diff --git a/phonelibs/capnp-cpp/aarch64/bin/capnpc b/phonelibs/capnp-cpp/aarch64/bin/capnpc deleted file mode 120000 index 5668473f09dafe..00000000000000 --- a/phonelibs/capnp-cpp/aarch64/bin/capnpc +++ /dev/null @@ -1 +0,0 @@ -capnp \ No newline at end of file diff --git a/phonelibs/capnp-cpp/aarch64/bin/capnpc b/phonelibs/capnp-cpp/aarch64/bin/capnpc new file mode 100755 index 00000000000000..e60fe7fe3c1ba0 Binary files /dev/null and b/phonelibs/capnp-cpp/aarch64/bin/capnpc differ diff --git a/phonelibs/libyuv/x64/include b/phonelibs/libyuv/x64/include deleted file mode 120000 index f5030fe8899824..00000000000000 --- a/phonelibs/libyuv/x64/include +++ /dev/null @@ -1 +0,0 @@ -../include \ No newline at end of file diff --git a/phonelibs/libyuv/x64/include/libyuv.h b/phonelibs/libyuv/x64/include/libyuv.h new file mode 100644 index 00000000000000..aeffd5ef7a4ca2 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv.h @@ -0,0 +1,32 @@ +/* + * 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_H_ +#define INCLUDE_LIBYUV_H_ + +#include "libyuv/basic_types.h" +#include "libyuv/compare.h" +#include "libyuv/convert.h" +#include "libyuv/convert_argb.h" +#include "libyuv/convert_from.h" +#include "libyuv/convert_from_argb.h" +#include "libyuv/cpu_id.h" +#include "libyuv/mjpeg_decoder.h" +#include "libyuv/planar_functions.h" +#include "libyuv/rotate.h" +#include "libyuv/rotate_argb.h" +#include "libyuv/row.h" +#include "libyuv/scale.h" +#include "libyuv/scale_argb.h" +#include "libyuv/scale_row.h" +#include "libyuv/version.h" +#include "libyuv/video_common.h" + +#endif // INCLUDE_LIBYUV_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/basic_types.h b/phonelibs/libyuv/x64/include/libyuv/basic_types.h new file mode 100644 index 00000000000000..5b760ee0d4d4a2 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/basic_types.h @@ -0,0 +1,118 @@ +/* + * 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_BASIC_TYPES_H_ +#define INCLUDE_LIBYUV_BASIC_TYPES_H_ + +#include // for NULL, size_t + +#if defined(_MSC_VER) && (_MSC_VER < 1600) +#include // for uintptr_t on x86 +#else +#include // for uintptr_t +#endif + +#ifndef GG_LONGLONG +#ifndef INT_TYPES_DEFINED +#define INT_TYPES_DEFINED +#ifdef COMPILER_MSVC +typedef unsigned __int64 uint64; +typedef __int64 int64; +#ifndef INT64_C +#define INT64_C(x) x ## I64 +#endif +#ifndef UINT64_C +#define UINT64_C(x) x ## UI64 +#endif +#define INT64_F "I64" +#else // COMPILER_MSVC +#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 +#ifndef UINT64_C +#define UINT64_C(x) x ## UL +#endif +#define INT64_F "l" +#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 +#ifndef UINT64_C +#define UINT64_C(x) x ## ULL +#endif +#define INT64_F "ll" +#endif // __LP64__ +#endif // COMPILER_MSVC +typedef unsigned int uint32; +typedef int int32; +typedef unsigned short uint16; // NOLINT +typedef short int16; // NOLINT +typedef unsigned char uint8; +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) \ + (reinterpret_cast(((reinterpret_cast(p) + \ + ((t) - 1)) & ~((t) - 1)))) +#else +#define ALIGNP(p, t) \ + ((uint8*)((((uintptr_t)(p) + ((t) - 1)) & ~((t) - 1)))) /* NOLINT */ +#endif +#endif + +#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_ diff --git a/phonelibs/libyuv/x64/include/libyuv/compare.h b/phonelibs/libyuv/x64/include/libyuv/compare.h new file mode 100644 index 00000000000000..550712de6e589b --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/compare.h @@ -0,0 +1,78 @@ +/* + * 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_COMPARE_H_ +#define INCLUDE_LIBYUV_COMPARE_H_ + +#include "libyuv/basic_types.h" + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +// Compute a hash for specified memory. Seed of 5381 recommended. +LIBYUV_API +uint32 HashDjb2(const uint8* src, uint64 count, uint32 seed); + +// Scan an opaque argb image and return fourcc based on alpha offset. +// Returns FOURCC_ARGB, FOURCC_BGRA, or 0 if unknown. +LIBYUV_API +uint32 ARGBDetect(const uint8* argb, int stride_argb, int width, int height); + +// Sum Square Error - used to compute Mean Square Error or PSNR. +LIBYUV_API +uint64 ComputeSumSquareError(const uint8* src_a, + const uint8* src_b, int count); + +LIBYUV_API +uint64 ComputeSumSquareErrorPlane(const uint8* src_a, int stride_a, + const uint8* src_b, int stride_b, + int width, int height); + +static const int kMaxPsnr = 128; + +LIBYUV_API +double SumSquareErrorToPsnr(uint64 sse, uint64 count); + +LIBYUV_API +double CalcFramePsnr(const uint8* src_a, int stride_a, + const uint8* src_b, int stride_b, + int width, int height); + +LIBYUV_API +double I420Psnr(const uint8* src_y_a, int stride_y_a, + const uint8* src_u_a, int stride_u_a, + const uint8* src_v_a, int stride_v_a, + const uint8* src_y_b, int stride_y_b, + const uint8* src_u_b, int stride_u_b, + const uint8* src_v_b, int stride_v_b, + int width, int height); + +LIBYUV_API +double CalcFrameSsim(const uint8* src_a, int stride_a, + const uint8* src_b, int stride_b, + int width, int height); + +LIBYUV_API +double I420Ssim(const uint8* src_y_a, int stride_y_a, + const uint8* src_u_a, int stride_u_a, + const uint8* src_v_a, int stride_v_a, + const uint8* src_y_b, int stride_y_b, + const uint8* src_u_b, int stride_u_b, + const uint8* src_v_b, int stride_v_b, + int width, int height); + +#ifdef __cplusplus +} // extern "C" +} // namespace libyuv +#endif + +#endif // INCLUDE_LIBYUV_COMPARE_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/compare_row.h b/phonelibs/libyuv/x64/include/libyuv/compare_row.h new file mode 100644 index 00000000000000..781cad3e65aba0 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/compare_row.h @@ -0,0 +1,84 @@ +/* + * 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_COMPARE_ROW_H_ +#define INCLUDE_LIBYUV_COMPARE_ROW_H_ + +#include "libyuv/basic_types.h" + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +#if defined(__pnacl__) || defined(__CLR_VER) || \ + (defined(__i386__) && !defined(__SSE2__)) +#define LIBYUV_DISABLE_X86 +#endif +// MemorySanitizer does not support assembly code yet. http://crbug.com/344505 +#if defined(__has_feature) +#if __has_feature(memory_sanitizer) +#define LIBYUV_DISABLE_X86 +#endif +#endif + +// Visual C 2012 required for AVX2. +#if defined(_M_IX86) && !defined(__clang__) && \ + defined(_MSC_VER) && _MSC_VER >= 1700 +#define VISUALC_HAS_AVX2 1 +#endif // VisualStudio >= 2012 + +// clang >= 3.4.0 required for AVX2. +#if defined(__clang__) && (defined(__x86_64__) || defined(__i386__)) +#if (__clang_major__ > 3) || (__clang_major__ == 3 && (__clang_minor__ >= 4)) +#define CLANG_HAS_AVX2 1 +#endif // clang >= 3.4 +#endif // __clang__ + +#if !defined(LIBYUV_DISABLE_X86) && \ + defined(_M_IX86) && (defined(VISUALC_HAS_AVX2) || defined(CLANG_HAS_AVX2)) +#define HAS_HASHDJB2_AVX2 +#endif + +// The following are available for Visual C and GCC: +#if !defined(LIBYUV_DISABLE_X86) && \ + (defined(__x86_64__) || (defined(__i386__) || defined(_M_IX86))) +#define HAS_HASHDJB2_SSE41 +#define HAS_SUMSQUAREERROR_SSE2 +#endif + +// The following are available for Visual C and clangcl 32 bit: +#if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && \ + (defined(VISUALC_HAS_AVX2) || defined(CLANG_HAS_AVX2)) +#define HAS_HASHDJB2_AVX2 +#define HAS_SUMSQUAREERROR_AVX2 +#endif + +// The following are available for Neon: +#if !defined(LIBYUV_DISABLE_NEON) && \ + (defined(__ARM_NEON__) || defined(LIBYUV_NEON) || defined(__aarch64__)) +#define HAS_SUMSQUAREERROR_NEON +#endif + +uint32 SumSquareError_C(const uint8* src_a, const uint8* src_b, int count); +uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b, int count); +uint32 SumSquareError_AVX2(const uint8* src_a, const uint8* src_b, int count); +uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b, int count); + +uint32 HashDjb2_C(const uint8* src, int count, uint32 seed); +uint32 HashDjb2_SSE41(const uint8* src, int count, uint32 seed); +uint32 HashDjb2_AVX2(const uint8* src, int count, uint32 seed); + +#ifdef __cplusplus +} // extern "C" +} // namespace libyuv +#endif + +#endif // INCLUDE_LIBYUV_COMPARE_ROW_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/convert.h b/phonelibs/libyuv/x64/include/libyuv/convert.h new file mode 100644 index 00000000000000..d44485847be1d6 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/convert.h @@ -0,0 +1,259 @@ +/* + * 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_CONVERT_H_ +#define INCLUDE_LIBYUV_CONVERT_H_ + +#include "libyuv/basic_types.h" + +#include "libyuv/rotate.h" // For enum RotationMode. + +// TODO(fbarchard): fix WebRTC source to include following libyuv headers: +#include "libyuv/convert_argb.h" // For WebRTC I420ToARGB. b/620 +#include "libyuv/convert_from.h" // For WebRTC ConvertFromI420. b/620 +#include "libyuv/planar_functions.h" // For WebRTC I420Rect, CopyPlane. b/618 + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +// Convert I444 to I420. +LIBYUV_API +int I444ToI420(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 I422 to I420. +LIBYUV_API +int I422ToI420(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 I411 to I420. +LIBYUV_API +int I411ToI420(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 I420 to I420. +#define I420ToI420 I420Copy +LIBYUV_API +int I420Copy(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 I400 (grey) to I420. +LIBYUV_API +int I400ToI420(const uint8* src_y, int src_stride_y, + 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); + +#define J400ToJ420 I400ToI420 + +// Convert NV12 to I420. +LIBYUV_API +int NV12ToI420(const uint8* src_y, int src_stride_y, + const uint8* src_uv, int src_stride_uv, + 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 NV21 to I420. +LIBYUV_API +int NV21ToI420(const uint8* src_y, int src_stride_y, + const uint8* src_vu, int src_stride_vu, + 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 I420. +LIBYUV_API +int YUY2ToI420(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 I420. +LIBYUV_API +int UYVYToI420(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 M420 to I420. +LIBYUV_API +int M420ToI420(const uint8* src_m420, int src_stride_m420, + 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 Android420 to I420. +LIBYUV_API +int Android420ToI420(const uint8* src_y, int src_stride_y, + const uint8* src_u, int src_stride_u, + const uint8* src_v, int src_stride_v, + int pixel_stride_uv, + 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); + +// ARGB little endian (bgra in memory) to I420. +LIBYUV_API +int ARGBToI420(const uint8* src_frame, int src_stride_frame, + 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); + +// BGRA little endian (argb in memory) to I420. +LIBYUV_API +int BGRAToI420(const uint8* src_frame, int src_stride_frame, + 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); + +// ABGR little endian (rgba in memory) to I420. +LIBYUV_API +int ABGRToI420(const uint8* src_frame, int src_stride_frame, + 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); + +// RGBA little endian (abgr in memory) to I420. +LIBYUV_API +int RGBAToI420(const uint8* src_frame, int src_stride_frame, + 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); + +// RGB little endian (bgr in memory) to I420. +LIBYUV_API +int RGB24ToI420(const uint8* src_frame, int src_stride_frame, + 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); + +// RGB big endian (rgb in memory) to I420. +LIBYUV_API +int RAWToI420(const uint8* src_frame, int src_stride_frame, + 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); + +// RGB16 (RGBP fourcc) little endian to I420. +LIBYUV_API +int RGB565ToI420(const uint8* src_frame, int src_stride_frame, + 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); + +// RGB15 (RGBO fourcc) little endian to I420. +LIBYUV_API +int ARGB1555ToI420(const uint8* src_frame, int src_stride_frame, + 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); + +// RGB12 (R444 fourcc) little endian to I420. +LIBYUV_API +int ARGB4444ToI420(const uint8* src_frame, int src_stride_frame, + 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); + +#ifdef HAVE_JPEG +// src_width/height provided by capture. +// dst_width/height for clipping determine final size. +LIBYUV_API +int MJPGToI420(const uint8* sample, size_t sample_size, + uint8* dst_y, int dst_stride_y, + uint8* dst_u, int dst_stride_u, + uint8* dst_v, int dst_stride_v, + int src_width, int src_height, + int dst_width, int dst_height); + +// Query size of MJPG in pixels. +LIBYUV_API +int MJPGSize(const uint8* sample, size_t sample_size, + int* width, int* height); +#endif + +// Convert camera sample to I420 with cropping, rotation and vertical flip. +// "src_size" is needed to parse MJPG. +// "dst_stride_y" number of bytes in a row of the dst_y plane. +// Normally this would be the same as dst_width, with recommended alignment +// to 16 bytes for better efficiency. +// If rotation of 90 or 270 is used, stride is affected. The caller should +// allocate the I420 buffer according to rotation. +// "dst_stride_u" number of bytes in a row of the dst_u plane. +// Normally this would be the same as (dst_width + 1) / 2, with +// recommended alignment to 16 bytes for better efficiency. +// If rotation of 90 or 270 is used, stride is affected. +// "crop_x" and "crop_y" are starting position for cropping. +// To center, crop_x = (src_width - dst_width) / 2 +// crop_y = (src_height - dst_height) / 2 +// "src_width" / "src_height" is size of src_frame in pixels. +// "src_height" can be negative indicating a vertically flipped image source. +// "crop_width" / "crop_height" is the size to crop the src to. +// Must be less than or equal to src_width/src_height +// Cropping parameters are pre-rotation. +// "rotation" can be 0, 90, 180 or 270. +// "format" is a fourcc. ie 'I420', 'YUY2' +// Returns 0 for successful; -1 for invalid parameter. Non-zero for failure. +LIBYUV_API +int ConvertToI420(const uint8* src_frame, size_t src_size, + uint8* dst_y, int dst_stride_y, + uint8* dst_u, int dst_stride_u, + uint8* dst_v, int dst_stride_v, + int crop_x, int crop_y, + int src_width, int src_height, + int crop_width, int crop_height, + enum RotationMode rotation, + uint32 format); + +#ifdef __cplusplus +} // extern "C" +} // namespace libyuv +#endif + +#endif // INCLUDE_LIBYUV_CONVERT_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/convert_argb.h b/phonelibs/libyuv/x64/include/libyuv/convert_argb.h new file mode 100644 index 00000000000000..dc03ac8d5dc4e1 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/convert_argb.h @@ -0,0 +1,319 @@ +/* + * Copyright 2012 The LibYuv Project Authors. All rights reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef INCLUDE_LIBYUV_CONVERT_ARGB_H_ +#define INCLUDE_LIBYUV_CONVERT_ARGB_H_ + +#include "libyuv/basic_types.h" + +#include "libyuv/rotate.h" // For enum RotationMode. + +// TODO(fbarchard): This set of functions should exactly match convert.h +// TODO(fbarchard): Add tests. Create random content of right size and convert +// with C vs Opt and or to I420 and compare. +// TODO(fbarchard): Some of these functions lack parameter setting. + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +// Alias. +#define ARGBToARGB ARGBCopy + +// 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); + +// Convert I420 to ARGB. +LIBYUV_API +int I420ToARGB(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_argb, int dst_stride_argb, + int width, int height); + +// Duplicate prototype for function in convert_from.h for remoting. +LIBYUV_API +int I420ToABGR(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_argb, int dst_stride_argb, + int width, int height); + +// Convert I422 to ARGB. +LIBYUV_API +int I422ToARGB(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_argb, int dst_stride_argb, + int width, int height); + +// Convert I444 to ARGB. +LIBYUV_API +int I444ToARGB(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_argb, int dst_stride_argb, + int width, int height); + +// Convert J444 to ARGB. +LIBYUV_API +int J444ToARGB(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_argb, int dst_stride_argb, + int width, int height); + +// Convert I444 to ABGR. +LIBYUV_API +int I444ToABGR(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 I411 to ARGB. +LIBYUV_API +int I411ToARGB(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_argb, int dst_stride_argb, + int width, int height); + +// Convert I420 with Alpha to preattenuated ARGB. +LIBYUV_API +int I420AlphaToARGB(const uint8* src_y, int src_stride_y, + const uint8* src_u, int src_stride_u, + const uint8* src_v, int src_stride_v, + const uint8* src_a, int src_stride_a, + uint8* dst_argb, int dst_stride_argb, + int width, int height, int attenuate); + +// Convert I420 with Alpha to preattenuated ABGR. +LIBYUV_API +int I420AlphaToABGR(const uint8* src_y, int src_stride_y, + const uint8* src_u, int src_stride_u, + const uint8* src_v, int src_stride_v, + const uint8* src_a, int src_stride_a, + uint8* dst_abgr, int dst_stride_abgr, + int width, int height, int attenuate); + +// Convert I400 (grey) to ARGB. Reverse of ARGBToI400. +LIBYUV_API +int I400ToARGB(const uint8* src_y, int src_stride_y, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// Convert J400 (jpeg grey) to ARGB. +LIBYUV_API +int J400ToARGB(const uint8* src_y, int src_stride_y, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// Alias. +#define YToARGB I400ToARGB + +// Convert NV12 to ARGB. +LIBYUV_API +int NV12ToARGB(const uint8* src_y, int src_stride_y, + const uint8* src_uv, int src_stride_uv, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// Convert NV21 to ARGB. +LIBYUV_API +int NV21ToARGB(const uint8* src_y, int src_stride_y, + const uint8* src_vu, int src_stride_vu, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// Convert M420 to ARGB. +LIBYUV_API +int M420ToARGB(const uint8* src_m420, int src_stride_m420, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// Convert YUY2 to ARGB. +LIBYUV_API +int YUY2ToARGB(const uint8* src_yuy2, int src_stride_yuy2, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// Convert UYVY to ARGB. +LIBYUV_API +int UYVYToARGB(const uint8* src_uyvy, int src_stride_uyvy, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// Convert J420 to ARGB. +LIBYUV_API +int J420ToARGB(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_argb, int dst_stride_argb, + int width, int height); + +// Convert J422 to ARGB. +LIBYUV_API +int J422ToARGB(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_argb, int dst_stride_argb, + int width, int height); + +// Convert J420 to ABGR. +LIBYUV_API +int J420ToABGR(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 J422 to ABGR. +LIBYUV_API +int J422ToABGR(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 H420 to ARGB. +LIBYUV_API +int H420ToARGB(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_argb, int dst_stride_argb, + int width, int height); + +// Convert H422 to ARGB. +LIBYUV_API +int H422ToARGB(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_argb, int dst_stride_argb, + int width, int height); + +// Convert H420 to ABGR. +LIBYUV_API +int H420ToABGR(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 H422 to ABGR. +LIBYUV_API +int H422ToABGR(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); + +// BGRA little endian (argb in memory) to ARGB. +LIBYUV_API +int BGRAToARGB(const uint8* src_frame, int src_stride_frame, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// ABGR little endian (rgba in memory) to ARGB. +LIBYUV_API +int ABGRToARGB(const uint8* src_frame, int src_stride_frame, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// RGBA little endian (abgr in memory) to ARGB. +LIBYUV_API +int RGBAToARGB(const uint8* src_frame, int src_stride_frame, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// Deprecated function name. +#define BG24ToARGB RGB24ToARGB + +// RGB little endian (bgr in memory) to ARGB. +LIBYUV_API +int RGB24ToARGB(const uint8* src_frame, int src_stride_frame, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// RGB big endian (rgb in memory) to ARGB. +LIBYUV_API +int RAWToARGB(const uint8* src_frame, int src_stride_frame, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// RGB16 (RGBP fourcc) little endian to ARGB. +LIBYUV_API +int RGB565ToARGB(const uint8* src_frame, int src_stride_frame, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// RGB15 (RGBO fourcc) little endian to ARGB. +LIBYUV_API +int ARGB1555ToARGB(const uint8* src_frame, int src_stride_frame, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// RGB12 (R444 fourcc) little endian to ARGB. +LIBYUV_API +int ARGB4444ToARGB(const uint8* src_frame, int src_stride_frame, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +#ifdef HAVE_JPEG +// src_width/height provided by capture +// dst_width/height for clipping determine final size. +LIBYUV_API +int MJPGToARGB(const uint8* sample, size_t sample_size, + uint8* dst_argb, int dst_stride_argb, + int src_width, int src_height, + int dst_width, int dst_height); +#endif + +// Convert camera sample to ARGB with cropping, rotation and vertical flip. +// "src_size" is needed to parse MJPG. +// "dst_stride_argb" number of bytes in a row of the dst_argb plane. +// Normally this would be the same as dst_width, with recommended alignment +// to 16 bytes for better efficiency. +// If rotation of 90 or 270 is used, stride is affected. The caller should +// allocate the I420 buffer according to rotation. +// "dst_stride_u" number of bytes in a row of the dst_u plane. +// Normally this would be the same as (dst_width + 1) / 2, with +// recommended alignment to 16 bytes for better efficiency. +// If rotation of 90 or 270 is used, stride is affected. +// "crop_x" and "crop_y" are starting position for cropping. +// To center, crop_x = (src_width - dst_width) / 2 +// crop_y = (src_height - dst_height) / 2 +// "src_width" / "src_height" is size of src_frame in pixels. +// "src_height" can be negative indicating a vertically flipped image source. +// "crop_width" / "crop_height" is the size to crop the src to. +// Must be less than or equal to src_width/src_height +// Cropping parameters are pre-rotation. +// "rotation" can be 0, 90, 180 or 270. +// "format" is a fourcc. ie 'I420', 'YUY2' +// Returns 0 for successful; -1 for invalid parameter. Non-zero for failure. +LIBYUV_API +int ConvertToARGB(const uint8* src_frame, size_t src_size, + uint8* dst_argb, int dst_stride_argb, + int crop_x, int crop_y, + int src_width, int src_height, + int crop_width, int crop_height, + enum RotationMode rotation, + uint32 format); + +#ifdef __cplusplus +} // extern "C" +} // namespace libyuv +#endif + +#endif // INCLUDE_LIBYUV_CONVERT_ARGB_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/convert_from.h b/phonelibs/libyuv/x64/include/libyuv/convert_from.h new file mode 100644 index 00000000000000..59c40474f1ef24 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/convert_from.h @@ -0,0 +1,179 @@ +/* + * 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_CONVERT_FROM_H_ +#define INCLUDE_LIBYUV_CONVERT_FROM_H_ + +#include "libyuv/basic_types.h" +#include "libyuv/rotate.h" + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +// See Also convert.h for conversions from formats to I420. + +// I420Copy in convert to I420ToI420. + +LIBYUV_API +int I420ToI422(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); + +LIBYUV_API +int I420ToI444(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); + +LIBYUV_API +int I420ToI411(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 to I400. Source can be I420, I422, I444, I400, NV12 or NV21. +LIBYUV_API +int I400Copy(const uint8* src_y, int src_stride_y, + uint8* dst_y, int dst_stride_y, + int width, int height); + +LIBYUV_API +int I420ToNV12(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_uv, int dst_stride_uv, + int width, int height); + +LIBYUV_API +int I420ToNV21(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_vu, int dst_stride_vu, + int width, int height); + +LIBYUV_API +int I420ToYUY2(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); + +LIBYUV_API +int I420ToUYVY(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); + +LIBYUV_API +int I420ToARGB(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_argb, int dst_stride_argb, + int width, int height); + +LIBYUV_API +int I420ToBGRA(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_argb, int dst_stride_argb, + int width, int height); + +LIBYUV_API +int I420ToABGR(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_argb, int dst_stride_argb, + int width, int height); + +LIBYUV_API +int I420ToRGBA(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); + +LIBYUV_API +int I420ToRGB24(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); + +LIBYUV_API +int I420ToRAW(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); + +LIBYUV_API +int I420ToRGB565(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 I420 To RGB565 with 4x4 dither matrix (16 bytes). +// Values in dither matrix from 0 to 7 recommended. +// The order of the dither matrix is first byte is upper left. + +LIBYUV_API +int I420ToRGB565Dither(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, + const uint8* dither4x4, int width, int height); + +LIBYUV_API +int I420ToARGB1555(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); + +LIBYUV_API +int I420ToARGB4444(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 I420 to specified format. +// "dst_sample_stride" is bytes in a row for the destination. Pass 0 if the +// buffer has contiguous rows. Can be negative. A multiple of 16 is optimal. +LIBYUV_API +int ConvertFromI420(const uint8* y, int y_stride, + const uint8* u, int u_stride, + const uint8* v, int v_stride, + uint8* dst_sample, int dst_sample_stride, + int width, int height, + uint32 format); + +#ifdef __cplusplus +} // extern "C" +} // namespace libyuv +#endif + +#endif // INCLUDE_LIBYUV_CONVERT_FROM_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/convert_from_argb.h b/phonelibs/libyuv/x64/include/libyuv/convert_from_argb.h new file mode 100644 index 00000000000000..8d7f02f8c4dde6 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/convert_from_argb.h @@ -0,0 +1,190 @@ +/* + * Copyright 2012 The LibYuv Project Authors. All rights reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef INCLUDE_LIBYUV_CONVERT_FROM_ARGB_H_ +#define INCLUDE_LIBYUV_CONVERT_FROM_ARGB_H_ + +#include "libyuv/basic_types.h" + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +// Copy ARGB to ARGB. +#define ARGBToARGB ARGBCopy +LIBYUV_API +int ARGBCopy(const uint8* src_argb, int src_stride_argb, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// Convert ARGB To BGRA. +LIBYUV_API +int ARGBToBGRA(const uint8* src_argb, int src_stride_argb, + uint8* dst_bgra, int dst_stride_bgra, + int width, int height); + +// Convert ARGB To ABGR. +LIBYUV_API +int ARGBToABGR(const uint8* src_argb, int src_stride_argb, + uint8* dst_abgr, int dst_stride_abgr, + int width, int height); + +// Convert ARGB To RGBA. +LIBYUV_API +int ARGBToRGBA(const uint8* src_argb, int src_stride_argb, + uint8* dst_rgba, int dst_stride_rgba, + int width, int height); + +// Convert ARGB To RGB24. +LIBYUV_API +int ARGBToRGB24(const uint8* src_argb, int src_stride_argb, + uint8* dst_rgb24, int dst_stride_rgb24, + int width, int height); + +// Convert ARGB To RAW. +LIBYUV_API +int ARGBToRAW(const uint8* src_argb, int src_stride_argb, + uint8* dst_rgb, int dst_stride_rgb, + int width, int height); + +// Convert ARGB To RGB565. +LIBYUV_API +int ARGBToRGB565(const uint8* src_argb, int src_stride_argb, + uint8* dst_rgb565, int dst_stride_rgb565, + int width, int height); + +// Convert ARGB To RGB565 with 4x4 dither matrix (16 bytes). +// Values in dither matrix from 0 to 7 recommended. +// The order of the dither matrix is first byte is upper left. +// TODO(fbarchard): Consider pointer to 2d array for dither4x4. +// const uint8(*dither)[4][4]; +LIBYUV_API +int ARGBToRGB565Dither(const uint8* src_argb, int src_stride_argb, + uint8* dst_rgb565, int dst_stride_rgb565, + const uint8* dither4x4, int width, int height); + +// Convert ARGB To ARGB1555. +LIBYUV_API +int ARGBToARGB1555(const uint8* src_argb, int src_stride_argb, + uint8* dst_argb1555, int dst_stride_argb1555, + int width, int height); + +// Convert ARGB To ARGB4444. +LIBYUV_API +int ARGBToARGB4444(const uint8* src_argb, int src_stride_argb, + uint8* dst_argb4444, int dst_stride_argb4444, + int width, int height); + +// Convert ARGB To I444. +LIBYUV_API +int ARGBToI444(const uint8* src_argb, int src_stride_argb, + uint8* dst_y, int dst_stride_y, + uint8* dst_u, int dst_stride_u, + uint8* dst_v, int dst_stride_v, + int width, int height); + +// Convert ARGB To I422. +LIBYUV_API +int ARGBToI422(const uint8* src_argb, int src_stride_argb, + uint8* dst_y, int dst_stride_y, + uint8* dst_u, int dst_stride_u, + uint8* dst_v, int dst_stride_v, + int width, int height); + +// Convert ARGB To I420. (also in convert.h) +LIBYUV_API +int ARGBToI420(const uint8* src_argb, int src_stride_argb, + uint8* dst_y, int dst_stride_y, + uint8* dst_u, int dst_stride_u, + uint8* dst_v, int dst_stride_v, + int width, int height); + +// Convert ARGB to J420. (JPeg full range I420). +LIBYUV_API +int ARGBToJ420(const uint8* src_argb, int src_stride_argb, + uint8* dst_yj, int dst_stride_yj, + uint8* dst_u, int dst_stride_u, + uint8* dst_v, int dst_stride_v, + int width, int height); + +// Convert ARGB to J422. +LIBYUV_API +int ARGBToJ422(const uint8* src_argb, int src_stride_argb, + uint8* dst_yj, int dst_stride_yj, + uint8* dst_u, int dst_stride_u, + uint8* dst_v, int dst_stride_v, + int width, int height); + +// Convert ARGB To I411. +LIBYUV_API +int ARGBToI411(const uint8* src_argb, int src_stride_argb, + uint8* dst_y, int dst_stride_y, + uint8* dst_u, int dst_stride_u, + uint8* dst_v, int dst_stride_v, + int width, int height); + +// Convert ARGB to J400. (JPeg full range). +LIBYUV_API +int ARGBToJ400(const uint8* src_argb, int src_stride_argb, + uint8* dst_yj, int dst_stride_yj, + int width, int height); + +// Convert ARGB to I400. +LIBYUV_API +int ARGBToI400(const uint8* src_argb, int src_stride_argb, + uint8* dst_y, int dst_stride_y, + int width, int height); + +// Convert ARGB to G. (Reverse of J400toARGB, which replicates G back to ARGB) +LIBYUV_API +int ARGBToG(const uint8* src_argb, int src_stride_argb, + uint8* dst_g, int dst_stride_g, + int width, int height); + +// Convert ARGB To NV12. +LIBYUV_API +int ARGBToNV12(const uint8* src_argb, int src_stride_argb, + uint8* dst_y, int dst_stride_y, + uint8* dst_uv, int dst_stride_uv, + int width, int height); + +// Convert ARGB To NV21. +LIBYUV_API +int ARGBToNV21(const uint8* src_argb, int src_stride_argb, + uint8* dst_y, int dst_stride_y, + uint8* dst_vu, int dst_stride_vu, + int width, int height); + +// Convert ARGB To NV21. +LIBYUV_API +int ARGBToNV21(const uint8* src_argb, int src_stride_argb, + uint8* dst_y, int dst_stride_y, + uint8* dst_vu, int dst_stride_vu, + int width, int height); + +// Convert ARGB To YUY2. +LIBYUV_API +int ARGBToYUY2(const uint8* src_argb, int src_stride_argb, + uint8* dst_yuy2, int dst_stride_yuy2, + int width, int height); + +// Convert ARGB To UYVY. +LIBYUV_API +int ARGBToUYVY(const uint8* src_argb, int src_stride_argb, + uint8* dst_uyvy, int dst_stride_uyvy, + int width, int height); + +#ifdef __cplusplus +} // extern "C" +} // namespace libyuv +#endif + +#endif // INCLUDE_LIBYUV_CONVERT_FROM_ARGB_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/cpu_id.h b/phonelibs/libyuv/x64/include/libyuv/cpu_id.h new file mode 100644 index 00000000000000..7c6c9aeb00515d --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/cpu_id.h @@ -0,0 +1,81 @@ +/* + * 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_CPU_ID_H_ +#define INCLUDE_LIBYUV_CPU_ID_H_ + +#include "libyuv/basic_types.h" + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +// Internal flag to indicate cpuid requires initialization. +static const int kCpuInitialized = 0x1; + +// These flags are only valid on ARM processors. +static const int kCpuHasARM = 0x2; +static const int kCpuHasNEON = 0x4; +// 0x8 reserved for future ARM flag. + +// 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; +static const int kCpuHasAVX3 = 0x2000; +// 0x2000, 0x4000, 0x8000 reserved for future X86 flags. + +// These flags are only valid on MIPS processors. +static const int kCpuHasMIPS = 0x10000; +static const int kCpuHasDSPR2 = 0x20000; +static const int kCpuHasMSA = 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. +// returns non-zero if instruction set is detected +static __inline int TestCpuFlag(int test_flag) { + LIBYUV_API extern int cpu_info_; + return (!cpu_info_ ? InitCpuFlags() : cpu_info_) & test_flag; +} + +// For testing, allow CPU flags to be disabled. +// ie MaskCpuFlags(~kCpuHasSSSE3) to disable SSSE3. +// MaskCpuFlags(-1) to enable all cpu specific optimizations. +// MaskCpuFlags(1) 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_ diff --git a/phonelibs/libyuv/x64/include/libyuv/macros_msa.h b/phonelibs/libyuv/x64/include/libyuv/macros_msa.h new file mode 100644 index 00000000000000..92ed21c38532d4 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/macros_msa.h @@ -0,0 +1,76 @@ +/* + * Copyright 2016 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_MACROS_MSA_H_ +#define INCLUDE_LIBYUV_MACROS_MSA_H_ + +#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa) +#include +#include + +#define LD_B(RTYPE, psrc) *((RTYPE*)(psrc)) /* NOLINT */ +#define LD_UB(...) LD_B(v16u8, __VA_ARGS__) + +#define ST_B(RTYPE, in, pdst) *((RTYPE*)(pdst)) = (in) /* NOLINT */ +#define ST_UB(...) ST_B(v16u8, __VA_ARGS__) + +/* Description : Load two vectors with 16 'byte' sized elements + Arguments : Inputs - psrc, stride + Outputs - out0, out1 + Return Type - as per RTYPE + Details : Load 16 byte elements in 'out0' from (psrc) + Load 16 byte elements in 'out1' from (psrc + stride) +*/ +#define LD_B2(RTYPE, psrc, stride, out0, out1) { \ + out0 = LD_B(RTYPE, (psrc)); \ + out1 = LD_B(RTYPE, (psrc) + stride); \ +} +#define LD_UB2(...) LD_B2(v16u8, __VA_ARGS__) + +#define LD_B4(RTYPE, psrc, stride, out0, out1, out2, out3) { \ + LD_B2(RTYPE, (psrc), stride, out0, out1); \ + LD_B2(RTYPE, (psrc) + 2 * stride , stride, out2, out3); \ +} +#define LD_UB4(...) LD_B4(v16u8, __VA_ARGS__) + +/* Description : Store two vectors with stride each having 16 'byte' sized + elements + Arguments : Inputs - in0, in1, pdst, stride + Details : Store 16 byte elements from 'in0' to (pdst) + Store 16 byte elements from 'in1' to (pdst + stride) +*/ +#define ST_B2(RTYPE, in0, in1, pdst, stride) { \ + ST_B(RTYPE, in0, (pdst)); \ + ST_B(RTYPE, in1, (pdst) + stride); \ +} +#define ST_UB2(...) ST_B2(v16u8, __VA_ARGS__) +# +#define ST_B4(RTYPE, in0, in1, in2, in3, pdst, stride) { \ + ST_B2(RTYPE, in0, in1, (pdst), stride); \ + ST_B2(RTYPE, in2, in3, (pdst) + 2 * stride, stride); \ +} +#define ST_UB4(...) ST_B4(v16u8, __VA_ARGS__) +# +/* Description : Shuffle byte vector elements as per mask vector + Arguments : Inputs - in0, in1, in2, in3, mask0, mask1 + Outputs - out0, out1 + Return Type - as per RTYPE + Details : Byte elements from 'in0' & 'in1' are copied selectively to + 'out0' as per control vector 'mask0' +*/ +#define VSHF_B2(RTYPE, in0, in1, in2, in3, mask0, mask1, out0, out1) { \ + out0 = (RTYPE) __msa_vshf_b((v16i8) mask0, (v16i8) in1, (v16i8) in0); \ + out1 = (RTYPE) __msa_vshf_b((v16i8) mask1, (v16i8) in3, (v16i8) in2); \ +} +#define VSHF_B2_UB(...) VSHF_B2(v16u8, __VA_ARGS__) + +#endif /* !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa) */ + +#endif // INCLUDE_LIBYUV_MACROS_MSA_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/mjpeg_decoder.h b/phonelibs/libyuv/x64/include/libyuv/mjpeg_decoder.h new file mode 100644 index 00000000000000..4975bae5b76ef1 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/mjpeg_decoder.h @@ -0,0 +1,192 @@ +/* + * Copyright 2012 The LibYuv Project Authors. All rights reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef INCLUDE_LIBYUV_MJPEG_DECODER_H_ +#define INCLUDE_LIBYUV_MJPEG_DECODER_H_ + +#include "libyuv/basic_types.h" + +#ifdef __cplusplus +// NOTE: For a simplified public API use convert.h MJPGToI420(). + +struct jpeg_common_struct; +struct jpeg_decompress_struct; +struct jpeg_source_mgr; + +namespace libyuv { + +#ifdef __cplusplus +extern "C" { +#endif + +LIBYUV_BOOL ValidateJpeg(const uint8* sample, size_t sample_size); + +#ifdef __cplusplus +} // extern "C" +#endif + +static const uint32 kUnknownDataSize = 0xFFFFFFFF; + +enum JpegSubsamplingType { + kJpegYuv420, + kJpegYuv422, + kJpegYuv411, + kJpegYuv444, + kJpegYuv400, + kJpegUnknown +}; + +struct Buffer { + const uint8* data; + int len; +}; + +struct BufferVector { + Buffer* buffers; + int len; + int pos; +}; + +struct SetJmpErrorMgr; + +// MJPEG ("Motion JPEG") is a pseudo-standard video codec where the frames are +// simply independent JPEG images with a fixed huffman table (which is omitted). +// It is rarely used in video transmission, but is common as a camera capture +// format, especially in Logitech devices. This class implements a decoder for +// MJPEG frames. +// +// See http://tools.ietf.org/html/rfc2435 +class LIBYUV_API MJpegDecoder { + public: + typedef void (*CallbackFunction)(void* opaque, + const uint8* const* data, + const int* strides, + int rows); + + static const int kColorSpaceUnknown; + static const int kColorSpaceGrayscale; + static const int kColorSpaceRgb; + static const int kColorSpaceYCbCr; + static const int kColorSpaceCMYK; + static const int kColorSpaceYCCK; + + MJpegDecoder(); + ~MJpegDecoder(); + + // Loads a new frame, reads its headers, and determines the uncompressed + // image format. + // Returns LIBYUV_TRUE if image looks valid and format is supported. + // If return value is LIBYUV_TRUE, then the values for all the following + // getters are populated. + // src_len is the size of the compressed mjpeg frame in bytes. + LIBYUV_BOOL LoadFrame(const uint8* src, size_t src_len); + + // Returns width of the last loaded frame in pixels. + int GetWidth(); + + // Returns height of the last loaded frame in pixels. + int GetHeight(); + + // Returns format of the last loaded frame. The return value is one of the + // kColorSpace* constants. + int GetColorSpace(); + + // Number of color components in the color space. + int GetNumComponents(); + + // Sample factors of the n-th component. + int GetHorizSampFactor(int component); + + int GetVertSampFactor(int component); + + int GetHorizSubSampFactor(int component); + + int GetVertSubSampFactor(int component); + + // Public for testability. + int GetImageScanlinesPerImcuRow(); + + // Public for testability. + int GetComponentScanlinesPerImcuRow(int component); + + // Width of a component in bytes. + int GetComponentWidth(int component); + + // Height of a component. + int GetComponentHeight(int component); + + // Width of a component in bytes with padding for DCTSIZE. Public for testing. + int GetComponentStride(int component); + + // Size of a component in bytes. + int GetComponentSize(int component); + + // Call this after LoadFrame() if you decide you don't want to decode it + // after all. + LIBYUV_BOOL UnloadFrame(); + + // Decodes the entire image into a one-buffer-per-color-component format. + // dst_width must match exactly. dst_height must be <= to image height; if + // less, the image is cropped. "planes" must have size equal to at least + // GetNumComponents() and they must point to non-overlapping buffers of size + // at least GetComponentSize(i). The pointers in planes are incremented + // to point to after the end of the written data. + // TODO(fbarchard): Add dst_x, dst_y to allow specific rect to be decoded. + LIBYUV_BOOL DecodeToBuffers(uint8** planes, int dst_width, int dst_height); + + // Decodes the entire image and passes the data via repeated calls to a + // callback function. Each call will get the data for a whole number of + // image scanlines. + // TODO(fbarchard): Add dst_x, dst_y to allow specific rect to be decoded. + LIBYUV_BOOL DecodeToCallback(CallbackFunction fn, void* opaque, + int dst_width, int dst_height); + + // The helper function which recognizes the jpeg sub-sampling type. + static JpegSubsamplingType JpegSubsamplingTypeHelper( + int* subsample_x, int* subsample_y, int number_of_components); + + private: + void AllocOutputBuffers(int num_outbufs); + void DestroyOutputBuffers(); + + LIBYUV_BOOL StartDecode(); + LIBYUV_BOOL FinishDecode(); + + void SetScanlinePointers(uint8** data); + LIBYUV_BOOL DecodeImcuRow(); + + int GetComponentScanlinePadding(int component); + + // A buffer holding the input data for a frame. + Buffer buf_; + BufferVector buf_vec_; + + jpeg_decompress_struct* decompress_struct_; + jpeg_source_mgr* source_mgr_; + SetJmpErrorMgr* error_mgr_; + + // LIBYUV_TRUE iff at least one component has scanline padding. (i.e., + // GetComponentScanlinePadding() != 0.) + LIBYUV_BOOL has_scanline_padding_; + + // Temporaries used to point to scanline outputs. + int num_outbufs_; // Outermost size of all arrays below. + uint8*** scanlines_; + int* scanlines_sizes_; + // Temporary buffer used for decoding when we can't decode directly to the + // output buffers. Large enough for just one iMCU row. + uint8** databuf_; + int* databuf_strides_; +}; + +} // namespace libyuv + +#endif // __cplusplus +#endif // INCLUDE_LIBYUV_MJPEG_DECODER_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/planar_functions.h b/phonelibs/libyuv/x64/include/libyuv/planar_functions.h new file mode 100644 index 00000000000000..1b57b29261eeb3 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/planar_functions.h @@ -0,0 +1,529 @@ +/* + * 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_ +#define INCLUDE_LIBYUV_PLANAR_FUNCTIONS_H_ + +#include "libyuv/basic_types.h" + +// TODO(fbarchard): Remove the following headers includes. +#include "libyuv/convert.h" +#include "libyuv/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); + +// Split interleaved UV plane into separate U and V planes. +LIBYUV_API +void SplitUVPlane(const uint8* src_uv, int src_stride_uv, + uint8* dst_u, int dst_stride_u, + uint8* dst_v, int dst_stride_v, + int width, int height); + +// Merge separate U and V planes into one interleaved UV plane. +LIBYUV_API +void MergeUVPlane(const uint8* src_u, int src_stride_u, + const uint8* src_v, int src_stride_v, + uint8* dst_uv, int dst_stride_uv, + int width, int height); + +// 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); + +#define J400ToJ400 I400ToI400 + +// 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); + +LIBYUV_API +int YUY2ToNV12(const uint8* src_yuy2, int src_stride_yuy2, + uint8* dst_y, int dst_stride_y, + uint8* dst_uv, int dst_stride_uv, + int width, int height); + +LIBYUV_API +int UYVYToNV12(const uint8* src_uyvy, int src_stride_uyvy, + uint8* dst_y, int dst_stride_y, + uint8* dst_uv, int dst_stride_uv, + 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 J420ToJ400 I420ToI400 +#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); + +// 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); + +// Alias +#define RGB24ToRAW RAWToRGB24 + +LIBYUV_API +int RAWToRGB24(const uint8* src_raw, int src_stride_raw, + uint8* dst_rgb24, int dst_stride_rgb24, + 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); + +// Convert plane of 16 bit shorts to half floats. +// Source values are multiplied by scale before storing as half float. +LIBYUV_API +int HalfFloatPlane(const uint16* src_y, int src_stride_y, + uint16* dst_y, int dst_stride_y, + float scale, + 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 Alpha channel of ARGB to alpha of ARGB. +LIBYUV_API +int ARGBCopyAlpha(const uint8* src_argb, int src_stride_argb, + uint8* dst_argb, int dst_stride_argb, + int width, int height); + +// Extract the alpha channel from ARGB. +LIBYUV_API +int ARGBExtractAlpha(const uint8* src_argb, int src_stride_argb, + uint8* dst_a, int dst_stride_a, + int width, int height); + +// Copy Y channel to Alpha of 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. +// Source is pre-multiplied by alpha using ARGBAttenuate. +// 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); + +// Alpha Blend plane and store to destination. +// Source is not pre-multiplied by alpha. +LIBYUV_API +int BlendPlane(const uint8* src_y0, int src_stride_y0, + const uint8* src_y1, int src_stride_y1, + const uint8* alpha, int alpha_stride, + uint8* dst_y, int dst_stride_y, + int width, int height); + +// Alpha Blend YUV images and store to destination. +// Source is not pre-multiplied by alpha. +// Alpha is full width x height and subsampled to half size to apply to UV. +LIBYUV_API +int I420Blend(const uint8* src_y0, int src_stride_y0, + const uint8* src_u0, int src_stride_u0, + const uint8* src_v0, int src_stride_v0, + const uint8* src_y1, int src_stride_y1, + const uint8* src_u1, int src_stride_u1, + const uint8* src_v1, int src_stride_v1, + const uint8* alpha, int alpha_stride, + 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); + +// 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); + +// 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 images using specified amount of interpolation +// (0 to 255) and store to destination. +// 'interpolation' is specified as 8 bit fraction where 0 means 100% src0 +// and 255 means 1% src0 and 99% src1. +LIBYUV_API +int InterpolatePlane(const uint8* src0, int src_stride0, + const uint8* src1, int src_stride1, + uint8* dst, int dst_stride, + int width, int height, int interpolation); + +// Interpolate between two ARGB images using specified amount of interpolation +// Internally calls InterpolatePlane with width * 4 (bpp). +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); + +// Interpolate between two YUV images using specified amount of interpolation +// Internally calls InterpolatePlane on each plane where the U and V planes +// are half width and half height. +LIBYUV_API +int I420Interpolate(const uint8* src0_y, int src0_stride_y, + const uint8* src0_u, int src0_stride_u, + const uint8* src0_v, int src0_stride_v, + const uint8* src1_y, int src1_stride_y, + const uint8* src1_u, int src1_stride_u, + const uint8* src1_v, int src1_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, int interpolation); + +#if defined(__pnacl__) || defined(__CLR_VER) || \ + (defined(__i386__) && !defined(__SSE2__)) +#define LIBYUV_DISABLE_X86 +#endif +// MemorySanitizer does not support assembly code yet. http://crbug.com/344505 +#if defined(__has_feature) +#if __has_feature(memory_sanitizer) +#define LIBYUV_DISABLE_X86 +#endif +#endif +// The following are available on all x86 platforms: +#if !defined(LIBYUV_DISABLE_X86) && \ + (defined(_M_IX86) || defined(__x86_64__) || defined(__i386__)) +#define HAS_ARGBAFFINEROW_SSE2 +#endif + +// Row function for copying 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); +LIBYUV_API +void ARGBAffineRow_SSE2(const uint8* src_argb, int src_argb_stride, + uint8* dst_argb, const float* uv_dudv, int width); + +// 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_ diff --git a/phonelibs/libyuv/x64/include/libyuv/rotate.h b/phonelibs/libyuv/x64/include/libyuv/rotate.h new file mode 100644 index 00000000000000..8a2da9a5aad5c5 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/rotate.h @@ -0,0 +1,117 @@ +/* + * 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_ROTATE_H_ +#define INCLUDE_LIBYUV_ROTATE_H_ + +#include "libyuv/basic_types.h" + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +// Supported rotation. +typedef enum RotationMode { + kRotate0 = 0, // No rotation. + kRotate90 = 90, // Rotate 90 degrees clockwise. + kRotate180 = 180, // Rotate 180 degrees. + kRotate270 = 270, // Rotate 270 degrees clockwise. + + // Deprecated. + kRotateNone = 0, + kRotateClockwise = 90, + kRotateCounterClockwise = 270, +} RotationModeEnum; + +// Rotate I420 frame. +LIBYUV_API +int I420Rotate(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 src_width, int src_height, enum RotationMode mode); + +// Rotate NV12 input and store in I420. +LIBYUV_API +int NV12ToI420Rotate(const uint8* src_y, int src_stride_y, + const uint8* src_uv, int src_stride_uv, + uint8* dst_y, int dst_stride_y, + uint8* dst_u, int dst_stride_u, + uint8* dst_v, int dst_stride_v, + int src_width, int src_height, enum RotationMode mode); + +// Rotate a plane by 0, 90, 180, or 270. +LIBYUV_API +int RotatePlane(const uint8* src, int src_stride, + uint8* dst, int dst_stride, + int src_width, int src_height, enum RotationMode mode); + +// Rotate planes by 90, 180, 270. Deprecated. +LIBYUV_API +void RotatePlane90(const uint8* src, int src_stride, + uint8* dst, int dst_stride, + int width, int height); + +LIBYUV_API +void RotatePlane180(const uint8* src, int src_stride, + uint8* dst, int dst_stride, + int width, int height); + +LIBYUV_API +void RotatePlane270(const uint8* src, int src_stride, + uint8* dst, int dst_stride, + int width, int height); + +LIBYUV_API +void RotateUV90(const uint8* src, int src_stride, + uint8* dst_a, int dst_stride_a, + uint8* dst_b, int dst_stride_b, + int width, int height); + +// Rotations for when U and V are interleaved. +// These functions take one input pointer and +// split the data into two buffers while +// rotating them. Deprecated. +LIBYUV_API +void RotateUV180(const uint8* src, int src_stride, + uint8* dst_a, int dst_stride_a, + uint8* dst_b, int dst_stride_b, + int width, int height); + +LIBYUV_API +void RotateUV270(const uint8* src, int src_stride, + uint8* dst_a, int dst_stride_a, + uint8* dst_b, int dst_stride_b, + int width, int height); + +// The 90 and 270 functions are based on transposes. +// Doing a transpose with reversing the read/write +// order will result in a rotation by +- 90 degrees. +// Deprecated. +LIBYUV_API +void TransposePlane(const uint8* src, int src_stride, + uint8* dst, int dst_stride, + int width, int height); + +LIBYUV_API +void TransposeUV(const uint8* src, int src_stride, + uint8* dst_a, int dst_stride_a, + uint8* dst_b, int dst_stride_b, + int width, int height); + +#ifdef __cplusplus +} // extern "C" +} // namespace libyuv +#endif + +#endif // INCLUDE_LIBYUV_ROTATE_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/rotate_argb.h b/phonelibs/libyuv/x64/include/libyuv/rotate_argb.h new file mode 100644 index 00000000000000..21fe7e1807c334 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/rotate_argb.h @@ -0,0 +1,33 @@ +/* + * Copyright 2012 The LibYuv Project Authors. All rights reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef INCLUDE_LIBYUV_ROTATE_ARGB_H_ +#define INCLUDE_LIBYUV_ROTATE_ARGB_H_ + +#include "libyuv/basic_types.h" +#include "libyuv/rotate.h" // For RotationMode. + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +// Rotate ARGB frame +LIBYUV_API +int ARGBRotate(const uint8* src_argb, int src_stride_argb, + uint8* dst_argb, int dst_stride_argb, + int src_width, int src_height, enum RotationMode mode); + +#ifdef __cplusplus +} // extern "C" +} // namespace libyuv +#endif + +#endif // INCLUDE_LIBYUV_ROTATE_ARGB_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/rotate_row.h b/phonelibs/libyuv/x64/include/libyuv/rotate_row.h new file mode 100644 index 00000000000000..6abd201677460b --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/rotate_row.h @@ -0,0 +1,121 @@ +/* + * 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_ROTATE_ROW_H_ +#define INCLUDE_LIBYUV_ROTATE_ROW_H_ + +#include "libyuv/basic_types.h" + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +#if defined(__pnacl__) || defined(__CLR_VER) || \ + (defined(__i386__) && !defined(__SSE2__)) +#define LIBYUV_DISABLE_X86 +#endif +// MemorySanitizer does not support assembly code yet. http://crbug.com/344505 +#if defined(__has_feature) +#if __has_feature(memory_sanitizer) +#define LIBYUV_DISABLE_X86 +#endif +#endif +// The following are available for Visual C and clangcl 32 bit: +#if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) +#define HAS_TRANSPOSEWX8_SSSE3 +#define HAS_TRANSPOSEUVWX8_SSE2 +#endif + +// The following are available for GCC 32 or 64 bit but not NaCL for 64 bit: +#if !defined(LIBYUV_DISABLE_X86) && \ + (defined(__i386__) || (defined(__x86_64__) && !defined(__native_client__))) +#define HAS_TRANSPOSEWX8_SSSE3 +#endif + +// The following are available for 64 bit GCC but not NaCL: +#if !defined(LIBYUV_DISABLE_X86) && !defined(__native_client__) && \ + defined(__x86_64__) +#define HAS_TRANSPOSEWX8_FAST_SSSE3 +#define HAS_TRANSPOSEUVWX8_SSE2 +#endif + +#if !defined(LIBYUV_DISABLE_NEON) && !defined(__native_client__) && \ + (defined(__ARM_NEON__) || defined(LIBYUV_NEON) || defined(__aarch64__)) +#define HAS_TRANSPOSEWX8_NEON +#define HAS_TRANSPOSEUVWX8_NEON +#endif + +#if !defined(LIBYUV_DISABLE_MIPS) && !defined(__native_client__) && \ + defined(__mips__) && \ + defined(__mips_dsp) && (__mips_dsp_rev >= 2) +#define HAS_TRANSPOSEWX8_DSPR2 +#define HAS_TRANSPOSEUVWX8_DSPR2 +#endif // defined(__mips__) + +void TransposeWxH_C(const uint8* src, int src_stride, + uint8* dst, int dst_stride, int width, int height); + +void TransposeWx8_C(const uint8* src, int src_stride, + uint8* dst, int dst_stride, int width); +void TransposeWx8_NEON(const uint8* src, int src_stride, + uint8* dst, int dst_stride, int width); +void TransposeWx8_SSSE3(const uint8* src, int src_stride, + uint8* dst, int dst_stride, int width); +void TransposeWx8_Fast_SSSE3(const uint8* src, int src_stride, + uint8* dst, int dst_stride, int width); +void TransposeWx8_DSPR2(const uint8* src, int src_stride, + uint8* dst, int dst_stride, int width); +void TransposeWx8_Fast_DSPR2(const uint8* src, int src_stride, + uint8* dst, int dst_stride, int width); + +void TransposeWx8_Any_NEON(const uint8* src, int src_stride, + uint8* dst, int dst_stride, int width); +void TransposeWx8_Any_SSSE3(const uint8* src, int src_stride, + uint8* dst, int dst_stride, int width); +void TransposeWx8_Fast_Any_SSSE3(const uint8* src, int src_stride, + uint8* dst, int dst_stride, int width); +void TransposeWx8_Any_DSPR2(const uint8* src, int src_stride, + uint8* dst, int dst_stride, int width); + +void TransposeUVWxH_C(const uint8* src, int src_stride, + uint8* dst_a, int dst_stride_a, + uint8* dst_b, int dst_stride_b, + int width, int height); + +void TransposeUVWx8_C(const uint8* src, int src_stride, + uint8* dst_a, int dst_stride_a, + uint8* dst_b, int dst_stride_b, int width); +void TransposeUVWx8_SSE2(const uint8* src, int src_stride, + uint8* dst_a, int dst_stride_a, + uint8* dst_b, int dst_stride_b, int width); +void TransposeUVWx8_NEON(const uint8* src, int src_stride, + uint8* dst_a, int dst_stride_a, + uint8* dst_b, int dst_stride_b, int width); +void TransposeUVWx8_DSPR2(const uint8* src, int src_stride, + uint8* dst_a, int dst_stride_a, + uint8* dst_b, int dst_stride_b, int width); + +void TransposeUVWx8_Any_SSE2(const uint8* src, int src_stride, + uint8* dst_a, int dst_stride_a, + uint8* dst_b, int dst_stride_b, int width); +void TransposeUVWx8_Any_NEON(const uint8* src, int src_stride, + uint8* dst_a, int dst_stride_a, + uint8* dst_b, int dst_stride_b, int width); +void TransposeUVWx8_Any_DSPR2(const uint8* src, int src_stride, + uint8* dst_a, int dst_stride_a, + uint8* dst_b, int dst_stride_b, int width); + +#ifdef __cplusplus +} // extern "C" +} // namespace libyuv +#endif + +#endif // INCLUDE_LIBYUV_ROTATE_ROW_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/row.h b/phonelibs/libyuv/x64/include/libyuv/row.h new file mode 100644 index 00000000000000..b810221ec7a787 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/row.h @@ -0,0 +1,1963 @@ +/* + * 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_ROW_H_ +#define INCLUDE_LIBYUV_ROW_H_ + +#include // For malloc. + +#include "libyuv/basic_types.h" + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +#define IS_ALIGNED(p, a) (!((uintptr_t)(p) & ((a) - 1))) + +#define align_buffer_64(var, size) \ + uint8* var##_mem = (uint8*)(malloc((size) + 63)); /* NOLINT */ \ + uint8* var = (uint8*)(((intptr_t)(var##_mem) + 63) & ~63) /* NOLINT */ + +#define free_aligned_buffer_64(var) \ + free(var##_mem); \ + var = 0 + +#if defined(__pnacl__) || defined(__CLR_VER) || \ + (defined(__i386__) && !defined(__SSE2__)) +#define LIBYUV_DISABLE_X86 +#endif +// MemorySanitizer does not support assembly code yet. http://crbug.com/344505 +#if defined(__has_feature) +#if __has_feature(memory_sanitizer) +#define LIBYUV_DISABLE_X86 +#endif +#endif +// True if compiling for SSSE3 as a requirement. +#if defined(__SSSE3__) || (defined(_M_IX86_FP) && (_M_IX86_FP >= 3)) +#define LIBYUV_SSSE3_ONLY +#endif + +#if defined(__native_client__) +#define LIBYUV_DISABLE_NEON +#endif +// clang >= 3.5.0 required for Arm64. +#if defined(__clang__) && defined(__aarch64__) && !defined(LIBYUV_DISABLE_NEON) +#if (__clang_major__ < 3) || (__clang_major__ == 3 && (__clang_minor__ < 5)) +#define LIBYUV_DISABLE_NEON +#endif // clang >= 3.5 +#endif // __clang__ + +// GCC >= 4.7.0 required for AVX2. +#if defined(__GNUC__) && (defined(__x86_64__) || defined(__i386__)) +#if (__GNUC__ > 4) || (__GNUC__ == 4 && (__GNUC_MINOR__ >= 7)) +#define GCC_HAS_AVX2 1 +#endif // GNUC >= 4.7 +#endif // __GNUC__ + +// clang >= 3.4.0 required for AVX2. +#if defined(__clang__) && (defined(__x86_64__) || defined(__i386__)) +#if (__clang_major__ > 3) || (__clang_major__ == 3 && (__clang_minor__ >= 4)) +#define CLANG_HAS_AVX2 1 +#endif // clang >= 3.4 +#endif // __clang__ + +// Visual C 2012 required for AVX2. +#if defined(_M_IX86) && !defined(__clang__) && \ + defined(_MSC_VER) && _MSC_VER >= 1700 +#define VISUALC_HAS_AVX2 1 +#endif // VisualStudio >= 2012 + +// The following are available on all x86 platforms: +#if !defined(LIBYUV_DISABLE_X86) && \ + (defined(_M_IX86) || defined(__x86_64__) || defined(__i386__)) +// Conversions: +#define HAS_ABGRTOUVROW_SSSE3 +#define HAS_ABGRTOYROW_SSSE3 +#define HAS_ARGB1555TOARGBROW_SSE2 +#define HAS_ARGB4444TOARGBROW_SSE2 +#define HAS_ARGBSETROW_X86 +#define HAS_ARGBSHUFFLEROW_SSE2 +#define HAS_ARGBSHUFFLEROW_SSSE3 +#define HAS_ARGBTOARGB1555ROW_SSE2 +#define HAS_ARGBTOARGB4444ROW_SSE2 +#define HAS_ARGBTORAWROW_SSSE3 +#define HAS_ARGBTORGB24ROW_SSSE3 +#define HAS_ARGBTORGB565DITHERROW_SSE2 +#define HAS_ARGBTORGB565ROW_SSE2 +#define HAS_ARGBTOUV444ROW_SSSE3 +#define HAS_ARGBTOUVJROW_SSSE3 +#define HAS_ARGBTOUVROW_SSSE3 +#define HAS_ARGBTOYJROW_SSSE3 +#define HAS_ARGBTOYROW_SSSE3 +#define HAS_ARGBEXTRACTALPHAROW_SSE2 +#define HAS_BGRATOUVROW_SSSE3 +#define HAS_BGRATOYROW_SSSE3 +#define HAS_COPYROW_ERMS +#define HAS_COPYROW_SSE2 +#define HAS_H422TOARGBROW_SSSE3 +#define HAS_I400TOARGBROW_SSE2 +#define HAS_I422TOARGB1555ROW_SSSE3 +#define HAS_I422TOARGB4444ROW_SSSE3 +#define HAS_I422TOARGBROW_SSSE3 +#define HAS_I422TORGB24ROW_SSSE3 +#define HAS_I422TORGB565ROW_SSSE3 +#define HAS_I422TORGBAROW_SSSE3 +#define HAS_I422TOUYVYROW_SSE2 +#define HAS_I422TOYUY2ROW_SSE2 +#define HAS_I444TOARGBROW_SSSE3 +#define HAS_J400TOARGBROW_SSE2 +#define HAS_J422TOARGBROW_SSSE3 +#define HAS_MERGEUVROW_SSE2 +#define HAS_MIRRORROW_SSSE3 +#define HAS_MIRRORUVROW_SSSE3 +#define HAS_NV12TOARGBROW_SSSE3 +#define HAS_NV12TORGB565ROW_SSSE3 +#define HAS_NV21TOARGBROW_SSSE3 +#define HAS_RAWTOARGBROW_SSSE3 +#define HAS_RAWTORGB24ROW_SSSE3 +#define HAS_RAWTOYROW_SSSE3 +#define HAS_RGB24TOARGBROW_SSSE3 +#define HAS_RGB24TOYROW_SSSE3 +#define HAS_RGB565TOARGBROW_SSE2 +#define HAS_RGBATOUVROW_SSSE3 +#define HAS_RGBATOYROW_SSSE3 +#define HAS_SETROW_ERMS +#define HAS_SETROW_X86 +#define HAS_SPLITUVROW_SSE2 +#define HAS_UYVYTOARGBROW_SSSE3 +#define HAS_UYVYTOUV422ROW_SSE2 +#define HAS_UYVYTOUVROW_SSE2 +#define HAS_UYVYTOYROW_SSE2 +#define HAS_YUY2TOARGBROW_SSSE3 +#define HAS_YUY2TOUV422ROW_SSE2 +#define HAS_YUY2TOUVROW_SSE2 +#define HAS_YUY2TOYROW_SSE2 + +// Effects: +#define HAS_ARGBADDROW_SSE2 +#define HAS_ARGBAFFINEROW_SSE2 +#define HAS_ARGBATTENUATEROW_SSSE3 +#define HAS_ARGBBLENDROW_SSSE3 +#define HAS_ARGBCOLORMATRIXROW_SSSE3 +#define HAS_ARGBCOLORTABLEROW_X86 +#define HAS_ARGBCOPYALPHAROW_SSE2 +#define HAS_ARGBCOPYYTOALPHAROW_SSE2 +#define HAS_ARGBGRAYROW_SSSE3 +#define HAS_ARGBLUMACOLORTABLEROW_SSSE3 +#define HAS_ARGBMIRRORROW_SSE2 +#define HAS_ARGBMULTIPLYROW_SSE2 +#define HAS_ARGBPOLYNOMIALROW_SSE2 +#define HAS_ARGBQUANTIZEROW_SSE2 +#define HAS_ARGBSEPIAROW_SSSE3 +#define HAS_ARGBSHADEROW_SSE2 +#define HAS_ARGBSUBTRACTROW_SSE2 +#define HAS_ARGBUNATTENUATEROW_SSE2 +#define HAS_BLENDPLANEROW_SSSE3 +#define HAS_COMPUTECUMULATIVESUMROW_SSE2 +#define HAS_CUMULATIVESUMTOAVERAGEROW_SSE2 +#define HAS_INTERPOLATEROW_SSSE3 +#define HAS_RGBCOLORTABLEROW_X86 +#define HAS_SOBELROW_SSE2 +#define HAS_SOBELTOPLANEROW_SSE2 +#define HAS_SOBELXROW_SSE2 +#define HAS_SOBELXYROW_SSE2 +#define HAS_SOBELYROW_SSE2 + +// The following functions fail on gcc/clang 32 bit with fpic and framepointer. +// caveat: clangcl uses row_win.cc which works. +#if defined(NDEBUG) || !(defined(_DEBUG) && defined(__i386__)) || \ + !defined(__i386__) || defined(_MSC_VER) +// TODO(fbarchard): fix build error on x86 debug +// https://code.google.com/p/libyuv/issues/detail?id=524 +#define HAS_I411TOARGBROW_SSSE3 +// TODO(fbarchard): fix build error on android_full_debug=1 +// https://code.google.com/p/libyuv/issues/detail?id=517 +#define HAS_I422ALPHATOARGBROW_SSSE3 +#endif +#endif + +// The following are available on all x86 platforms, but +// require VS2012, clang 3.4 or gcc 4.7. +// The code supports NaCL but requires a new compiler and validator. +#if !defined(LIBYUV_DISABLE_X86) && (defined(VISUALC_HAS_AVX2) || \ + defined(CLANG_HAS_AVX2) || defined(GCC_HAS_AVX2)) +#define HAS_ARGBCOPYALPHAROW_AVX2 +#define HAS_ARGBCOPYYTOALPHAROW_AVX2 +#define HAS_ARGBMIRRORROW_AVX2 +#define HAS_ARGBPOLYNOMIALROW_AVX2 +#define HAS_ARGBSHUFFLEROW_AVX2 +#define HAS_ARGBTORGB565DITHERROW_AVX2 +#define HAS_ARGBTOUVJROW_AVX2 +#define HAS_ARGBTOUVROW_AVX2 +#define HAS_ARGBTOYJROW_AVX2 +#define HAS_ARGBTOYROW_AVX2 +#define HAS_COPYROW_AVX +#define HAS_H422TOARGBROW_AVX2 +#define HAS_I400TOARGBROW_AVX2 +#if !(defined(_DEBUG) && defined(__i386__)) +// TODO(fbarchard): fix build error on android_full_debug=1 +// https://code.google.com/p/libyuv/issues/detail?id=517 +#define HAS_I422ALPHATOARGBROW_AVX2 +#endif +#define HAS_I411TOARGBROW_AVX2 +#define HAS_I422TOARGB1555ROW_AVX2 +#define HAS_I422TOARGB4444ROW_AVX2 +#define HAS_I422TOARGBROW_AVX2 +#define HAS_I422TORGB24ROW_AVX2 +#define HAS_I422TORGB565ROW_AVX2 +#define HAS_I422TORGBAROW_AVX2 +#define HAS_I444TOARGBROW_AVX2 +#define HAS_INTERPOLATEROW_AVX2 +#define HAS_J422TOARGBROW_AVX2 +#define HAS_MERGEUVROW_AVX2 +#define HAS_MIRRORROW_AVX2 +#define HAS_NV12TOARGBROW_AVX2 +#define HAS_NV12TORGB565ROW_AVX2 +#define HAS_NV21TOARGBROW_AVX2 +#define HAS_SPLITUVROW_AVX2 +#define HAS_UYVYTOARGBROW_AVX2 +#define HAS_UYVYTOUV422ROW_AVX2 +#define HAS_UYVYTOUVROW_AVX2 +#define HAS_UYVYTOYROW_AVX2 +#define HAS_YUY2TOARGBROW_AVX2 +#define HAS_YUY2TOUV422ROW_AVX2 +#define HAS_YUY2TOUVROW_AVX2 +#define HAS_YUY2TOYROW_AVX2 +#define HAS_HALFFLOATROW_AVX2 + +// Effects: +#define HAS_ARGBADDROW_AVX2 +#define HAS_ARGBATTENUATEROW_AVX2 +#define HAS_ARGBMULTIPLYROW_AVX2 +#define HAS_ARGBSUBTRACTROW_AVX2 +#define HAS_ARGBUNATTENUATEROW_AVX2 +#define HAS_BLENDPLANEROW_AVX2 +#endif + +// The following are available for AVX2 Visual C and clangcl 32 bit: +// TODO(fbarchard): Port to gcc. +#if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && \ + (defined(VISUALC_HAS_AVX2) || defined(CLANG_HAS_AVX2)) +#define HAS_ARGB1555TOARGBROW_AVX2 +#define HAS_ARGB4444TOARGBROW_AVX2 +#define HAS_ARGBTOARGB1555ROW_AVX2 +#define HAS_ARGBTOARGB4444ROW_AVX2 +#define HAS_ARGBTORGB565ROW_AVX2 +#define HAS_J400TOARGBROW_AVX2 +#define HAS_RGB565TOARGBROW_AVX2 +#endif + +// The following are also available on x64 Visual C. +#if !defined(LIBYUV_DISABLE_X86) && defined(_MSC_VER) && defined(_M_X64) && \ + (!defined(__clang__) || defined(__SSSE3__)) +#define HAS_I422ALPHATOARGBROW_SSSE3 +#define HAS_I422TOARGBROW_SSSE3 +#endif + +// The following are available on gcc x86 platforms: +// TODO(fbarchard): Port to Visual C. +#if !defined(LIBYUV_DISABLE_X86) && \ + (defined(__x86_64__) || (defined(__i386__) && !defined(_MSC_VER))) +#define HAS_HALFFLOATROW_SSE2 +#endif + +// The following are available on Neon platforms: +#if !defined(LIBYUV_DISABLE_NEON) && \ + (defined(__aarch64__) || defined(__ARM_NEON__) || defined(LIBYUV_NEON)) +#define HAS_ABGRTOUVROW_NEON +#define HAS_ABGRTOYROW_NEON +#define HAS_ARGB1555TOARGBROW_NEON +#define HAS_ARGB1555TOUVROW_NEON +#define HAS_ARGB1555TOYROW_NEON +#define HAS_ARGB4444TOARGBROW_NEON +#define HAS_ARGB4444TOUVROW_NEON +#define HAS_ARGB4444TOYROW_NEON +#define HAS_ARGBSETROW_NEON +#define HAS_ARGBTOARGB1555ROW_NEON +#define HAS_ARGBTOARGB4444ROW_NEON +#define HAS_ARGBTORAWROW_NEON +#define HAS_ARGBTORGB24ROW_NEON +#define HAS_ARGBTORGB565DITHERROW_NEON +#define HAS_ARGBTORGB565ROW_NEON +#define HAS_ARGBTOUV411ROW_NEON +#define HAS_ARGBTOUV444ROW_NEON +#define HAS_ARGBTOUVJROW_NEON +#define HAS_ARGBTOUVROW_NEON +#define HAS_ARGBTOYJROW_NEON +#define HAS_ARGBTOYROW_NEON +#define HAS_ARGBEXTRACTALPHAROW_NEON +#define HAS_BGRATOUVROW_NEON +#define HAS_BGRATOYROW_NEON +#define HAS_COPYROW_NEON +#define HAS_I400TOARGBROW_NEON +#define HAS_I411TOARGBROW_NEON +#define HAS_I422ALPHATOARGBROW_NEON +#define HAS_I422TOARGB1555ROW_NEON +#define HAS_I422TOARGB4444ROW_NEON +#define HAS_I422TOARGBROW_NEON +#define HAS_I422TORGB24ROW_NEON +#define HAS_I422TORGB565ROW_NEON +#define HAS_I422TORGBAROW_NEON +#define HAS_I422TOUYVYROW_NEON +#define HAS_I422TOYUY2ROW_NEON +#define HAS_I444TOARGBROW_NEON +#define HAS_J400TOARGBROW_NEON +#define HAS_MERGEUVROW_NEON +#define HAS_MIRRORROW_NEON +#define HAS_MIRRORUVROW_NEON +#define HAS_NV12TOARGBROW_NEON +#define HAS_NV12TORGB565ROW_NEON +#define HAS_NV21TOARGBROW_NEON +#define HAS_RAWTOARGBROW_NEON +#define HAS_RAWTORGB24ROW_NEON +#define HAS_RAWTOUVROW_NEON +#define HAS_RAWTOYROW_NEON +#define HAS_RGB24TOARGBROW_NEON +#define HAS_RGB24TOUVROW_NEON +#define HAS_RGB24TOYROW_NEON +#define HAS_RGB565TOARGBROW_NEON +#define HAS_RGB565TOUVROW_NEON +#define HAS_RGB565TOYROW_NEON +#define HAS_RGBATOUVROW_NEON +#define HAS_RGBATOYROW_NEON +#define HAS_SETROW_NEON +#define HAS_SPLITUVROW_NEON +#define HAS_UYVYTOARGBROW_NEON +#define HAS_UYVYTOUV422ROW_NEON +#define HAS_UYVYTOUVROW_NEON +#define HAS_UYVYTOYROW_NEON +#define HAS_YUY2TOARGBROW_NEON +#define HAS_YUY2TOUV422ROW_NEON +#define HAS_YUY2TOUVROW_NEON +#define HAS_YUY2TOYROW_NEON + +// Effects: +#define HAS_ARGBADDROW_NEON +#define HAS_ARGBATTENUATEROW_NEON +#define HAS_ARGBBLENDROW_NEON +#define HAS_ARGBCOLORMATRIXROW_NEON +#define HAS_ARGBGRAYROW_NEON +#define HAS_ARGBMIRRORROW_NEON +#define HAS_ARGBMULTIPLYROW_NEON +#define HAS_ARGBQUANTIZEROW_NEON +#define HAS_ARGBSEPIAROW_NEON +#define HAS_ARGBSHADEROW_NEON +#define HAS_ARGBSHUFFLEROW_NEON +#define HAS_ARGBSUBTRACTROW_NEON +#define HAS_INTERPOLATEROW_NEON +#define HAS_SOBELROW_NEON +#define HAS_SOBELTOPLANEROW_NEON +#define HAS_SOBELXROW_NEON +#define HAS_SOBELXYROW_NEON +#define HAS_SOBELYROW_NEON +#endif + +// The following are available on Mips platforms: +#if !defined(LIBYUV_DISABLE_MIPS) && defined(__mips__) && \ + (_MIPS_SIM == _MIPS_SIM_ABI32) && (__mips_isa_rev < 6) +#define HAS_COPYROW_MIPS +#if defined(__mips_dsp) && (__mips_dsp_rev >= 2) +#define HAS_I422TOARGBROW_DSPR2 +#define HAS_INTERPOLATEROW_DSPR2 +#define HAS_MIRRORROW_DSPR2 +#define HAS_MIRRORUVROW_DSPR2 +#define HAS_SPLITUVROW_DSPR2 +#endif +#endif + +#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa) +#define HAS_MIRRORROW_MSA +#define HAS_ARGBMIRRORROW_MSA +#endif + +#if defined(_MSC_VER) && !defined(__CLR_VER) && !defined(__clang__) +#if defined(VISUALC_HAS_AVX2) +#define SIMD_ALIGNED(var) __declspec(align(32)) var +#else +#define SIMD_ALIGNED(var) __declspec(align(16)) var +#endif +typedef __declspec(align(16)) int16 vec16[8]; +typedef __declspec(align(16)) int32 vec32[4]; +typedef __declspec(align(16)) int8 vec8[16]; +typedef __declspec(align(16)) uint16 uvec16[8]; +typedef __declspec(align(16)) uint32 uvec32[4]; +typedef __declspec(align(16)) uint8 uvec8[16]; +typedef __declspec(align(32)) int16 lvec16[16]; +typedef __declspec(align(32)) int32 lvec32[8]; +typedef __declspec(align(32)) int8 lvec8[32]; +typedef __declspec(align(32)) uint16 ulvec16[16]; +typedef __declspec(align(32)) uint32 ulvec32[8]; +typedef __declspec(align(32)) uint8 ulvec8[32]; +#elif !defined(__pnacl__) && (defined(__GNUC__) || defined(__clang__)) +// Caveat GCC 4.2 to 4.7 have a known issue using vectors with const. +#if defined(CLANG_HAS_AVX2) || defined(GCC_HAS_AVX2) +#define SIMD_ALIGNED(var) var __attribute__((aligned(32))) +#else +#define SIMD_ALIGNED(var) var __attribute__((aligned(16))) +#endif +typedef int16 __attribute__((vector_size(16))) vec16; +typedef int32 __attribute__((vector_size(16))) vec32; +typedef int8 __attribute__((vector_size(16))) vec8; +typedef uint16 __attribute__((vector_size(16))) uvec16; +typedef uint32 __attribute__((vector_size(16))) uvec32; +typedef uint8 __attribute__((vector_size(16))) uvec8; +typedef int16 __attribute__((vector_size(32))) lvec16; +typedef int32 __attribute__((vector_size(32))) lvec32; +typedef int8 __attribute__((vector_size(32))) lvec8; +typedef uint16 __attribute__((vector_size(32))) ulvec16; +typedef uint32 __attribute__((vector_size(32))) ulvec32; +typedef uint8 __attribute__((vector_size(32))) ulvec8; +#else +#define SIMD_ALIGNED(var) var +typedef int16 vec16[8]; +typedef int32 vec32[4]; +typedef int8 vec8[16]; +typedef uint16 uvec16[8]; +typedef uint32 uvec32[4]; +typedef uint8 uvec8[16]; +typedef int16 lvec16[16]; +typedef int32 lvec32[8]; +typedef int8 lvec8[32]; +typedef uint16 ulvec16[16]; +typedef uint32 ulvec32[8]; +typedef uint8 ulvec8[32]; +#endif + +#if defined(__aarch64__) +// This struct is for Arm64 color conversion. +struct YuvConstants { + uvec16 kUVToRB; + uvec16 kUVToRB2; + uvec16 kUVToG; + uvec16 kUVToG2; + vec16 kUVBiasBGR; + vec32 kYToRgb; +}; +#elif defined(__arm__) +// This struct is for ArmV7 color conversion. +struct YuvConstants { + uvec8 kUVToRB; + uvec8 kUVToG; + vec16 kUVBiasBGR; + vec32 kYToRgb; +}; +#else +// This struct is for Intel color conversion. +struct YuvConstants { + int8 kUVToB[32]; + int8 kUVToG[32]; + int8 kUVToR[32]; + int16 kUVBiasB[16]; + int16 kUVBiasG[16]; + int16 kUVBiasR[16]; + int16 kYToRgb[16]; +}; + +// Offsets into YuvConstants structure +#define KUVTOB 0 +#define KUVTOG 32 +#define KUVTOR 64 +#define KUVBIASB 96 +#define KUVBIASG 128 +#define KUVBIASR 160 +#define KYTORGB 192 +#endif + +// Conversion matrix for YUV to RGB +extern const struct YuvConstants SIMD_ALIGNED(kYuvI601Constants); // BT.601 +extern const struct YuvConstants SIMD_ALIGNED(kYuvJPEGConstants); // JPeg +extern const struct YuvConstants SIMD_ALIGNED(kYuvH709Constants); // BT.709 + +// Conversion matrix for YVU to BGR +extern const struct YuvConstants SIMD_ALIGNED(kYvuI601Constants); // BT.601 +extern const struct YuvConstants SIMD_ALIGNED(kYvuJPEGConstants); // JPeg +extern const struct YuvConstants SIMD_ALIGNED(kYvuH709Constants); // BT.709 + +#if defined(__APPLE__) || defined(__x86_64__) || defined(__llvm__) +#define OMITFP +#else +#define OMITFP __attribute__((optimize("omit-frame-pointer"))) +#endif + +// NaCL macros for GCC x86 and x64. +#if defined(__native_client__) +#define LABELALIGN ".p2align 5\n" +#else +#define LABELALIGN +#endif +#if defined(__native_client__) && defined(__x86_64__) +// r14 is used for MEMOP macros. +#define NACL_R14 "r14", +#define BUNDLELOCK ".bundle_lock\n" +#define BUNDLEUNLOCK ".bundle_unlock\n" +#define MEMACCESS(base) "%%nacl:(%%r15,%q" #base ")" +#define MEMACCESS2(offset, base) "%%nacl:" #offset "(%%r15,%q" #base ")" +#define MEMLEA(offset, base) #offset "(%q" #base ")" +#define MEMLEA3(offset, index, scale) \ + #offset "(,%q" #index "," #scale ")" +#define MEMLEA4(offset, base, index, scale) \ + #offset "(%q" #base ",%q" #index "," #scale ")" +#define MEMMOVESTRING(s, d) "%%nacl:(%q" #s "),%%nacl:(%q" #d "), %%r15" +#define MEMSTORESTRING(reg, d) "%%" #reg ",%%nacl:(%q" #d "), %%r15" +#define MEMOPREG(opcode, offset, base, index, scale, reg) \ + BUNDLELOCK \ + "lea " #offset "(%q" #base ",%q" #index "," #scale "),%%r14d\n" \ + #opcode " (%%r15,%%r14),%%" #reg "\n" \ + BUNDLEUNLOCK +#define MEMOPMEM(opcode, reg, offset, base, index, scale) \ + BUNDLELOCK \ + "lea " #offset "(%q" #base ",%q" #index "," #scale "),%%r14d\n" \ + #opcode " %%" #reg ",(%%r15,%%r14)\n" \ + BUNDLEUNLOCK +#define MEMOPARG(opcode, offset, base, index, scale, arg) \ + BUNDLELOCK \ + "lea " #offset "(%q" #base ",%q" #index "," #scale "),%%r14d\n" \ + #opcode " (%%r15,%%r14),%" #arg "\n" \ + BUNDLEUNLOCK +#define VMEMOPREG(opcode, offset, base, index, scale, reg1, reg2) \ + BUNDLELOCK \ + "lea " #offset "(%q" #base ",%q" #index "," #scale "),%%r14d\n" \ + #opcode " (%%r15,%%r14),%%" #reg1 ",%%" #reg2 "\n" \ + BUNDLEUNLOCK +#define VEXTOPMEM(op, sel, reg, offset, base, index, scale) \ + BUNDLELOCK \ + "lea " #offset "(%q" #base ",%q" #index "," #scale "),%%r14d\n" \ + #op " $" #sel ",%%" #reg ",(%%r15,%%r14)\n" \ + BUNDLEUNLOCK +#else // defined(__native_client__) && defined(__x86_64__) +#define NACL_R14 +#define BUNDLEALIGN +#define MEMACCESS(base) "(%" #base ")" +#define MEMACCESS2(offset, base) #offset "(%" #base ")" +#define MEMLEA(offset, base) #offset "(%" #base ")" +#define MEMLEA3(offset, index, scale) \ + #offset "(,%" #index "," #scale ")" +#define MEMLEA4(offset, base, index, scale) \ + #offset "(%" #base ",%" #index "," #scale ")" +#define MEMMOVESTRING(s, d) +#define MEMSTORESTRING(reg, d) +#define MEMOPREG(opcode, offset, base, index, scale, reg) \ + #opcode " " #offset "(%" #base ",%" #index "," #scale "),%%" #reg "\n" +#define MEMOPMEM(opcode, reg, offset, base, index, scale) \ + #opcode " %%" #reg ","#offset "(%" #base ",%" #index "," #scale ")\n" +#define MEMOPARG(opcode, offset, base, index, scale, arg) \ + #opcode " " #offset "(%" #base ",%" #index "," #scale "),%" #arg "\n" +#define VMEMOPREG(opcode, offset, base, index, scale, reg1, reg2) \ + #opcode " " #offset "(%" #base ",%" #index "," #scale "),%%" #reg1 ",%%" \ + #reg2 "\n" +#define VEXTOPMEM(op, sel, reg, offset, base, index, scale) \ + #op " $" #sel ",%%" #reg ","#offset "(%" #base ",%" #index "," #scale ")\n" +#endif // defined(__native_client__) && defined(__x86_64__) + +#if defined(__arm__) || defined(__aarch64__) +#undef MEMACCESS +#if defined(__native_client__) +#define MEMACCESS(base) ".p2align 3\nbic %" #base ", #0xc0000000\n" +#else +#define MEMACCESS(base) +#endif +#endif + +void I444ToARGBRow_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGBRow_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422AlphaToARGBRow_NEON(const uint8* y_buf, + const uint8* u_buf, + const uint8* v_buf, + const uint8* a_buf, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGBRow_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I411ToARGBRow_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGBARow_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgba, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB24Row_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgb24, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB565Row_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgb565, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB1555Row_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb1555, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB4444Row_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb4444, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToARGBRow_NEON(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToRGB565Row_NEON(const uint8* src_y, + const uint8* src_uv, + uint8* dst_rgb565, + const struct YuvConstants* yuvconstants, + int width); +void NV21ToARGBRow_NEON(const uint8* src_y, + const uint8* src_vu, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void YUY2ToARGBRow_NEON(const uint8* src_yuy2, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void UYVYToARGBRow_NEON(const uint8* src_uyvy, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); + +void ARGBToYRow_AVX2(const uint8* src_argb, uint8* dst_y, int width); +void ARGBToYRow_Any_AVX2(const uint8* src_argb, uint8* dst_y, int width); +void ARGBToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int width); +void ARGBToYJRow_AVX2(const uint8* src_argb, uint8* dst_y, int width); +void ARGBToYJRow_Any_AVX2(const uint8* src_argb, uint8* dst_y, int width); +void ARGBToYJRow_SSSE3(const uint8* src_argb, uint8* dst_y, int width); +void BGRAToYRow_SSSE3(const uint8* src_bgra, uint8* dst_y, int width); +void ABGRToYRow_SSSE3(const uint8* src_abgr, uint8* dst_y, int width); +void RGBAToYRow_SSSE3(const uint8* src_rgba, uint8* dst_y, int width); +void RGB24ToYRow_SSSE3(const uint8* src_rgb24, uint8* dst_y, int width); +void RAWToYRow_SSSE3(const uint8* src_raw, uint8* dst_y, int width); +void ARGBToYRow_NEON(const uint8* src_argb, uint8* dst_y, int width); +void ARGBToYJRow_NEON(const uint8* src_argb, uint8* dst_y, int width); +void ARGBToUV444Row_NEON(const uint8* src_argb, uint8* dst_u, uint8* dst_v, + int width); +void ARGBToUV411Row_NEON(const uint8* src_argb, uint8* dst_u, uint8* dst_v, + int width); +void ARGBToUVRow_NEON(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUVJRow_NEON(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void BGRAToUVRow_NEON(const uint8* src_bgra, int src_stride_bgra, + uint8* dst_u, uint8* dst_v, int width); +void ABGRToUVRow_NEON(const uint8* src_abgr, int src_stride_abgr, + uint8* dst_u, uint8* dst_v, int width); +void RGBAToUVRow_NEON(const uint8* src_rgba, int src_stride_rgba, + uint8* dst_u, uint8* dst_v, int width); +void RGB24ToUVRow_NEON(const uint8* src_rgb24, int src_stride_rgb24, + uint8* dst_u, uint8* dst_v, int width); +void RAWToUVRow_NEON(const uint8* src_raw, int src_stride_raw, + uint8* dst_u, uint8* dst_v, int width); +void RGB565ToUVRow_NEON(const uint8* src_rgb565, int src_stride_rgb565, + uint8* dst_u, uint8* dst_v, int width); +void ARGB1555ToUVRow_NEON(const uint8* src_argb1555, int src_stride_argb1555, + uint8* dst_u, uint8* dst_v, int width); +void ARGB4444ToUVRow_NEON(const uint8* src_argb4444, int src_stride_argb4444, + uint8* dst_u, uint8* dst_v, int width); +void BGRAToYRow_NEON(const uint8* src_bgra, uint8* dst_y, int width); +void ABGRToYRow_NEON(const uint8* src_abgr, uint8* dst_y, int width); +void RGBAToYRow_NEON(const uint8* src_rgba, uint8* dst_y, int width); +void RGB24ToYRow_NEON(const uint8* src_rgb24, uint8* dst_y, int width); +void RAWToYRow_NEON(const uint8* src_raw, uint8* dst_y, int width); +void RGB565ToYRow_NEON(const uint8* src_rgb565, uint8* dst_y, int width); +void ARGB1555ToYRow_NEON(const uint8* src_argb1555, uint8* dst_y, int width); +void ARGB4444ToYRow_NEON(const uint8* src_argb4444, uint8* dst_y, int width); +void ARGBToYRow_C(const uint8* src_argb, uint8* dst_y, int width); +void ARGBToYJRow_C(const uint8* src_argb, uint8* dst_y, int width); +void BGRAToYRow_C(const uint8* src_bgra, uint8* dst_y, int width); +void ABGRToYRow_C(const uint8* src_abgr, uint8* dst_y, int width); +void RGBAToYRow_C(const uint8* src_rgba, uint8* dst_y, int width); +void RGB24ToYRow_C(const uint8* src_rgb24, uint8* dst_y, int width); +void RAWToYRow_C(const uint8* src_raw, uint8* dst_y, int width); +void RGB565ToYRow_C(const uint8* src_rgb565, uint8* dst_y, int width); +void ARGB1555ToYRow_C(const uint8* src_argb1555, uint8* dst_y, int width); +void ARGB4444ToYRow_C(const uint8* src_argb4444, uint8* dst_y, int width); +void ARGBToYRow_Any_SSSE3(const uint8* src_argb, uint8* dst_y, int width); +void ARGBToYJRow_Any_SSSE3(const uint8* src_argb, uint8* dst_y, int width); +void BGRAToYRow_Any_SSSE3(const uint8* src_bgra, uint8* dst_y, int width); +void ABGRToYRow_Any_SSSE3(const uint8* src_abgr, uint8* dst_y, int width); +void RGBAToYRow_Any_SSSE3(const uint8* src_rgba, uint8* dst_y, int width); +void RGB24ToYRow_Any_SSSE3(const uint8* src_rgb24, uint8* dst_y, int width); +void RAWToYRow_Any_SSSE3(const uint8* src_raw, uint8* dst_y, int width); +void ARGBToYRow_Any_NEON(const uint8* src_argb, uint8* dst_y, int width); +void ARGBToYJRow_Any_NEON(const uint8* src_argb, uint8* dst_y, int width); +void BGRAToYRow_Any_NEON(const uint8* src_bgra, uint8* dst_y, int width); +void ABGRToYRow_Any_NEON(const uint8* src_abgr, uint8* dst_y, int width); +void RGBAToYRow_Any_NEON(const uint8* src_rgba, uint8* dst_y, int width); +void RGB24ToYRow_Any_NEON(const uint8* src_rgb24, uint8* dst_y, int width); +void RAWToYRow_Any_NEON(const uint8* src_raw, uint8* dst_y, int width); +void RGB565ToYRow_Any_NEON(const uint8* src_rgb565, uint8* dst_y, int width); +void ARGB1555ToYRow_Any_NEON(const uint8* src_argb1555, uint8* dst_y, + int width); +void ARGB4444ToYRow_Any_NEON(const uint8* src_argb4444, uint8* dst_y, + int width); + +void ARGBToUVRow_AVX2(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUVJRow_AVX2(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUVRow_SSSE3(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUVJRow_SSSE3(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void BGRAToUVRow_SSSE3(const uint8* src_bgra, int src_stride_bgra, + uint8* dst_u, uint8* dst_v, int width); +void ABGRToUVRow_SSSE3(const uint8* src_abgr, int src_stride_abgr, + uint8* dst_u, uint8* dst_v, int width); +void RGBAToUVRow_SSSE3(const uint8* src_rgba, int src_stride_rgba, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUVRow_Any_AVX2(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUVJRow_Any_AVX2(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUVRow_Any_SSSE3(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUVJRow_Any_SSSE3(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void BGRAToUVRow_Any_SSSE3(const uint8* src_bgra, int src_stride_bgra, + uint8* dst_u, uint8* dst_v, int width); +void ABGRToUVRow_Any_SSSE3(const uint8* src_abgr, int src_stride_abgr, + uint8* dst_u, uint8* dst_v, int width); +void RGBAToUVRow_Any_SSSE3(const uint8* src_rgba, int src_stride_rgba, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUV444Row_Any_NEON(const uint8* src_argb, uint8* dst_u, uint8* dst_v, + int width); +void ARGBToUV411Row_Any_NEON(const uint8* src_argb, uint8* dst_u, uint8* dst_v, + int width); +void ARGBToUVRow_Any_NEON(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUVJRow_Any_NEON(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void BGRAToUVRow_Any_NEON(const uint8* src_bgra, int src_stride_bgra, + uint8* dst_u, uint8* dst_v, int width); +void ABGRToUVRow_Any_NEON(const uint8* src_abgr, int src_stride_abgr, + uint8* dst_u, uint8* dst_v, int width); +void RGBAToUVRow_Any_NEON(const uint8* src_rgba, int src_stride_rgba, + uint8* dst_u, uint8* dst_v, int width); +void RGB24ToUVRow_Any_NEON(const uint8* src_rgb24, int src_stride_rgb24, + uint8* dst_u, uint8* dst_v, int width); +void RAWToUVRow_Any_NEON(const uint8* src_raw, int src_stride_raw, + uint8* dst_u, uint8* dst_v, int width); +void RGB565ToUVRow_Any_NEON(const uint8* src_rgb565, int src_stride_rgb565, + uint8* dst_u, uint8* dst_v, int width); +void ARGB1555ToUVRow_Any_NEON(const uint8* src_argb1555, + int src_stride_argb1555, + uint8* dst_u, uint8* dst_v, int width); +void ARGB4444ToUVRow_Any_NEON(const uint8* src_argb4444, + int src_stride_argb4444, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUVRow_C(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUVJRow_C(const uint8* src_argb, int src_stride_argb, + uint8* dst_u, uint8* dst_v, int width); +void BGRAToUVRow_C(const uint8* src_bgra, int src_stride_bgra, + uint8* dst_u, uint8* dst_v, int width); +void ABGRToUVRow_C(const uint8* src_abgr, int src_stride_abgr, + uint8* dst_u, uint8* dst_v, int width); +void RGBAToUVRow_C(const uint8* src_rgba, int src_stride_rgba, + uint8* dst_u, uint8* dst_v, int width); +void RGB24ToUVRow_C(const uint8* src_rgb24, int src_stride_rgb24, + uint8* dst_u, uint8* dst_v, int width); +void RAWToUVRow_C(const uint8* src_raw, int src_stride_raw, + uint8* dst_u, uint8* dst_v, int width); +void RGB565ToUVRow_C(const uint8* src_rgb565, int src_stride_rgb565, + uint8* dst_u, uint8* dst_v, int width); +void ARGB1555ToUVRow_C(const uint8* src_argb1555, int src_stride_argb1555, + uint8* dst_u, uint8* dst_v, int width); +void ARGB4444ToUVRow_C(const uint8* src_argb4444, int src_stride_argb4444, + uint8* dst_u, uint8* dst_v, int width); + +void ARGBToUV444Row_SSSE3(const uint8* src_argb, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUV444Row_Any_SSSE3(const uint8* src_argb, + uint8* dst_u, uint8* dst_v, int width); + +void ARGBToUV444Row_C(const uint8* src_argb, + uint8* dst_u, uint8* dst_v, int width); +void ARGBToUV411Row_C(const uint8* src_argb, + uint8* dst_u, uint8* dst_v, int width); + +void MirrorRow_AVX2(const uint8* src, uint8* dst, int width); +void MirrorRow_SSSE3(const uint8* src, uint8* dst, int width); +void MirrorRow_NEON(const uint8* src, uint8* dst, int width); +void MirrorRow_DSPR2(const uint8* src, uint8* dst, int width); +void MirrorRow_MSA(const uint8* src, uint8* dst, int width); +void MirrorRow_C(const uint8* src, uint8* dst, int width); +void MirrorRow_Any_AVX2(const uint8* src, uint8* dst, int width); +void MirrorRow_Any_SSSE3(const uint8* src, uint8* dst, int width); +void MirrorRow_Any_SSE2(const uint8* src, uint8* dst, int width); +void MirrorRow_Any_NEON(const uint8* src, uint8* dst, int width); +void MirrorRow_Any_MSA(const uint8* src, uint8* dst, int width); + +void MirrorUVRow_SSSE3(const uint8* src_uv, uint8* dst_u, uint8* dst_v, + int width); +void MirrorUVRow_NEON(const uint8* src_uv, uint8* dst_u, uint8* dst_v, + int width); +void MirrorUVRow_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, + int width); +void MirrorUVRow_C(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width); + +void ARGBMirrorRow_AVX2(const uint8* src, uint8* dst, int width); +void ARGBMirrorRow_SSE2(const uint8* src, uint8* dst, int width); +void ARGBMirrorRow_NEON(const uint8* src, uint8* dst, int width); +void ARGBMirrorRow_MSA(const uint8* src, uint8* dst, int width); +void ARGBMirrorRow_C(const uint8* src, uint8* dst, int width); +void ARGBMirrorRow_Any_AVX2(const uint8* src, uint8* dst, int width); +void ARGBMirrorRow_Any_SSE2(const uint8* src, uint8* dst, int width); +void ARGBMirrorRow_Any_NEON(const uint8* src, uint8* dst, int width); +void ARGBMirrorRow_Any_MSA(const uint8* src, uint8* dst, int width); + +void SplitUVRow_C(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width); +void SplitUVRow_SSE2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, + int width); +void SplitUVRow_AVX2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, + int width); +void SplitUVRow_NEON(const uint8* src_uv, uint8* dst_u, uint8* dst_v, + int width); +void SplitUVRow_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, + int width); +void SplitUVRow_Any_SSE2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, + int width); +void SplitUVRow_Any_AVX2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, + int width); +void SplitUVRow_Any_NEON(const uint8* src_uv, uint8* dst_u, uint8* dst_v, + int width); +void SplitUVRow_Any_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, + int width); + +void MergeUVRow_C(const uint8* src_u, const uint8* src_v, uint8* dst_uv, + int width); +void MergeUVRow_SSE2(const uint8* src_u, const uint8* src_v, uint8* dst_uv, + int width); +void MergeUVRow_AVX2(const uint8* src_u, const uint8* src_v, uint8* dst_uv, + int width); +void MergeUVRow_NEON(const uint8* src_u, const uint8* src_v, uint8* dst_uv, + int width); +void MergeUVRow_Any_SSE2(const uint8* src_u, const uint8* src_v, uint8* dst_uv, + int width); +void MergeUVRow_Any_AVX2(const uint8* src_u, const uint8* src_v, uint8* dst_uv, + int width); +void MergeUVRow_Any_NEON(const uint8* src_u, const uint8* src_v, uint8* dst_uv, + int width); + +void CopyRow_SSE2(const uint8* src, uint8* dst, int count); +void CopyRow_AVX(const uint8* src, uint8* dst, int count); +void CopyRow_ERMS(const uint8* src, uint8* dst, int count); +void CopyRow_NEON(const uint8* src, uint8* dst, int count); +void CopyRow_MIPS(const uint8* src, uint8* dst, int count); +void CopyRow_C(const uint8* src, uint8* dst, int count); +void CopyRow_Any_SSE2(const uint8* src, uint8* dst, int count); +void CopyRow_Any_AVX(const uint8* src, uint8* dst, int count); +void CopyRow_Any_NEON(const uint8* src, uint8* dst, int count); + +void CopyRow_16_C(const uint16* src, uint16* dst, int count); + +void ARGBCopyAlphaRow_C(const uint8* src_argb, uint8* dst_argb, int width); +void ARGBCopyAlphaRow_SSE2(const uint8* src_argb, uint8* dst_argb, int width); +void ARGBCopyAlphaRow_AVX2(const uint8* src_argb, uint8* dst_argb, int width); +void ARGBCopyAlphaRow_Any_SSE2(const uint8* src_argb, uint8* dst_argb, + int width); +void ARGBCopyAlphaRow_Any_AVX2(const uint8* src_argb, uint8* dst_argb, + int width); + +void ARGBExtractAlphaRow_C(const uint8* src_argb, uint8* dst_a, int width); +void ARGBExtractAlphaRow_SSE2(const uint8* src_argb, uint8* dst_a, int width); +void ARGBExtractAlphaRow_NEON(const uint8* src_argb, uint8* dst_a, int width); +void ARGBExtractAlphaRow_Any_SSE2(const uint8* src_argb, uint8* dst_a, + int width); +void ARGBExtractAlphaRow_Any_NEON(const uint8* src_argb, uint8* dst_a, + int width); + +void ARGBCopyYToAlphaRow_C(const uint8* src_y, uint8* dst_argb, int width); +void ARGBCopyYToAlphaRow_SSE2(const uint8* src_y, uint8* dst_argb, int width); +void ARGBCopyYToAlphaRow_AVX2(const uint8* src_y, uint8* dst_argb, int width); +void ARGBCopyYToAlphaRow_Any_SSE2(const uint8* src_y, uint8* dst_argb, + int width); +void ARGBCopyYToAlphaRow_Any_AVX2(const uint8* src_y, uint8* dst_argb, + int width); + +void SetRow_C(uint8* dst, uint8 v8, int count); +void SetRow_X86(uint8* dst, uint8 v8, int count); +void SetRow_ERMS(uint8* dst, uint8 v8, int count); +void SetRow_NEON(uint8* dst, uint8 v8, int count); +void SetRow_Any_X86(uint8* dst, uint8 v8, int count); +void SetRow_Any_NEON(uint8* dst, uint8 v8, int count); + +void ARGBSetRow_C(uint8* dst_argb, uint32 v32, int count); +void ARGBSetRow_X86(uint8* dst_argb, uint32 v32, int count); +void ARGBSetRow_NEON(uint8* dst_argb, uint32 v32, int count); +void ARGBSetRow_Any_NEON(uint8* dst_argb, uint32 v32, int count); + +// ARGBShufflers for BGRAToARGB etc. +void ARGBShuffleRow_C(const uint8* src_argb, uint8* dst_argb, + const uint8* shuffler, int width); +void ARGBShuffleRow_SSE2(const uint8* src_argb, uint8* dst_argb, + const uint8* shuffler, int width); +void ARGBShuffleRow_SSSE3(const uint8* src_argb, uint8* dst_argb, + const uint8* shuffler, int width); +void ARGBShuffleRow_AVX2(const uint8* src_argb, uint8* dst_argb, + const uint8* shuffler, int width); +void ARGBShuffleRow_NEON(const uint8* src_argb, uint8* dst_argb, + const uint8* shuffler, int width); +void ARGBShuffleRow_Any_SSE2(const uint8* src_argb, uint8* dst_argb, + const uint8* shuffler, int width); +void ARGBShuffleRow_Any_SSSE3(const uint8* src_argb, uint8* dst_argb, + const uint8* shuffler, int width); +void ARGBShuffleRow_Any_AVX2(const uint8* src_argb, uint8* dst_argb, + const uint8* shuffler, int width); +void ARGBShuffleRow_Any_NEON(const uint8* src_argb, uint8* dst_argb, + const uint8* shuffler, int width); + +void RGB24ToARGBRow_SSSE3(const uint8* src_rgb24, uint8* dst_argb, int width); +void RAWToARGBRow_SSSE3(const uint8* src_raw, uint8* dst_argb, int width); +void RAWToRGB24Row_SSSE3(const uint8* src_raw, uint8* dst_rgb24, int width); +void RGB565ToARGBRow_SSE2(const uint8* src_rgb565, uint8* dst_argb, int width); +void ARGB1555ToARGBRow_SSE2(const uint8* src_argb1555, uint8* dst_argb, + int width); +void ARGB4444ToARGBRow_SSE2(const uint8* src_argb4444, uint8* dst_argb, + int width); +void RGB565ToARGBRow_AVX2(const uint8* src_rgb565, uint8* dst_argb, int width); +void ARGB1555ToARGBRow_AVX2(const uint8* src_argb1555, uint8* dst_argb, + int width); +void ARGB4444ToARGBRow_AVX2(const uint8* src_argb4444, uint8* dst_argb, + int width); + +void RGB24ToARGBRow_NEON(const uint8* src_rgb24, uint8* dst_argb, int width); +void RAWToARGBRow_NEON(const uint8* src_raw, uint8* dst_argb, int width); +void RAWToRGB24Row_NEON(const uint8* src_raw, uint8* dst_rgb24, int width); +void RGB565ToARGBRow_NEON(const uint8* src_rgb565, uint8* dst_argb, int width); +void ARGB1555ToARGBRow_NEON(const uint8* src_argb1555, uint8* dst_argb, + int width); +void ARGB4444ToARGBRow_NEON(const uint8* src_argb4444, uint8* dst_argb, + int width); +void RGB24ToARGBRow_C(const uint8* src_rgb24, uint8* dst_argb, int width); +void RAWToARGBRow_C(const uint8* src_raw, uint8* dst_argb, int width); +void RAWToRGB24Row_C(const uint8* src_raw, uint8* dst_rgb24, int width); +void RGB565ToARGBRow_C(const uint8* src_rgb, uint8* dst_argb, int width); +void ARGB1555ToARGBRow_C(const uint8* src_argb, uint8* dst_argb, int width); +void ARGB4444ToARGBRow_C(const uint8* src_argb, uint8* dst_argb, int width); +void RGB24ToARGBRow_Any_SSSE3(const uint8* src_rgb24, uint8* dst_argb, + int width); +void RAWToARGBRow_Any_SSSE3(const uint8* src_raw, uint8* dst_argb, int width); +void RAWToRGB24Row_Any_SSSE3(const uint8* src_raw, uint8* dst_rgb24, int width); + +void RGB565ToARGBRow_Any_SSE2(const uint8* src_rgb565, uint8* dst_argb, + int width); +void ARGB1555ToARGBRow_Any_SSE2(const uint8* src_argb1555, uint8* dst_argb, + int width); +void ARGB4444ToARGBRow_Any_SSE2(const uint8* src_argb4444, uint8* dst_argb, + int width); +void RGB565ToARGBRow_Any_AVX2(const uint8* src_rgb565, uint8* dst_argb, + int width); +void ARGB1555ToARGBRow_Any_AVX2(const uint8* src_argb1555, uint8* dst_argb, + int width); +void ARGB4444ToARGBRow_Any_AVX2(const uint8* src_argb4444, uint8* dst_argb, + int width); + +void RGB24ToARGBRow_Any_NEON(const uint8* src_rgb24, uint8* dst_argb, + int width); +void RAWToARGBRow_Any_NEON(const uint8* src_raw, uint8* dst_argb, int width); +void RAWToRGB24Row_Any_NEON(const uint8* src_raw, uint8* dst_rgb24, int width); +void RGB565ToARGBRow_Any_NEON(const uint8* src_rgb565, uint8* dst_argb, + int width); +void ARGB1555ToARGBRow_Any_NEON(const uint8* src_argb1555, uint8* dst_argb, + int width); +void ARGB4444ToARGBRow_Any_NEON(const uint8* src_argb4444, uint8* dst_argb, + int width); + +void ARGBToRGB24Row_SSSE3(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToRAWRow_SSSE3(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToRGB565Row_SSE2(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToARGB1555Row_SSE2(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToARGB4444Row_SSE2(const uint8* src_argb, uint8* dst_rgb, int width); + +void ARGBToRGB565DitherRow_C(const uint8* src_argb, uint8* dst_rgb, + const uint32 dither4, int width); +void ARGBToRGB565DitherRow_SSE2(const uint8* src_argb, uint8* dst_rgb, + const uint32 dither4, int width); +void ARGBToRGB565DitherRow_AVX2(const uint8* src_argb, uint8* dst_rgb, + const uint32 dither4, int width); + +void ARGBToRGB565Row_AVX2(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToARGB1555Row_AVX2(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToARGB4444Row_AVX2(const uint8* src_argb, uint8* dst_rgb, int width); + +void ARGBToRGB24Row_NEON(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToRAWRow_NEON(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToRGB565Row_NEON(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToARGB1555Row_NEON(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToARGB4444Row_NEON(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToRGB565DitherRow_NEON(const uint8* src_argb, uint8* dst_rgb, + const uint32 dither4, int width); + +void ARGBToRGBARow_C(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToRGB24Row_C(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToRAWRow_C(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToRGB565Row_C(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToARGB1555Row_C(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToARGB4444Row_C(const uint8* src_argb, uint8* dst_rgb, int width); + +void J400ToARGBRow_SSE2(const uint8* src_y, uint8* dst_argb, int width); +void J400ToARGBRow_AVX2(const uint8* src_y, uint8* dst_argb, int width); +void J400ToARGBRow_NEON(const uint8* src_y, uint8* dst_argb, int width); +void J400ToARGBRow_C(const uint8* src_y, uint8* dst_argb, int width); +void J400ToARGBRow_Any_SSE2(const uint8* src_y, uint8* dst_argb, int width); +void J400ToARGBRow_Any_AVX2(const uint8* src_y, uint8* dst_argb, int width); +void J400ToARGBRow_Any_NEON(const uint8* src_y, uint8* dst_argb, int width); + +void I444ToARGBRow_C(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGBRow_C(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGBRow_C(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422AlphaToARGBRow_C(const uint8* y_buf, + const uint8* u_buf, + const uint8* v_buf, + const uint8* a_buf, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I411ToARGBRow_C(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToARGBRow_C(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToRGB565Row_C(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV21ToARGBRow_C(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void YUY2ToARGBRow_C(const uint8* src_yuy2, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void UYVYToARGBRow_C(const uint8* src_uyvy, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGBARow_C(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgba, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB24Row_C(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgb24, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB4444Row_C(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb4444, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB1555Row_C(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb4444, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB565Row_C(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgb565, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGBRow_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGBRow_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGBARow_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I444ToARGBRow_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I444ToARGBRow_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I444ToARGBRow_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I444ToARGBRow_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGBRow_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422AlphaToARGBRow_SSSE3(const uint8* y_buf, + const uint8* u_buf, + const uint8* v_buf, + const uint8* a_buf, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422AlphaToARGBRow_AVX2(const uint8* y_buf, + const uint8* u_buf, + const uint8* v_buf, + const uint8* a_buf, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGBRow_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I411ToARGBRow_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I411ToARGBRow_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToARGBRow_SSSE3(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToARGBRow_AVX2(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToRGB565Row_SSSE3(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToRGB565Row_AVX2(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV21ToARGBRow_SSSE3(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV21ToARGBRow_AVX2(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void YUY2ToARGBRow_SSSE3(const uint8* src_yuy2, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void UYVYToARGBRow_SSSE3(const uint8* src_uyvy, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void YUY2ToARGBRow_AVX2(const uint8* src_yuy2, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void UYVYToARGBRow_AVX2(const uint8* src_uyvy, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGBARow_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgba, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB4444Row_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB4444Row_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB1555Row_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB1555Row_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB565Row_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB565Row_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB24Row_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgb24, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB24Row_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgb24, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGBRow_Any_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGBARow_Any_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I444ToARGBRow_Any_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I444ToARGBRow_Any_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGBRow_Any_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422AlphaToARGBRow_Any_SSSE3(const uint8* y_buf, + const uint8* u_buf, + const uint8* v_buf, + const uint8* a_buf, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422AlphaToARGBRow_Any_AVX2(const uint8* y_buf, + const uint8* u_buf, + const uint8* v_buf, + const uint8* a_buf, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I411ToARGBRow_Any_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I411ToARGBRow_Any_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToARGBRow_Any_SSSE3(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToARGBRow_Any_AVX2(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV21ToARGBRow_Any_SSSE3(const uint8* src_y, + const uint8* src_vu, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV21ToARGBRow_Any_AVX2(const uint8* src_y, + const uint8* src_vu, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToRGB565Row_Any_SSSE3(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToRGB565Row_Any_AVX2(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void YUY2ToARGBRow_Any_SSSE3(const uint8* src_yuy2, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void UYVYToARGBRow_Any_SSSE3(const uint8* src_uyvy, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void YUY2ToARGBRow_Any_AVX2(const uint8* src_yuy2, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void UYVYToARGBRow_Any_AVX2(const uint8* src_uyvy, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGBARow_Any_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgba, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB4444Row_Any_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgba, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB4444Row_Any_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgba, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB1555Row_Any_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgba, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB1555Row_Any_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgba, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB565Row_Any_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgba, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB565Row_Any_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_rgba, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB24Row_Any_SSSE3(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB24Row_Any_AVX2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); + +void I400ToARGBRow_C(const uint8* src_y, uint8* dst_argb, int width); +void I400ToARGBRow_SSE2(const uint8* src_y, uint8* dst_argb, int width); +void I400ToARGBRow_AVX2(const uint8* src_y, uint8* dst_argb, int width); +void I400ToARGBRow_NEON(const uint8* src_y, uint8* dst_argb, int width); +void I400ToARGBRow_Any_SSE2(const uint8* src_y, uint8* dst_argb, int width); +void I400ToARGBRow_Any_AVX2(const uint8* src_y, uint8* dst_argb, int width); +void I400ToARGBRow_Any_NEON(const uint8* src_y, uint8* dst_argb, int width); + +// ARGB preattenuated alpha blend. +void ARGBBlendRow_SSSE3(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBBlendRow_NEON(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBBlendRow_C(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); + +// Unattenuated planar alpha blend. +void BlendPlaneRow_SSSE3(const uint8* src0, const uint8* src1, + const uint8* alpha, uint8* dst, int width); +void BlendPlaneRow_Any_SSSE3(const uint8* src0, const uint8* src1, + const uint8* alpha, uint8* dst, int width); +void BlendPlaneRow_AVX2(const uint8* src0, const uint8* src1, + const uint8* alpha, uint8* dst, int width); +void BlendPlaneRow_Any_AVX2(const uint8* src0, const uint8* src1, + const uint8* alpha, uint8* dst, int width); +void BlendPlaneRow_C(const uint8* src0, const uint8* src1, + const uint8* alpha, uint8* dst, int width); + +// ARGB multiply images. Same API as Blend, but these require +// pointer and width alignment for SSE2. +void ARGBMultiplyRow_C(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBMultiplyRow_SSE2(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBMultiplyRow_Any_SSE2(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBMultiplyRow_AVX2(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBMultiplyRow_Any_AVX2(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBMultiplyRow_NEON(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBMultiplyRow_Any_NEON(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); + +// ARGB add images. +void ARGBAddRow_C(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBAddRow_SSE2(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBAddRow_Any_SSE2(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBAddRow_AVX2(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBAddRow_Any_AVX2(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBAddRow_NEON(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBAddRow_Any_NEON(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); + +// ARGB subtract images. Same API as Blend, but these require +// pointer and width alignment for SSE2. +void ARGBSubtractRow_C(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBSubtractRow_SSE2(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBSubtractRow_Any_SSE2(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBSubtractRow_AVX2(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBSubtractRow_Any_AVX2(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBSubtractRow_NEON(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); +void ARGBSubtractRow_Any_NEON(const uint8* src_argb, const uint8* src_argb1, + uint8* dst_argb, int width); + +void ARGBToRGB24Row_Any_SSSE3(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToRAWRow_Any_SSSE3(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToRGB565Row_Any_SSE2(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToARGB1555Row_Any_SSE2(const uint8* src_argb, uint8* dst_rgb, + int width); +void ARGBToARGB4444Row_Any_SSE2(const uint8* src_argb, uint8* dst_rgb, + int width); + +void ARGBToRGB565DitherRow_Any_SSE2(const uint8* src_argb, uint8* dst_rgb, + const uint32 dither4, int width); +void ARGBToRGB565DitherRow_Any_AVX2(const uint8* src_argb, uint8* dst_rgb, + const uint32 dither4, int width); + +void ARGBToRGB565Row_Any_AVX2(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToARGB1555Row_Any_AVX2(const uint8* src_argb, uint8* dst_rgb, + int width); +void ARGBToARGB4444Row_Any_AVX2(const uint8* src_argb, uint8* dst_rgb, + int width); + +void ARGBToRGB24Row_Any_NEON(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToRAWRow_Any_NEON(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToRGB565Row_Any_NEON(const uint8* src_argb, uint8* dst_rgb, int width); +void ARGBToARGB1555Row_Any_NEON(const uint8* src_argb, uint8* dst_rgb, + int width); +void ARGBToARGB4444Row_Any_NEON(const uint8* src_argb, uint8* dst_rgb, + int width); +void ARGBToRGB565DitherRow_Any_NEON(const uint8* src_argb, uint8* dst_rgb, + const uint32 dither4, int width); + +void I444ToARGBRow_Any_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGBRow_Any_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422AlphaToARGBRow_Any_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + const uint8* src_a, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I411ToARGBRow_Any_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGBARow_Any_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB24Row_Any_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB4444Row_Any_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGB1555Row_Any_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToRGB565Row_Any_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToARGBRow_Any_NEON(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV21ToARGBRow_Any_NEON(const uint8* src_y, + const uint8* src_vu, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToRGB565Row_Any_NEON(const uint8* src_y, + const uint8* src_uv, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void YUY2ToARGBRow_Any_NEON(const uint8* src_yuy2, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void UYVYToARGBRow_Any_NEON(const uint8* src_uyvy, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGBRow_DSPR2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void I422ToARGBRow_DSPR2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_argb, + const struct YuvConstants* yuvconstants, + int width); + +void YUY2ToYRow_AVX2(const uint8* src_yuy2, uint8* dst_y, int width); +void YUY2ToUVRow_AVX2(const uint8* src_yuy2, int stride_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void YUY2ToUV422Row_AVX2(const uint8* src_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void YUY2ToYRow_SSE2(const uint8* src_yuy2, uint8* dst_y, int width); +void YUY2ToUVRow_SSE2(const uint8* src_yuy2, int stride_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void YUY2ToUV422Row_SSE2(const uint8* src_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void YUY2ToYRow_NEON(const uint8* src_yuy2, uint8* dst_y, int width); +void YUY2ToUVRow_NEON(const uint8* src_yuy2, int stride_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void YUY2ToUV422Row_NEON(const uint8* src_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void YUY2ToYRow_C(const uint8* src_yuy2, uint8* dst_y, int width); +void YUY2ToUVRow_C(const uint8* src_yuy2, int stride_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void YUY2ToUV422Row_C(const uint8* src_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void YUY2ToYRow_Any_AVX2(const uint8* src_yuy2, uint8* dst_y, int width); +void YUY2ToUVRow_Any_AVX2(const uint8* src_yuy2, int stride_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void YUY2ToUV422Row_Any_AVX2(const uint8* src_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void YUY2ToYRow_Any_SSE2(const uint8* src_yuy2, uint8* dst_y, int width); +void YUY2ToUVRow_Any_SSE2(const uint8* src_yuy2, int stride_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void YUY2ToUV422Row_Any_SSE2(const uint8* src_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void YUY2ToYRow_Any_NEON(const uint8* src_yuy2, uint8* dst_y, int width); +void YUY2ToUVRow_Any_NEON(const uint8* src_yuy2, int stride_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void YUY2ToUV422Row_Any_NEON(const uint8* src_yuy2, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToYRow_AVX2(const uint8* src_uyvy, uint8* dst_y, int width); +void UYVYToUVRow_AVX2(const uint8* src_uyvy, int stride_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToUV422Row_AVX2(const uint8* src_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToYRow_SSE2(const uint8* src_uyvy, uint8* dst_y, int width); +void UYVYToUVRow_SSE2(const uint8* src_uyvy, int stride_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToUV422Row_SSE2(const uint8* src_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToYRow_AVX2(const uint8* src_uyvy, uint8* dst_y, int width); +void UYVYToUVRow_AVX2(const uint8* src_uyvy, int stride_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToUV422Row_AVX2(const uint8* src_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToYRow_NEON(const uint8* src_uyvy, uint8* dst_y, int width); +void UYVYToUVRow_NEON(const uint8* src_uyvy, int stride_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToUV422Row_NEON(const uint8* src_uyvy, + uint8* dst_u, uint8* dst_v, int width); + +void UYVYToYRow_C(const uint8* src_uyvy, uint8* dst_y, int width); +void UYVYToUVRow_C(const uint8* src_uyvy, int stride_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToUV422Row_C(const uint8* src_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToYRow_Any_AVX2(const uint8* src_uyvy, uint8* dst_y, int width); +void UYVYToUVRow_Any_AVX2(const uint8* src_uyvy, int stride_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToUV422Row_Any_AVX2(const uint8* src_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToYRow_Any_SSE2(const uint8* src_uyvy, uint8* dst_y, int width); +void UYVYToUVRow_Any_SSE2(const uint8* src_uyvy, int stride_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToUV422Row_Any_SSE2(const uint8* src_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToYRow_Any_NEON(const uint8* src_uyvy, uint8* dst_y, int width); +void UYVYToUVRow_Any_NEON(const uint8* src_uyvy, int stride_uyvy, + uint8* dst_u, uint8* dst_v, int width); +void UYVYToUV422Row_Any_NEON(const uint8* src_uyvy, + uint8* dst_u, uint8* dst_v, int width); + +void I422ToYUY2Row_C(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_yuy2, int width); +void I422ToUYVYRow_C(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_uyvy, int width); +void I422ToYUY2Row_SSE2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_yuy2, int width); +void I422ToUYVYRow_SSE2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_uyvy, int width); +void I422ToYUY2Row_Any_SSE2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_yuy2, int width); +void I422ToUYVYRow_Any_SSE2(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_uyvy, int width); +void I422ToYUY2Row_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_yuy2, int width); +void I422ToUYVYRow_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_uyvy, int width); +void I422ToYUY2Row_Any_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_yuy2, int width); +void I422ToUYVYRow_Any_NEON(const uint8* src_y, + const uint8* src_u, + const uint8* src_v, + uint8* dst_uyvy, int width); + +// Effects related row functions. +void ARGBAttenuateRow_C(const uint8* src_argb, uint8* dst_argb, int width); +void ARGBAttenuateRow_SSSE3(const uint8* src_argb, uint8* dst_argb, int width); +void ARGBAttenuateRow_AVX2(const uint8* src_argb, uint8* dst_argb, int width); +void ARGBAttenuateRow_NEON(const uint8* src_argb, uint8* dst_argb, int width); +void ARGBAttenuateRow_Any_SSE2(const uint8* src_argb, uint8* dst_argb, + int width); +void ARGBAttenuateRow_Any_SSSE3(const uint8* src_argb, uint8* dst_argb, + int width); +void ARGBAttenuateRow_Any_AVX2(const uint8* src_argb, uint8* dst_argb, + int width); +void ARGBAttenuateRow_Any_NEON(const uint8* src_argb, uint8* dst_argb, + int width); + +// Inverse table for unattenuate, shared by C and SSE2. +extern const uint32 fixed_invtbl8[256]; +void ARGBUnattenuateRow_C(const uint8* src_argb, uint8* dst_argb, int width); +void ARGBUnattenuateRow_SSE2(const uint8* src_argb, uint8* dst_argb, int width); +void ARGBUnattenuateRow_AVX2(const uint8* src_argb, uint8* dst_argb, int width); +void ARGBUnattenuateRow_Any_SSE2(const uint8* src_argb, uint8* dst_argb, + int width); +void ARGBUnattenuateRow_Any_AVX2(const uint8* src_argb, uint8* dst_argb, + int width); + +void ARGBGrayRow_C(const uint8* src_argb, uint8* dst_argb, int width); +void ARGBGrayRow_SSSE3(const uint8* src_argb, uint8* dst_argb, int width); +void ARGBGrayRow_NEON(const uint8* src_argb, uint8* dst_argb, int width); + +void ARGBSepiaRow_C(uint8* dst_argb, int width); +void ARGBSepiaRow_SSSE3(uint8* dst_argb, int width); +void ARGBSepiaRow_NEON(uint8* dst_argb, int width); + +void ARGBColorMatrixRow_C(const uint8* src_argb, uint8* dst_argb, + const int8* matrix_argb, int width); +void ARGBColorMatrixRow_SSSE3(const uint8* src_argb, uint8* dst_argb, + const int8* matrix_argb, int width); +void ARGBColorMatrixRow_NEON(const uint8* src_argb, uint8* dst_argb, + const int8* matrix_argb, int width); + +void ARGBColorTableRow_C(uint8* dst_argb, const uint8* table_argb, int width); +void ARGBColorTableRow_X86(uint8* dst_argb, const uint8* table_argb, int width); + +void RGBColorTableRow_C(uint8* dst_argb, const uint8* table_argb, int width); +void RGBColorTableRow_X86(uint8* dst_argb, const uint8* table_argb, int width); + +void ARGBQuantizeRow_C(uint8* dst_argb, int scale, int interval_size, + int interval_offset, int width); +void ARGBQuantizeRow_SSE2(uint8* dst_argb, int scale, int interval_size, + int interval_offset, int width); +void ARGBQuantizeRow_NEON(uint8* dst_argb, int scale, int interval_size, + int interval_offset, int width); + +void ARGBShadeRow_C(const uint8* src_argb, uint8* dst_argb, int width, + uint32 value); +void ARGBShadeRow_SSE2(const uint8* src_argb, uint8* dst_argb, int width, + uint32 value); +void ARGBShadeRow_NEON(const uint8* src_argb, uint8* dst_argb, int width, + uint32 value); + +// Used for blur. +void CumulativeSumToAverageRow_SSE2(const int32* topleft, const int32* botleft, + int width, int area, uint8* dst, int count); +void ComputeCumulativeSumRow_SSE2(const uint8* row, int32* cumsum, + const int32* previous_cumsum, int width); + +void CumulativeSumToAverageRow_C(const int32* topleft, const int32* botleft, + int width, int area, uint8* dst, int count); +void ComputeCumulativeSumRow_C(const uint8* row, int32* cumsum, + const int32* previous_cumsum, int width); + +LIBYUV_API +void ARGBAffineRow_C(const uint8* src_argb, int src_argb_stride, + uint8* dst_argb, const float* uv_dudv, int width); +LIBYUV_API +void ARGBAffineRow_SSE2(const uint8* src_argb, int src_argb_stride, + uint8* dst_argb, const float* uv_dudv, int width); + +// Used for I420Scale, ARGBScale, and ARGBInterpolate. +void InterpolateRow_C(uint8* dst_ptr, const uint8* src_ptr, + ptrdiff_t src_stride_ptr, + int width, int source_y_fraction); +void InterpolateRow_SSSE3(uint8* dst_ptr, const uint8* src_ptr, + ptrdiff_t src_stride_ptr, int width, + int source_y_fraction); +void InterpolateRow_AVX2(uint8* dst_ptr, const uint8* src_ptr, + ptrdiff_t src_stride_ptr, int width, + int source_y_fraction); +void InterpolateRow_NEON(uint8* dst_ptr, const uint8* src_ptr, + ptrdiff_t src_stride_ptr, int width, + int source_y_fraction); +void InterpolateRow_DSPR2(uint8* dst_ptr, const uint8* src_ptr, + ptrdiff_t src_stride_ptr, int width, + int source_y_fraction); +void InterpolateRow_Any_NEON(uint8* dst_ptr, const uint8* src_ptr, + ptrdiff_t src_stride_ptr, int width, + int source_y_fraction); +void InterpolateRow_Any_SSSE3(uint8* dst_ptr, const uint8* src_ptr, + ptrdiff_t src_stride_ptr, int width, + int source_y_fraction); +void InterpolateRow_Any_AVX2(uint8* dst_ptr, const uint8* src_ptr, + ptrdiff_t src_stride_ptr, int width, + int source_y_fraction); +void InterpolateRow_Any_DSPR2(uint8* dst_ptr, const uint8* src_ptr, + ptrdiff_t src_stride_ptr, int width, + int source_y_fraction); + +void InterpolateRow_16_C(uint16* dst_ptr, const uint16* src_ptr, + ptrdiff_t src_stride_ptr, + int width, int source_y_fraction); + +// Sobel images. +void SobelXRow_C(const uint8* src_y0, const uint8* src_y1, const uint8* src_y2, + uint8* dst_sobelx, int width); +void SobelXRow_SSE2(const uint8* src_y0, const uint8* src_y1, + const uint8* src_y2, uint8* dst_sobelx, int width); +void SobelXRow_NEON(const uint8* src_y0, const uint8* src_y1, + const uint8* src_y2, uint8* dst_sobelx, int width); +void SobelYRow_C(const uint8* src_y0, const uint8* src_y1, + uint8* dst_sobely, int width); +void SobelYRow_SSE2(const uint8* src_y0, const uint8* src_y1, + uint8* dst_sobely, int width); +void SobelYRow_NEON(const uint8* src_y0, const uint8* src_y1, + uint8* dst_sobely, int width); +void SobelRow_C(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_argb, int width); +void SobelRow_SSE2(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_argb, int width); +void SobelRow_NEON(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_argb, int width); +void SobelToPlaneRow_C(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_y, int width); +void SobelToPlaneRow_SSE2(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_y, int width); +void SobelToPlaneRow_NEON(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_y, int width); +void SobelXYRow_C(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_argb, int width); +void SobelXYRow_SSE2(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_argb, int width); +void SobelXYRow_NEON(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_argb, int width); +void SobelRow_Any_SSE2(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_argb, int width); +void SobelRow_Any_NEON(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_argb, int width); +void SobelToPlaneRow_Any_SSE2(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_y, int width); +void SobelToPlaneRow_Any_NEON(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_y, int width); +void SobelXYRow_Any_SSE2(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_argb, int width); +void SobelXYRow_Any_NEON(const uint8* src_sobelx, const uint8* src_sobely, + uint8* dst_argb, int width); + +void ARGBPolynomialRow_C(const uint8* src_argb, + uint8* dst_argb, const float* poly, + int width); +void ARGBPolynomialRow_SSE2(const uint8* src_argb, + uint8* dst_argb, const float* poly, + int width); +void ARGBPolynomialRow_AVX2(const uint8* src_argb, + uint8* dst_argb, const float* poly, + int width); + +// Scale and convert to half float. +void HalfFloatRow_C(const uint16* src, uint16* dst, float scale, int width); +void HalfFloatRow_AVX2(const uint16* src, uint16* dst, float scale, int width); +void HalfFloatRow_Any_AVX2(const uint16* src, uint16* dst, float scale, + int width); +void HalfFloatRow_SSE2(const uint16* src, uint16* dst, float scale, int width); +void HalfFloatRow_Any_SSE2(const uint16* src, uint16* dst, float scale, + int width); + +void ARGBLumaColorTableRow_C(const uint8* src_argb, uint8* dst_argb, int width, + const uint8* luma, uint32 lumacoeff); +void ARGBLumaColorTableRow_SSSE3(const uint8* src_argb, uint8* dst_argb, + int width, + const uint8* luma, uint32 lumacoeff); + +#ifdef __cplusplus +} // extern "C" +} // namespace libyuv +#endif + +#endif // INCLUDE_LIBYUV_ROW_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/scale.h b/phonelibs/libyuv/x64/include/libyuv/scale.h new file mode 100644 index 00000000000000..ae14694598b2cc --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/scale.h @@ -0,0 +1,103 @@ +/* + * 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_SCALE_H_ +#define INCLUDE_LIBYUV_SCALE_H_ + +#include "libyuv/basic_types.h" + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +// 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); + +LIBYUV_API +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 +// used. This produces basic (blocky) quality at the fastest speed. +// If filtering is kFilterBilinear, interpolation is used to produce a better +// quality image, at the expense of speed. +// If filtering is kFilterBox, averaging is used to produce ever better +// 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, + int src_width, int src_height, + uint8* dst_y, int dst_stride_y, + uint8* dst_u, int dst_stride_u, + uint8* dst_v, int dst_stride_v, + int dst_width, int dst_height, + enum FilterMode filtering); + +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, + LIBYUV_BOOL 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 specialized scalers. +LIBYUV_API +void SetUseReferenceImpl(LIBYUV_BOOL use); +#endif // __cplusplus + +#ifdef __cplusplus +} // extern "C" +} // namespace libyuv +#endif + +#endif // INCLUDE_LIBYUV_SCALE_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/scale_argb.h b/phonelibs/libyuv/x64/include/libyuv/scale_argb.h new file mode 100644 index 00000000000000..35cd191c0f6123 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/scale_argb.h @@ -0,0 +1,56 @@ +/* + * Copyright 2012 The LibYuv Project Authors. All rights reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef INCLUDE_LIBYUV_SCALE_ARGB_H_ +#define INCLUDE_LIBYUV_SCALE_ARGB_H_ + +#include "libyuv/basic_types.h" +#include "libyuv/scale.h" // For FilterMode + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +LIBYUV_API +int ARGBScale(const uint8* src_argb, int src_stride_argb, + int src_width, int src_height, + uint8* dst_argb, int dst_stride_argb, + int dst_width, int dst_height, + enum FilterMode filtering); + +// Clipped scale takes destination rectangle coordinates for clip values. +LIBYUV_API +int ARGBScaleClip(const uint8* src_argb, int src_stride_argb, + int src_width, int src_height, + uint8* dst_argb, int dst_stride_argb, + int dst_width, int dst_height, + int clip_x, int clip_y, int clip_width, int clip_height, + enum FilterMode filtering); + +// Scale with YUV conversion to ARGB and clipping. +LIBYUV_API +int YUVToARGBScaleClip(const uint8* src_y, int src_stride_y, + const uint8* src_u, int src_stride_u, + const uint8* src_v, int src_stride_v, + uint32 src_fourcc, + int src_width, int src_height, + uint8* dst_argb, int dst_stride_argb, + uint32 dst_fourcc, + int dst_width, int dst_height, + int clip_x, int clip_y, int clip_width, int clip_height, + enum FilterMode filtering); + +#ifdef __cplusplus +} // extern "C" +} // namespace libyuv +#endif + +#endif // INCLUDE_LIBYUV_SCALE_ARGB_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/scale_row.h b/phonelibs/libyuv/x64/include/libyuv/scale_row.h new file mode 100644 index 00000000000000..791fbf7d0538fb --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/scale_row.h @@ -0,0 +1,503 @@ +/* + * 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_ +#define INCLUDE_LIBYUV_SCALE_ROW_H_ + +#include "libyuv/basic_types.h" +#include "libyuv/scale.h" + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +#if defined(__pnacl__) || defined(__CLR_VER) || \ + (defined(__i386__) && !defined(__SSE2__)) +#define LIBYUV_DISABLE_X86 +#endif +// MemorySanitizer does not support assembly code yet. http://crbug.com/344505 +#if defined(__has_feature) +#if __has_feature(memory_sanitizer) +#define LIBYUV_DISABLE_X86 +#endif +#endif + +// GCC >= 4.7.0 required for AVX2. +#if defined(__GNUC__) && (defined(__x86_64__) || defined(__i386__)) +#if (__GNUC__ > 4) || (__GNUC__ == 4 && (__GNUC_MINOR__ >= 7)) +#define GCC_HAS_AVX2 1 +#endif // GNUC >= 4.7 +#endif // __GNUC__ + +// clang >= 3.4.0 required for AVX2. +#if defined(__clang__) && (defined(__x86_64__) || defined(__i386__)) +#if (__clang_major__ > 3) || (__clang_major__ == 3 && (__clang_minor__ >= 4)) +#define CLANG_HAS_AVX2 1 +#endif // clang >= 3.4 +#endif // __clang__ + +// Visual C 2012 required for AVX2. +#if defined(_M_IX86) && !defined(__clang__) && \ + defined(_MSC_VER) && _MSC_VER >= 1700 +#define VISUALC_HAS_AVX2 1 +#endif // VisualStudio >= 2012 + +// The following are available on all x86 platforms: +#if !defined(LIBYUV_DISABLE_X86) && \ + (defined(_M_IX86) || defined(__x86_64__) || defined(__i386__)) +#define HAS_FIXEDDIV1_X86 +#define HAS_FIXEDDIV_X86 +#define HAS_SCALEARGBCOLS_SSE2 +#define HAS_SCALEARGBCOLSUP2_SSE2 +#define HAS_SCALEARGBFILTERCOLS_SSSE3 +#define HAS_SCALEARGBROWDOWN2_SSE2 +#define HAS_SCALEARGBROWDOWNEVEN_SSE2 +#define HAS_SCALECOLSUP2_SSE2 +#define HAS_SCALEFILTERCOLS_SSSE3 +#define HAS_SCALEROWDOWN2_SSSE3 +#define HAS_SCALEROWDOWN34_SSSE3 +#define HAS_SCALEROWDOWN38_SSSE3 +#define HAS_SCALEROWDOWN4_SSSE3 +#define HAS_SCALEADDROW_SSE2 +#endif + +// The following are available on all x86 platforms, but +// require VS2012, clang 3.4 or gcc 4.7. +// The code supports NaCL but requires a new compiler and validator. +#if !defined(LIBYUV_DISABLE_X86) && (defined(VISUALC_HAS_AVX2) || \ + defined(CLANG_HAS_AVX2) || defined(GCC_HAS_AVX2)) +#define HAS_SCALEADDROW_AVX2 +#define HAS_SCALEROWDOWN2_AVX2 +#define HAS_SCALEROWDOWN4_AVX2 +#endif + +// The following are available on Neon platforms: +#if !defined(LIBYUV_DISABLE_NEON) && !defined(__native_client__) && \ + (defined(__ARM_NEON__) || defined(LIBYUV_NEON) || defined(__aarch64__)) +#define HAS_SCALEARGBCOLS_NEON +#define HAS_SCALEARGBROWDOWN2_NEON +#define HAS_SCALEARGBROWDOWNEVEN_NEON +#define HAS_SCALEFILTERCOLS_NEON +#define HAS_SCALEROWDOWN2_NEON +#define HAS_SCALEROWDOWN34_NEON +#define HAS_SCALEROWDOWN38_NEON +#define HAS_SCALEROWDOWN4_NEON +#define HAS_SCALEARGBFILTERCOLS_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_DSPR2 +#define HAS_SCALEROWDOWN4_DSPR2 +#define HAS_SCALEROWDOWN34_DSPR2 +#define HAS_SCALEROWDOWN38_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_Odd_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 ScaleAddRow_C(const uint8* src_ptr, uint16* dst_ptr, int src_width); +void ScaleAddRow_16_C(const uint16* src_ptr, uint32* dst_ptr, int src_width); +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); + +// Specialized scalers for x86. +void ScaleRowDown2_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown2Linear_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown2Box_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown2_AVX2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown2Linear_AVX2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown2Box_AVX2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown4_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown4Box_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown4_AVX2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown4Box_AVX2(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 ScaleRowDown2_Any_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown2Linear_Any_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown2Box_Any_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown2Box_Odd_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown2_Any_AVX2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown2Linear_Any_AVX2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown2Box_Any_AVX2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown2Box_Odd_AVX2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown4_Any_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown4Box_Any_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown4_Any_AVX2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown4Box_Any_AVX2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); + +void ScaleRowDown34_Any_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown34_1_Box_Any_SSSE3(const uint8* src_ptr, + ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown34_0_Box_Any_SSSE3(const uint8* src_ptr, + ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown38_Any_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown38_3_Box_Any_SSSE3(const uint8* src_ptr, + ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown38_2_Box_Any_SSSE3(const uint8* src_ptr, + ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); + +void ScaleAddRow_SSE2(const uint8* src_ptr, uint16* dst_ptr, int src_width); +void ScaleAddRow_AVX2(const uint8* src_ptr, uint16* dst_ptr, int src_width); +void ScaleAddRow_Any_SSE2(const uint8* src_ptr, uint16* dst_ptr, int src_width); +void ScaleAddRow_Any_AVX2(const uint8* src_ptr, uint16* dst_ptr, int src_width); + +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); + + +// ARGB Column functions +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); +void ScaleARGBFilterCols_NEON(uint8* dst_argb, const uint8* src_argb, + int dst_width, int x, int dx); +void ScaleARGBCols_NEON(uint8* dst_argb, const uint8* src_argb, + int dst_width, int x, int dx); +void ScaleARGBFilterCols_Any_NEON(uint8* dst_argb, const uint8* src_argb, + int dst_width, int x, int dx); +void ScaleARGBCols_Any_NEON(uint8* dst_argb, const uint8* src_argb, + int dst_width, int x, int dx); + +// ARGB Row functions +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 ScaleARGBRowDown2_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, int dst_width); +void ScaleARGBRowDown2Linear_NEON(const uint8* src_argb, ptrdiff_t src_stride, + uint8* dst_argb, int dst_width); +void ScaleARGBRowDown2Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, int dst_width); +void ScaleARGBRowDown2_Any_SSE2(const uint8* src_argb, ptrdiff_t src_stride, + uint8* dst_argb, int dst_width); +void ScaleARGBRowDown2Linear_Any_SSE2(const uint8* src_argb, + ptrdiff_t src_stride, + uint8* dst_argb, int dst_width); +void ScaleARGBRowDown2Box_Any_SSE2(const uint8* src_argb, ptrdiff_t src_stride, + uint8* dst_argb, int dst_width); +void ScaleARGBRowDown2_Any_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, int dst_width); +void ScaleARGBRowDown2Linear_Any_NEON(const uint8* src_argb, + ptrdiff_t src_stride, + uint8* dst_argb, int dst_width); +void ScaleARGBRowDown2Box_Any_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, 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 ScaleARGBRowDownEven_NEON(const uint8* src_argb, ptrdiff_t src_stride, + int src_stepx, + uint8* dst_argb, int dst_width); +void ScaleARGBRowDownEvenBox_NEON(const uint8* src_argb, ptrdiff_t src_stride, + int src_stepx, + uint8* dst_argb, int dst_width); +void ScaleARGBRowDownEven_Any_SSE2(const uint8* src_argb, ptrdiff_t src_stride, + int src_stepx, + uint8* dst_argb, int dst_width); +void ScaleARGBRowDownEvenBox_Any_SSE2(const uint8* src_argb, + ptrdiff_t src_stride, + int src_stepx, + uint8* dst_argb, int dst_width); +void ScaleARGBRowDownEven_Any_NEON(const uint8* src_argb, ptrdiff_t src_stride, + int src_stepx, + uint8* dst_argb, int dst_width); +void ScaleARGBRowDownEvenBox_Any_NEON(const uint8* src_argb, + ptrdiff_t src_stride, + int src_stepx, + uint8* dst_argb, 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 ScaleRowDown2Linear_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_Any_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, int dst_width); +void ScaleRowDown2Linear_Any_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, int dst_width); +void ScaleRowDown2Box_Any_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, int dst_width); +void ScaleRowDown2Box_Odd_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, int dst_width); +void ScaleRowDown4_Any_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown4Box_Any_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown34_Any_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown34_0_Box_Any_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown34_1_Box_Any_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +// 32 -> 12 +void ScaleRowDown38_Any_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +// 32x3 -> 12x1 +void ScaleRowDown38_3_Box_Any_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +// 32x2 -> 12x1 +void ScaleRowDown38_2_Box_Any_NEON(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); + +void ScaleAddRow_NEON(const uint8* src_ptr, uint16* dst_ptr, int src_width); +void ScaleAddRow_Any_NEON(const uint8* src_ptr, uint16* dst_ptr, int src_width); + +void ScaleFilterCols_NEON(uint8* dst_ptr, const uint8* src_ptr, + int dst_width, int x, int dx); + +void ScaleFilterCols_Any_NEON(uint8* dst_ptr, const uint8* src_ptr, + int dst_width, int x, int dx); + +void ScaleRowDown2_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, int dst_width); +void ScaleRowDown2Box_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, int dst_width); +void ScaleRowDown4_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, int dst_width); +void ScaleRowDown4Box_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, int dst_width); +void ScaleRowDown34_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, int dst_width); +void ScaleRowDown34_0_Box_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* d, int dst_width); +void ScaleRowDown34_1_Box_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* d, int dst_width); +void ScaleRowDown38_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst, int dst_width); +void ScaleRowDown38_2_Box_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride, + uint8* dst_ptr, int dst_width); +void ScaleRowDown38_3_Box_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_ diff --git a/phonelibs/libyuv/x64/include/libyuv/version.h b/phonelibs/libyuv/x64/include/libyuv/version.h new file mode 100644 index 00000000000000..3a8f6337ca22cd --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/version.h @@ -0,0 +1,16 @@ +/* + * Copyright 2012 The LibYuv Project Authors. All rights reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef INCLUDE_LIBYUV_VERSION_H_ +#define INCLUDE_LIBYUV_VERSION_H_ + +#define LIBYUV_VERSION 1622 + +#endif // INCLUDE_LIBYUV_VERSION_H_ diff --git a/phonelibs/libyuv/x64/include/libyuv/video_common.h b/phonelibs/libyuv/x64/include/libyuv/video_common.h new file mode 100644 index 00000000000000..cb425426a256e0 --- /dev/null +++ b/phonelibs/libyuv/x64/include/libyuv/video_common.h @@ -0,0 +1,184 @@ +/* + * 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. + */ + +// Common definitions for video, including fourcc and VideoFormat. + +#ifndef INCLUDE_LIBYUV_VIDEO_COMMON_H_ +#define INCLUDE_LIBYUV_VIDEO_COMMON_H_ + +#include "libyuv/basic_types.h" + +#ifdef __cplusplus +namespace libyuv { +extern "C" { +#endif + +////////////////////////////////////////////////////////////////////////////// +// Definition of FourCC codes +////////////////////////////////////////////////////////////////////////////// + +// Convert four characters to a FourCC code. +// Needs to be a macro otherwise the OS X compiler complains when the kFormat* +// constants are used in a switch. +#ifdef __cplusplus +#define FOURCC(a, b, c, d) ( \ + (static_cast(a)) | (static_cast(b) << 8) | \ + (static_cast(c) << 16) | (static_cast(d) << 24)) +#else +#define FOURCC(a, b, c, d) ( \ + ((uint32)(a)) | ((uint32)(b) << 8) | /* NOLINT */ \ + ((uint32)(c) << 16) | ((uint32)(d) << 24)) /* NOLINT */ +#endif + +// Some pages discussing FourCC codes: +// http://www.fourcc.org/yuv.php +// http://v4l2spec.bytesex.org/spec/book1.htm +// http://developer.apple.com/quicktime/icefloe/dispatch020.html +// http://msdn.microsoft.com/library/windows/desktop/dd206750.aspx#nv12 +// http://people.xiph.org/~xiphmont/containers/nut/nut4cc.txt + +// FourCC codes grouped according to implementation efficiency. +// Primary formats should convert in 1 efficient step. +// Secondary formats are converted in 2 steps. +// Auxilliary formats call primary converters. +enum FourCC { + // 9 Primary YUV formats: 5 planar, 2 biplanar, 2 packed. + FOURCC_I420 = FOURCC('I', '4', '2', '0'), + FOURCC_I422 = FOURCC('I', '4', '2', '2'), + FOURCC_I444 = FOURCC('I', '4', '4', '4'), + FOURCC_I411 = FOURCC('I', '4', '1', '1'), + FOURCC_I400 = FOURCC('I', '4', '0', '0'), + FOURCC_NV21 = FOURCC('N', 'V', '2', '1'), + FOURCC_NV12 = FOURCC('N', 'V', '1', '2'), + FOURCC_YUY2 = FOURCC('Y', 'U', 'Y', '2'), + FOURCC_UYVY = FOURCC('U', 'Y', 'V', 'Y'), + + // 2 Secondary YUV formats: row biplanar. + FOURCC_M420 = FOURCC('M', '4', '2', '0'), + FOURCC_Q420 = FOURCC('Q', '4', '2', '0'), // deprecated. + + // 9 Primary RGB formats: 4 32 bpp, 2 24 bpp, 3 16 bpp. + FOURCC_ARGB = FOURCC('A', 'R', 'G', 'B'), + FOURCC_BGRA = FOURCC('B', 'G', 'R', 'A'), + FOURCC_ABGR = FOURCC('A', 'B', 'G', 'R'), + FOURCC_24BG = FOURCC('2', '4', 'B', 'G'), + FOURCC_RAW = FOURCC('r', 'a', 'w', ' '), + FOURCC_RGBA = FOURCC('R', 'G', 'B', 'A'), + FOURCC_RGBP = FOURCC('R', 'G', 'B', 'P'), // rgb565 LE. + FOURCC_RGBO = FOURCC('R', 'G', 'B', 'O'), // argb1555 LE. + FOURCC_R444 = FOURCC('R', '4', '4', '4'), // argb4444 LE. + + // 4 Secondary RGB formats: 4 Bayer Patterns. deprecated. + FOURCC_RGGB = FOURCC('R', 'G', 'G', 'B'), + FOURCC_BGGR = FOURCC('B', 'G', 'G', 'R'), + FOURCC_GRBG = FOURCC('G', 'R', 'B', 'G'), + FOURCC_GBRG = FOURCC('G', 'B', 'R', 'G'), + + // 1 Primary Compressed YUV format. + FOURCC_MJPG = FOURCC('M', 'J', 'P', 'G'), + + // 5 Auxiliary YUV variations: 3 with U and V planes are swapped, 1 Alias. + FOURCC_YV12 = FOURCC('Y', 'V', '1', '2'), + FOURCC_YV16 = FOURCC('Y', 'V', '1', '6'), + FOURCC_YV24 = FOURCC('Y', 'V', '2', '4'), + FOURCC_YU12 = FOURCC('Y', 'U', '1', '2'), // Linux version of I420. + FOURCC_J420 = FOURCC('J', '4', '2', '0'), + FOURCC_J400 = FOURCC('J', '4', '0', '0'), // unofficial fourcc + FOURCC_H420 = FOURCC('H', '4', '2', '0'), // unofficial fourcc + + // 14 Auxiliary aliases. CanonicalFourCC() maps these to canonical fourcc. + FOURCC_IYUV = FOURCC('I', 'Y', 'U', 'V'), // Alias for I420. + FOURCC_YU16 = FOURCC('Y', 'U', '1', '6'), // Alias for I422. + FOURCC_YU24 = FOURCC('Y', 'U', '2', '4'), // Alias for I444. + FOURCC_YUYV = FOURCC('Y', 'U', 'Y', 'V'), // Alias for YUY2. + FOURCC_YUVS = FOURCC('y', 'u', 'v', 's'), // Alias for YUY2 on Mac. + FOURCC_HDYC = FOURCC('H', 'D', 'Y', 'C'), // Alias for UYVY. + FOURCC_2VUY = FOURCC('2', 'v', 'u', 'y'), // Alias for UYVY on Mac. + FOURCC_JPEG = FOURCC('J', 'P', 'E', 'G'), // Alias for MJPG. + FOURCC_DMB1 = FOURCC('d', 'm', 'b', '1'), // Alias for MJPG on Mac. + FOURCC_BA81 = FOURCC('B', 'A', '8', '1'), // Alias for BGGR. + FOURCC_RGB3 = FOURCC('R', 'G', 'B', '3'), // Alias for RAW. + FOURCC_BGR3 = FOURCC('B', 'G', 'R', '3'), // Alias for 24BG. + FOURCC_CM32 = FOURCC(0, 0, 0, 32), // Alias for BGRA kCMPixelFormat_32ARGB + FOURCC_CM24 = FOURCC(0, 0, 0, 24), // Alias for RAW kCMPixelFormat_24RGB + FOURCC_L555 = FOURCC('L', '5', '5', '5'), // Alias for RGBO. + FOURCC_L565 = FOURCC('L', '5', '6', '5'), // Alias for RGBP. + FOURCC_5551 = FOURCC('5', '5', '5', '1'), // Alias for RGBO. + + // 1 Auxiliary compressed YUV format set aside for capturer. + FOURCC_H264 = FOURCC('H', '2', '6', '4'), + + // Match any fourcc. + FOURCC_ANY = -1, +}; + +enum FourCCBpp { + // Canonical fourcc codes used in our code. + FOURCC_BPP_I420 = 12, + FOURCC_BPP_I422 = 16, + FOURCC_BPP_I444 = 24, + FOURCC_BPP_I411 = 12, + FOURCC_BPP_I400 = 8, + FOURCC_BPP_NV21 = 12, + FOURCC_BPP_NV12 = 12, + FOURCC_BPP_YUY2 = 16, + FOURCC_BPP_UYVY = 16, + FOURCC_BPP_M420 = 12, + FOURCC_BPP_Q420 = 12, + FOURCC_BPP_ARGB = 32, + FOURCC_BPP_BGRA = 32, + FOURCC_BPP_ABGR = 32, + FOURCC_BPP_RGBA = 32, + FOURCC_BPP_24BG = 24, + FOURCC_BPP_RAW = 24, + FOURCC_BPP_RGBP = 16, + FOURCC_BPP_RGBO = 16, + FOURCC_BPP_R444 = 16, + FOURCC_BPP_RGGB = 8, + FOURCC_BPP_BGGR = 8, + FOURCC_BPP_GRBG = 8, + FOURCC_BPP_GBRG = 8, + FOURCC_BPP_YV12 = 12, + FOURCC_BPP_YV16 = 16, + FOURCC_BPP_YV24 = 24, + FOURCC_BPP_YU12 = 12, + FOURCC_BPP_J420 = 12, + FOURCC_BPP_J400 = 8, + FOURCC_BPP_H420 = 12, + FOURCC_BPP_MJPG = 0, // 0 means unknown. + FOURCC_BPP_H264 = 0, + FOURCC_BPP_IYUV = 12, + FOURCC_BPP_YU16 = 16, + FOURCC_BPP_YU24 = 24, + FOURCC_BPP_YUYV = 16, + FOURCC_BPP_YUVS = 16, + FOURCC_BPP_HDYC = 16, + FOURCC_BPP_2VUY = 16, + FOURCC_BPP_JPEG = 1, + FOURCC_BPP_DMB1 = 1, + FOURCC_BPP_BA81 = 8, + FOURCC_BPP_RGB3 = 24, + FOURCC_BPP_BGR3 = 24, + FOURCC_BPP_CM32 = 32, + FOURCC_BPP_CM24 = 24, + + // Match any fourcc. + FOURCC_BPP_ANY = 0, // 0 means unknown. +}; + +// Converts fourcc aliases into canonical ones. +LIBYUV_API uint32 CanonicalFourCC(uint32 fourcc); + +#ifdef __cplusplus +} // extern "C" +} // namespace libyuv +#endif + +#endif // INCLUDE_LIBYUV_VIDEO_COMMON_H_ diff --git a/phonelibs/openblas/libopenblas.so b/phonelibs/openblas/libopenblas.so deleted file mode 120000 index 7a792bc9071885..00000000000000 --- a/phonelibs/openblas/libopenblas.so +++ /dev/null @@ -1 +0,0 @@ -libopenblas_armv8p-r0.2.19.so \ No newline at end of file diff --git a/phonelibs/openblas/libopenblas.so b/phonelibs/openblas/libopenblas.so new file mode 100755 index 00000000000000..ace58c8a13122f Binary files /dev/null and b/phonelibs/openblas/libopenblas.so differ diff --git a/phonelibs/zmq/aarch64-linux/lib/libczmq.so b/phonelibs/zmq/aarch64-linux/lib/libczmq.so deleted file mode 120000 index db9aa60f9b5868..00000000000000 --- a/phonelibs/zmq/aarch64-linux/lib/libczmq.so +++ /dev/null @@ -1 +0,0 @@ -libczmq.so.4.0.2 \ No newline at end of file diff --git a/phonelibs/zmq/aarch64-linux/lib/libczmq.so b/phonelibs/zmq/aarch64-linux/lib/libczmq.so new file mode 100755 index 00000000000000..bb9c2d52ebdaea Binary files /dev/null and b/phonelibs/zmq/aarch64-linux/lib/libczmq.so differ diff --git a/phonelibs/zmq/aarch64-linux/lib/libczmq.so.4 b/phonelibs/zmq/aarch64-linux/lib/libczmq.so.4 deleted file mode 120000 index db9aa60f9b5868..00000000000000 --- a/phonelibs/zmq/aarch64-linux/lib/libczmq.so.4 +++ /dev/null @@ -1 +0,0 @@ -libczmq.so.4.0.2 \ No newline at end of file diff --git a/phonelibs/zmq/aarch64-linux/lib/libczmq.so.4 b/phonelibs/zmq/aarch64-linux/lib/libczmq.so.4 new file mode 100755 index 00000000000000..bb9c2d52ebdaea Binary files /dev/null and b/phonelibs/zmq/aarch64-linux/lib/libczmq.so.4 differ diff --git a/phonelibs/zmq/aarch64-linux/lib/libzmq.so b/phonelibs/zmq/aarch64-linux/lib/libzmq.so deleted file mode 120000 index ef44cafc6ac861..00000000000000 --- a/phonelibs/zmq/aarch64-linux/lib/libzmq.so +++ /dev/null @@ -1 +0,0 @@ -libzmq.so.5.1.2 \ No newline at end of file diff --git a/phonelibs/zmq/aarch64-linux/lib/libzmq.so b/phonelibs/zmq/aarch64-linux/lib/libzmq.so new file mode 100755 index 00000000000000..e96750a8ef4365 Binary files /dev/null and b/phonelibs/zmq/aarch64-linux/lib/libzmq.so differ diff --git a/phonelibs/zmq/aarch64-linux/lib/libzmq.so.5 b/phonelibs/zmq/aarch64-linux/lib/libzmq.so.5 deleted file mode 120000 index ef44cafc6ac861..00000000000000 --- a/phonelibs/zmq/aarch64-linux/lib/libzmq.so.5 +++ /dev/null @@ -1 +0,0 @@ -libzmq.so.5.1.2 \ No newline at end of file diff --git a/phonelibs/zmq/aarch64-linux/lib/libzmq.so.5 b/phonelibs/zmq/aarch64-linux/lib/libzmq.so.5 new file mode 100755 index 00000000000000..e96750a8ef4365 Binary files /dev/null and b/phonelibs/zmq/aarch64-linux/lib/libzmq.so.5 differ diff --git a/selfdrive/controls/lib/long_mpc.py b/selfdrive/controls/lib/long_mpc.py index 19fb706debb2d7..7bb4c4cbb03e01 100644 --- a/selfdrive/controls/lib/long_mpc.py +++ b/selfdrive/controls/lib/long_mpc.py @@ -90,7 +90,14 @@ def update(self, CS, lead, v_cruise_setpoint): # Calculate mpc t = sec_since_boot() - n_its = self.libmpc.run_mpc(self.cur_state, self.mpc_solution, self.a_lead_tau, a_lead) + + # Openpilot Buttons -- https://github.com/rhinodavid/OpenpilotButtons + # hardcode follow time for now + follow_time = 1.8 + distance_cost = 0.1 + follow_time_value = scale_time_gap(v_ego, follow_time) + + n_its = self.libmpc.run_mpc(self.cur_state, self.mpc_solution, self.a_lead_tau, a_lead, follow_time_value, distance_cost) duration = int((sec_since_boot() - t) * 1e9) self.send_mpc_solution(n_its, duration) diff --git a/selfdrive/controls/lib/longitudinal_mpc/Makefile b/selfdrive/controls/lib/longitudinal_mpc/Makefile index 17f78eada990df..4e5bfd3807d39c 100644 --- a/selfdrive/controls/lib/longitudinal_mpc/Makefile +++ b/selfdrive/controls/lib/longitudinal_mpc/Makefile @@ -13,12 +13,10 @@ QPOASES_FLAGS = -I$(PHONELIBS)/qpoases -I$(PHONELIBS)/qpoases/INCLUDE -I$(PHONEL ACADO_FLAGS = -I$(PHONELIBS)/acado/include -I$(PHONELIBS)/acado/include/acado -ifeq ($(UNAME_S),Darwin) - ACADO_LIBS := -lacado_toolkit_s -else ifeq ($(UNAME_M),aarch64) - ACADO_LIBS := -L $(PHONELIBS)/acado/aarch64/lib -l:libacado_toolkit.a -l:libacado_casadi.a -l:libacado_csparse.a +ifeq ($(UNAME_M),aarch64) + ACADO_LIBS := -L $(PHONELIBS)/acado/aarch64/lib -l:libacado_toolkit_s.so -l:libacado_casadi.a -l:libacado_csparse.a else - ACADO_LIBS := -L $(PHONELIBS)/acado/x64/lib -l:libacado_toolkit.a -l:libacado_casadi.a -l:libacado_csparse.a + ACADO_LIBS := -L $(PHONELIBS)/acado/x64/lib -l:libacado_toolkit_s.so -l:libacado_casadi.a -l:libacado_csparse.a endif OBJS = \ @@ -80,7 +78,7 @@ generator: generator.cpp .PHONY: generate generate: generator - ./generator + LD_LIBRARY_PATH=../../../../phonelibs/acado/x64/lib ./generator .PHONY: clean clean: diff --git a/selfdrive/controls/lib/longitudinal_mpc/generator.cpp b/selfdrive/controls/lib/longitudinal_mpc/generator.cpp index 1f0e28d0057081..7b92be8bc394c3 100644 --- a/selfdrive/controls/lib/longitudinal_mpc/generator.cpp +++ b/selfdrive/controls/lib/longitudinal_mpc/generator.cpp @@ -5,10 +5,9 @@ const int controlHorizon = 50; using namespace std; #define G 9.81 -#define TR 1.8 -#define RW(v_ego, v_l) (v_ego * TR - (v_l - v_ego) * TR + v_ego*v_ego/(2*G) - v_l*v_l / (2*G)) -#define NORM_RW_ERROR(v_ego, v_l, p) ((RW(v_ego, v_l) + 4.0 - p)/(sqrt(v_ego + 0.5) + 0.1)) +#define RW(v_ego, v_l, time_gap) (v_ego * time_gap - (v_l - v_ego) * time_gap + v_ego*v_ego/(2*G) - v_l*v_l / (2*G)) +#define NORM_RW_ERROR(v_ego, v_l, p, time_gap) ((RW(v_ego, v_l, time_gap) + 4.0 - p)/(sqrt(v_ego + 0.5) + 0.1)) int main( ) { @@ -22,7 +21,12 @@ int main( ) Control j_ego; - auto desired = 4.0 + RW(v_ego, v_l); + // follow distance expressed as a stopping distance in seconds + // see https://github.com/rhinodavid/CommaButtons + // see https://github.com/acado/acado/issues/54 for a discussion of `OnlineData` + OnlineData time_gap; + + auto desired = 4.0 + RW(v_ego, v_l, time_gap); auto d_l = x_l - x_ego; // Equations of motion @@ -32,7 +36,7 @@ int main( ) // Running cost Function h; - h << exp(0.3 * NORM_RW_ERROR(v_ego, v_l, d_l)); + h << exp(0.3 * NORM_RW_ERROR(v_ego, v_l, d_l, time_gap)); h << (d_l - desired) / (0.05 * v_ego + 0.5); h << a_ego * (0.1 * v_ego + 1.0); h << j_ego * (0.1 * v_ego + 1.0); @@ -42,7 +46,7 @@ int main( ) // Terminal cost Function hN; - hN << exp(0.3 * NORM_RW_ERROR(v_ego, v_l, d_l)); + hN << exp(0.3 * NORM_RW_ERROR(v_ego, v_l, d_l, time_gap)); hN << (d_l - desired) / (0.05 * v_ego + 0.5); hN << a_ego * (0.1 * v_ego + 1.0); @@ -70,7 +74,7 @@ int main( ) ocp.minimizeLSQEndTerm(QN, hN); ocp.subjectTo( 0.0 <= v_ego); - ocp.setNOD(2); + ocp.setNOD(3); OCPexport mpc(ocp); mpc.set( HESSIAN_APPROXIMATION, GAUSS_NEWTON ); diff --git a/selfdrive/controls/lib/longitudinal_mpc/lib_mpc_export/acado_common.h b/selfdrive/controls/lib/longitudinal_mpc/lib_mpc_export/acado_common.h index bf64887f19c092..e47b2eed40dc70 100644 --- a/selfdrive/controls/lib/longitudinal_mpc/lib_mpc_export/acado_common.h +++ b/selfdrive/controls/lib/longitudinal_mpc/lib_mpc_export/acado_common.h @@ -64,7 +64,7 @@ extern "C" /** Number of control/estimation intervals. */ #define ACADO_N 20 /** Number of online data values. */ -#define ACADO_NOD 2 +#define ACADO_NOD 3 /** Number of path constraints. */ #define ACADO_NPAC 0 /** Number of control variables. */ @@ -83,6 +83,8 @@ extern "C" #define ACADO_QP_NV 23 /** Number of Runge-Kutta stages per integration step. */ #define ACADO_RK_NSTAGES 4 +/** Single versus double precision data type representation. */ +#define ACADO_SINGLE_PRECISION 0 /** Providing interface for arrival cost. */ #define ACADO_USE_ARRIVAL_COST 0 /** Indicator for usage of non-hard-coded linear terms in the objective. */ @@ -114,11 +116,11 @@ real_t x[ 63 ]; */ real_t u[ 20 ]; -/** Matrix of size: 21 x 2 (row major format) +/** Matrix of size: 21 x 3 (row major format) * * Matrix containing 21 online data vectors. */ -real_t od[ 42 ]; +real_t od[ 63 ]; /** Column vector of size: 80 * @@ -157,14 +159,14 @@ typedef struct ACADOworkspace_ { real_t rk_ttt; -/** Row vector of size: 18 */ -real_t rk_xxx[ 18 ]; +/** Row vector of size: 19 */ +real_t rk_xxx[ 19 ]; /** Matrix of size: 4 x 15 (row major format) */ real_t rk_kkk[ 60 ]; -/** Row vector of size: 18 */ -real_t state[ 18 ]; +/** Row vector of size: 19 */ +real_t state[ 19 ]; /** Column vector of size: 60 */ real_t d[ 60 ]; @@ -184,8 +186,8 @@ real_t evGu[ 60 ]; /** Column vector of size: 15 */ real_t objAuxVar[ 15 ]; -/** Row vector of size: 6 */ -real_t objValueIn[ 6 ]; +/** Row vector of size: 7 */ +real_t objValueIn[ 7 ]; /** Row vector of size: 20 */ real_t objValueOut[ 20 ]; diff --git a/selfdrive/controls/lib/longitudinal_mpc/lib_mpc_export/acado_integrator.c b/selfdrive/controls/lib/longitudinal_mpc/lib_mpc_export/acado_integrator.c index a0bac9681c5293..55dbe71fbfaef3 100644 --- a/selfdrive/controls/lib/longitudinal_mpc/lib_mpc_export/acado_integrator.c +++ b/selfdrive/controls/lib/longitudinal_mpc/lib_mpc_export/acado_integrator.c @@ -67,6 +67,7 @@ rk_eta[14] = 0.0000000000000000e+00; acadoWorkspace.rk_xxx[15] = rk_eta[15]; acadoWorkspace.rk_xxx[16] = rk_eta[16]; acadoWorkspace.rk_xxx[17] = rk_eta[17]; +acadoWorkspace.rk_xxx[18] = rk_eta[18]; for (run1 = 0; run1 < 1; ++run1) { diff --git a/selfdrive/controls/lib/longitudinal_mpc/lib_mpc_export/acado_solver.c b/selfdrive/controls/lib/longitudinal_mpc/lib_mpc_export/acado_solver.c index e4bf392f98fb30..e3d382f4cdf945 100644 --- a/selfdrive/controls/lib/longitudinal_mpc/lib_mpc_export/acado_solver.c +++ b/selfdrive/controls/lib/longitudinal_mpc/lib_mpc_export/acado_solver.c @@ -42,8 +42,9 @@ acadoWorkspace.state[1] = acadoVariables.x[lRun1 * 3 + 1]; acadoWorkspace.state[2] = acadoVariables.x[lRun1 * 3 + 2]; acadoWorkspace.state[15] = acadoVariables.u[lRun1]; -acadoWorkspace.state[16] = acadoVariables.od[lRun1 * 2]; -acadoWorkspace.state[17] = acadoVariables.od[lRun1 * 2 + 1]; +acadoWorkspace.state[16] = acadoVariables.od[lRun1 * 3]; +acadoWorkspace.state[17] = acadoVariables.od[lRun1 * 3 + 1]; +acadoWorkspace.state[18] = acadoVariables.od[lRun1 * 3 + 2]; ret = acado_integrate(acadoWorkspace.state, 1, lRun1); @@ -78,15 +79,15 @@ real_t* a = acadoWorkspace.objAuxVar; /* Compute intermediate quantities: */ a[0] = (sqrt((xd[1]+(real_t)(5.0000000000000000e-01)))); -a[1] = (exp(((real_t)(2.9999999999999999e-01)*(((((((xd[1]*(real_t)(1.8000000000000000e+00))-((od[1]-xd[1])*(real_t)(1.8000000000000000e+00)))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))+(real_t)(4.0000000000000000e+00))-(od[0]-xd[0]))/(a[0]+(real_t)(1.0000000000000001e-01)))))); +a[1] = (exp(((real_t)(2.9999999999999999e-01)*(((((((xd[1]*od[2])-((od[1]-xd[1])*od[2]))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))+(real_t)(4.0000000000000000e+00))-(od[0]-xd[0]))/(a[0]+(real_t)(1.0000000000000001e-01)))))); a[2] = ((real_t)(1.0000000000000000e+00)/(a[0]+(real_t)(1.0000000000000001e-01))); -a[3] = (exp(((real_t)(2.9999999999999999e-01)*(((((((xd[1]*(real_t)(1.8000000000000000e+00))-((od[1]-xd[1])*(real_t)(1.8000000000000000e+00)))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))+(real_t)(4.0000000000000000e+00))-(od[0]-xd[0]))/(a[0]+(real_t)(1.0000000000000001e-01)))))); +a[3] = (exp(((real_t)(2.9999999999999999e-01)*(((((((xd[1]*od[2])-((od[1]-xd[1])*od[2]))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))+(real_t)(4.0000000000000000e+00))-(od[0]-xd[0]))/(a[0]+(real_t)(1.0000000000000001e-01)))))); a[4] = (((real_t)(2.9999999999999999e-01)*(((real_t)(0.0000000000000000e+00)-((real_t)(0.0000000000000000e+00)-(real_t)(1.0000000000000000e+00)))*a[2]))*a[3]); a[5] = ((real_t)(1.0000000000000000e+00)/(real_t)(1.9620000000000001e+01)); a[6] = (1.0/sqrt((xd[1]+(real_t)(5.0000000000000000e-01)))); a[7] = (a[6]*(real_t)(5.0000000000000000e-01)); a[8] = (a[2]*a[2]); -a[9] = (((real_t)(2.9999999999999999e-01)*(((((real_t)(1.8000000000000000e+00)-((real_t)(-1.8000000000000000e+00)))+((xd[1]+xd[1])*a[5]))*a[2])-((((((((xd[1]*(real_t)(1.8000000000000000e+00))-((od[1]-xd[1])*(real_t)(1.8000000000000000e+00)))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))+(real_t)(4.0000000000000000e+00))-(od[0]-xd[0]))*a[7])*a[8])))*a[3]); +a[9] = (((real_t)(2.9999999999999999e-01)*((((od[2]-(((real_t)(0.0000000000000000e+00)-(real_t)(1.0000000000000000e+00))*od[2]))+((xd[1]+xd[1])*a[5]))*a[2])-((((((((xd[1]*od[2])-((od[1]-xd[1])*od[2]))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))+(real_t)(4.0000000000000000e+00))-(od[0]-xd[0]))*a[7])*a[8])))*a[3]); a[10] = (real_t)(0.0000000000000000e+00); a[11] = ((real_t)(1.0000000000000000e+00)/(((real_t)(5.0000000000000003e-02)*xd[1])+(real_t)(5.0000000000000000e-01))); a[12] = ((real_t)(1.0000000000000000e+00)/(real_t)(1.9620000000000001e+01)); @@ -95,14 +96,14 @@ a[14] = (real_t)(0.0000000000000000e+00); /* Compute outputs: */ out[0] = a[1]; -out[1] = (((od[0]-xd[0])-((real_t)(4.0000000000000000e+00)+((((xd[1]*(real_t)(1.8000000000000000e+00))-((od[1]-xd[1])*(real_t)(1.8000000000000000e+00)))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))))/(((real_t)(5.0000000000000003e-02)*xd[1])+(real_t)(5.0000000000000000e-01))); +out[1] = (((od[0]-xd[0])-((real_t)(4.0000000000000000e+00)+((((xd[1]*od[2])-((od[1]-xd[1])*od[2]))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))))/(((real_t)(5.0000000000000003e-02)*xd[1])+(real_t)(5.0000000000000000e-01))); out[2] = (xd[2]*(((real_t)(1.0000000000000001e-01)*xd[1])+(real_t)(1.0000000000000000e+00))); out[3] = (u[0]*(((real_t)(1.0000000000000001e-01)*xd[1])+(real_t)(1.0000000000000000e+00))); out[4] = a[4]; out[5] = a[9]; out[6] = a[10]; out[7] = (((real_t)(0.0000000000000000e+00)-(real_t)(1.0000000000000000e+00))*a[11]); -out[8] = ((((real_t)(0.0000000000000000e+00)-(((real_t)(1.8000000000000000e+00)-((real_t)(-1.8000000000000000e+00)))+((xd[1]+xd[1])*a[12])))*a[11])-((((od[0]-xd[0])-((real_t)(4.0000000000000000e+00)+((((xd[1]*(real_t)(1.8000000000000000e+00))-((od[1]-xd[1])*(real_t)(1.8000000000000000e+00)))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))))*(real_t)(5.0000000000000003e-02))*a[13])); +out[8] = ((((real_t)(0.0000000000000000e+00)-((od[2]-(((real_t)(0.0000000000000000e+00)-(real_t)(1.0000000000000000e+00))*od[2]))+((xd[1]+xd[1])*a[12])))*a[11])-((((od[0]-xd[0])-((real_t)(4.0000000000000000e+00)+((((xd[1]*od[2])-((od[1]-xd[1])*od[2]))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))))*(real_t)(5.0000000000000003e-02))*a[13])); out[9] = (real_t)(0.0000000000000000e+00); out[10] = (real_t)(0.0000000000000000e+00); out[11] = (xd[2]*(real_t)(1.0000000000000001e-01)); @@ -125,15 +126,15 @@ real_t* a = acadoWorkspace.objAuxVar; /* Compute intermediate quantities: */ a[0] = (sqrt((xd[1]+(real_t)(5.0000000000000000e-01)))); -a[1] = (exp(((real_t)(2.9999999999999999e-01)*(((((((xd[1]*(real_t)(1.8000000000000000e+00))-((od[1]-xd[1])*(real_t)(1.8000000000000000e+00)))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))+(real_t)(4.0000000000000000e+00))-(od[0]-xd[0]))/(a[0]+(real_t)(1.0000000000000001e-01)))))); +a[1] = (exp(((real_t)(2.9999999999999999e-01)*(((((((xd[1]*od[2])-((od[1]-xd[1])*od[2]))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))+(real_t)(4.0000000000000000e+00))-(od[0]-xd[0]))/(a[0]+(real_t)(1.0000000000000001e-01)))))); a[2] = ((real_t)(1.0000000000000000e+00)/(a[0]+(real_t)(1.0000000000000001e-01))); -a[3] = (exp(((real_t)(2.9999999999999999e-01)*(((((((xd[1]*(real_t)(1.8000000000000000e+00))-((od[1]-xd[1])*(real_t)(1.8000000000000000e+00)))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))+(real_t)(4.0000000000000000e+00))-(od[0]-xd[0]))/(a[0]+(real_t)(1.0000000000000001e-01)))))); +a[3] = (exp(((real_t)(2.9999999999999999e-01)*(((((((xd[1]*od[2])-((od[1]-xd[1])*od[2]))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))+(real_t)(4.0000000000000000e+00))-(od[0]-xd[0]))/(a[0]+(real_t)(1.0000000000000001e-01)))))); a[4] = (((real_t)(2.9999999999999999e-01)*(((real_t)(0.0000000000000000e+00)-((real_t)(0.0000000000000000e+00)-(real_t)(1.0000000000000000e+00)))*a[2]))*a[3]); a[5] = ((real_t)(1.0000000000000000e+00)/(real_t)(1.9620000000000001e+01)); a[6] = (1.0/sqrt((xd[1]+(real_t)(5.0000000000000000e-01)))); a[7] = (a[6]*(real_t)(5.0000000000000000e-01)); a[8] = (a[2]*a[2]); -a[9] = (((real_t)(2.9999999999999999e-01)*(((((real_t)(1.8000000000000000e+00)-((real_t)(-1.8000000000000000e+00)))+((xd[1]+xd[1])*a[5]))*a[2])-((((((((xd[1]*(real_t)(1.8000000000000000e+00))-((od[1]-xd[1])*(real_t)(1.8000000000000000e+00)))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))+(real_t)(4.0000000000000000e+00))-(od[0]-xd[0]))*a[7])*a[8])))*a[3]); +a[9] = (((real_t)(2.9999999999999999e-01)*((((od[2]-(((real_t)(0.0000000000000000e+00)-(real_t)(1.0000000000000000e+00))*od[2]))+((xd[1]+xd[1])*a[5]))*a[2])-((((((((xd[1]*od[2])-((od[1]-xd[1])*od[2]))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))+(real_t)(4.0000000000000000e+00))-(od[0]-xd[0]))*a[7])*a[8])))*a[3]); a[10] = (real_t)(0.0000000000000000e+00); a[11] = ((real_t)(1.0000000000000000e+00)/(((real_t)(5.0000000000000003e-02)*xd[1])+(real_t)(5.0000000000000000e-01))); a[12] = ((real_t)(1.0000000000000000e+00)/(real_t)(1.9620000000000001e+01)); @@ -141,13 +142,13 @@ a[13] = (a[11]*a[11]); /* Compute outputs: */ out[0] = a[1]; -out[1] = (((od[0]-xd[0])-((real_t)(4.0000000000000000e+00)+((((xd[1]*(real_t)(1.8000000000000000e+00))-((od[1]-xd[1])*(real_t)(1.8000000000000000e+00)))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))))/(((real_t)(5.0000000000000003e-02)*xd[1])+(real_t)(5.0000000000000000e-01))); +out[1] = (((od[0]-xd[0])-((real_t)(4.0000000000000000e+00)+((((xd[1]*od[2])-((od[1]-xd[1])*od[2]))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))))/(((real_t)(5.0000000000000003e-02)*xd[1])+(real_t)(5.0000000000000000e-01))); out[2] = (xd[2]*(((real_t)(1.0000000000000001e-01)*xd[1])+(real_t)(1.0000000000000000e+00))); out[3] = a[4]; out[4] = a[9]; out[5] = a[10]; out[6] = (((real_t)(0.0000000000000000e+00)-(real_t)(1.0000000000000000e+00))*a[11]); -out[7] = ((((real_t)(0.0000000000000000e+00)-(((real_t)(1.8000000000000000e+00)-((real_t)(-1.8000000000000000e+00)))+((xd[1]+xd[1])*a[12])))*a[11])-((((od[0]-xd[0])-((real_t)(4.0000000000000000e+00)+((((xd[1]*(real_t)(1.8000000000000000e+00))-((od[1]-xd[1])*(real_t)(1.8000000000000000e+00)))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))))*(real_t)(5.0000000000000003e-02))*a[13])); +out[7] = ((((real_t)(0.0000000000000000e+00)-((od[2]-(((real_t)(0.0000000000000000e+00)-(real_t)(1.0000000000000000e+00))*od[2]))+((xd[1]+xd[1])*a[12])))*a[11])-((((od[0]-xd[0])-((real_t)(4.0000000000000000e+00)+((((xd[1]*od[2])-((od[1]-xd[1])*od[2]))+((xd[1]*xd[1])/(real_t)(1.9620000000000001e+01)))-((od[1]*od[1])/(real_t)(1.9620000000000001e+01)))))*(real_t)(5.0000000000000003e-02))*a[13])); out[8] = (real_t)(0.0000000000000000e+00); out[9] = (real_t)(0.0000000000000000e+00); out[10] = (xd[2]*(real_t)(1.0000000000000001e-01)); @@ -219,8 +220,9 @@ acadoWorkspace.objValueIn[0] = acadoVariables.x[runObj * 3]; acadoWorkspace.objValueIn[1] = acadoVariables.x[runObj * 3 + 1]; acadoWorkspace.objValueIn[2] = acadoVariables.x[runObj * 3 + 2]; acadoWorkspace.objValueIn[3] = acadoVariables.u[runObj]; -acadoWorkspace.objValueIn[4] = acadoVariables.od[runObj * 2]; -acadoWorkspace.objValueIn[5] = acadoVariables.od[runObj * 2 + 1]; +acadoWorkspace.objValueIn[4] = acadoVariables.od[runObj * 3]; +acadoWorkspace.objValueIn[5] = acadoVariables.od[runObj * 3 + 1]; +acadoWorkspace.objValueIn[6] = acadoVariables.od[runObj * 3 + 2]; acado_evaluateLSQ( acadoWorkspace.objValueIn, acadoWorkspace.objValueOut ); acadoWorkspace.Dy[runObj * 4] = acadoWorkspace.objValueOut[0]; @@ -236,8 +238,9 @@ acado_setObjR1R2( &(acadoWorkspace.objValueOut[ 16 ]), &(acadoVariables.W[ runOb acadoWorkspace.objValueIn[0] = acadoVariables.x[60]; acadoWorkspace.objValueIn[1] = acadoVariables.x[61]; acadoWorkspace.objValueIn[2] = acadoVariables.x[62]; -acadoWorkspace.objValueIn[3] = acadoVariables.od[40]; -acadoWorkspace.objValueIn[4] = acadoVariables.od[41]; +acadoWorkspace.objValueIn[3] = acadoVariables.od[60]; +acadoWorkspace.objValueIn[4] = acadoVariables.od[61]; +acadoWorkspace.objValueIn[5] = acadoVariables.od[62]; acado_evaluateLSQEndTerm( acadoWorkspace.objValueIn, acadoWorkspace.objValueOut ); acadoWorkspace.DyN[0] = acadoWorkspace.objValueOut[0]; @@ -4636,8 +4639,9 @@ acadoWorkspace.state[0] = acadoVariables.x[index * 3]; acadoWorkspace.state[1] = acadoVariables.x[index * 3 + 1]; acadoWorkspace.state[2] = acadoVariables.x[index * 3 + 2]; acadoWorkspace.state[15] = acadoVariables.u[index]; -acadoWorkspace.state[16] = acadoVariables.od[index * 2]; -acadoWorkspace.state[17] = acadoVariables.od[index * 2 + 1]; +acadoWorkspace.state[16] = acadoVariables.od[index * 3]; +acadoWorkspace.state[17] = acadoVariables.od[index * 3 + 1]; +acadoWorkspace.state[18] = acadoVariables.od[index * 3 + 2]; acado_integrate(acadoWorkspace.state, index == 0, index); @@ -4676,8 +4680,9 @@ else { acadoWorkspace.state[15] = acadoVariables.u[19]; } -acadoWorkspace.state[16] = acadoVariables.od[40]; -acadoWorkspace.state[17] = acadoVariables.od[41]; +acadoWorkspace.state[16] = acadoVariables.od[60]; +acadoWorkspace.state[17] = acadoVariables.od[61]; +acadoWorkspace.state[18] = acadoVariables.od[62]; acado_integrate(acadoWorkspace.state, 1, 19); @@ -4746,8 +4751,9 @@ acadoWorkspace.objValueIn[0] = acadoVariables.x[lRun1 * 3]; acadoWorkspace.objValueIn[1] = acadoVariables.x[lRun1 * 3 + 1]; acadoWorkspace.objValueIn[2] = acadoVariables.x[lRun1 * 3 + 2]; acadoWorkspace.objValueIn[3] = acadoVariables.u[lRun1]; -acadoWorkspace.objValueIn[4] = acadoVariables.od[lRun1 * 2]; -acadoWorkspace.objValueIn[5] = acadoVariables.od[lRun1 * 2 + 1]; +acadoWorkspace.objValueIn[4] = acadoVariables.od[lRun1 * 3]; +acadoWorkspace.objValueIn[5] = acadoVariables.od[lRun1 * 3 + 1]; +acadoWorkspace.objValueIn[6] = acadoVariables.od[lRun1 * 3 + 2]; acado_evaluateLSQ( acadoWorkspace.objValueIn, acadoWorkspace.objValueOut ); acadoWorkspace.Dy[lRun1 * 4] = acadoWorkspace.objValueOut[0] - acadoVariables.y[lRun1 * 4]; @@ -4758,8 +4764,9 @@ acadoWorkspace.Dy[lRun1 * 4 + 3] = acadoWorkspace.objValueOut[3] - acadoVariable acadoWorkspace.objValueIn[0] = acadoVariables.x[60]; acadoWorkspace.objValueIn[1] = acadoVariables.x[61]; acadoWorkspace.objValueIn[2] = acadoVariables.x[62]; -acadoWorkspace.objValueIn[3] = acadoVariables.od[40]; -acadoWorkspace.objValueIn[4] = acadoVariables.od[41]; +acadoWorkspace.objValueIn[3] = acadoVariables.od[60]; +acadoWorkspace.objValueIn[4] = acadoVariables.od[61]; +acadoWorkspace.objValueIn[5] = acadoVariables.od[62]; acado_evaluateLSQEndTerm( acadoWorkspace.objValueIn, acadoWorkspace.objValueOut ); acadoWorkspace.DyN[0] = acadoWorkspace.objValueOut[0] - acadoVariables.yN[0]; acadoWorkspace.DyN[1] = acadoWorkspace.objValueOut[1] - acadoVariables.yN[1]; diff --git a/selfdrive/controls/lib/longitudinal_mpc/libmpc_py.py b/selfdrive/controls/lib/longitudinal_mpc/libmpc_py.py index 354a4a493eaa91..d71c21d9fe5334 100644 --- a/selfdrive/controls/lib/longitudinal_mpc/libmpc_py.py +++ b/selfdrive/controls/lib/longitudinal_mpc/libmpc_py.py @@ -37,10 +37,10 @@ def _get_libmpc(mpc_id): double cost; } log_t; - void init(double ttcCost, double distanceCost, double accelerationCost, double jerkCost); + void init(double ttcCost, double defaultDistanceCost, double accelerationCost, double jerkCost); void init_with_simulation(double v_ego, double x_l, double v_l, double a_l, double l); int run_mpc(state_t * x0, log_t * solution, - double l, double a_l_0); + double l, double a_l_0, double time_gap, double distance_cost); """) return (ffi, ffi.dlopen(libmpc_fn)) @@ -49,3 +49,4 @@ def _get_libmpc(mpc_id): def get_libmpc(mpc_id): return mpcs[mpc_id - 1] + \ No newline at end of file diff --git a/selfdrive/controls/lib/longitudinal_mpc/longitudinal_mpc.c b/selfdrive/controls/lib/longitudinal_mpc/longitudinal_mpc.c index e903ccc758c397..926cc278aebd86 100644 --- a/selfdrive/controls/lib/longitudinal_mpc/longitudinal_mpc.c +++ b/selfdrive/controls/lib/longitudinal_mpc/longitudinal_mpc.c @@ -14,6 +14,8 @@ #define N ACADO_N /* Number of intervals in the horizon. */ +const int STEP_MULTIPLIER = 3; + ACADOvariables acadoVariables; ACADOworkspace acadoWorkspace; @@ -34,10 +36,13 @@ typedef struct { double cost; } log_t; -void init(double ttcCost, double distanceCost, double accelerationCost, double jerkCost){ +void init(double ttcCost, double defaultDistanceCost, double accelerationCost, double jerkCost){ + // Comma Buttons https://github.com/rhinodavid/CommaButtons + // set the defaultFollowDistance cost here, but we'll add it as a parameter of the update + // function in order to change it as we change the time gap + acado_initializeSolver(); int i; - const int STEP_MULTIPLIER = 3; /* Initialize the states and controls. */ for (i = 0; i < NX * (N + 1); ++i) acadoVariables.x[ i ] = 0.0; @@ -57,12 +62,12 @@ void init(double ttcCost, double distanceCost, double accelerationCost, double j f = STEP_MULTIPLIER; } acadoVariables.W[16 * i + 0] = ttcCost * f; // exponential cost for time-to-collision (ttc) - acadoVariables.W[16 * i + 5] = distanceCost * f; // desired distance + acadoVariables.W[16 * i + 5] = defaultDistanceCost * f; // desired distance acadoVariables.W[16 * i + 10] = accelerationCost * f; // acceleration acadoVariables.W[16 * i + 15] = jerkCost * f; // jerk } acadoVariables.WN[0] = ttcCost * STEP_MULTIPLIER; // exponential cost for danger zone - acadoVariables.WN[4] = distanceCost * STEP_MULTIPLIER; // desired distance + acadoVariables.WN[4] = defaultDistanceCost * STEP_MULTIPLIER; // desired distance acadoVariables.WN[8] = accelerationCost * STEP_MULTIPLIER; // acceleration } @@ -111,8 +116,8 @@ void init_with_simulation(double v_ego, double x_l_0, double v_l_0, double a_l_0 for (i = 0; i < NYN; ++i) acadoVariables.yN[ i ] = 0.0; } -int run_mpc(state_t * x0, log_t * solution, double l, double a_l_0){ - // Calculate lead vehicle predictions +int run_mpc(state_t * x0, log_t * solution, double l, double a_l_0, double time_gap, double distance_cost){ + // Calculate lead vehicle predictions int i; double t = 0.; double dt = 0.2; @@ -130,6 +135,9 @@ int run_mpc(state_t * x0, log_t * solution, double l, double a_l_0){ acadoVariables.od[i*NOD] = x_l; acadoVariables.od[i*NOD+1] = v_l; + // For CommaButtons https://github.com/rhinodavid/CommaButtons + // Update the time gap + acadoVariables.od[i*NOD+2] = time_gap; solution->x_l[i] = x_l; solution->v_l[i] = v_l; @@ -146,6 +154,17 @@ int run_mpc(state_t * x0, log_t * solution, double l, double a_l_0){ t += dt; } + + // For CommaButtons https://github.com/rhinodavid/CommaButtons + // Update the distance_cost as we change the time gap + for (i = 0; i < N; i++) { + int f = 1; + if (i > 4) { + f = STEP_MULTIPLIER; + } + acadoVariables.W[16 * i + 5] = distance_cost * f; // desired distance + } + acadoVariables.WN[4] = distance_cost * STEP_MULTIPLIER; // desired distance acadoVariables.x[0] = acadoVariables.x0[0] = x0->x_ego; acadoVariables.x[1] = acadoVariables.x0[1] = x0->v_ego; diff --git a/selfdrive/controls/lib/planner.py b/selfdrive/controls/lib/planner.py index c26711f57e8b21..a0d51690add025 100755 --- a/selfdrive/controls/lib/planner.py +++ b/selfdrive/controls/lib/planner.py @@ -60,7 +60,208 @@ def limit_accel_in_turns(v_ego, angle_steers, a_target, CP): a_target[1] = min(a_target[1], a_x_allowed) return a_target +def scale_time_gap(v_ego, commanded_time_gap): + """ + Raise the time gap as we slow down. Use the full time gap at 30mph and increase + it to 1.8s at 25mph + Keyword arguments: + v_ego -- the vehicle speed in m/s + commanded_time_gap -- the time gap the set by the user in s + See Openpilot Buttons -- https://github.com/rhinodavid/OpenpilotButtons + """ + ms30 = 30. * CV.MPH_TO_MS + ms20 = 20. * CV.MPH_TO_MS + if v_ego > ms30: + return commanded_time_gap + if v_ego < ms20: + return 1.8 + return round(float(np.interp(v_ego, [ms20, ms30], [1.8, commanded_time_gap])), 2) + +<<<<<<< HEAD +======= +class FCWChecker(object): + def __init__(self): + self.reset_lead(0.0) + + def reset_lead(self, cur_time): + self.last_fcw_a = 0.0 + self.v_lead_max = 0.0 + self.lead_seen_t = cur_time + self.last_fcw_time = 0.0 + self.last_min_a = 0.0 + + self.counters = defaultdict(lambda: 0) + + @staticmethod + def calc_ttc(v_ego, a_ego, x_lead, v_lead, a_lead): + max_ttc = 5.0 + + v_rel = v_ego - v_lead + a_rel = a_ego - a_lead + + # assuming that closing gap ARel comes from lead vehicle decel, + # then limit ARel so that v_lead will get to zero in no sooner than t_decel. + # This helps underweighting ARel when v_lead is close to zero. + t_decel = 2. + a_rel = np.minimum(a_rel, v_lead / t_decel) + + # delta of the quadratic equation to solve for ttc + delta = v_rel**2 + 2 * x_lead * a_rel + + # assign an arbitrary high ttc value if there is no solution to ttc + if delta < 0.1 or (np.sqrt(delta) + v_rel < 0.1): + ttc = max_ttc + else: + ttc = np.minimum(2 * x_lead / (np.sqrt(delta) + v_rel), max_ttc) + return ttc + + def update(self, mpc_solution, cur_time, v_ego, a_ego, x_lead, v_lead, a_lead, y_lead, vlat_lead, fcw_lead, blinkers): + mpc_solution_a = list(mpc_solution[0].a_ego) + self.last_min_a = min(mpc_solution_a) + self.v_lead_max = max(self.v_lead_max, v_lead) + + if (fcw_lead > 0.99): + ttc = self.calc_ttc(v_ego, a_ego, x_lead, v_lead, a_lead) + self.counters['v_ego'] = self.counters['v_ego'] + 1 if v_ego > 5.0 else 0 + self.counters['ttc'] = self.counters['ttc'] + 1 if ttc < 2.5 else 0 + self.counters['v_lead_max'] = self.counters['v_lead_max'] + 1 if self.v_lead_max > 2.5 else 0 + self.counters['v_ego_lead'] = self.counters['v_ego_lead'] + 1 if v_ego > v_lead else 0 + self.counters['lead_seen'] = self.counters['lead_seen'] + 0.33 + self.counters['y_lead'] = self.counters['y_lead'] + 1 if abs(y_lead) < 1.0 else 0 + self.counters['vlat_lead'] = self.counters['vlat_lead'] + 1 if abs(vlat_lead) < 0.4 else 0 + self.counters['blinkers'] = self.counters['blinkers'] + 10.0 / (20 * 3.0) if not blinkers else 0 + + a_thr = interp(v_lead, _FCW_A_ACT_BP, _FCW_A_ACT_V) + a_delta = min(mpc_solution_a[:15]) - min(0.0, a_ego) + + fcw_allowed = all(c >= 10 for c in self.counters.values()) + if (self.last_min_a < -3.0 or a_delta < a_thr) and fcw_allowed and self.last_fcw_time + 5.0 < cur_time: + self.last_fcw_time = cur_time + self.last_fcw_a = self.last_min_a + return True + + return False + + +class LongitudinalMpc(object): + def __init__(self, mpc_id, live_longitudinal_mpc): + self.live_longitudinal_mpc = live_longitudinal_mpc + self.mpc_id = mpc_id + + self.setup_mpc() + self.v_mpc = 0.0 + self.v_mpc_future = 0.0 + self.a_mpc = 0.0 + self.v_cruise = 0.0 + self.prev_lead_status = False + self.prev_lead_x = 0.0 + self.new_lead = False + + self.last_cloudlog_t = 0.0 + + def send_mpc_solution(self, qp_iterations, calculation_time): + qp_iterations = max(0, qp_iterations) + dat = messaging.new_message() + dat.init('liveLongitudinalMpc') + dat.liveLongitudinalMpc.xEgo = list(self.mpc_solution[0].x_ego) + dat.liveLongitudinalMpc.vEgo = list(self.mpc_solution[0].v_ego) + dat.liveLongitudinalMpc.aEgo = list(self.mpc_solution[0].a_ego) + dat.liveLongitudinalMpc.xLead = list(self.mpc_solution[0].x_l) + dat.liveLongitudinalMpc.vLead = list(self.mpc_solution[0].v_l) + dat.liveLongitudinalMpc.cost = self.mpc_solution[0].cost + dat.liveLongitudinalMpc.aLeadTau = self.a_lead_tau + dat.liveLongitudinalMpc.qpIterations = qp_iterations + dat.liveLongitudinalMpc.mpcId = self.mpc_id + dat.liveLongitudinalMpc.calculationTime = calculation_time + self.live_longitudinal_mpc.send(dat.to_bytes()) + + def setup_mpc(self): + ffi, self.libmpc = libmpc_py.get_libmpc(self.mpc_id) + self.libmpc.init(MPC_COST_LONG.TTC, MPC_COST_LONG.DISTANCE, + MPC_COST_LONG.ACCELERATION, MPC_COST_LONG.JERK) + + self.mpc_solution = ffi.new("log_t *") + self.cur_state = ffi.new("state_t *") + self.cur_state[0].v_ego = 0 + self.cur_state[0].a_ego = 0 + self.a_lead_tau = _LEAD_ACCEL_TAU + + def set_cur_state(self, v, a): + self.cur_state[0].v_ego = v + self.cur_state[0].a_ego = a + + def update(self, CS, lead, v_cruise_setpoint): + v_ego = CS.carState.vEgo + # Setup current mpc state + self.cur_state[0].x_ego = 0.0 + + if lead is not None and lead.status: + x_lead = lead.dRel + v_lead = max(0.0, lead.vLead) + a_lead = lead.aLeadK + + if (v_lead < 0.1 or -a_lead / 2.0 > v_lead): + v_lead = 0.0 + a_lead = 0.0 + + self.a_lead_tau = max(lead.aLeadTau, (a_lead**2 * math.pi) / (2 * (v_lead + 0.01)**2)) + self.new_lead = False + if not self.prev_lead_status or abs(x_lead - self.prev_lead_x) > 2.5: + self.libmpc.init_with_simulation(self.v_mpc, x_lead, v_lead, a_lead, self.a_lead_tau) + self.new_lead = True + + self.prev_lead_status = True + self.prev_lead_x = x_lead + self.cur_state[0].x_l = x_lead + self.cur_state[0].v_l = v_lead + else: + self.prev_lead_status = False + # Fake a fast lead car, so mpc keeps running + self.cur_state[0].x_l = 50.0 + self.cur_state[0].v_l = v_ego + 10.0 + a_lead = 0.0 + self.a_lead_tau = _LEAD_ACCEL_TAU + + # Calculate mpc + t = sec_since_boot() + + # hardcode follow time for now + follow_time = 1.8 + distance_cost = 0.1 + follow_time_value = scale_time_gap(v_ego, follow_time) + + n_its = self.libmpc.run_mpc(self.cur_state, self.mpc_solution, self.a_lead_tau, a_lead, follow_time_value, distance_cost) + duration = int((sec_since_boot() - t) * 1e9) + self.send_mpc_solution(n_its, duration) + + # Get solution. MPC timestep is 0.2 s, so interpolation to 0.05 s is needed + self.v_mpc = self.mpc_solution[0].v_ego[1] + self.a_mpc = self.mpc_solution[0].a_ego[1] + self.v_mpc_future = self.mpc_solution[0].v_ego[10] + + # Reset if NaN or goes through lead car + dls = np.array(list(self.mpc_solution[0].x_l)) - np.array(list(self.mpc_solution[0].x_ego)) + crashing = min(dls) < -50.0 + nans = np.any(np.isnan(list(self.mpc_solution[0].v_ego))) + backwards = min(list(self.mpc_solution[0].v_ego)) < -0.01 + + if ((backwards or crashing) and self.prev_lead_status) or nans: + if t > self.last_cloudlog_t + 5.0: + self.last_cloudlog_t = t + cloudlog.warning("Longitudinal mpc %d reset - backwards: %s crashing: %s nan: %s" % ( + self.mpc_id, backwards, crashing, nans)) + + self.libmpc.init(MPC_COST_LONG.TTC, MPC_COST_LONG.DISTANCE, + MPC_COST_LONG.ACCELERATION, MPC_COST_LONG.JERK) + self.cur_state[0].v_ego = v_ego + self.cur_state[0].a_ego = 0.0 + self.v_mpc = v_ego + self.a_mpc = CS.carState.aEgo + self.prev_lead_status = False + + +>>>>>>> Add time gap and distance cost parameters to Longitudinal MPC (#23) class Planner(object): def __init__(self, CP, fcw_enabled): context = zmq.Context()