $t6, $t6, 8 \n"
"shra.ph $t7, $t7, 8 \n"
"shra.ph $t2, $t2, 8 \n"
"shra.ph $t3, $t3, 8 \n"
"precr.qb.ph $t6, $t6, $t7 \n"
"precr.qb.ph $t2, $t2, $t3 \n"
"addiu %[src_ptr], %[src_ptr], 8 \n"
"addiu %[src_ptr1], %[src_ptr1], 8 \n"
"addiu %[dst_width], %[dst_width], -8 \n"
"sw $t6, 0(%[dst_ptr]) \n"
"sw $t2, 4(%[dst_ptr]) \n"
"bgtz %[dst_width], 1b \n"
" addiu %[dst_ptr], %[dst_ptr], 8 \n"
".set pop \n"
: [dst_ptr] "+r" (dst_ptr),
[src_ptr1] "+r" (src_ptr1),
[src_ptr] "+r" (src_ptr),
[dst_width] "+r" (dst_width)
: [source_y_fraction] "r" (source_y_fraction),
[y0_fraction] "r" (y0_fraction),
[src_stride] "r" (src_stride)
: "t0", "t1", "t2", "t3", "t4", "t5",
"t6", "t7", "t8", "t9"
);
}
#endif // __mips_dsp_rev >= 2
#endif // defined(__mips__)
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif

2844
third_party/libyuv/source/row_neon.cc vendored Normal file
View File

File diff suppressed because it is too large Load Diff

6443
third_party/libyuv/source/row_posix.cc vendored Normal file
View File

File diff suppressed because it is too large Load Diff

7284
third_party/libyuv/source/row_win.cc vendored Normal file
View File

File diff suppressed because it is too large Load Diff

3884
third_party/libyuv/source/scale.c vendored
View File

File diff suppressed because it is too large Load Diff

1716
third_party/libyuv/source/scale.cc vendored Normal file
View File

File diff suppressed because it is too large Load Diff

1165
third_party/libyuv/source/scale_common.cc vendored Normal file
View File

File diff suppressed because it is too large Load Diff

653
third_party/libyuv/source/scale_mips.cc vendored Normal file
View File

@ -0,0 +1,653 @@
/*
* 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 "third_party/libyuv/include/libyuv/basic_types.h"
#include "third_party/libyuv/include/libyuv/row.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// This module is for GCC MIPS DSPR2
#if !defined(LIBYUV_DISABLE_MIPS) && \
defined(__mips_dsp) && (__mips_dsp_rev >= 2)
void ScaleRowDown2_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) {
__asm__ __volatile__(
".set push \n"
".set noreorder \n"
"srl $t9, %[dst_width], 4 \n" // iterations -> by 16
"beqz $t9, 2f \n"
" nop \n"
".p2align 2 \n"
"1: \n"
"lw $t0, 0(%[src_ptr]) \n" // |3|2|1|0|
"lw $t1, 4(%[src_ptr]) \n" // |7|6|5|4|
"lw $t2, 8(%[src_ptr]) \n" // |11|10|9|8|
"lw $t3, 12(%[src_ptr]) \n" // |15|14|13|12|
"lw $t4, 16(%[src_ptr]) \n" // |19|18|17|16|
"lw $t5, 20(%[src_ptr]) \n" // |23|22|21|20|
"lw $t6, 24(%[src_ptr]) \n" // |27|26|25|24|
"lw $t7, 28(%[src_ptr]) \n" // |31|30|29|28|
// TODO(fbarchard): Use odd pixels instead of even.
"precr.qb.ph $t8, $t1, $t0 \n" // |6|4|2|0|
"precr.qb.ph $t0, $t3, $t2 \n" // |14|12|10|8|
"precr.qb.ph $t1, $t5, $t4 \n" // |22|20|18|16|
"precr.qb.ph $t2, $t7, $t6 \n" // |30|28|26|24|
"addiu %[src_ptr], %[src_ptr], 32 \n"
"addiu $t9, $t9, -1 \n"
"sw $t8, 0(%[dst]) \n"
"sw $t0, 4(%[dst]) \n"
"sw $t1, 8(%[dst]) \n"
"sw $t2, 12(%[dst]) \n"
"bgtz $t9, 1b \n"
" addiu %[dst], %[dst], 16 \n"
"2: \n"
"andi $t9, %[dst_width], 0xf \n" // residue
"beqz $t9, 3f \n"
" nop \n"
"21: \n"
"lbu $t0, 0(%[src_ptr]) \n"
"addiu %[src_ptr], %[src_ptr], 2 \n"
"addiu $t9, $t9, -1 \n"
"sb $t0, 0(%[dst]) \n"
"bgtz $t9, 21b \n"
" addiu %[dst], %[dst], 1 \n"
"3: \n"
".set pop \n"
: [src_ptr] "+r" (src_ptr),
[dst] "+r" (dst)
: [dst_width] "r" (dst_width)
: "t0", "t1", "t2", "t3", "t4", "t5",
"t6", "t7", "t8", "t9"
);
}
void ScaleRowDown2Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) {
const uint8* t = src_ptr + src_stride;
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"srl $t9, %[dst_width], 3 \n" // iterations -> step 8
"bltz $t9, 2f \n"
" nop \n"
".p2align 2 \n"
"1: \n"
"lw $t0, 0(%[src_ptr]) \n" // |3|2|1|0|
"lw $t1, 4(%[src_ptr]) \n" // |7|6|5|4|
"lw $t2, 8(%[src_ptr]) \n" // |11|10|9|8|
"lw $t3, 12(%[src_ptr]) \n" // |15|14|13|12|
"lw $t4, 0(%[t]) \n" // |19|18|17|16|
"lw $t5, 4(%[t]) \n" // |23|22|21|20|
"lw $t6, 8(%[t]) \n" // |27|26|25|24|
"lw $t7, 12(%[t]) \n" // |31|30|29|28|
"addiu $t9, $t9, -1 \n"
"srl $t8, $t0, 16 \n" // |X|X|3|2|
"ins $t0, $t4, 16, 16 \n" // |17|16|1|0|
"ins $t4, $t8, 0, 16 \n" // |19|18|3|2|
"raddu.w.qb $t0, $t0 \n" // |17+16+1+0|
"raddu.w.qb $t4, $t4 \n" // |19+18+3+2|
"shra_r.w $t0, $t0, 2 \n" // |t0+2|>>2
"shra_r.w $t4, $t4, 2 \n" // |t4+2|>>2
"srl $t8, $t1, 16 \n" // |X|X|7|6|
"ins $t1, $t5, 16, 16 \n" // |21|20|5|4|
"ins $t5, $t8, 0, 16 \n" // |22|23|7|6|
"raddu.w.qb $t1, $t1 \n" // |21+20+5+4|
"raddu.w.qb $t5, $t5 \n" // |23+22+7+6|
"shra_r.w $t1, $t1, 2 \n" // |t1+2|>>2
"shra_r.w $t5, $t5, 2 \n" // |t5+2|>>2
"srl $t8, $t2, 16 \n" // |X|X|11|10|
"ins $t2, $t6, 16, 16 \n" // |25|24|9|8|
"ins $t6, $t8, 0, 16 \n" // |27|26|11|10|
"raddu.w.qb $t2, $t2 \n" // |25+24+9+8|
"raddu.w.qb $t6, $t6 \n" // |27+26+11+10|
"shra_r.w $t2, $t2, 2 \n" // |t2+2|>>2
"shra_r.w $t6, $t6, 2 \n" // |t5+2|>>2
"srl $t8, $t3, 16 \n" // |X|X|15|14|
"ins $t3, $t7, 16, 16 \n" // |29|28|13|12|
"ins $t7, $t8, 0, 16 \n" // |31|30|15|14|
"raddu.w.qb $t3, $t3 \n" // |29+28+13+12|
"raddu.w.qb $t7, $t7 \n" // |31+30+15+14|
"shra_r.w $t3, $t3, 2 \n" // |t3+2|>>2
"shra_r.w $t7, $t7, 2 \n" // |t7+2|>>2
"addiu %[src_ptr], %[src_ptr], 16 \n"
"addiu %[t], %[t], 16 \n"
"sb $t0, 0(%[dst]) \n"
"sb $t4, 1(%[dst]) \n"
"sb $t1, 2(%[dst]) \n"
"sb $t5, 3(%[dst]) \n"
"sb $t2, 4(%[dst]) \n"
"sb $t6, 5(%[dst]) \n"
"sb $t3, 6(%[dst]) \n"
"sb $t7, 7(%[dst]) \n"
"bgtz $t9, 1b \n"
" addiu %[dst], %[dst], 8 \n"
"2: \n"
"andi $t9, %[dst_width], 0x7 \n" // x = residue
"beqz $t9, 3f \n"
" nop \n"
"21: \n"
"lwr $t1, 0(%[src_ptr]) \n"
"lwl $t1, 3(%[src_ptr]) \n"
"lwr $t2, 0(%[t]) \n"
"lwl $t2, 3(%[t]) \n"
"srl $t8, $t1, 16 \n"
"ins $t1, $t2, 16, 16 \n"
"ins $t2, $t8, 0, 16 \n"
"raddu.w.qb $t1, $t1 \n"
"raddu.w.qb $t2, $t2 \n"
"shra_r.w $t1, $t1, 2 \n"
"shra_r.w $t2, $t2, 2 \n"
"sb $t1, 0(%[dst]) \n"
"sb $t2, 1(%[dst]) \n"
"addiu %[src_ptr], %[src_ptr], 4 \n"
"addiu $t9, $t9, -2 \n"
"addiu %[t], %[t], 4 \n"
"bgtz $t9, 21b \n"
" addiu %[dst], %[dst], 2 \n"
"3: \n"
".set pop \n"
: [src_ptr] "+r" (src_ptr),
[dst] "+r" (dst), [t] "+r" (t)
: [dst_width] "r" (dst_width)
: "t0", "t1", "t2", "t3", "t4", "t5",
"t6", "t7", "t8", "t9"
);
}
void ScaleRowDown4_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) {
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"srl $t9, %[dst_width], 3 \n"
"beqz $t9, 2f \n"
" nop \n"
".p2align 2 \n"
"1: \n"
"lw $t1, 0(%[src_ptr]) \n" // |3|2|1|0|
"lw $t2, 4(%[src_ptr]) \n" // |7|6|5|4|
"lw $t3, 8(%[src_ptr]) \n" // |11|10|9|8|
"lw $t4, 12(%[src_ptr]) \n" // |15|14|13|12|
"lw $t5, 16(%[src_ptr]) \n" // |19|18|17|16|
"lw $t6, 20(%[src_ptr]) \n" // |23|22|21|20|
"lw $t7, 24(%[src_ptr]) \n" // |27|26|25|24|
"lw $t8, 28(%[src_ptr]) \n" // |31|30|29|28|
"precr.qb.ph $t1, $t2, $t1 \n" // |6|4|2|0|
"precr.qb.ph $t2, $t4, $t3 \n" // |14|12|10|8|
"precr.qb.ph $t5, $t6, $t5 \n" // |22|20|18|16|
"precr.qb.ph $t6, $t8, $t7 \n" // |30|28|26|24|
"precr.qb.ph $t1, $t2, $t1 \n" // |12|8|4|0|
"precr.qb.ph $t5, $t6, $t5 \n" // |28|24|20|16|
"addiu %[src_ptr], %[src_ptr], 32 \n"
"addiu $t9, $t9, -1 \n"
"sw $t1, 0(%[dst]) \n"
"sw $t5, 4(%[dst]) \n"
"bgtz $t9, 1b \n"
" addiu %[dst], %[dst], 8 \n"
"2: \n"
"andi $t9, %[dst_width], 7 \n" // residue
"beqz $t9, 3f \n"
" nop \n"
"21: \n"
"lbu $t1, 0(%[src_ptr]) \n"
"addiu %[src_ptr], %[src_ptr], 4 \n"
"addiu $t9, $t9, -1 \n"
"sb $t1, 0(%[dst]) \n"
"bgtz $t9, 21b \n"
" addiu %[dst], %[dst], 1 \n"
"3: \n"
".set pop \n"
: [src_ptr] "+r" (src_ptr),
[dst] "+r" (dst)
: [dst_width] "r" (dst_width)
: "t1", "t2", "t3", "t4", "t5",
"t6", "t7", "t8", "t9"
);
}
void ScaleRowDown4Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) {
intptr_t stride = src_stride;
const uint8* s1 = src_ptr + stride;
const uint8* s2 = s1 + stride;
const uint8* s3 = s2 + stride;
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"srl $t9, %[dst_width], 1 \n"
"andi $t8, %[dst_width], 1 \n"
".p2align 2 \n"
"1: \n"
"lw $t0, 0(%[src_ptr]) \n" // |3|2|1|0|
"lw $t1, 0(%[s1]) \n" // |7|6|5|4|
"lw $t2, 0(%[s2]) \n" // |11|10|9|8|
"lw $t3, 0(%[s3]) \n" // |15|14|13|12|
"lw $t4, 4(%[src_ptr]) \n" // |19|18|17|16|
"lw $t5, 4(%[s1]) \n" // |23|22|21|20|
"lw $t6, 4(%[s2]) \n" // |27|26|25|24|
"lw $t7, 4(%[s3]) \n" // |31|30|29|28|
"raddu.w.qb $t0, $t0 \n" // |3 + 2 + 1 + 0|
"raddu.w.qb $t1, $t1 \n" // |7 + 6 + 5 + 4|
"raddu.w.qb $t2, $t2 \n" // |11 + 10 + 9 + 8|
"raddu.w.qb $t3, $t3 \n" // |15 + 14 + 13 + 12|
"raddu.w.qb $t4, $t4 \n" // |19 + 18 + 17 + 16|
"raddu.w.qb $t5, $t5 \n" // |23 + 22 + 21 + 20|
"raddu.w.qb $t6, $t6 \n" // |27 + 26 + 25 + 24|
"raddu.w.qb $t7, $t7 \n" // |31 + 30 + 29 + 28|
"add $t0, $t0, $t1 \n"
"add $t1, $t2, $t3 \n"
"add $t0, $t0, $t1 \n"
"add $t4, $t4, $t5 \n"
"add $t6, $t6, $t7 \n"
"add $t4, $t4, $t6 \n"
"shra_r.w $t0, $t0, 4 \n"
"shra_r.w $t4, $t4, 4 \n"
"sb $t0, 0(%[dst]) \n"
"sb $t4, 1(%[dst]) \n"
"addiu %[src_ptr], %[src_ptr], 8 \n"
"addiu %[s1], %[s1], 8 \n"
"addiu %[s2], %[s2], 8 \n"
"addiu %[s3], %[s3], 8 \n"
"addiu $t9, $t9, -1 \n"
"bgtz $t9, 1b \n"
" addiu %[dst], %[dst], 2 \n"
"beqz $t8, 2f \n"
" nop \n"
"lw $t0, 0(%[src_ptr]) \n" // |3|2|1|0|
"lw $t1, 0(%[s1]) \n" // |7|6|5|4|
"lw $t2, 0(%[s2]) \n" // |11|10|9|8|
"lw $t3, 0(%[s3]) \n" // |15|14|13|12|
"raddu.w.qb $t0, $t0 \n" // |3 + 2 + 1 + 0|
"raddu.w.qb $t1, $t1 \n" // |7 + 6 + 5 + 4|
"raddu.w.qb $t2, $t2 \n" // |11 + 10 + 9 + 8|
"raddu.w.qb $t3, $t3 \n" // |15 + 14 + 13 + 12|
"add $t0, $t0, $t1 \n"
"add $t1, $t2, $t3 \n"
"add $t0, $t0, $t1 \n"
"shra_r.w $t0, $t0, 4 \n"
"sb $t0, 0(%[dst]) \n"
"2: \n"
".set pop \n"
: [src_ptr] "+r" (src_ptr),
[dst] "+r" (dst),
[s1] "+r" (s1),
[s2] "+r" (s2),
[s3] "+r" (s3)
: [dst_width] "r" (dst_width)
: "t0", "t1", "t2", "t3", "t4", "t5",
"t6","t7", "t8", "t9"
);
}
void ScaleRowDown34_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) {
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
".p2align 2 \n"
"1: \n"
"lw $t1, 0(%[src_ptr]) \n" // |3|2|1|0|
"lw $t2, 4(%[src_ptr]) \n" // |7|6|5|4|
"lw $t3, 8(%[src_ptr]) \n" // |11|10|9|8|
"lw $t4, 12(%[src_ptr]) \n" // |15|14|13|12|
"lw $t5, 16(%[src_ptr]) \n" // |19|18|17|16|
"lw $t6, 20(%[src_ptr]) \n" // |23|22|21|20|
"lw $t7, 24(%[src_ptr]) \n" // |27|26|25|24|
"lw $t8, 28(%[src_ptr]) \n" // |31|30|29|28|
"precrq.qb.ph $t0, $t2, $t4 \n" // |7|5|15|13|
"precrq.qb.ph $t9, $t6, $t8 \n" // |23|21|31|30|
"addiu %[dst_width], %[dst_width], -24 \n"
"ins $t1, $t1, 8, 16 \n" // |3|1|0|X|
"ins $t4, $t0, 8, 16 \n" // |X|15|13|12|
"ins $t5, $t5, 8, 16 \n" // |19|17|16|X|
"ins $t8, $t9, 8, 16 \n" // |X|31|29|28|
"addiu %[src_ptr], %[src_ptr], 32 \n"
"packrl.ph $t0, $t3, $t0 \n" // |9|8|7|5|
"packrl.ph $t9, $t7, $t9 \n" // |25|24|23|21|
"prepend $t1, $t2, 8 \n" // |4|3|1|0|
"prepend $t3, $t4, 24 \n" // |15|13|12|11|
"prepend $t5, $t6, 8 \n" // |20|19|17|16|
"prepend $t7, $t8, 24 \n" // |31|29|28|27|
"sw $t1, 0(%[dst]) \n"
"sw $t0, 4(%[dst]) \n"
"sw $t3, 8(%[dst]) \n"
"sw $t5, 12(%[dst]) \n"
"sw $t9, 16(%[dst]) \n"
"sw $t7, 20(%[dst]) \n"
"bnez %[dst_width], 1b \n"
" addiu %[dst], %[dst], 24 \n"
".set pop \n"
: [src_ptr] "+r" (src_ptr),
[dst] "+r" (dst),
[dst_width] "+r" (dst_width)
:
: "t0", "t1", "t2", "t3", "t4", "t5",
"t6","t7", "t8", "t9"
);
}
void ScaleRowDown34_0_Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* d, int dst_width) {
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"repl.ph $t3, 3 \n" // 0x00030003
".p2align 2 \n"
"1: \n"
"lw $t0, 0(%[src_ptr]) \n" // |S3|S2|S1|S0|
"lwx $t1, %[src_stride](%[src_ptr]) \n" // |T3|T2|T1|T0|
"rotr $t2, $t0, 8 \n" // |S0|S3|S2|S1|
"rotr $t6, $t1, 8 \n" // |T0|T3|T2|T1|
"muleu_s.ph.qbl $t4, $t2, $t3 \n" // |S0*3|S3*3|
"muleu_s.ph.qbl $t5, $t6, $t3 \n" // |T0*3|T3*3|
"andi $t0, $t2, 0xFFFF \n" // |0|0|S2|S1|
"andi $t1, $t6, 0xFFFF \n" // |0|0|T2|T1|
"raddu.w.qb $t0, $t0 \n"
"raddu.w.qb $t1, $t1 \n"
"shra_r.w $t0, $t0, 1 \n"
"shra_r.w $t1, $t1, 1 \n"
"preceu.ph.qbr $t2, $t2 \n" // |0|S2|0|S1|
"preceu.ph.qbr $t6, $t6 \n" // |0|T2|0|T1|
"rotr $t2, $t2, 16 \n" // |0|S1|0|S2|
"rotr $t6, $t6, 16 \n" // |0|T1|0|T2|
"addu.ph $t2, $t2, $t4 \n"
"addu.ph $t6, $t6, $t5 \n"
"sll $t5, $t0, 1 \n"
"add $t0, $t5, $t0 \n"
"shra_r.ph $t2, $t2, 2 \n"
"shra_r.ph $t6, $t6, 2 \n"
"shll.ph $t4, $t2, 1 \n"
"addq.ph $t4, $t4, $t2 \n"
"addu $t0, $t0, $t1 \n"
"addiu %[src_ptr], %[src_ptr], 4 \n"
"shra_r.w $t0, $t0, 2 \n"
"addu.ph $t6, $t6, $t4 \n"
"shra_r.ph $t6, $t6, 2 \n"
"srl $t1, $t6, 16 \n"
"addiu %[dst_width], %[dst_width], -3 \n"
"sb $t1, 0(%[d]) \n"
"sb $t0, 1(%[d]) \n"
"sb $t6, 2(%[d]) \n"
"bgtz %[dst_width], 1b \n"
" addiu %[d], %[d], 3 \n"
"3: \n"
".set pop \n"
: [src_ptr] "+r" (src_ptr),
[src_stride] "+r" (src_stride),
[d] "+r" (d),
[dst_width] "+r" (dst_width)
:
: "t0", "t1", "t2", "t3",
"t4", "t5", "t6"
);
}
void ScaleRowDown34_1_Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* d, int dst_width) {
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"repl.ph $t2, 3 \n" // 0x00030003
".p2align 2 \n"
"1: \n"
"lw $t0, 0(%[src_ptr]) \n" // |S3|S2|S1|S0|
"lwx $t1, %[src_stride](%[src_ptr]) \n" // |T3|T2|T1|T0|
"rotr $t4, $t0, 8 \n" // |S0|S3|S2|S1|
"rotr $t6, $t1, 8 \n" // |T0|T3|T2|T1|
"muleu_s.ph.qbl $t3, $t4, $t2 \n" // |S0*3|S3*3|
"muleu_s.ph.qbl $t5, $t6, $t2 \n" // |T0*3|T3*3|
"andi $t0, $t4, 0xFFFF \n" // |0|0|S2|S1|
"andi $t1, $t6, 0xFFFF \n" // |0|0|T2|T1|
"raddu.w.qb $t0, $t0 \n"
"raddu.w.qb $t1, $t1 \n"
"shra_r.w $t0, $t0, 1 \n"
"shra_r.w $t1, $t1, 1 \n"
"preceu.ph.qbr $t4, $t4 \n" // |0|S2|0|S1|
"preceu.ph.qbr $t6, $t6 \n" // |0|T2|0|T1|
"rotr $t4, $t4, 16 \n" // |0|S1|0|S2|
"rotr $t6, $t6, 16 \n" // |0|T1|0|T2|
"addu.ph $t4, $t4, $t3 \n"
"addu.ph $t6, $t6, $t5 \n"
"shra_r.ph $t6, $t6, 2 \n"
"shra_r.ph $t4, $t4, 2 \n"
"addu.ph $t6, $t6, $t4 \n"
"addiu %[src_ptr], %[src_ptr], 4 \n"
"shra_r.ph $t6, $t6, 1 \n"
"addu $t0, $t0, $t1 \n"
"addiu %[dst_width], %[dst_width], -3 \n"
"shra_r.w $t0, $t0, 1 \n"
"srl $t1, $t6, 16 \n"
"sb $t1, 0(%[d]) \n"
"sb $t0, 1(%[d]) \n"
"sb $t6, 2(%[d]) \n"
"bgtz %[dst_width], 1b \n"
" addiu %[d], %[d], 3 \n"
"3: \n"
".set pop \n"
: [src_ptr] "+r" (src_ptr),
[src_stride] "+r" (src_stride),
[d] "+r" (d),
[dst_width] "+r" (dst_width)
:
: "t0", "t1", "t2", "t3",
"t4", "t5", "t6"
);
}
void ScaleRowDown38_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) {
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
".p2align 2 \n"
"1: \n"
"lw $t0, 0(%[src_ptr]) \n" // |3|2|1|0|
"lw $t1, 4(%[src_ptr]) \n" // |7|6|5|4|
"lw $t2, 8(%[src_ptr]) \n" // |11|10|9|8|
"lw $t3, 12(%[src_ptr]) \n" // |15|14|13|12|
"lw $t4, 16(%[src_ptr]) \n" // |19|18|17|16|
"lw $t5, 20(%[src_ptr]) \n" // |23|22|21|20|
"lw $t6, 24(%[src_ptr]) \n" // |27|26|25|24|
"lw $t7, 28(%[src_ptr]) \n" // |31|30|29|28|
"wsbh $t0, $t0 \n" // |2|3|0|1|
"wsbh $t6, $t6 \n" // |26|27|24|25|
"srl $t0, $t0, 8 \n" // |X|2|3|0|
"srl $t3, $t3, 16 \n" // |X|X|15|14|
"srl $t5, $t5, 16 \n" // |X|X|23|22|
"srl $t7, $t7, 16 \n" // |X|X|31|30|
"ins $t1, $t2, 24, 8 \n" // |8|6|5|4|
"ins $t6, $t5, 0, 8 \n" // |26|27|24|22|
"ins $t1, $t0, 0, 16 \n" // |8|6|3|0|
"ins $t6, $t7, 24, 8 \n" // |30|27|24|22|
"prepend $t2, $t3, 24 \n" // |X|15|14|11|
"ins $t4, $t4, 16, 8 \n" // |19|16|17|X|
"ins $t4, $t2, 0, 16 \n" // |19|16|14|11|
"addiu %[src_ptr], %[src_ptr], 32 \n"
"addiu %[dst_width], %[dst_width], -12 \n"
"addiu $t8,%[dst_width], -12 \n"
"sw $t1, 0(%[dst]) \n"
"sw $t4, 4(%[dst]) \n"
"sw $t6, 8(%[dst]) \n"
"bgez $t8, 1b \n"
" addiu %[dst], %[dst], 12 \n"
".set pop \n"
: [src_ptr] "+r" (src_ptr),
[dst] "+r" (dst),
[dst_width] "+r" (dst_width)
:
: "t0", "t1", "t2", "t3", "t4",
"t5", "t6", "t7", "t8"
);
}
void ScaleRowDown38_2_Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) {
intptr_t stride = src_stride;
const uint8* t = src_ptr + stride;
const int c = 0x2AAA;
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
".p2align 2 \n"
"1: \n"
"lw $t0, 0(%[src_ptr]) \n" // |S3|S2|S1|S0|
"lw $t1, 4(%[src_ptr]) \n" // |S7|S6|S5|S4|
"lw $t2, 0(%[t]) \n" // |T3|T2|T1|T0|
"lw $t3, 4(%[t]) \n" // |T7|T6|T5|T4|
"rotr $t1, $t1, 16 \n" // |S5|S4|S7|S6|
"packrl.ph $t4, $t1, $t3 \n" // |S7|S6|T7|T6|
"packrl.ph $t5, $t3, $t1 \n" // |T5|T4|S5|S4|
"raddu.w.qb $t4, $t4 \n" // S7+S6+T7+T6
"raddu.w.qb $t5, $t5 \n" // T5+T4+S5+S4
"precrq.qb.ph $t6, $t0, $t2 \n" // |S3|S1|T3|T1|
"precrq.qb.ph $t6, $t6, $t6 \n" // |S3|T3|S3|T3|
"srl $t4, $t4, 2 \n" // t4 / 4
"srl $t6, $t6, 16 \n" // |0|0|S3|T3|
"raddu.w.qb $t6, $t6 \n" // 0+0+S3+T3
"addu $t6, $t5, $t6 \n"
"mul $t6, $t6, %[c] \n" // t6 * 0x2AAA
"sll $t0, $t0, 8 \n" // |S2|S1|S0|0|
"sll $t2, $t2, 8 \n" // |T2|T1|T0|0|
"raddu.w.qb $t0, $t0 \n" // S2+S1+S0+0
"raddu.w.qb $t2, $t2 \n" // T2+T1+T0+0
"addu $t0, $t0, $t2 \n"
"mul $t0, $t0, %[c] \n" // t0 * 0x2AAA
"addiu %[src_ptr], %[src_ptr], 8 \n"
"addiu %[t], %[t], 8 \n"
"addiu %[dst_width], %[dst_width], -3 \n"
"addiu %[dst_ptr], %[dst_ptr], 3 \n"
"srl $t6, $t6, 16 \n"
"srl $t0, $t0, 16 \n"
"sb $t4, -1(%[dst_ptr]) \n"
"sb $t6, -2(%[dst_ptr]) \n"
"bgtz %[dst_width], 1b \n"
" sb $t0, -3(%[dst_ptr]) \n"
".set pop \n"
: [src_ptr] "+r" (src_ptr),
[dst_ptr] "+r" (dst_ptr),
[t] "+r" (t),
[dst_width] "+r" (dst_width)
: [c] "r" (c)
: "t0", "t1", "t2", "t3", "t4", "t5", "t6"
);
}
void ScaleRowDown38_3_Box_MIPS_DSPR2(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) {
intptr_t stride = src_stride;
const uint8* s1 = src_ptr + stride;
stride += stride;
const uint8* s2 = src_ptr + stride;
const int c1 = 0x1C71;
const int c2 = 0x2AAA;
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
".p2align 2 \n"
"1: \n"
"lw $t0, 0(%[src_ptr]) \n" // |S3|S2|S1|S0|
"lw $t1, 4(%[src_ptr]) \n" // |S7|S6|S5|S4|
"lw $t2, 0(%[s1]) \n" // |T3|T2|T1|T0|
"lw $t3, 4(%[s1]) \n" // |T7|T6|T5|T4|
"lw $t4, 0(%[s2]) \n" // |R3|R2|R1|R0|
"lw $t5, 4(%[s2]) \n" // |R7|R6|R5|R4|
"rotr $t1, $t1, 16 \n" // |S5|S4|S7|S6|
"packrl.ph $t6, $t1, $t3 \n" // |S7|S6|T7|T6|
"raddu.w.qb $t6, $t6 \n" // S7+S6+T7+T6
"packrl.ph $t7, $t3, $t1 \n" // |T5|T4|S5|S4|
"raddu.w.qb $t7, $t7 \n" // T5+T4+S5+S4
"sll $t8, $t5, 16 \n" // |R5|R4|0|0|
"raddu.w.qb $t8, $t8 \n" // R5+R4
"addu $t7, $t7, $t8 \n"
"srl $t8, $t5, 16 \n" // |0|0|R7|R6|
"raddu.w.qb $t8, $t8 \n" // R7 + R6
"addu $t6, $t6, $t8 \n"
"mul $t6, $t6, %[c2] \n" // t6 * 0x2AAA
"precrq.qb.ph $t8, $t0, $t2 \n" // |S3|S1|T3|T1|
"precrq.qb.ph $t8, $t8, $t4 \n" // |S3|T3|R3|R1|
"srl $t8, $t8, 8 \n" // |0|S3|T3|R3|
"raddu.w.qb $t8, $t8 \n" // S3 + T3 + R3
"addu $t7, $t7, $t8 \n"
"mul $t7, $t7, %[c1] \n" // t7 * 0x1C71
"sll $t0, $t0, 8 \n" // |S2|S1|S0|0|
"sll $t2, $t2, 8 \n" // |T2|T1|T0|0|
"sll $t4, $t4, 8 \n" // |R2|R1|R0|0|
"raddu.w.qb $t0, $t0 \n"
"raddu.w.qb $t2, $t2 \n"
"raddu.w.qb $t4, $t4 \n"
"addu $t0, $t0, $t2 \n"
"addu $t0, $t0, $t4 \n"
"mul $t0, $t0, %[c1] \n" // t0 * 0x1C71
"addiu %[src_ptr], %[src_ptr], 8 \n"
"addiu %[s1], %[s1], 8 \n"
"addiu %[s2], %[s2], 8 \n"
"addiu %[dst_width], %[dst_width], -3 \n"
"addiu %[dst_ptr], %[dst_ptr], 3 \n"
"srl $t6, $t6, 16 \n"
"srl $t7, $t7, 16 \n"
"srl $t0, $t0, 16 \n"
"sb $t6, -1(%[dst_ptr]) \n"
"sb $t7, -2(%[dst_ptr]) \n"
"bgtz %[dst_width], 1b \n"
" sb $t0, -3(%[dst_ptr]) \n"
".set pop \n"
: [src_ptr] "+r" (src_ptr),
[dst_ptr] "+r" (dst_ptr),
[s1] "+r" (s1),
[s2] "+r" (s2),
[dst_width] "+r" (dst_width)
: [c1] "r" (c1), [c2] "r" (c2)
: "t0", "t1", "t2", "t3", "t4",
"t5", "t6", "t7", "t8"
);
}
#endif // defined(__mips_dsp) && (__mips_dsp_rev >= 2)
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif

684
third_party/libyuv/source/scale_neon.cc vendored Normal file
View File

@ -0,0 +1,684 @@
/*
* Copyright 2011 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 "third_party/libyuv/include/libyuv/row.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// This module is for GCC Neon.
#if !defined(LIBYUV_DISABLE_NEON) && defined(__ARM_NEON__)
// NEON downscalers with interpolation.
// Provided by Fritz Koenig
// Read 32x1 throw away even pixels, and write 16x1.
void ScaleRowDown2_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) {
asm volatile (
".p2align 2 \n"
"1: \n"
// load even pixels into q0, odd into q1
"vld2.8 {q0, q1}, [%0]! \n"
"subs %2, %2, #16 \n" // 16 processed per loop
"vst1.8 {q1}, [%1]! \n" // store odd pixels
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(dst), // %1
"+r"(dst_width) // %2
:
: "q0", "q1" // Clobber List
);
}
// Read 32x2 average down and write 16x1.
void ScaleRowDown2Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) {
asm volatile (
// change the stride to row 2 pointer
"add %1, %0 \n"
".p2align 2 \n"
"1: \n"
"vld1.8 {q0, q1}, [%0]! \n" // load row 1 and post inc
"vld1.8 {q2, q3}, [%1]! \n" // load row 2 and post inc
"subs %3, %3, #16 \n" // 16 processed per loop
"vpaddl.u8 q0, q0 \n" // row 1 add adjacent
"vpaddl.u8 q1, q1 \n"
"vpadal.u8 q0, q2 \n" // row 2 add adjacent + row1
"vpadal.u8 q1, q3 \n"
"vrshrn.u16 d0, q0, #2 \n" // downshift, round and pack
"vrshrn.u16 d1, q1, #2 \n"
"vst1.8 {q0}, [%2]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(src_stride), // %1
"+r"(dst), // %2
"+r"(dst_width) // %3
:
: "q0", "q1", "q2", "q3" // Clobber List
);
}
void ScaleRowDown4_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) {
asm volatile (
".p2align 2 \n"
"1: \n"
"vld4.8 {d0, d1, d2, d3}, [%0]! \n" // src line 0
"subs %2, %2, #8 \n" // 8 processed per loop
"vst1.8 {d2}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1
"+r"(dst_width) // %2
:
: "q0", "q1", "memory", "cc"
);
}
void ScaleRowDown4Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) {
asm volatile (
"add r4, %0, %3 \n"
"add r5, r4, %3 \n"
"add %3, r5, %3 \n"
".p2align 2 \n"
"1: \n"
"vld1.8 {q0}, [%0]! \n" // load up 16x4
"vld1.8 {q1}, [r4]! \n"
"vld1.8 {q2}, [r5]! \n"
"vld1.8 {q3}, [%3]! \n"
"subs %2, %2, #4 \n"
"vpaddl.u8 q0, q0 \n"
"vpadal.u8 q0, q1 \n"
"vpadal.u8 q0, q2 \n"
"vpadal.u8 q0, q3 \n"
"vpaddl.u16 q0, q0 \n"
"vrshrn.u32 d0, q0, #4 \n" // divide by 16 w/rounding
"vmovn.u16 d0, q0 \n"
"vst1.32 {d0[0]}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1
"+r"(dst_width) // %2
: "r"(src_stride) // %3
: "r4", "r5", "q0", "q1", "q2", "q3", "memory", "cc"
);
}
// Down scale from 4 to 3 pixels. Use the neon multilane read/write
// to load up the every 4th pixel into a 4 different registers.
// Point samples 32 pixels to 24 pixels.
void ScaleRowDown34_NEON(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) {
asm volatile (
".p2align 2 \n"
"1: \n"
"vld4.8 {d0, d1, d2, d3}, [%0]! \n" // src line 0
"subs %2, %2, #24 \n"
"vmov d2, d3 \n" // order d0, d1, d2
"vst3.8 {d0, d1, d2}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1
"+r"(dst_width) // %2
:
: "d0", "d1", "d2", "d3", "memory", "cc"
);
}
void ScaleRowDown34_0_Box_NEON(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) {
asm volatile (
"vmov.u8 d24, #3 \n"
"add %3, %0 \n"
".p2align 2 \n"
"1: \n"
"vld4.8 {d0, d1, d2, d3}, [%0]! \n" // src line 0
"vld4.8 {d4, d5, d6, d7}, [%3]! \n" // src line 1
"subs %2, %2, #24 \n"
// filter src line 0 with src line 1
// expand chars to shorts to allow for room
// when adding lines together
"vmovl.u8 q8, d4 \n"
"vmovl.u8 q9, d5 \n"
"vmovl.u8 q10, d6 \n"
"vmovl.u8 q11, d7 \n"
// 3 * line_0 + line_1
"vmlal.u8 q8, d0, d24 \n"
"vmlal.u8 q9, d1, d24 \n"
"vmlal.u8 q10, d2, d24 \n"
"vmlal.u8 q11, d3, d24 \n"
// (3 * line_0 + line_1) >> 2
"vqrshrn.u16 d0, q8, #2 \n"
"vqrshrn.u16 d1, q9, #2 \n"
"vqrshrn.u16 d2, q10, #2 \n"
"vqrshrn.u16 d3, q11, #2 \n"
// a0 = (src[0] * 3 + s[1] * 1) >> 2
"vmovl.u8 q8, d1 \n"
"vmlal.u8 q8, d0, d24 \n"
"vqrshrn.u16 d0, q8, #2 \n"
// a1 = (src[1] * 1 + s[2] * 1) >> 1
"vrhadd.u8 d1, d1, d2 \n"
// a2 = (src[2] * 1 + s[3] * 3) >> 2
"vmovl.u8 q8, d2 \n"
"vmlal.u8 q8, d3, d24 \n"
"vqrshrn.u16 d2, q8, #2 \n"
"vst3.8 {d0, d1, d2}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1
"+r"(dst_width), // %2
"+r"(src_stride) // %3
:
: "q0", "q1", "q2", "q3", "q8", "q9", "q10", "q11", "d24", "memory", "cc"
);
}
void ScaleRowDown34_1_Box_NEON(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) {
asm volatile (
"vmov.u8 d24, #3 \n"
"add %3, %0 \n"
".p2align 2 \n"
"1: \n"
"vld4.8 {d0, d1, d2, d3}, [%0]! \n" // src line 0
"vld4.8 {d4, d5, d6, d7}, [%3]! \n" // src line 1
"subs %2, %2, #24 \n"
// average src line 0 with src line 1
"vrhadd.u8 q0, q0, q2 \n"
"vrhadd.u8 q1, q1, q3 \n"
// a0 = (src[0] * 3 + s[1] * 1) >> 2
"vmovl.u8 q3, d1 \n"
"vmlal.u8 q3, d0, d24 \n"
"vqrshrn.u16 d0, q3, #2 \n"
// a1 = (src[1] * 1 + s[2] * 1) >> 1
"vrhadd.u8 d1, d1, d2 \n"
// a2 = (src[2] * 1 + s[3] * 3) >> 2
"vmovl.u8 q3, d2 \n"
"vmlal.u8 q3, d3, d24 \n"
"vqrshrn.u16 d2, q3, #2 \n"
"vst3.8 {d0, d1, d2}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1
"+r"(dst_width), // %2
"+r"(src_stride) // %3
:
: "r4", "q0", "q1", "q2", "q3", "d24", "memory", "cc"
);
}
#define HAS_SCALEROWDOWN38_NEON
static uvec8 kShuf38 =
{ 0, 3, 6, 8, 11, 14, 16, 19, 22, 24, 27, 30, 0, 0, 0, 0 };
static uvec8 kShuf38_2 =
{ 0, 8, 16, 2, 10, 17, 4, 12, 18, 6, 14, 19, 0, 0, 0, 0 };
static vec16 kMult38_Div6 =
{ 65536 / 12, 65536 / 12, 65536 / 12, 65536 / 12,
65536 / 12, 65536 / 12, 65536 / 12, 65536 / 12 };
static vec16 kMult38_Div9 =
{ 65536 / 18, 65536 / 18, 65536 / 18, 65536 / 18,
65536 / 18, 65536 / 18, 65536 / 18, 65536 / 18 };
// 32 -> 12
void ScaleRowDown38_NEON(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) {
asm volatile (
"vld1.8 {q3}, [%3] \n"
".p2align 2 \n"
"1: \n"
"vld1.8 {d0, d1, d2, d3}, [%0]! \n"
"subs %2, %2, #12 \n"
"vtbl.u8 d4, {d0, d1, d2, d3}, d6 \n"
"vtbl.u8 d5, {d0, d1, d2, d3}, d7 \n"
"vst1.8 {d4}, [%1]! \n"
"vst1.32 {d5[0]}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1
"+r"(dst_width) // %2
: "r"(&kShuf38) // %3
: "d0", "d1", "d2", "d3", "d4", "d5", "memory", "cc"
);
}
// 32x3 -> 12x1
void OMITFP ScaleRowDown38_3_Box_NEON(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) {
asm volatile (
"vld1.16 {q13}, [%4] \n"
"vld1.8 {q14}, [%5] \n"
"vld1.8 {q15}, [%6] \n"
"add r4, %0, %3, lsl #1 \n"
"add %3, %0 \n"
".p2align 2 \n"
"1: \n"
// d0 = 00 40 01 41 02 42 03 43
// d1 = 10 50 11 51 12 52 13 53
// d2 = 20 60 21 61 22 62 23 63
// d3 = 30 70 31 71 32 72 33 73
"vld4.8 {d0, d1, d2, d3}, [%0]! \n"
"vld4.8 {d4, d5, d6, d7}, [%3]! \n"
"vld4.8 {d16, d17, d18, d19}, [r4]! \n"
"subs %2, %2, #12 \n"
// Shuffle the input data around to get align the data
// so adjacent data can be added. 0,1 - 2,3 - 4,5 - 6,7
// d0 = 00 10 01 11 02 12 03 13
// d1 = 40 50 41 51 42 52 43 53
"vtrn.u8 d0, d1 \n"
"vtrn.u8 d4, d5 \n"
"vtrn.u8 d16, d17 \n"
// d2 = 20 30 21 31 22 32 23 33
// d3 = 60 70 61 71 62 72 63 73
"vtrn.u8 d2, d3 \n"
"vtrn.u8 d6, d7 \n"
"vtrn.u8 d18, d19 \n"
// d0 = 00+10 01+11 02+12 03+13
// d2 = 40+50 41+51 42+52 43+53
"vpaddl.u8 q0, q0 \n"
"vpaddl.u8 q2, q2 \n"
"vpaddl.u8 q8, q8 \n"
// d3 = 60+70 61+71 62+72 63+73
"vpaddl.u8 d3, d3 \n"
"vpaddl.u8 d7, d7 \n"
"vpaddl.u8 d19, d19 \n"
// combine source lines
"vadd.u16 q0, q2 \n"
"vadd.u16 q0, q8 \n"
"vadd.u16 d4, d3, d7 \n"
"vadd.u16 d4, d19 \n"
// dst_ptr[3] = (s[6 + st * 0] + s[7 + st * 0]
// + s[6 + st * 1] + s[7 + st * 1]
// + s[6 + st * 2] + s[7 + st * 2]) / 6
"vqrdmulh.s16 q2, q2, q13 \n"
"vmovn.u16 d4, q2 \n"
// Shuffle 2,3 reg around so that 2 can be added to the
// 0,1 reg and 3 can be added to the 4,5 reg. This
// requires expanding from u8 to u16 as the 0,1 and 4,5
// registers are already expanded. Then do transposes
// to get aligned.
// q2 = xx 20 xx 30 xx 21 xx 31 xx 22 xx 32 xx 23 xx 33
"vmovl.u8 q1, d2 \n"
"vmovl.u8 q3, d6 \n"
"vmovl.u8 q9, d18 \n"
// combine source lines
"vadd.u16 q1, q3 \n"
"vadd.u16 q1, q9 \n"
// d4 = xx 20 xx 30 xx 22 xx 32
// d5 = xx 21 xx 31 xx 23 xx 33
"vtrn.u32 d2, d3 \n"
// d4 = xx 20 xx 21 xx 22 xx 23
// d5 = xx 30 xx 31 xx 32 xx 33
"vtrn.u16 d2, d3 \n"
// 0+1+2, 3+4+5
"vadd.u16 q0, q1 \n"
// Need to divide, but can't downshift as the the value
// isn't a power of 2. So multiply by 65536 / n
// and take the upper 16 bits.
"vqrdmulh.s16 q0, q0, q15 \n"
// Align for table lookup, vtbl requires registers to
// be adjacent
"vmov.u8 d2, d4 \n"
"vtbl.u8 d3, {d0, d1, d2}, d28 \n"
"vtbl.u8 d4, {d0, d1, d2}, d29 \n"
"vst1.8 {d3}, [%1]! \n"
"vst1.32 {d4[0]}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1
"+r"(dst_width), // %2
"+r"(src_stride) // %3
: "r"(&kMult38_Div6), // %4
"r"(&kShuf38_2), // %5
"r"(&kMult38_Div9) // %6
: "r4", "q0", "q1", "q2", "q3", "q8", "q9",
"q13", "q14", "q15", "memory", "cc"
);
}
// 32x2 -> 12x1
void ScaleRowDown38_2_Box_NEON(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) {
asm volatile (
"vld1.16 {q13}, [%4] \n"
"vld1.8 {q14}, [%5] \n"
"add %3, %0 \n"
".p2align 2 \n"
"1: \n"
// d0 = 00 40 01 41 02 42 03 43
// d1 = 10 50 11 51 12 52 13 53
// d2 = 20 60 21 61 22 62 23 63
// d3 = 30 70 31 71 32 72 33 73
"vld4.8 {d0, d1, d2, d3}, [%0]! \n"
"vld4.8 {d4, d5, d6, d7}, [%3]! \n"
"subs %2, %2, #12 \n"
// Shuffle the input data around to get align the data
// so adjacent data can be added. 0,1 - 2,3 - 4,5 - 6,7
// d0 = 00 10 01 11 02 12 03 13
// d1 = 40 50 41 51 42 52 43 53
"vtrn.u8 d0, d1 \n"
"vtrn.u8 d4, d5 \n"
// d2 = 20 30 21 31 22 32 23 33
// d3 = 60 70 61 71 62 72 63 73
"vtrn.u8 d2, d3 \n"
"vtrn.u8 d6, d7 \n"
// d0 = 00+10 01+11 02+12 03+13
// d2 = 40+50 41+51 42+52 43+53
"vpaddl.u8 q0, q0 \n"
"vpaddl.u8 q2, q2 \n"
// d3 = 60+70 61+71 62+72 63+73
"vpaddl.u8 d3, d3 \n"
"vpaddl.u8 d7, d7 \n"
// combine source lines
"vadd.u16 q0, q2 \n"
"vadd.u16 d4, d3, d7 \n"
// dst_ptr[3] = (s[6] + s[7] + s[6+st] + s[7+st]) / 4
"vqrshrn.u16 d4, q2, #2 \n"
// Shuffle 2,3 reg around so that 2 can be added to the
// 0,1 reg and 3 can be added to the 4,5 reg. This
// requires expanding from u8 to u16 as the 0,1 and 4,5
// registers are already expanded. Then do transposes
// to get aligned.
// q2 = xx 20 xx 30 xx 21 xx 31 xx 22 xx 32 xx 23 xx 33
"vmovl.u8 q1, d2 \n"
"vmovl.u8 q3, d6 \n"
// combine source lines
"vadd.u16 q1, q3 \n"
// d4 = xx 20 xx 30 xx 22 xx 32
// d5 = xx 21 xx 31 xx 23 xx 33
"vtrn.u32 d2, d3 \n"
// d4 = xx 20 xx 21 xx 22 xx 23
// d5 = xx 30 xx 31 xx 32 xx 33
"vtrn.u16 d2, d3 \n"
// 0+1+2, 3+4+5
"vadd.u16 q0, q1 \n"
// Need to divide, but can't downshift as the the value
// isn't a power of 2. So multiply by 65536 / n
// and take the upper 16 bits.
"vqrdmulh.s16 q0, q0, q13 \n"
// Align for table lookup, vtbl requires registers to
// be adjacent
"vmov.u8 d2, d4 \n"
"vtbl.u8 d3, {d0, d1, d2}, d28 \n"
"vtbl.u8 d4, {d0, d1, d2}, d29 \n"
"vst1.8 {d3}, [%1]! \n"
"vst1.32 {d4[0]}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1
"+r"(dst_width), // %2
"+r"(src_stride) // %3
: "r"(&kMult38_Div6), // %4
"r"(&kShuf38_2) // %5
: "q0", "q1", "q2", "q3", "q13", "q14", "memory", "cc"
);
}
// 16x2 -> 16x1
void ScaleFilterRows_NEON(uint8* dst_ptr,
const uint8* src_ptr, ptrdiff_t src_stride,
int dst_width, int source_y_fraction) {
asm volatile (
"cmp %4, #0 \n"
"beq 100f \n"
"add %2, %1 \n"
"cmp %4, #64 \n"
"beq 75f \n"
"cmp %4, #128 \n"
"beq 50f \n"
"cmp %4, #192 \n"
"beq 25f \n"
"vdup.8 d5, %4 \n"
"rsb %4, #256 \n"
"vdup.8 d4, %4 \n"
// General purpose row blend.
"1: \n"
"vld1.8 {q0}, [%1]! \n"
"vld1.8 {q1}, [%2]! \n"
"subs %3, %3, #16 \n"
"vmull.u8 q13, d0, d4 \n"
"vmull.u8 q14, d1, d4 \n"
"vmlal.u8 q13, d2, d5 \n"
"vmlal.u8 q14, d3, d5 \n"
"vrshrn.u16 d0, q13, #8 \n"
"vrshrn.u16 d1, q14, #8 \n"
"vst1.8 {q0}, [%0]! \n"
"bgt 1b \n"
"b 99f \n"
// Blend 25 / 75.
"25: \n"
"vld1.8 {q0}, [%1]! \n"
"vld1.8 {q1}, [%2]! \n"
"subs %3, %3, #16 \n"
"vrhadd.u8 q0, q1 \n"
"vrhadd.u8 q0, q1 \n"
"vst1.8 {q0}, [%0]! \n"
"bgt 25b \n"
"b 99f \n"
// Blend 50 / 50.
"50: \n"
"vld1.8 {q0}, [%1]! \n"
"vld1.8 {q1}, [%2]! \n"
"subs %3, %3, #16 \n"
"vrhadd.u8 q0, q1 \n"
"vst1.8 {q0}, [%0]! \n"
"bgt 50b \n"
"b 99f \n"
// Blend 75 / 25.
"75: \n"
"vld1.8 {q1}, [%1]! \n"
"vld1.8 {q0}, [%2]! \n"
"subs %3, %3, #16 \n"
"vrhadd.u8 q0, q1 \n"
"vrhadd.u8 q0, q1 \n"
"vst1.8 {q0}, [%0]! \n"
"bgt 75b \n"
"b 99f \n"
// Blend 100 / 0 - Copy row unchanged.
"100: \n"
"vld1.8 {q0}, [%1]! \n"
"subs %3, %3, #16 \n"
"vst1.8 {q0}, [%0]! \n"
"bgt 100b \n"
"99: \n"
"vst1.8 {d1[7]}, [%0] \n"
: "+r"(dst_ptr), // %0
"+r"(src_ptr), // %1
"+r"(src_stride), // %2
"+r"(dst_width), // %3
"+r"(source_y_fraction) // %4
:
: "q0", "q1", "d4", "d5", "q13", "q14", "memory", "cc"
);
}
void ScaleARGBRowDown2_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) {
asm volatile (
".p2align 2 \n"
"1: \n"
// load even pixels into q0, odd into q1
"vld2.32 {q0, q1}, [%0]! \n"
"vld2.32 {q2, q3}, [%0]! \n"
"subs %2, %2, #8 \n" // 8 processed per loop
"vst1.8 {q1}, [%1]! \n" // store odd pixels
"vst1.8 {q3}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(dst), // %1
"+r"(dst_width) // %2
:
: "memory", "cc", "q0", "q1", "q2", "q3" // Clobber List
);
}
void ScaleARGBRowDown2Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) {
asm volatile (
// change the stride to row 2 pointer
"add %1, %1, %0 \n"
".p2align 2 \n"
"1: \n"
"vld4.8 {d0, d2, d4, d6}, [%0]! \n" // load 8 ARGB pixels.
"vld4.8 {d1, d3, d5, d7}, [%0]! \n" // load next 8 ARGB pixels.
"subs %3, %3, #8 \n" // 8 processed per loop.
"vpaddl.u8 q0, q0 \n" // B 16 bytes -> 8 shorts.
"vpaddl.u8 q1, q1 \n" // G 16 bytes -> 8 shorts.
"vpaddl.u8 q2, q2 \n" // R 16 bytes -> 8 shorts.
"vpaddl.u8 q3, q3 \n" // A 16 bytes -> 8 shorts.
"vld4.8 {d16, d18, d20, d22}, [%1]! \n" // load 8 more ARGB pixels.
"vld4.8 {d17, d19, d21, d23}, [%1]! \n" // load last 8 ARGB pixels.
"vpadal.u8 q0, q8 \n" // B 16 bytes -> 8 shorts.
"vpadal.u8 q1, q9 \n" // G 16 bytes -> 8 shorts.
"vpadal.u8 q2, q10 \n" // R 16 bytes -> 8 shorts.
"vpadal.u8 q3, q11 \n" // A 16 bytes -> 8 shorts.
"vrshrn.u16 d0, q0, #2 \n" // downshift, round and pack
"vrshrn.u16 d1, q1, #2 \n"
"vrshrn.u16 d2, q2, #2 \n"
"vrshrn.u16 d3, q3, #2 \n"
"vst4.8 {d0, d1, d2, d3}, [%2]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(src_stride), // %1
"+r"(dst), // %2
"+r"(dst_width) // %3
:
: "memory", "cc", "q0", "q1", "q2", "q3", "q8", "q9", "q10", "q11"
);
}
// Reads 4 pixels at a time.
// Alignment requirement: src_argb 4 byte aligned.
void ScaleARGBRowDownEven_NEON(const uint8* src_argb, ptrdiff_t src_stride,
int src_stepx, uint8* dst_argb, int dst_width) {
asm volatile (
"mov r12, %3, lsl #2 \n"
".p2align 2 \n"
"1: \n"
"vld1.32 {d0[0]}, [%0], r12 \n"
"vld1.32 {d0[1]}, [%0], r12 \n"
"vld1.32 {d1[0]}, [%0], r12 \n"
"vld1.32 {d1[1]}, [%0], r12 \n"
"subs %2, %2, #4 \n" // 4 pixels per loop.
"vst1.8 {q0}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_argb), // %0
"+r"(dst_argb), // %1
"+r"(dst_width) // %2
: "r"(src_stepx) // %3
: "memory", "cc", "r12", "q0"
);
}
// Reads 4 pixels at a time.
// Alignment requirement: src_argb 4 byte aligned.
void ScaleARGBRowDownEvenBox_NEON(const uint8* src_argb, ptrdiff_t src_stride,
int src_stepx,
uint8* dst_argb, int dst_width) {
asm volatile (
"mov r12, %4, lsl #2 \n"
"add %1, %1, %0 \n"
".p2align 2 \n"
"1: \n"
"vld1.8 {d0}, [%0], r12 \n" // Read 4 2x2 blocks -> 2x1
"vld1.8 {d1}, [%1], r12 \n"
"vld1.8 {d2}, [%0], r12 \n"
"vld1.8 {d3}, [%1], r12 \n"
"vld1.8 {d4}, [%0], r12 \n"
"vld1.8 {d5}, [%1], r12 \n"
"vld1.8 {d6}, [%0], r12 \n"
"vld1.8 {d7}, [%1], r12 \n"
"vaddl.u8 q0, d0, d1 \n"
"vaddl.u8 q1, d2, d3 \n"
"vaddl.u8 q2, d4, d5 \n"
"vaddl.u8 q3, d6, d7 \n"
"vswp.8 d1, d2 \n" // ab_cd -> ac_bd
"vswp.8 d5, d6 \n" // ef_gh -> eg_fh
"vadd.u16 q0, q0, q1 \n" // (a+b)_(c+d)
"vadd.u16 q2, q2, q3 \n" // (e+f)_(g+h)
"vrshrn.u16 d0, q0, #2 \n" // first 2 pixels.
"vrshrn.u16 d1, q2, #2 \n" // next 2 pixels.
"subs %3, %3, #4 \n" // 4 pixels per loop.
"vst1.8 {q0}, [%2]! \n"
"bgt 1b \n"
: "+r"(src_argb), // %0
"+r"(src_stride), // %1
"+r"(dst_argb), // %2
"+r"(dst_width) // %3
: "r"(src_stepx) // %4
: "memory", "cc", "r12", "q0", "q1", "q2", "q3"
);
}
#endif // __ARM_NEON__
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif

1315
third_party/libyuv/source/scale_posix.cc vendored Normal file
View File

File diff suppressed because it is too large Load Diff

1320
third_party/libyuv/source/scale_win.cc vendored Normal file
View File

File diff suppressed because it is too large Load Diff

8
vpxdec.c
View File

@ -873,8 +873,16 @@ int main_loop(int argc, const char **argv_) {
}
if (img->d_w != scaled_img->d_w || img->d_h != scaled_img->d_h) {
#if CONFIG_LIBYUV
vpx_image_scale(img, scaled_img, kFilterBox);
img = scaled_img;
#else
fprintf(stderr, "Failed to scale output frame: %s.\n"
"Scaling is disabled in this configuration. "
"To enable scaling, configure with --enable-libyuv\n",
vpx_codec_error(&decoder));
return EXIT_FAILURE;
#endif
}
}

10
vpxenc.c
View File

@ -1268,6 +1268,7 @@ static void encode_frame(struct stream_state *stream,
fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
exit(EXIT_FAILURE);
}
#if CONFIG_LIBYUV
if (!stream->img)
stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
cfg->g_w, cfg->g_h, 16);
@ -1283,8 +1284,15 @@ static void encode_frame(struct stream_state *stream,
stream->img->stride[VPX_PLANE_V],
stream->img->d_w, stream->img->d_h,
kFilterBox);
img = stream->img;
#else
stream->encoder.err = 1;
ctx_exit_on_error(&stream->encoder,
"Stream %d: Failed to encode frame.\n"
"Scaling disabled in this configuration. \n"
"To enable, configure with --enable-libyuv\n",
stream->index);
#endif
}
vpx_usec_timer_start(&timer);