coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2025-12-08 01:36:47 +08:00
Code Issues Packages Projects Releases Wiki Activity

refactor I420ToABGR to use I420ToARGBRow

Browse Source 
Using a transposed conversion matrix, I420ToARGB can output ABGR.

R=harryjin@google.com, xhwang@chromium.org
BUG=libyuv:473

Review URL: https://codereview.chromium.org/1413573010 .
This commit is contained in:
Frank Barchard 2015-10-30 11:56:57 -07:00
parent 254ef01551
commit 5d97b93369
10 changed files with 557 additions and 1359 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

90
include/libyuv/row.h
View File

@ -112,7 +112,6 @@ extern "C" {
// https://code.google.com/p/libyuv/issues/detail?id=517
#define HAS_I422ALPHATOARGBROW_SSSE3
#endif
#define HAS_I422TOABGRROW_SSSE3
#define HAS_I422TOARGB1555ROW_SSSE3
#define HAS_I422TOARGB4444ROW_SSSE3
#define HAS_I422TOARGBROW_SSSE3
@ -207,7 +206,6 @@ extern "C" {
// https://code.google.com/p/libyuv/issues/detail?id=517
#define HAS_I422ALPHATOARGBROW_AVX2
#endif
#define HAS_I422TOABGRROW_AVX2
#define HAS_I422TOARGBROW_AVX2
#define HAS_I422TOBGRAROW_AVX2
#define HAS_I422TORAWROW_AVX2
@ -261,7 +259,6 @@ extern "C" {
#if !defined(LIBYUV_DISABLE_X86) && defined (_M_X64) && \
(!defined(__clang__) || defined(__SSSE3__))
#define HAS_I422ALPHATOARGBROW_SSSE3
#define HAS_I422TOABGRROW_SSSE3
#define HAS_I422TOARGBROW_SSSE3
#endif
@ -295,7 +292,6 @@ extern "C" {
#define HAS_COPYROW_NEON
#define HAS_I400TOARGBROW_NEON
#define HAS_I411TOARGBROW_NEON
#define HAS_I422TOABGRROW_NEON
#define HAS_I422TOARGB1555ROW_NEON
#define HAS_I422TOARGB4444ROW_NEON
#define HAS_I422TOARGBROW_NEON
@ -362,7 +358,6 @@ extern "C" {
(_MIPS_SIM == _MIPS_SIM_ABI32) && (__mips_isa_rev < 6)
#define HAS_COPYROW_MIPS
#if defined(__mips_dsp) && (__mips_dsp_rev >= 2)
#define HAS_I422TOABGRROW_MIPS_DSPR2
#define HAS_I422TOARGBROW_MIPS_DSPR2
#define HAS_I422TOBGRAROW_MIPS_DSPR2
#define HAS_INTERPOLATEROW_MIPS_DSPR2
@ -460,10 +455,15 @@ struct YuvConstants {
#define KYTORGB 192
#endif
// Conversion matrix for YUV to RGB
extern const struct YuvConstants kYuvIConstants; // BT.601
extern const struct YuvConstants kYuvJConstants; // JPeg color space
extern const struct YuvConstants kYuvHConstants; // BT.709
extern const struct YuvConstants kYvuIConstants; // YVU to BGR BT.601
// Conversion matrix for YVU to BGR.
extern const struct YuvConstants kYvuIConstants; // BT.601
extern const struct YuvConstants kYvuJConstants; // JPeg color space
extern const struct YuvConstants kYvuHConstants; // BT.709
#if defined(__APPLE__) || defined(__x86_64__) || defined(__llvm__)
#define OMITFP
@ -568,12 +568,6 @@ void I422ToARGBRow_NEON(const uint8* src_y,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToABGRRow_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,
@ -586,12 +580,6 @@ void I422ToBGRARow_NEON(const uint8* src_y,
uint8* dst_bgra,
const struct YuvConstants* yuvconstants,
int width);
void I422ToABGRRow_NEON(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
uint8* dst_abgr,
const struct YuvConstants* yuvconstants,
int width);
void I422ToRGBARow_NEON(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
@ -1058,12 +1046,6 @@ void I422AlphaToARGBRow_C(const uint8* y_buf,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToABGRRow_C(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I411ToARGBRow_C(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
@ -1099,12 +1081,6 @@ void I422ToBGRARow_C(const uint8* src_y,
uint8* dst_bgra,
const struct YuvConstants* yuvconstants,
int width);
void I422ToABGRRow_C(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
uint8* dst_abgr,
const struct YuvConstants* yuvconstants,
int width);
void I422ToRGBARow_C(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
@ -1153,12 +1129,6 @@ void I422ToARGBRow_AVX2(const uint8* src_y,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToABGRRow_AVX2(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToBGRARow_AVX2(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
@ -1171,12 +1141,6 @@ void I422ToRGBARow_AVX2(const uint8* src_y,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToABGRRow_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,
@ -1227,12 +1191,6 @@ void I422ToARGBRow_SSSE3(const uint8* src_y,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToABGRRow_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,
@ -1297,12 +1255,6 @@ void I422ToBGRARow_SSSE3(const uint8* src_y,
uint8* dst_bgra,
const struct YuvConstants* yuvconstants,
int width);
void I422ToABGRRow_SSSE3(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
uint8* dst_abgr,
const struct YuvConstants* yuvconstants,
int width);
void I422ToRGBARow_SSSE3(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
@ -1387,12 +1339,6 @@ void I422ToRGBARow_Any_AVX2(const uint8* src_y,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToABGRRow_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,
@ -1489,12 +1435,6 @@ void I422ToBGRARow_Any_SSSE3(const uint8* src_y,
uint8* dst_bgra,
const struct YuvConstants* yuvconstants,
int width);
void I422ToABGRRow_Any_SSSE3(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
uint8* dst_abgr,
const struct YuvConstants* yuvconstants,
int width);
void I422ToRGBARow_Any_SSSE3(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
@ -1681,12 +1621,6 @@ void I422ToBGRARow_Any_NEON(const uint8* src_y,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToABGRRow_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,
@ -1758,12 +1692,6 @@ void I422ToBGRARow_MIPS_DSPR2(const uint8* src_y,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToABGRRow_MIPS_DSPR2(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToARGBRow_MIPS_DSPR2(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
@ -1776,12 +1704,6 @@ void I422ToBGRARow_MIPS_DSPR2(const uint8* src_y,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToABGRRow_MIPS_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,

1004
source/convert_argb.cc
View File

File diff suppressed because it is too large Load Diff

131
source/convert_from.cc
View File

@ -450,76 +450,6 @@ int I420ToNV21(const uint8* src_y, int src_stride_y,
width, 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) {
int y;
void (*I422ToARGBRow)(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
const struct YuvConstants* yuvconstants,
int width) = I422ToARGBRow_C;
if (!src_y || !src_u || !src_v || !dst_argb ||
width <= 0 || height == 0) {
return -1;
}
// Negative height means invert the image.
if (height < 0) {
height = -height;
dst_argb = dst_argb + (height - 1) * dst_stride_argb;
dst_stride_argb = -dst_stride_argb;
}
#if defined(HAS_I422TOARGBROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
I422ToARGBRow = I422ToARGBRow_Any_SSSE3;
if (IS_ALIGNED(width, 8)) {
I422ToARGBRow = I422ToARGBRow_SSSE3;
}
}
#endif
#if defined(HAS_I422TOARGBROW_AVX2)
if (TestCpuFlag(kCpuHasAVX2)) {
I422ToARGBRow = I422ToARGBRow_Any_AVX2;
if (IS_ALIGNED(width, 16)) {
I422ToARGBRow = I422ToARGBRow_AVX2;
}
}
#endif
#if defined(HAS_I422TOARGBROW_NEON)
if (TestCpuFlag(kCpuHasNEON)) {
I422ToARGBRow = I422ToARGBRow_Any_NEON;
if (IS_ALIGNED(width, 8)) {
I422ToARGBRow = I422ToARGBRow_NEON;
}
}
#endif
#if defined(HAS_I422TOARGBROW_MIPS_DSPR2)
if (TestCpuFlag(kCpuHasMIPS_DSPR2) && IS_ALIGNED(width, 4) &&
IS_ALIGNED(src_y, 4) && IS_ALIGNED(src_stride_y, 4) &&
IS_ALIGNED(src_u, 2) && IS_ALIGNED(src_stride_u, 2) &&
IS_ALIGNED(src_v, 2) && IS_ALIGNED(src_stride_v, 2) &&
IS_ALIGNED(dst_argb, 4) && IS_ALIGNED(dst_stride_argb, 4)) {
I422ToARGBRow = I422ToARGBRow_MIPS_DSPR2;
}
#endif
for (y = 0; y < height; ++y) {
I422ToARGBRow(src_y, src_u, src_v, dst_argb, &kYuvIConstants, width);
dst_argb += dst_stride_argb;
src_y += src_stride_y;
if (y & 1) {
src_u += src_stride_u;
src_v += src_stride_v;
}
}
return 0;
}
// Convert I420 to BGRA.
LIBYUV_API
int I420ToBGRA(const uint8* src_y, int src_stride_y,
@ -590,67 +520,6 @@ int I420ToBGRA(const uint8* src_y, int src_stride_y,
return 0;
}
// Convert I420 to ABGR.
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_abgr, int dst_stride_abgr,
int width, int height) {
int y;
void (*I422ToABGRRow)(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
const struct YuvConstants* yuvconstants,
int width) = I422ToABGRRow_C;
if (!src_y || !src_u || !src_v || !dst_abgr ||
width <= 0 || height == 0) {
return -1;
}
// Negative height means invert the image.
if (height < 0) {
height = -height;
dst_abgr = dst_abgr + (height - 1) * dst_stride_abgr;
dst_stride_abgr = -dst_stride_abgr;
}
#if defined(HAS_I422TOABGRROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
I422ToABGRRow = I422ToABGRRow_Any_SSSE3;
if (IS_ALIGNED(width, 8)) {
I422ToABGRRow = I422ToABGRRow_SSSE3;
}
}
#endif
#if defined(HAS_I422TOABGRROW_AVX2)
if (TestCpuFlag(kCpuHasAVX2)) {
I422ToABGRRow = I422ToABGRRow_Any_AVX2;
if (IS_ALIGNED(width, 16)) {
I422ToABGRRow = I422ToABGRRow_AVX2;
}
}
#endif
#if defined(HAS_I422TOABGRROW_NEON)
if (TestCpuFlag(kCpuHasNEON)) {
I422ToABGRRow = I422ToABGRRow_Any_NEON;
if (IS_ALIGNED(width, 8)) {
I422ToABGRRow = I422ToABGRRow_NEON;
}
}
#endif
for (y = 0; y < height; ++y) {
I422ToABGRRow(src_y, src_u, src_v, dst_abgr, &kYuvIConstants, width);
dst_abgr += dst_stride_abgr;
src_y += src_stride_y;
if (y & 1) {
src_u += src_stride_u;
src_v += src_stride_v;
}
}
return 0;
}
// Convert I420 to RGBA.
LIBYUV_API
int I420ToRGBA(const uint8* src_y, int src_stride_y,

69
source/planar_functions.cc
View File

@ -843,75 +843,6 @@ int I422ToBGRA(const uint8* src_y, int src_stride_y,
return 0;
}
// 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) {
int y;
void (*I422ToABGRRow)(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
const struct YuvConstants* yuvconstants,
int width) = I422ToABGRRow_C;
if (!src_y || !src_u || !src_v ||
!dst_abgr ||
width <= 0 || height == 0) {
return -1;
}
// Negative height means invert the image.
if (height < 0) {
height = -height;
dst_abgr = dst_abgr + (height - 1) * dst_stride_abgr;
dst_stride_abgr = -dst_stride_abgr;
}
// Coalesce rows.
if (src_stride_y == width &&
src_stride_u * 2 == width &&
src_stride_v * 2 == width &&
dst_stride_abgr == width * 4) {
width *= height;
height = 1;
src_stride_y = src_stride_u = src_stride_v = dst_stride_abgr = 0;
}
#if defined(HAS_I422TOABGRROW_NEON)
if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
I422ToABGRRow = I422ToABGRRow_Any_NEON;
if (IS_ALIGNED(width, 8)) {
I422ToABGRRow = I422ToABGRRow_NEON;
}
}
#endif
#if defined(HAS_I422TOABGRROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
I422ToABGRRow = I422ToABGRRow_Any_SSSE3;
if (IS_ALIGNED(width, 8)) {
I422ToABGRRow = I422ToABGRRow_SSSE3;
}
}
#endif
#if defined(HAS_I422TOABGRROW_AVX2)
if (TestCpuFlag(kCpuHasAVX2)) {
I422ToABGRRow = I422ToABGRRow_Any_AVX2;
if (IS_ALIGNED(width, 16)) {
I422ToABGRRow = I422ToABGRRow_AVX2;
}
}
#endif
for (y = 0; y < height; ++y) {
I422ToABGRRow(src_y, src_u, src_v, dst_abgr, &kYuvIConstants, width);
dst_abgr += dst_stride_abgr;
src_y += src_stride_y;
src_u += src_stride_u;
src_v += src_stride_v;
}
return 0;
}
// Convert I422 to RGBA.
LIBYUV_API
int I422ToRGBA(const uint8* src_y, int src_stride_y,

5
source/row_any.cc
View File

@ -111,7 +111,6 @@ ANY31(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, 1, 1, 4, 15)
#ifdef HAS_I422TOARGBROW_SSSE3
ANY31C(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_SSSE3, 1, 0, 4, 7)
ANY31C(I422ToABGRRow_Any_SSSE3, I422ToABGRRow_SSSE3, 1, 0, 4, 7)
#endif
#ifdef HAS_I444TOARGBROW_SSSE3
ANY31C(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_SSSE3, 0, 0, 4, 7)
@ -139,9 +138,6 @@ ANY31C(I422ToBGRARow_Any_AVX2, I422ToBGRARow_AVX2, 1, 0, 4, 15)
#ifdef HAS_I422TORGBAROW_AVX2
ANY31C(I422ToRGBARow_Any_AVX2, I422ToRGBARow_AVX2, 1, 0, 4, 15)
#endif
#ifdef HAS_I422TOABGRROW_AVX2
ANY31C(I422ToABGRRow_Any_AVX2, I422ToABGRRow_AVX2, 1, 0, 4, 15)
#endif
#ifdef HAS_I444TOARGBROW_AVX2
ANY31C(I444ToARGBRow_Any_AVX2, I444ToARGBRow_AVX2, 0, 0, 4, 15)
#endif
@ -162,7 +158,6 @@ ANY31C(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, 0, 0, 4, 7)
ANY31C(I422ToARGBRow_Any_NEON, I422ToARGBRow_NEON, 1, 0, 4, 7)
ANY31C(I411ToARGBRow_Any_NEON, I411ToARGBRow_NEON, 2, 0, 4, 7)
ANY31C(I422ToBGRARow_Any_NEON, I422ToBGRARow_NEON, 1, 0, 4, 7)
ANY31C(I422ToABGRRow_Any_NEON, I422ToABGRRow_NEON, 1, 0, 4, 7)
ANY31C(I422ToRGBARow_Any_NEON, I422ToRGBARow_NEON, 1, 0, 4, 7)
ANY31C(I422ToRGB24Row_Any_NEON, I422ToRGB24Row_NEON, 1, 0, 3, 7)
ANY31C(I422ToRAWRow_Any_NEON, I422ToRAWRow_NEON, 1, 0, 3, 7)

431
source/row_common.cc
View File

@ -995,6 +995,9 @@ void J400ToARGBRow_C(const uint8* src_y, uint8* dst_argb, int width) {
}
}
// TODO(fbarchard): Unify these structures to be platform independent.
// TODO(fbarchard): Generate SIMD structures from float matrix.
// BT.601 YUV to RGB reference
// R = (Y - 16) * 1.164 - V * -1.596
// G = (Y - 16) * 1.164 - U * 0.391 - V * 0.813
@ -1015,9 +1018,6 @@ void J400ToARGBRow_C(const uint8* src_y, uint8* dst_argb, int width) {
#define BG (UG * 128 + VG * 128 + YGB)
#define BR (VR * 128 + YGB)
// BT.601 constants for YUV to RGB.
// TODO(fbarchard): Unify these structures to be platform independent.
// TODO(fbarchard): Generate SIMD structures from float matrix.
#if defined(__aarch64__)
const YuvConstants SIMD_ALIGNED(kYuvIConstants) = {
{ -UB, -VR, -UB, -VR, -UB, -VR, -UB, -VR },
@ -1027,6 +1027,14 @@ const YuvConstants SIMD_ALIGNED(kYuvIConstants) = {
{ BB, BG, BR, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
const YuvConstants SIMD_ALIGNED(kYvuIConstants) = {
{ -VR, -UB, -VR, -UB, -VR, -UB, -VR, -UB },
{ -VR, -UB, -VR, -UB, -VR, -UB, -VR, -UB },
{ VG, UG, VG, UG, VG, UG, VG, UG },
{ VG, UG, VG, UG, VG, UG, VG, UG },
{ BR, BG, BB, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
#elif defined(__arm__)
const YuvConstants SIMD_ALIGNED(kYuvIConstants) = {
{ -UB, -UB, -UB, -UB, -VR, -VR, -VR, -VR, 0, 0, 0, 0, 0, 0, 0, 0 },
@ -1034,6 +1042,12 @@ const YuvConstants SIMD_ALIGNED(kYuvIConstants) = {
{ BB, BG, BR, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
const YuvConstants SIMD_ALIGNED(kYvuIConstants) = {
{ -VR, -VR, -VR, -VR, -UB, -UB, -UB, -UB, 0, 0, 0, 0, 0, 0, 0, 0 },
{ VG, VG, VG, VG, UG, UG, UG, UG, 0, 0, 0, 0, 0, 0, 0, 0 },
{ BR, BG, BB, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
#else
const YuvConstants SIMD_ALIGNED(kYuvIConstants) = {
{ UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0,
@ -1047,8 +1061,205 @@ const YuvConstants SIMD_ALIGNED(kYuvIConstants) = {
{ BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR },
{ YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG }
};
const YuvConstants SIMD_ALIGNED(kYvuIConstants) = {
{ VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0,
VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, UB, 0 },
{ VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG,
VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG },
{ 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB,
0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, VR },
{ BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR },
{ BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG },
{ BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB },
{ YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG }
};
#endif
#undef BB
#undef BG
#undef BR
#undef YGB
#undef UB
#undef UG
#undef VG
#undef VR
#undef YG
// JPEG YUV to RGB reference
// * R = Y - V * -1.40200
// * G = Y - U * 0.34414 - V * 0.71414
// * B = Y - U * -1.77200
// Y contribution to R,G,B. Scale and bias.
#define YG 16320 /* round(1.000 * 64 * 256 * 256 / 257) */
#define YGB 32 /* 64 / 2 */
// U and V contributions to R,G,B.
#define UB -113 /* round(-1.77200 * 64) */
#define UG 22 /* round(0.34414 * 64) */
#define VG 46 /* round(0.71414 * 64) */
#define VR -90 /* round(-1.40200 * 64) */
// Bias values to round, and subtract 128 from U and V.
#define BB (UB * 128 + YGB)
#define BG (UG * 128 + VG * 128 + YGB)
#define BR (VR * 128 + YGB)
#if defined(__aarch64__)
const YuvConstants SIMD_ALIGNED(kYuvJConstants) = {
{ -UB, -VR, -UB, -VR, -UB, -VR, -UB, -VR },
{ -UB, -VR, -UB, -VR, -UB, -VR, -UB, -VR },
{ UG, VG, UG, VG, UG, VG, UG, VG },
{ UG, VG, UG, VG, UG, VG, UG, VG },
{ BB, BG, BR, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
const YuvConstants SIMD_ALIGNED(kYvuJConstants) = {
{ -VR, -UB, -VR, -UB, -VR, -UB, -VR, -UB },
{ -VR, -UB, -VR, -UB, -VR, -UB, -VR, -UB },
{ VG, UG, VG, UG, VG, UG, VG, UG },
{ VG, UG, VG, UG, VG, UG, VG, UG },
{ BR, BG, BB, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
#elif defined(__arm__)
const YuvConstants SIMD_ALIGNED(kYuvJConstants) = {
{ -UB, -UB, -UB, -UB, -VR, -VR, -VR, -VR, 0, 0, 0, 0, 0, 0, 0, 0 },
{ UG, UG, UG, UG, VG, VG, VG, VG, 0, 0, 0, 0, 0, 0, 0, 0 },
{ BB, BG, BR, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
const YuvConstants SIMD_ALIGNED(kYvuJConstants) = {
{ -VR, -VR, -VR, -VR, -UB, -UB, -UB, -UB, 0, 0, 0, 0, 0, 0, 0, 0 },
{ VG, VG, VG, VG, UG, UG, UG, UG, 0, 0, 0, 0, 0, 0, 0, 0 },
{ BR, BG, BB, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
#else
const YuvConstants SIMD_ALIGNED(kYuvJConstants) = {
{ UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0,
UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0 },
{ UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG,
UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG },
{ 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR,
0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR },
{ BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB },
{ BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG },
{ BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR },
{ YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG }
};
const YuvConstants SIMD_ALIGNED(kYvuJConstants) = {
{ VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0,
VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, UB, 0 },
{ VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG,
VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG },
{ 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB,
0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, VR },
{ BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR },
{ BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG },
{ BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB },
{ YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG }
};
#endif
#undef BB
#undef BG
#undef BR
#undef YGB
#undef UB
#undef UG
#undef VG
#undef VR
#undef YG
// BT.709 YUV to RGB reference
// * R = Y - V * -1.28033
// * G = Y - U * 0.21482 - V * 0.38059
// * B = Y - U * -2.12798
// Y contribution to R,G,B. Scale and bias.
#define YG 16320 /* round(1.000 * 64 * 256 * 256 / 257) */
#define YGB 32 /* 64 / 2 */
// TODO(fbarchard): Find way to express 2.12 instead of 2.0.
// U and V contributions to R,G,B.
#define UB -128 /* max(-128, round(-2.12798 * 64)) */
#define UG 14 /* round(0.21482 * 64) */
#define VG 24 /* round(0.38059 * 64) */
#define VR -82 /* round(-1.28033 * 64) */
// Bias values to round, and subtract 128 from U and V.
#define BB (UB * 128 + YGB)
#define BG (UG * 128 + VG * 128 + YGB)
#define BR (VR * 128 + YGB)
#if defined(__aarch64__)
const YuvConstants SIMD_ALIGNED(kYuvHConstants) = {
{ -UB, -VR, -UB, -VR, -UB, -VR, -UB, -VR },
{ -UB, -VR, -UB, -VR, -UB, -VR, -UB, -VR },
{ UG, VG, UG, VG, UG, VG, UG, VG },
{ UG, VG, UG, VG, UG, VG, UG, VG },
{ BB, BG, BR, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
const YuvConstants SIMD_ALIGNED(kYvuHConstants) = {
{ -VR, -UB, -VR, -UB, -VR, -UB, -VR, -UB },
{ -VR, -UB, -VR, -UB, -VR, -UB, -VR, -UB },
{ VG, UG, VG, UG, VG, UG, VG, UG },
{ VG, UG, VG, UG, VG, UG, VG, UG },
{ BR, BG, BB, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
#elif defined(__arm__)
const YuvConstants SIMD_ALIGNED(kYuvHConstants) = {
{ -UB, -UB, -UB, -UB, -VR, -VR, -VR, -VR, 0, 0, 0, 0, 0, 0, 0, 0 },
{ UG, UG, UG, UG, VG, VG, VG, VG, 0, 0, 0, 0, 0, 0, 0, 0 },
{ BB, BG, BR, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
const YuvConstants SIMD_ALIGNED(kYvuHConstants) = {
{ -VR, -VR, -VR, -VR, -UB, -UB, -UB, -UB, 0, 0, 0, 0, 0, 0, 0, 0 },
{ VG, VG, VG, VG, UG, UG, UG, UG, 0, 0, 0, 0, 0, 0, 0, 0 },
{ BR, BG, BB, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
#else
const YuvConstants SIMD_ALIGNED(kYuvHConstants) = {
{ UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0,
UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0 },
{ UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG,
UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG },
{ 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR,
0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR },
{ BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB },
{ BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG },
{ BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR },
{ YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG }
};
const YuvConstants SIMD_ALIGNED(kYvuHConstants) = {
{ VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0,
VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, UB, 0 },
{ VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG,
VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG },
{ 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB,
0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, VR },
{ BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR },
{ BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG },
{ BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB },
{ YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG }
};
#endif
#undef BB
#undef BG
#undef BR
#undef YGB
#undef UB
#undef UG
#undef VG
#undef VR
#undef YG
// C reference code that mimics the YUV assembly.
static __inline void YuvPixel(uint8 y, uint8 u, uint8 v,
uint8* b, uint8* g, uint8* r,
@ -1088,6 +1299,10 @@ static __inline void YuvPixel(uint8 y, uint8 u, uint8 v,
*r = Clamp((int32)(-( v * vr) + y1 + br) >> 6);
}
// Y contribution to R,G,B. Scale and bias.
#define YG 18997 /* round(1.164 * 64 * 256 * 256 / 257) */
#define YGB -1160 /* 1.164 * 64 * -16 + 64 / 2 */
// C reference code that mimics the YUV assembly.
static __inline void YPixel(uint8 y, uint8* b, uint8* g, uint8* r) {
uint32 y1 = (uint32)(y * 0x0101 * YG) >> 16;
@ -1096,190 +1311,8 @@ static __inline void YPixel(uint8 y, uint8* b, uint8* g, uint8* r) {
*r = Clamp((int32)(y1 + YGB) >> 6);
}
// BT.601 constants for YVU to BGR.
// Allows YUV TO RGB code to implement YUV to BGR by swapping UV and using this
// matrix.
#if defined(__aarch64__)
const YuvConstants SIMD_ALIGNED(kYvuIConstants) = {
{ -VR, -UB, -VR, -UB, -VR, -UB, -VR, -UB },
{ -VR, -UB, -VR, -UB, -VR, -UB, -VR, -UB },
{ VG, UG, VG, UG, VG, UG, VG, UG },
{ VG, UG, VG, UG, VG, UG, VG, UG },
{ BR, BG, BB, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
#elif defined(__arm__)
const YuvConstants SIMD_ALIGNED(kYvuIConstants) = {
{ -VR, -VR, -VR, -VR, -UB, -UB, -UB, -UB, 0, 0, 0, 0, 0, 0, 0, 0 },
{ VG, VG, VG, VG, UG, UG, UG, UG, 0, 0, 0, 0, 0, 0, 0, 0 },
{ BR, BG, BB, 0, 0, 0, 0, 0 },
{ 0x0101 * YG, 0, 0, 0 }
};
#else
const YuvConstants SIMD_ALIGNED(kYvuIConstants) = {
{ VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0,
VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, VR, 0, UB, 0 },
{ VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG,
VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG, VG, UG },
{ 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB,
0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, UB, 0, VR },
{ BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR, BR },
{ BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG, BG },
{ BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB, BB },
{ YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG, YG }
};
#endif
#undef BB
#undef BG
#undef BR
#undef YGB
#undef UB
#undef UG
#undef VG
#undef VR
#undef YG
// JPEG YUV to RGB reference
// * R = Y - V * -1.40200
// * G = Y - U * 0.34414 - V * 0.71414
// * B = Y - U * -1.77200
// Y contribution to R,G,B. Scale and bias.
#define YGJ 16320 /* round(1.000 * 64 * 256 * 256 / 257) */
#define YGBJ 32 /* 64 / 2 */
// U and V contributions to R,G,B.
#define UBJ -113 /* round(-1.77200 * 64) */
#define UGJ 22 /* round(0.34414 * 64) */
#define VGJ 46 /* round(0.71414 * 64) */
#define VRJ -90 /* round(-1.40200 * 64) */
// Bias values to round, and subtract 128 from U and V.
#define BBJ (UBJ * 128 + YGBJ)
#define BGJ (UGJ * 128 + VGJ * 128 + YGBJ)
#define BRJ (VRJ * 128 + YGBJ)
// JPEG constants for YUV to RGB.
#if defined(__aarch64__)
const YuvConstants SIMD_ALIGNED(kYuvJConstants) = {
{ -UBJ, -VRJ, -UBJ, -VRJ, -UBJ, -VRJ, -UBJ, -VRJ },
{ -UBJ, -VRJ, -UBJ, -VRJ, -UBJ, -VRJ, -UBJ, -VRJ },
{ UGJ, VGJ, UGJ, VGJ, UGJ, VGJ, UGJ, VGJ },
{ UGJ, VGJ, UGJ, VGJ, UGJ, VGJ, UGJ, VGJ },
{ BBJ, BGJ, BRJ, 0, 0, 0, 0, 0 },
{ 0x0101 * YGJ, 0, 0, 0 }
};
#elif defined(__arm__)
const YuvConstants SIMD_ALIGNED(kYuvJConstants) = {
{ -UBJ, -UBJ, -UBJ, -UBJ, -VRJ, -VRJ, -VRJ, -VRJ, 0, 0, 0, 0, 0, 0, 0, 0 },
{ UGJ, UGJ, UGJ, UGJ, VGJ, VGJ, VGJ, VGJ, 0, 0, 0, 0, 0, 0, 0, 0 },
{ BBJ, BGJ, BRJ, 0, 0, 0, 0, 0 },
{ 0x0101 * YGJ, 0, 0, 0 }
};
#else
const YuvConstants SIMD_ALIGNED(kYuvJConstants) = {
{ UBJ, 0, UBJ, 0, UBJ, 0, UBJ, 0, UBJ, 0, UBJ, 0, UBJ, 0, UBJ, 0,
UBJ, 0, UBJ, 0, UBJ, 0, UBJ, 0, UBJ, 0, UBJ, 0, UBJ, 0, UBJ, 0 },
{ UGJ, VGJ, UGJ, VGJ, UGJ, VGJ, UGJ, VGJ,
UGJ, VGJ, UGJ, VGJ, UGJ, VGJ, UGJ, VGJ,
UGJ, VGJ, UGJ, VGJ, UGJ, VGJ, UGJ, VGJ,
UGJ, VGJ, UGJ, VGJ, UGJ, VGJ, UGJ, VGJ },
{ 0, VRJ, 0, VRJ, 0, VRJ, 0, VRJ, 0, VRJ, 0, VRJ, 0, VRJ, 0, VRJ,
0, VRJ, 0, VRJ, 0, VRJ, 0, VRJ, 0, VRJ, 0, VRJ, 0, VRJ, 0, VRJ },
{ BBJ, BBJ, BBJ, BBJ, BBJ, BBJ, BBJ, BBJ,
BBJ, BBJ, BBJ, BBJ, BBJ, BBJ, BBJ, BBJ },
{ BGJ, BGJ, BGJ, BGJ, BGJ, BGJ, BGJ, BGJ,
BGJ, BGJ, BGJ, BGJ, BGJ, BGJ, BGJ, BGJ },
{ BRJ, BRJ, BRJ, BRJ, BRJ, BRJ, BRJ, BRJ,
BRJ, BRJ, BRJ, BRJ, BRJ, BRJ, BRJ, BRJ },
{ YGJ, YGJ, YGJ, YGJ, YGJ, YGJ, YGJ, YGJ,
YGJ, YGJ, YGJ, YGJ, YGJ, YGJ, YGJ, YGJ }
};
#endif
#undef YGJ
#undef YGBJ
#undef UBJ
#undef UGJ
#undef VGJ
#undef VRJ
#undef BBJ
#undef BGJ
#undef BRJ
// BT.709 YUV to RGB reference
// * R = Y - V * -1.28033
// * G = Y - U * 0.21482 - V * 0.38059
// * B = Y - U * -2.12798
// Y contribution to R,G,B. Scale and bias.
#define YGH 16320 /* round(1.000 * 64 * 256 * 256 / 257) */
#define YGBH 32 /* 64 / 2 */
// TODO(fbarchard): Find way to express 2.12 instead of 2.0.
// U and V contributions to R,G,B.
#define UBH -128 /* max(-128, round(-2.12798 * 64)) */
#define UGH 14 /* round(0.21482 * 64) */
#define VGH 24 /* round(0.38059 * 64) */
#define VRH -82 /* round(-1.28033 * 64) */
// Bias values to round, and subtract 128 from U and V.
#define BBH (UBH * 128 + YGBH)
#define BGH (UGH * 128 + VGH * 128 + YGBH)
#define BRH (VRH * 128 + YGBH)
// BT.709 constants for YUV to RGB.
#if defined(__aarch64__)
const YuvConstants SIMD_ALIGNED(kYuvHConstants) = {
{ -UBH, -VRH, -UBH, -VRH, -UBH, -VRH, -UBH, -VRH },
{ -UBH, -VRH, -UBH, -VRH, -UBH, -VRH, -UBH, -VRH },
{ UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH },
{ UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH },
{ BBH, BGH, BRH, 0, 0, 0, 0, 0 },
{ 0x0101 * YGH, 0, 0, 0 }
};
#elif defined(__arm__)
const YuvConstants SIMD_ALIGNED(kYuvHConstants) = {
{ -UBH, -UBH, -UBH, -UBH, -VRH, -VRH, -VRH, -VRH, 0, 0, 0, 0, 0, 0, 0, 0 },
{ UGH, UGH, UGH, UGH, VGH, VGH, VGH, VGH, 0, 0, 0, 0, 0, 0, 0, 0 },
{ BBH, BGH, BRH, 0, 0, 0, 0, 0 },
{ 0x0101 * YGH, 0, 0, 0 }
};
#else
const YuvConstants SIMD_ALIGNED(kYuvHConstants) = {
{ UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0,
UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0 },
{ UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH,
UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH,
UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH,
UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH },
{ 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH,
0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH },
{ BBH, BBH, BBH, BBH, BBH, BBH, BBH, BBH,
BBH, BBH, BBH, BBH, BBH, BBH, BBH, BBH },
{ BGH, BGH, BGH, BGH, BGH, BGH, BGH, BGH,
BGH, BGH, BGH, BGH, BGH, BGH, BGH, BGH },
{ BRH, BRH, BRH, BRH, BRH, BRH, BRH, BRH,
BRH, BRH, BRH, BRH, BRH, BRH, BRH, BRH },
{ YGH, YGH, YGH, YGH, YGH, YGH, YGH, YGH,
YGH, YGH, YGH, YGH, YGH, YGH, YGH, YGH }
};
#endif
#undef YGH
#undef YGBH
#undef UBH
#undef UGH
#undef VGH
#undef VRH
#undef BBH
#undef BGH
#undef BRH
#undef YGB
#if !defined(LIBYUV_DISABLE_NEON) && \
(defined(__ARM_NEON__) || defined(__aarch64__) || defined(LIBYUV_NEON))
@ -1387,32 +1420,6 @@ void I422AlphaToARGBRow_C(const uint8* src_y,
}
}
void I422ToABGRRow_C(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
uint8* rgb_buf,
const struct YuvConstants* yuvconstants,
int width) {
int x;
for (x = 0; x < width - 1; x += 2) {
YuvPixel(src_y[0], src_u[0], src_v[0],
rgb_buf + 2, rgb_buf + 1, rgb_buf + 0, yuvconstants);
rgb_buf[3] = 255;
YuvPixel(src_y[1], src_u[0], src_v[0],
rgb_buf + 6, rgb_buf + 5, rgb_buf + 4, yuvconstants);
rgb_buf[7] = 255;
src_y += 2;
src_u += 1;
src_v += 1;
rgb_buf += 8; // Advance 2 pixels.
}
if (width & 1) {
YuvPixel(src_y[0], src_u[0], src_v[0],
rgb_buf + 2, rgb_buf + 1, rgb_buf + 0, yuvconstants);
rgb_buf[3] = 255;
}
}
void I422ToRGB24Row_C(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,

62
source/row_mips.cc
View File

@ -715,68 +715,6 @@ void I422ToARGBRow_MIPS_DSPR2(const uint8* y_buf,
);
}
void I422ToABGRRow_MIPS_DSPR2(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
const struct YuvConstants* yuvconstants,
int width) {
__asm__ __volatile__ (
".set push \n"
".set noreorder \n"
"beqz %[width], 2f \n"
" repl.ph $s0, 74 \n" // |YG|YG| = |74|74|
"repl.ph $s1, -25 \n" // |UG|UG| = |-25|-25|
"repl.ph $s2, -52 \n" // |VG|VG| = |-52|-52|
"repl.ph $s3, 102 \n" // |VR|VR| = |102|102|
"repl.ph $s4, 16 \n" // |0|16|0|16|
"repl.ph $s5, 128 \n" // |128|128|
"lui $s6, 0xff00 \n"
"ori $s6, 0xff00 \n" // |ff|00|ff|00|
"1: \n"
YUVTORGB
// Arranging into abgr format
"precr.qb.ph $t0, $t8, $t1 \n" // |G1|g1|R1|r1|
"precr.qb.ph $t3, $t9, $t2 \n" // |G0|g0|R0|r0|
"precrq.qb.ph $t8, $t0, $t3 \n" // |G1|R1|G0|R0|
"precr.qb.ph $t9, $t0, $t3 \n" // |g1|r1|g0|r0|
"precr.qb.ph $t2, $t4, $t5 \n" // |B1|b1|B0|b0|
"addiu %[width], -4 \n"
"addiu %[y_buf], 4 \n"
"preceu.ph.qbla $t1, $t2 \n" // |0 |B1|0 |B0|
"preceu.ph.qbra $t2, $t2 \n" // |0 |b1|0 |b0|
"or $t1, $t1, $s6 \n" // |ff|B1|ff|B0|
"or $t2, $t2, $s6 \n" // |ff|b1|ff|b0|
"precrq.ph.w $t0, $t2, $t9 \n" // |ff|b1|g1|r1|
"precrq.ph.w $t3, $t1, $t8 \n" // |ff|B1|G1|R1|
"sll $t9, $t9, 16 \n"
"sll $t8, $t8, 16 \n"
"packrl.ph $t2, $t2, $t9 \n" // |ff|b0|g0|r0|
"packrl.ph $t1, $t1, $t8 \n" // |ff|B0|G0|R0|
// Store results.
"sw $t2, 0(%[rgb_buf]) \n"
"sw $t0, 4(%[rgb_buf]) \n"
"sw $t1, 8(%[rgb_buf]) \n"
"sw $t3, 12(%[rgb_buf]) \n"
"bnez %[width], 1b \n"
" addiu %[rgb_buf], 16 \n"
"2: \n"
".set pop \n"
:[y_buf] "+r" (y_buf),
[u_buf] "+r" (u_buf),
[v_buf] "+r" (v_buf),
[width] "+r" (width),
[rgb_buf] "+r" (rgb_buf)
:
: "t0", "t1", "t2", "t3", "t4", "t5",
"t6", "t7", "t8", "t9",
"s0", "s1", "s2", "s3",
"s4", "s5", "s6"
);
}
void I422ToBGRARow_MIPS_DSPR2(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,

31
source/row_neon.cc
View File

@ -255,37 +255,6 @@ void I422ToBGRARow_NEON(const uint8* src_y,
);
}
void I422ToABGRRow_NEON(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
uint8* dst_abgr,
const struct YuvConstants* yuvconstants,
int width) {
asm volatile (
YUVTORGB_SETUP
"1: \n"
READYUV422
YUVTORGB
"subs %4, %4, #8 \n"
"vswp.u8 d20, d22 \n"
"vmov.u8 d23, #255 \n"
MEMACCESS(3)
"vst4.8 {d20, d21, d22, d23}, [%3]! \n"
"bgt 1b \n"
: "+r"(src_y), // %0
"+r"(src_u), // %1
"+r"(src_v), // %2
"+r"(dst_abgr), // %3
"+r"(width) // %4
: [kUVToRB]"r"(&yuvconstants->kUVToRB),
[kUVToG]"r"(&yuvconstants->kUVToG),
[kUVBiasBGR]"r"(&yuvconstants->kUVBiasBGR),
[kYToRgb]"r"(&yuvconstants->kYToRgb)
: "cc", "memory", "q0", "q1", "q2", "q3", "q4",
"q8", "q9", "q10", "q11", "q12", "q13", "q14", "q15"
);
}
void I422ToRGBARow_NEON(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,

33
source/row_neon64.cc
View File

@ -265,39 +265,6 @@ void I422ToBGRARow_NEON(const uint8* src_y,
}
#endif // HAS_I422TOBGRAROW_NEON
// TODO(fbarchard): Switch to Matrix version of this function.
#ifdef HAS_I422TOABGRROW_NEON
void I422ToABGRRow_NEON(const uint8* src_y,
const uint8* src_u,
const uint8* src_v,
uint8* dst_abgr,
const struct YuvConstants* yuvconstants,
int width) {
asm volatile (
YUVTORGB_SETUP
"1: \n"
READYUV422
YUVTORGB(v20, v21, v22)
"subs %w4, %w4, #8 \n"
"movi v23.8b, #255 \n" /* A */
MEMACCESS(3)
"st4 {v20.8b,v21.8b,v22.8b,v23.8b}, [%3], #32 \n"
"b.gt 1b \n"
: "+r"(src_y), // %0
"+r"(src_u), // %1
"+r"(src_v), // %2
"+r"(dst_abgr), // %3
"+r"(width) // %4
: [kUVToRB]"r"(&yuvconstants->kUVToRB),
[kUVToG]"r"(&yuvconstants->kUVToG),
[kUVBiasBGR]"r"(&yuvconstants->kUVBiasBGR),
[kYToRgb]"r"(&yuvconstants->kYToRgb)
: "cc", "memory", "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v20",
"v21", "v22", "v23", "v24", "v25", "v26", "v27", "v28", "v29", "v30"
);
}
#endif // HAS_I422TOABGRROW_NEON
#ifdef HAS_I422TORGBAROW_NEON
void I422ToRGBARow_NEON(const uint8* src_y,
const uint8* src_u,

60
source/row_win.cc
View File

@ -115,25 +115,6 @@ void I422ToARGBRow_SSSE3(const uint8* y_buf,
}
#endif
#if defined(HAS_I422TOABGRROW_SSSE3)
void I422ToABGRRow_SSSE3(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* dst_abgr,
const struct YuvConstants* yuvconstants,
int width) {
__m128i xmm0, xmm1, xmm2, xmm4;
const __m128i xmm5 = _mm_set1_epi8(-1);
const ptrdiff_t offset = (uint8*)v_buf - (uint8*)u_buf;
while (width > 0) {
READYUV422
YUVTORGB(yuvconstants)
STOREABGR
width -= 8;
}
}
#endif
#if defined(HAS_I422ALPHATOARGBROW_SSSE3)
void I422AlphaToARGBRow_SSSE3(const uint8* y_buf,
const uint8* u_buf,
@ -2455,48 +2436,9 @@ void I422ToRGBARow_AVX2(const uint8* y_buf,
}
#endif // HAS_I422TORGBAROW_AVX2
#ifdef HAS_I422TOABGRROW_AVX2
// 16 pixels
// 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 ABGR (64 bytes).
__declspec(naked)
void I422ToABGRRow_AVX2(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
__asm {
push esi
push edi
push ebx
mov eax, [esp + 12 + 4] // Y
mov esi, [esp + 12 + 8] // U
mov edi, [esp + 12 + 12] // V
mov edx, [esp + 12 + 16] // argb
mov ebx, [esp + 12 + 20] // yuvconstants
mov ecx, [esp + 12 + 24] // width
sub edi, esi
vpcmpeqb ymm5, ymm5, ymm5 // generate 0xffffffffffffffff for alpha
convertloop:
READYUV422_AVX2
YUVTORGB_AVX2(ebx)
STOREABGR_AVX2
sub ecx, 16
jg convertloop
pop ebx
pop edi
pop esi
vzeroupper
ret
}
}
#endif // HAS_I422TOABGRROW_AVX2
#if defined(HAS_I422TOARGBROW_SSSE3)
// TODO(fbarchard): Read that does half size on Y and treats 420 as 444.
// Allows a conversion with half size scaling.
// Read 8 UV from 444.
#define READYUV444 __asm { \