coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2026-02-16 23:29:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add I422AlphaToARGB and I444AlphaToARGB

Browse Source 
Bug: libyuv:878
Change-Id: I64c314326ac7ae5242acc64e20016e30adc6d17f
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2639439
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
This commit is contained in:
Yuan Tong 2021-01-18 11:53:26 +08:00 committed by Frank Barchard
parent 93b1b332cd
commit 08d0dce5fc
10 changed files with 950 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

102
include/libyuv/convert_argb.h
View File

@ -49,6 +49,10 @@ LIBYUV_API extern const struct YuvConstants kYvu2020Constants; // BT.2020
NV12ToRGB24Matrix(a, b, c, d, e, f, g##VU, h, i) NV12ToRGB24Matrix(a, b, c, d, e, f, g##VU, h, i)
#define I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, k, l, m, n) \ #define I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, k, l, m, n) \
I420AlphaToARGBMatrix(a, b, e, f, c, d, g, h, i, j, k##VU, l, m, n) I420AlphaToARGBMatrix(a, b, e, f, c, d, g, h, i, j, k##VU, l, m, n)
#define I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, k, l, m, n) \
I422AlphaToARGBMatrix(a, b, e, f, c, d, g, h, i, j, k##VU, l, m, n)
#define I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, k, l, m, n) \
I444AlphaToARGBMatrix(a, b, e, f, c, d, g, h, i, j, k##VU, l, m, n)
// Alias. // Alias.
#define ARGBToARGB ARGBCopy #define ARGBToARGB ARGBCopy
@ -562,6 +566,70 @@ int I420AlphaToABGR(const uint8_t* src_y,
int height, int height,
int attenuate); int attenuate);
// Convert I422 with Alpha to preattenuated ARGB.
LIBYUV_API
int I422AlphaToARGB(const uint8_t* src_y,
int src_stride_y,
const uint8_t* src_u,
int src_stride_u,
const uint8_t* src_v,
int src_stride_v,
const uint8_t* src_a,
int src_stride_a,
uint8_t* dst_argb,
int dst_stride_argb,
int width,
int height,
int attenuate);
// Convert I422 with Alpha to preattenuated ABGR.
LIBYUV_API
int I422AlphaToABGR(const uint8_t* src_y,
int src_stride_y,
const uint8_t* src_u,
int src_stride_u,
const uint8_t* src_v,
int src_stride_v,
const uint8_t* src_a,
int src_stride_a,
uint8_t* dst_abgr,
int dst_stride_abgr,
int width,
int height,
int attenuate);
// Convert I444 with Alpha to preattenuated ARGB.
LIBYUV_API
int I444AlphaToARGB(const uint8_t* src_y,
int src_stride_y,
const uint8_t* src_u,
int src_stride_u,
const uint8_t* src_v,
int src_stride_v,
const uint8_t* src_a,
int src_stride_a,
uint8_t* dst_argb,
int dst_stride_argb,
int width,
int height,
int attenuate);
// Convert I444 with Alpha to preattenuated ABGR.
LIBYUV_API
int I444AlphaToABGR(const uint8_t* src_y,
int src_stride_y,
const uint8_t* src_u,
int src_stride_u,
const uint8_t* src_v,
int src_stride_v,
const uint8_t* src_a,
int src_stride_a,
uint8_t* dst_abgr,
int dst_stride_abgr,
int width,
int height,
int attenuate);
// Convert I400 (grey) to ARGB. Reverse of ARGBToI400. // Convert I400 (grey) to ARGB. Reverse of ARGBToI400.
LIBYUV_API LIBYUV_API
int I400ToARGB(const uint8_t* src_y, int I400ToARGB(const uint8_t* src_y,
@ -1412,6 +1480,40 @@ int I420AlphaToARGBMatrix(const uint8_t* src_y,
int height, int height,
int attenuate); int attenuate);
// Convert I422 with Alpha to preattenuated ARGB with matrix.
LIBYUV_API
int I422AlphaToARGBMatrix(const uint8_t* src_y,
int src_stride_y,
const uint8_t* src_u,
int src_stride_u,
const uint8_t* src_v,
int src_stride_v,
const uint8_t* src_a,
int src_stride_a,
uint8_t* dst_argb,
int dst_stride_argb,
const struct YuvConstants* yuvconstants,
int width,
int height,
int attenuate);
// Convert I444 with Alpha to preattenuated ARGB with matrix.
LIBYUV_API
int I444AlphaToARGBMatrix(const uint8_t* src_y,
int src_stride_y,
const uint8_t* src_u,
int src_stride_u,
const uint8_t* src_v,
int src_stride_v,
const uint8_t* src_a,
int src_stride_a,
uint8_t* dst_argb,
int dst_stride_argb,
const struct YuvConstants* yuvconstants,
int width,
int height,
int attenuate);
// Convert NV12 to ARGB with matrix. // Convert NV12 to ARGB with matrix.
LIBYUV_API LIBYUV_API
int NV12ToARGBMatrix(const uint8_t* src_y, int NV12ToARGBMatrix(const uint8_t* src_y,

72
include/libyuv/row.h
View File

@ -175,6 +175,7 @@ extern "C" {
defined(_MSC_VER) defined(_MSC_VER)
// TODO(fbarchard): fix build error on android_full_debug=1 // TODO(fbarchard): fix build error on android_full_debug=1
// https://code.google.com/p/libyuv/issues/detail?id=517 // https://code.google.com/p/libyuv/issues/detail?id=517
#define HAS_I444ALPHATOARGBROW_SSSE3
#define HAS_I422ALPHATOARGBROW_SSSE3 #define HAS_I422ALPHATOARGBROW_SSSE3
#endif #endif
#endif #endif
@ -239,6 +240,7 @@ extern "C" {
defined(_MSC_VER) defined(_MSC_VER)
// TODO(fbarchard): fix build error on android_full_debug=1 // TODO(fbarchard): fix build error on android_full_debug=1
// https://code.google.com/p/libyuv/issues/detail?id=517 // https://code.google.com/p/libyuv/issues/detail?id=517
#define HAS_I444ALPHATOARGBROW_AVX2
#define HAS_I422ALPHATOARGBROW_AVX2 #define HAS_I422ALPHATOARGBROW_AVX2
#endif #endif
#endif #endif
@ -259,6 +261,8 @@ extern "C" {
// The following are also available on x64 Visual C. // The following are also available on x64 Visual C.
#if !defined(LIBYUV_DISABLE_X86) && defined(_MSC_VER) && defined(_M_X64) && \ #if !defined(LIBYUV_DISABLE_X86) && defined(_MSC_VER) && defined(_M_X64) && \
(!defined(__clang__) || defined(__SSSE3__)) (!defined(__clang__) || defined(__SSSE3__))
#define HAS_I444ALPHATOARGBROW_SSSE3
#define HAS_I444TOARGBROW_SSSE3
#define HAS_I422ALPHATOARGBROW_SSSE3 #define HAS_I422ALPHATOARGBROW_SSSE3
#define HAS_I422TOARGBROW_SSSE3 #define HAS_I422TOARGBROW_SSSE3
#endif #endif
@ -357,6 +361,7 @@ extern "C" {
#define HAS_HALFFLOATROW_NEON #define HAS_HALFFLOATROW_NEON
#define HAS_HALFMERGEUVROW_NEON #define HAS_HALFMERGEUVROW_NEON
#define HAS_I400TOARGBROW_NEON #define HAS_I400TOARGBROW_NEON
#define HAS_I444ALPHATOARGBROW_NEON
#define HAS_I422ALPHATOARGBROW_NEON #define HAS_I422ALPHATOARGBROW_NEON
#define HAS_I422TOARGB1555ROW_NEON #define HAS_I422TOARGB1555ROW_NEON
#define HAS_I422TOARGB4444ROW_NEON #define HAS_I422TOARGB4444ROW_NEON
@ -795,6 +800,13 @@ void I422ToARGBRow_NEON(const uint8_t* src_y,
uint8_t* dst_argb, uint8_t* dst_argb,
const struct YuvConstants* yuvconstants, const struct YuvConstants* yuvconstants,
int width); int width);
void I444AlphaToARGBRow_NEON(const uint8_t* src_y,
const uint8_t* src_u,
const uint8_t* src_v,
const uint8_t* src_a,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422AlphaToARGBRow_NEON(const uint8_t* src_y, void I422AlphaToARGBRow_NEON(const uint8_t* src_y,
const uint8_t* src_u, const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,
@ -802,12 +814,6 @@ void I422AlphaToARGBRow_NEON(const uint8_t* src_y,
uint8_t* dst_argb, uint8_t* dst_argb,
const struct YuvConstants* yuvconstants, const struct YuvConstants* yuvconstants,
int width); int width);
void I422ToARGBRow_NEON(const uint8_t* src_y,
const uint8_t* src_u,
const uint8_t* src_v,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToRGBARow_NEON(const uint8_t* src_y, void I422ToRGBARow_NEON(const uint8_t* src_y,
const uint8_t* src_u, const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,
@ -2306,6 +2312,13 @@ void I210ToARGBRow_C(const uint16_t* src_y,
uint8_t* rgb_buf, uint8_t* rgb_buf,
const struct YuvConstants* yuvconstants, const struct YuvConstants* yuvconstants,
int width); int width);
void I444AlphaToARGBRow_C(const uint8_t* src_y,
const uint8_t* src_u,
const uint8_t* src_v,
const uint8_t* src_a,
uint8_t* rgb_buf,
const struct YuvConstants* yuvconstants,
int width);
void I422AlphaToARGBRow_C(const uint8_t* src_y, void I422AlphaToARGBRow_C(const uint8_t* src_y,
const uint8_t* src_u, const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,
@ -2404,18 +2417,6 @@ void I444ToARGBRow_AVX2(const uint8_t* y_buf,
uint8_t* dst_argb, uint8_t* dst_argb,
const struct YuvConstants* yuvconstants, const struct YuvConstants* yuvconstants,
int width); int width);
void I444ToARGBRow_SSSE3(const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I444ToARGBRow_AVX2(const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToARGBRow_SSSE3(const uint8_t* y_buf, void I422ToARGBRow_SSSE3(const uint8_t* y_buf,
const uint8_t* u_buf, const uint8_t* u_buf,
const uint8_t* v_buf, const uint8_t* v_buf,
@ -2459,6 +2460,20 @@ void I210ToAR30Row_AVX2(const uint16_t* y_buf,
uint8_t* dst_ar30, uint8_t* dst_ar30,
const struct YuvConstants* yuvconstants, const struct YuvConstants* yuvconstants,
int width); int width);
void I444AlphaToARGBRow_SSSE3(const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
const uint8_t* a_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I444AlphaToARGBRow_AVX2(const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
const uint8_t* a_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422AlphaToARGBRow_SSSE3(const uint8_t* y_buf, void I422AlphaToARGBRow_SSSE3(const uint8_t* y_buf,
const uint8_t* u_buf, const uint8_t* u_buf,
const uint8_t* v_buf, const uint8_t* v_buf,
@ -2663,6 +2678,20 @@ void I210ToAR30Row_Any_AVX2(const uint16_t* y_buf,
uint8_t* dst_ptr, uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants, const struct YuvConstants* yuvconstants,
int width); int width);
void I444AlphaToARGBRow_Any_SSSE3(const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
const uint8_t* a_buf,
uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants,
int width);
void I444AlphaToARGBRow_Any_AVX2(const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
const uint8_t* a_buf,
uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants,
int width);
void I422AlphaToARGBRow_Any_SSSE3(const uint8_t* y_buf, void I422AlphaToARGBRow_Any_SSSE3(const uint8_t* y_buf,
const uint8_t* u_buf, const uint8_t* u_buf,
const uint8_t* v_buf, const uint8_t* v_buf,
@ -3164,6 +3193,13 @@ void I422ToARGBRow_Any_NEON(const uint8_t* y_buf,
uint8_t* dst_ptr, uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants, const struct YuvConstants* yuvconstants,
int width); int width);
void I444AlphaToARGBRow_Any_NEON(const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
const uint8_t* a_buf,
uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants,
int width);
void I422AlphaToARGBRow_Any_NEON(const uint8_t* y_buf, void I422AlphaToARGBRow_Any_NEON(const uint8_t* y_buf,
const uint8_t* u_buf, const uint8_t* u_buf,
const uint8_t* v_buf, const uint8_t* v_buf,

344
source/convert_argb.cc
View File

@ -1516,6 +1516,264 @@ int I420AlphaToARGBMatrix(const uint8_t* src_y,
return 0; return 0;
} }
// Convert I422 with Alpha to preattenuated ARGB with matrix.
LIBYUV_API
int I422AlphaToARGBMatrix(const uint8_t* src_y,
int src_stride_y,
const uint8_t* src_u,
int src_stride_u,
const uint8_t* src_v,
int src_stride_v,
const uint8_t* src_a,
int src_stride_a,
uint8_t* dst_argb,
int dst_stride_argb,
const struct YuvConstants* yuvconstants,
int width,
int height,
int attenuate) {
int y;
void (*I422AlphaToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
const uint8_t* v_buf, const uint8_t* a_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) = I422AlphaToARGBRow_C;
void (*ARGBAttenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
int width) = ARGBAttenuateRow_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_I422ALPHATOARGBROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_Any_SSSE3;
if (IS_ALIGNED(width, 8)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_SSSE3;
}
}
#endif
#if defined(HAS_I422ALPHATOARGBROW_AVX2)
if (TestCpuFlag(kCpuHasAVX2)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_Any_AVX2;
if (IS_ALIGNED(width, 16)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_AVX2;
}
}
#endif
#if defined(HAS_I422ALPHATOARGBROW_NEON)
if (TestCpuFlag(kCpuHasNEON)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_Any_NEON;
if (IS_ALIGNED(width, 8)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_NEON;
}
}
#endif
#if defined(HAS_I422ALPHATOARGBROW_MMI)
if (TestCpuFlag(kCpuHasMMI)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_Any_MMI;
if (IS_ALIGNED(width, 4)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_MMI;
}
}
#endif
#if defined(HAS_I422ALPHATOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_MSA;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;
if (IS_ALIGNED(width, 4)) {
ARGBAttenuateRow = ARGBAttenuateRow_SSSE3;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_AVX2)
if (TestCpuFlag(kCpuHasAVX2)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_AVX2;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_AVX2;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_NEON)
if (TestCpuFlag(kCpuHasNEON)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_NEON;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_NEON;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MMI)
if (TestCpuFlag(kCpuHasMMI)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MMI;
if (IS_ALIGNED(width, 2)) {
ARGBAttenuateRow = ARGBAttenuateRow_MMI;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_MSA;
}
}
#endif
for (y = 0; y < height; ++y) {
I422AlphaToARGBRow(src_y, src_u, src_v, src_a, dst_argb, yuvconstants,
width);
if (attenuate) {
ARGBAttenuateRow(dst_argb, dst_argb, width);
}
dst_argb += dst_stride_argb;
src_a += src_stride_a;
src_y += src_stride_y;
src_u += src_stride_u;
src_v += src_stride_v;
}
return 0;
}
// Convert I444 with Alpha to preattenuated ARGB with matrix.
LIBYUV_API
int I444AlphaToARGBMatrix(const uint8_t* src_y,
int src_stride_y,
const uint8_t* src_u,
int src_stride_u,
const uint8_t* src_v,
int src_stride_v,
const uint8_t* src_a,
int src_stride_a,
uint8_t* dst_argb,
int dst_stride_argb,
const struct YuvConstants* yuvconstants,
int width,
int height,
int attenuate) {
int y;
void (*I444AlphaToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
const uint8_t* v_buf, const uint8_t* a_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) = I444AlphaToARGBRow_C;
void (*ARGBAttenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
int width) = ARGBAttenuateRow_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_I444ALPHATOARGBROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_Any_SSSE3;
if (IS_ALIGNED(width, 8)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_SSSE3;
}
}
#endif
#if defined(HAS_I444ALPHATOARGBROW_AVX2)
if (TestCpuFlag(kCpuHasAVX2)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_Any_AVX2;
if (IS_ALIGNED(width, 16)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_AVX2;
}
}
#endif
#if defined(HAS_I444ALPHATOARGBROW_NEON)
if (TestCpuFlag(kCpuHasNEON)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_Any_NEON;
if (IS_ALIGNED(width, 8)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_NEON;
}
}
#endif
#if defined(HAS_I444ALPHATOARGBROW_MMI)
if (TestCpuFlag(kCpuHasMMI)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_Any_MMI;
if (IS_ALIGNED(width, 4)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_MMI;
}
}
#endif
#if defined(HAS_I444ALPHATOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_MSA;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;
if (IS_ALIGNED(width, 4)) {
ARGBAttenuateRow = ARGBAttenuateRow_SSSE3;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_AVX2)
if (TestCpuFlag(kCpuHasAVX2)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_AVX2;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_AVX2;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_NEON)
if (TestCpuFlag(kCpuHasNEON)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_NEON;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_NEON;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MMI)
if (TestCpuFlag(kCpuHasMMI)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MMI;
if (IS_ALIGNED(width, 2)) {
ARGBAttenuateRow = ARGBAttenuateRow_MMI;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_MSA;
}
}
#endif
for (y = 0; y < height; ++y) {
I444AlphaToARGBRow(src_y, src_u, src_v, src_a, dst_argb, yuvconstants,
width);
if (attenuate) {
ARGBAttenuateRow(dst_argb, dst_argb, width);
}
dst_argb += dst_stride_argb;
src_a += src_stride_a;
src_y += src_stride_y;
src_u += src_stride_u;
src_v += src_stride_v;
}
return 0;
}
// Convert I420 with Alpha to ARGB. // Convert I420 with Alpha to ARGB.
LIBYUV_API LIBYUV_API
int I420AlphaToARGB(const uint8_t* src_y, int I420AlphaToARGB(const uint8_t* src_y,
@ -1559,6 +1817,92 @@ int I420AlphaToABGR(const uint8_t* src_y,
width, height, attenuate); width, height, attenuate);
} }
// Convert I422 with Alpha to ARGB.
LIBYUV_API
int I422AlphaToARGB(const uint8_t* src_y,
int src_stride_y,
const uint8_t* src_u,
int src_stride_u,
const uint8_t* src_v,
int src_stride_v,
const uint8_t* src_a,
int src_stride_a,
uint8_t* dst_argb,
int dst_stride_argb,
int width,
int height,
int attenuate) {
return I422AlphaToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
src_stride_v, src_a, src_stride_a, dst_argb,
dst_stride_argb, &kYuvI601Constants, width,
height, attenuate);
}
// Convert I422 with Alpha to ABGR.
LIBYUV_API
int I422AlphaToABGR(const uint8_t* src_y,
int src_stride_y,
const uint8_t* src_u,
int src_stride_u,
const uint8_t* src_v,
int src_stride_v,
const uint8_t* src_a,
int src_stride_a,
uint8_t* dst_abgr,
int dst_stride_abgr,
int width,
int height,
int attenuate) {
return I422AlphaToARGBMatrix(
src_y, src_stride_y, src_v, src_stride_v, // Swap U and V
src_u, src_stride_u, src_a, src_stride_a, dst_abgr, dst_stride_abgr,
&kYvuI601Constants, // Use Yvu matrix
width, height, attenuate);
}
// Convert I444 with Alpha to ARGB.
LIBYUV_API
int I444AlphaToARGB(const uint8_t* src_y,
int src_stride_y,
const uint8_t* src_u,
int src_stride_u,
const uint8_t* src_v,
int src_stride_v,
const uint8_t* src_a,
int src_stride_a,
uint8_t* dst_argb,
int dst_stride_argb,
int width,
int height,
int attenuate) {
return I444AlphaToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
src_stride_v, src_a, src_stride_a, dst_argb,
dst_stride_argb, &kYuvI601Constants, width,
height, attenuate);
}
// Convert I444 with Alpha to ABGR.
LIBYUV_API
int I444AlphaToABGR(const uint8_t* src_y,
int src_stride_y,
const uint8_t* src_u,
int src_stride_u,
const uint8_t* src_v,
int src_stride_v,
const uint8_t* src_a,
int src_stride_a,
uint8_t* dst_abgr,
int dst_stride_abgr,
int width,
int height,
int attenuate) {
return I444AlphaToARGBMatrix(
src_y, src_stride_y, src_v, src_stride_v, // Swap U and V
src_u, src_stride_u, src_a, src_stride_a, dst_abgr, dst_stride_abgr,
&kYvuI601Constants, // Use Yvu matrix
width, height, attenuate);
}
// Convert I400 to ARGB with matrix. // Convert I400 to ARGB with matrix.
LIBYUV_API LIBYUV_API
int I400ToARGBMatrix(const uint8_t* src_y, int I400ToARGBMatrix(const uint8_t* src_y,

25
source/row_any.cc
View File

@ -52,18 +52,33 @@ extern "C" {
SS(r, DUVSHIFT) * BPP); \ SS(r, DUVSHIFT) * BPP); \
} }
#ifdef HAS_I444ALPHATOARGBROW_SSSE3
ANY41C(I444AlphaToARGBRow_Any_SSSE3, I444AlphaToARGBRow_SSSE3, 0, 0, 4, 7)
#endif
#ifdef HAS_I444ALPHATOARGBROW_AVX2
ANY41C(I444AlphaToARGBRow_Any_AVX2, I444AlphaToARGBRow_AVX2, 0, 0, 4, 15)
#endif
#ifdef HAS_I422ALPHATOARGBROW_SSSE3 #ifdef HAS_I422ALPHATOARGBROW_SSSE3
ANY41C(I422AlphaToARGBRow_Any_SSSE3, I422AlphaToARGBRow_SSSE3, 1, 0, 4, 7) ANY41C(I422AlphaToARGBRow_Any_SSSE3, I422AlphaToARGBRow_SSSE3, 1, 0, 4, 7)
#endif #endif
#ifdef HAS_I422ALPHATOARGBROW_AVX2 #ifdef HAS_I422ALPHATOARGBROW_AVX2
ANY41C(I422AlphaToARGBRow_Any_AVX2, I422AlphaToARGBRow_AVX2, 1, 0, 4, 15) ANY41C(I422AlphaToARGBRow_Any_AVX2, I422AlphaToARGBRow_AVX2, 1, 0, 4, 15)
#endif #endif
#ifdef HAS_I444ALPHATOARGBROW_NEON
ANY41C(I444AlphaToARGBRow_Any_NEON, I444AlphaToARGBRow_NEON, 0, 0, 4, 7)
#endif
#ifdef HAS_I422ALPHATOARGBROW_NEON #ifdef HAS_I422ALPHATOARGBROW_NEON
ANY41C(I422AlphaToARGBRow_Any_NEON, I422AlphaToARGBRow_NEON, 1, 0, 4, 7) ANY41C(I422AlphaToARGBRow_Any_NEON, I422AlphaToARGBRow_NEON, 1, 0, 4, 7)
#endif #endif
#ifdef HAS_I444ALPHATOARGBROW_MSA
ANY41C(I444AlphaToARGBRow_Any_MSA, I444AlphaToARGBRow_MSA, 0, 0, 4, 7)
#endif
#ifdef HAS_I422ALPHATOARGBROW_MSA #ifdef HAS_I422ALPHATOARGBROW_MSA
ANY41C(I422AlphaToARGBRow_Any_MSA, I422AlphaToARGBRow_MSA, 1, 0, 4, 7) ANY41C(I422AlphaToARGBRow_Any_MSA, I422AlphaToARGBRow_MSA, 1, 0, 4, 7)
#endif #endif
#ifdef HAS_I444ALPHATOARGBROW_MMI
ANY41C(I444AlphaToARGBRow_Any_MMI, I444AlphaToARGBRow_MMI, 0, 0, 4, 7)
#endif
#ifdef HAS_I422ALPHATOARGBROW_MMI #ifdef HAS_I422ALPHATOARGBROW_MMI
ANY41C(I422AlphaToARGBRow_Any_MMI, I422AlphaToARGBRow_MMI, 1, 0, 4, 7) ANY41C(I422AlphaToARGBRow_Any_MMI, I422AlphaToARGBRow_MMI, 1, 0, 4, 7)
#endif #endif
@ -164,6 +179,11 @@ ANY31(BlendPlaneRow_Any_MMI, BlendPlaneRow_MMI, 0, 0, 1, 7)
#ifdef HAS_I422TOARGBROW_SSSE3 #ifdef HAS_I422TOARGBROW_SSSE3
ANY31C(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_SSSE3, 1, 0, 4, 7) ANY31C(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_SSSE3, 1, 0, 4, 7)
ANY31C(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_SSSE3, 1, 0, 4, 7)
ANY31C(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, 1, 0, 2, 7)
ANY31C(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, 1, 0, 2, 7)
ANY31C(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, 1, 0, 2, 7)
ANY31C(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, 1, 0, 3, 15)
#endif #endif
#ifdef HAS_I422TOAR30ROW_SSSE3 #ifdef HAS_I422TOAR30ROW_SSSE3
ANY31C(I422ToAR30Row_Any_SSSE3, I422ToAR30Row_SSSE3, 1, 0, 4, 7) ANY31C(I422ToAR30Row_Any_SSSE3, I422ToAR30Row_SSSE3, 1, 0, 4, 7)
@ -173,11 +193,6 @@ ANY31C(I422ToAR30Row_Any_AVX2, I422ToAR30Row_AVX2, 1, 0, 4, 15)
#endif #endif
#ifdef HAS_I444TOARGBROW_SSSE3 #ifdef HAS_I444TOARGBROW_SSSE3
ANY31C(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_SSSE3, 0, 0, 4, 7) ANY31C(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_SSSE3, 0, 0, 4, 7)
ANY31C(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_SSSE3, 1, 0, 4, 7)
ANY31C(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, 1, 0, 2, 7)
ANY31C(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, 1, 0, 2, 7)
ANY31C(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, 1, 0, 2, 7)
ANY31C(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, 1, 0, 3, 15)
#endif // HAS_I444TOARGBROW_SSSE3 #endif // HAS_I444TOARGBROW_SSSE3
#ifdef HAS_I422TORGB24ROW_AVX2 #ifdef HAS_I422TORGB24ROW_AVX2
ANY31C(I422ToRGB24Row_Any_AVX2, I422ToRGB24Row_AVX2, 1, 0, 3, 31) ANY31C(I422ToRGB24Row_Any_AVX2, I422ToRGB24Row_AVX2, 1, 0, 3, 31)

20
source/row_common.cc
View File

@ -2016,6 +2016,26 @@ void I422ToAR30Row_C(const uint8_t* src_y,
} }
} }
void I444AlphaToARGBRow_C(const uint8_t* src_y,
const uint8_t* src_u,
const uint8_t* src_v,
const uint8_t* src_a,
uint8_t* rgb_buf,
const struct YuvConstants* yuvconstants,
int width) {
int x;
for (x = 0; x < width; ++x) {
YuvPixel(src_y[0], src_u[0], src_v[0], rgb_buf + 0, rgb_buf + 1,
rgb_buf + 2, yuvconstants);
rgb_buf[3] = src_a[0];
src_y += 1;
src_u += 1;
src_v += 1;
src_a += 1;
rgb_buf += 4; // Advance 1 pixel.
}
}
void I422AlphaToARGBRow_C(const uint8_t* src_y, void I422AlphaToARGBRow_C(const uint8_t* src_y,
const uint8_t* src_u, const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,

107
source/row_gcc.cc
View File

@ -1964,6 +1964,18 @@ void RGBAToUVRow_SSSE3(const uint8_t* src_rgba0,
"movq (%[a_buf]),%%xmm5 \n" \ "movq (%[a_buf]),%%xmm5 \n" \
"lea 0x8(%[a_buf]),%[a_buf] \n" "lea 0x8(%[a_buf]),%[a_buf] \n"
// Read 8 UV from 444. With 8 Alpha.
#define READYUVA444 \
"movq (%[u_buf]),%%xmm0 \n" \
"movq 0x00(%[u_buf],%[v_buf],1),%%xmm1 \n" \
"lea 0x8(%[u_buf]),%[u_buf] \n" \
"punpcklbw %%xmm1,%%xmm0 \n" \
"movq (%[y_buf]),%%xmm4 \n" \
"punpcklbw %%xmm4,%%xmm4 \n" \
"lea 0x8(%[y_buf]),%[y_buf] \n" \
"movq (%[a_buf]),%%xmm5 \n" \
"lea 0x8(%[a_buf]),%[a_buf] \n"
// Read 4 UV from NV12, upsample to 8 UV // Read 4 UV from NV12, upsample to 8 UV
#define READNV12 \ #define READNV12 \
"movq (%[uv_buf]),%%xmm0 \n" \ "movq (%[uv_buf]),%%xmm0 \n" \
@ -2138,6 +2150,44 @@ void OMITFP I444ToARGBRow_SSSE3(const uint8_t* y_buf,
); );
} }
#ifdef HAS_I444ALPHATOARGBROW_SSSE3
void OMITFP I444AlphaToARGBRow_SSSE3(const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
const uint8_t* a_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
// clang-format off
asm volatile (
YUVTORGB_SETUP(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
LABELALIGN
"1: \n"
READYUVA444
YUVTORGB(yuvconstants)
STOREARGB
"subl $0x8,%[width] \n"
"jg 1b \n"
: [y_buf]"+r"(y_buf), // %[y_buf]
[u_buf]"+r"(u_buf), // %[u_buf]
[v_buf]"+r"(v_buf), // %[v_buf]
[a_buf]"+r"(a_buf), // %[a_buf]
[dst_argb]"+r"(dst_argb), // %[dst_argb]
#if defined(__i386__)
[width]"+m"(width) // %[width]
#else
[width]"+rm"(width) // %[width]
#endif
: [yuvconstants]"r"(yuvconstants) // %[yuvconstants]
: "memory", "cc", YUVTORGB_REGS
"xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
);
// clang-format on
}
#endif // HAS_I444ALPHATOARGBROW_SSSE3
void OMITFP I422ToRGB24Row_SSSE3(const uint8_t* y_buf, void OMITFP I422ToRGB24Row_SSSE3(const uint8_t* y_buf,
const uint8_t* u_buf, const uint8_t* u_buf,
const uint8_t* v_buf, const uint8_t* v_buf,
@ -2537,6 +2587,22 @@ void OMITFP I422ToRGBARow_SSSE3(const uint8_t* y_buf,
"vpsllw $0x6,%%ymm4,%%ymm4 \n" \ "vpsllw $0x6,%%ymm4,%%ymm4 \n" \
"lea 0x20(%[y_buf]),%[y_buf] \n" "lea 0x20(%[y_buf]),%[y_buf] \n"
// Read 16 UV from 444. With 16 Alpha.
#define READYUVA444_AVX2 \
"vmovdqu (%[u_buf]),%%xmm0 \n" \
"vmovdqu 0x00(%[u_buf],%[v_buf],1),%%xmm1 \n" \
"lea 0x10(%[u_buf]),%[u_buf] \n" \
"vpermq $0xd8,%%ymm0,%%ymm0 \n" \
"vpermq $0xd8,%%ymm1,%%ymm1 \n" \
"vpunpcklbw %%ymm1,%%ymm0,%%ymm0 \n" \
"vmovdqu (%[y_buf]),%%xmm4 \n" \
"vpermq $0xd8,%%ymm4,%%ymm4 \n" \
"vpunpcklbw %%ymm4,%%ymm4,%%ymm4 \n" \
"lea 0x10(%[y_buf]),%[y_buf] \n" \
"vmovdqu (%[a_buf]),%%xmm5 \n" \
"vpermq $0xd8,%%ymm5,%%ymm5 \n" \
"lea 0x10(%[a_buf]),%[a_buf] \n"
// Read 8 UV from 422, upsample to 16 UV. With 16 Alpha. // Read 8 UV from 422, upsample to 16 UV. With 16 Alpha.
#define READYUVA422_AVX2 \ #define READYUVA422_AVX2 \
"vmovq (%[u_buf]),%%xmm0 \n" \ "vmovq (%[u_buf]),%%xmm0 \n" \
@ -2868,6 +2934,47 @@ void OMITFP I210ToAR30Row_AVX2(const uint16_t* y_buf,
} }
#endif // HAS_I210TOAR30ROW_AVX2 #endif // HAS_I210TOAR30ROW_AVX2
#if defined(HAS_I444ALPHATOARGBROW_AVX2)
// 16 pixels
// 16 UV values with 16 Y and 16 A producing 16 ARGB.
void OMITFP I444AlphaToARGBRow_AVX2(const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
const uint8_t* a_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
// clang-format off
asm volatile (
YUVTORGB_SETUP_AVX2(yuvconstants)
"sub %[u_buf],%[v_buf] \n"
LABELALIGN
"1: \n"
READYUVA444_AVX2
YUVTORGB_AVX2(yuvconstants)
STOREARGB_AVX2
"subl $0x10,%[width] \n"
"jg 1b \n"
"vzeroupper \n"
: [y_buf]"+r"(y_buf), // %[y_buf]
[u_buf]"+r"(u_buf), // %[u_buf]
[v_buf]"+r"(v_buf), // %[v_buf]
[a_buf]"+r"(a_buf), // %[a_buf]
[dst_argb]"+r"(dst_argb), // %[dst_argb]
#if defined(__i386__)
[width]"+m"(width) // %[width]
#else
[width]"+rm"(width) // %[width]
#endif
: [yuvconstants]"r"(yuvconstants) // %[yuvconstants]
: "memory", "cc", YUVTORGB_REGS_AVX2
"xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
);
// clang-format on
}
#endif // HAS_I444ALPHATOARGBROW_AVX2
#if defined(HAS_I422ALPHATOARGBROW_AVX2) #if defined(HAS_I422ALPHATOARGBROW_AVX2)
// 16 pixels // 16 pixels
// 8 UV values upsampled to 16 UV, mixed with 16 Y and 16 A producing 16 ARGB. // 8 UV values upsampled to 16 UV, mixed with 16 Y and 16 A producing 16 ARGB.

28
source/row_neon.cc
View File

@ -158,6 +158,34 @@ void I422ToARGBRow_NEON(const uint8_t* src_y,
"q12", "q13", "q14", "q15"); "q12", "q13", "q14", "q15");
} }
void I444AlphaToARGBRow_NEON(const uint8_t* src_y,
const uint8_t* src_u,
const uint8_t* src_v,
const uint8_t* src_a,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm volatile(
YUVTORGB_SETUP
"1: \n" READYUV444 YUVTORGB
"subs %5, %5, #8 \n"
"vld1.8 {d23}, [%3]! \n"
"vst4.8 {d20, d21, d22, d23}, [%4]! \n"
"bgt 1b \n"
: "+r"(src_y), // %0
"+r"(src_u), // %1
"+r"(src_v), // %2
"+r"(src_a), // %3
"+r"(dst_argb), // %4
"+r"(width) // %5
: [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 I422AlphaToARGBRow_NEON(const uint8_t* src_y, void I422AlphaToARGBRow_NEON(const uint8_t* src_y,
const uint8_t* src_u, const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,

35
source/row_neon64.cc
View File

@ -171,6 +171,41 @@ void I422ToARGBRow_NEON(const uint8_t* src_y,
); );
} }
void I444AlphaToARGBRow_NEON(const uint8_t* src_y,
const uint8_t* src_u,
const uint8_t* src_v,
const uint8_t* src_a,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
asm volatile (
YUVTORGB_SETUP
"1: \n"
READYUV444
"prfm pldl1keep, [%0, 448] \n"
YUVTORGB(v22, v21, v20)
"ld1 {v23.8b}, [%3], #8 \n"
"prfm pldl1keep, [%1, 128] \n"
"prfm pldl1keep, [%2, 128] \n"
"prfm pldl1keep, [%3, 448] \n"
"subs %w5, %w5, #8 \n"
"st4 {v20.8b,v21.8b,v22.8b,v23.8b}, [%4], #32 \n"
"b.gt 1b \n"
: "+r"(src_y), // %0
"+r"(src_u), // %1
"+r"(src_v), // %2
"+r"(src_a), // %3
"+r"(dst_argb), // %4
"+r"(width) // %5
: [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"
);
}
void I422AlphaToARGBRow_NEON(const uint8_t* src_y, void I422AlphaToARGBRow_NEON(const uint8_t* src_y,
const uint8_t* src_u, const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,

172
source/row_win.cc
View File

@ -27,6 +27,28 @@ extern "C" {
// 64 bit // 64 bit
#if defined(_M_X64) #if defined(_M_X64)
// Read 8 UV from 444
#define READYUV444 \
xmm0 = _mm_loadl_epi64((__m128i*)u_buf); \
xmm1 = _mm_loadl_epi64((__m128i*)(u_buf + offset)); \
xmm0 = _mm_unpacklo_epi8(xmm0, xmm1); \
u_buf += 8; \
xmm4 = _mm_loadl_epi64((__m128i*)y_buf); \
xmm4 = _mm_unpacklo_epi8(xmm4, xmm4); \
y_buf += 8;
// Read 8 UV from 444, With 8 Alpha.
#define READYUVA444 \
xmm0 = _mm_loadl_epi64((__m128i*)u_buf); \
xmm1 = _mm_loadl_epi64((__m128i*)(u_buf + offset)); \
xmm0 = _mm_unpacklo_epi8(xmm0, xmm1); \
u_buf += 8; \
xmm4 = _mm_loadl_epi64((__m128i*)y_buf); \
xmm4 = _mm_unpacklo_epi8(xmm4, xmm4); \
y_buf += 8; \
xmm5 = _mm_loadl_epi64((__m128i*)a_buf); \
a_buf += 8;
// Read 4 UV from 422, upsample to 8 UV. // Read 4 UV from 422, upsample to 8 UV.
#define READYUV422 \ #define READYUV422 \
xmm0 = _mm_cvtsi32_si128(*(uint32_t*)u_buf); \ xmm0 = _mm_cvtsi32_si128(*(uint32_t*)u_buf); \
@ -121,6 +143,44 @@ void I422AlphaToARGBRow_SSSE3(const uint8_t* y_buf,
} }
#endif #endif
#if defined(HAS_I444TOARGBROW_SSSE3)
void I444ToARGBRow_SSSE3(const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
__m128i xmm0, xmm1, xmm2, xmm4;
const __m128i xmm5 = _mm_set1_epi8(-1);
const ptrdiff_t offset = (uint8_t*)v_buf - (uint8_t*)u_buf;
while (width > 0) {
READYUV444
YUVTORGB(yuvconstants)
STOREARGB
width -= 8;
}
}
#endif
#if defined(HAS_I444ALPHATOARGBROW_SSSE3)
void I444AlphaToARGBRow_SSSE3(const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
const uint8_t* a_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
__m128i xmm0, xmm1, xmm2, xmm4, xmm5;
const ptrdiff_t offset = (uint8_t*)v_buf - (uint8_t*)u_buf;
while (width > 0) {
READYUVA444
YUVTORGB(yuvconstants)
STOREARGB
width -= 8;
}
}
#endif
// 32 bit // 32 bit
#else // defined(_M_X64) #else // defined(_M_X64)
#ifdef HAS_ARGBTOYROW_SSSE3 #ifdef HAS_ARGBTOYROW_SSSE3
@ -1938,6 +1998,23 @@ __declspec(naked) void RGBAToUVRow_SSSE3(const uint8_t* src_argb0,
__asm vpunpcklbw ymm4, ymm4, ymm4 \ __asm vpunpcklbw ymm4, ymm4, ymm4 \
__asm lea eax, [eax + 16]} __asm lea eax, [eax + 16]}
// Read 16 UV from 444. With 16 Alpha.
#define READYUVA444_AVX2 \
__asm { \
__asm vmovdqu xmm0, [esi] /* U */ \
__asm vmovdqu xmm1, [esi + edi] /* V */ \
__asm lea esi, [esi + 16] \
__asm vpermq ymm0, ymm0, 0xd8 \
__asm vpermq ymm1, ymm1, 0xd8 \
__asm vpunpcklbw ymm0, ymm0, ymm1 /* UV */ \
__asm vmovdqu xmm4, [eax] /* Y */ \
__asm vpermq ymm4, ymm4, 0xd8 \
__asm vpunpcklbw ymm4, ymm4, ymm4 \
__asm lea eax, [eax + 16] \
__asm vmovdqu xmm5, [ebp] /* A */ \
__asm vpermq ymm5, ymm5, 0xd8 \
__asm lea ebp, [ebp + 16]}
// Read 8 UV from 422, upsample to 16 UV. // Read 8 UV from 422, upsample to 16 UV.
#define READYUV422_AVX2 \ #define READYUV422_AVX2 \
__asm { \ __asm { \
@ -2183,6 +2260,48 @@ __declspec(naked) void I444ToARGBRow_AVX2(
} }
#endif // HAS_I444TOARGBROW_AVX2 #endif // HAS_I444TOARGBROW_AVX2
#ifdef HAS_I444ALPHATOARGBROW_AVX2
// 16 pixels
// 16 UV values with 16 Y producing 16 ARGB (64 bytes).
__declspec(naked) void I444AlphaToARGBRow_AVX2(
const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
const uint8_t* a_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
__asm {
push esi
push edi
push ebx
push ebp
mov eax, [esp + 16 + 4] // Y
mov esi, [esp + 16 + 8] // U
mov edi, [esp + 16 + 12] // V
mov ebp, [esp + 16 + 16] // A
mov edx, [esp + 16 + 20] // argb
mov ebx, [esp + 16 + 24] // yuvconstants
mov ecx, [esp + 16 + 28] // width
sub edi, esi
convertloop:
READYUVA444_AVX2
YUVTORGB_AVX2(ebx)
STOREARGB_AVX2
sub ecx, 16
jg convertloop
pop ebp
pop ebx
pop edi
pop esi
vzeroupper
ret
}
}
#endif // HAS_I444AlphaTOARGBROW_AVX2
#ifdef HAS_NV12TOARGBROW_AVX2 #ifdef HAS_NV12TOARGBROW_AVX2
// 16 pixels. // 16 pixels.
// 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 ARGB (64 bytes). // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 ARGB (64 bytes).
@ -2370,6 +2489,19 @@ __declspec(naked) void I422ToRGBARow_AVX2(
__asm punpcklbw xmm4, xmm4 \ __asm punpcklbw xmm4, xmm4 \
__asm lea eax, [eax + 8]} __asm lea eax, [eax + 8]}
// Read 4 UV from 444. With 8 Alpha.
#define READYUVA444 \
__asm { \
__asm movq xmm0, qword ptr [esi] /* U */ \
__asm movq xmm1, qword ptr [esi + edi] /* V */ \
__asm lea esi, [esi + 8] \
__asm punpcklbw xmm0, xmm1 /* UV */ \
__asm movq xmm4, qword ptr [eax] \
__asm punpcklbw xmm4, xmm4 \
__asm lea eax, [eax + 8] \
__asm movq xmm5, qword ptr [ebp] /* A */ \
__asm lea ebp, [ebp + 8]}
// Read 4 UV from 422, upsample to 8 UV. // Read 4 UV from 422, upsample to 8 UV.
#define READYUV422 \ #define READYUV422 \
__asm { \ __asm { \
@ -2585,6 +2717,46 @@ __declspec(naked) void I444ToARGBRow_SSSE3(
} }
} }
// 8 pixels.
// 8 UV values, mixed with 8 Y and 8A producing 8 ARGB (32 bytes).
__declspec(naked) void I444AlphaToARGBRow_SSSE3(
const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
const uint8_t* a_buf,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width) {
__asm {
push esi
push edi
push ebx
push ebp
mov eax, [esp + 16 + 4] // Y
mov esi, [esp + 16 + 8] // U
mov edi, [esp + 16 + 12] // V
mov ebp, [esp + 16 + 16] // A
mov edx, [esp + 16 + 20] // argb
mov ebx, [esp + 16 + 24] // yuvconstants
mov ecx, [esp + 16 + 28] // width
sub edi, esi
convertloop:
READYUVA444
YUVTORGB(ebx)
STOREARGB
sub ecx, 8
jg convertloop
pop ebp
pop ebx
pop edi
pop esi
ret
}
}
// 8 pixels. // 8 pixels.
// 4 UV values upsampled to 8 UV, mixed with 8 Y producing 8 RGB24 (24 bytes). // 4 UV values upsampled to 8 UV, mixed with 8 Y producing 8 RGB24 (24 bytes).
__declspec(naked) void I422ToRGB24Row_SSSE3( __declspec(naked) void I422ToRGB24Row_SSSE3(

68
unit_test/convert_test.cc
View File

@ -738,6 +738,42 @@ TESTPLANARTOB(H420, 2, 2, AR30, 4, 4, 1)
#define U420AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ #define U420AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \
I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \ I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \
l, m) l, m)
#define J422AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \
I422AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \
l, m)
#define J422AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \
I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \
l, m)
#define H422AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \
I422AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \
l, m)
#define H422AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \
I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \
l, m)
#define U422AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \
I422AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \
l, m)
#define U422AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \
I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \
l, m)
#define J444AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \
I444AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \
l, m)
#define J444AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \
I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \
l, m)
#define H444AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \
I444AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \
l, m)
#define H444AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \
I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \
l, m)
#define U444AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \
I444AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \
l, m)
#define U444AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \
I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \
l, m)
TESTQPLANARTOB(I420Alpha, 2, 2, ARGB, 4, 4, 1) TESTQPLANARTOB(I420Alpha, 2, 2, ARGB, 4, 4, 1)
TESTQPLANARTOB(I420Alpha, 2, 2, ABGR, 4, 4, 1) TESTQPLANARTOB(I420Alpha, 2, 2, ABGR, 4, 4, 1)
@ -747,6 +783,22 @@ TESTQPLANARTOB(H420Alpha, 2, 2, ARGB, 4, 4, 1)
TESTQPLANARTOB(H420Alpha, 2, 2, ABGR, 4, 4, 1) TESTQPLANARTOB(H420Alpha, 2, 2, ABGR, 4, 4, 1)
TESTQPLANARTOB(U420Alpha, 2, 2, ARGB, 4, 4, 1) TESTQPLANARTOB(U420Alpha, 2, 2, ARGB, 4, 4, 1)
TESTQPLANARTOB(U420Alpha, 2, 2, ABGR, 4, 4, 1) TESTQPLANARTOB(U420Alpha, 2, 2, ABGR, 4, 4, 1)
TESTQPLANARTOB(I422Alpha, 2, 1, ARGB, 4, 4, 1)
TESTQPLANARTOB(I422Alpha, 2, 1, ABGR, 4, 4, 1)
TESTQPLANARTOB(J422Alpha, 2, 1, ARGB, 4, 4, 1)
TESTQPLANARTOB(J422Alpha, 2, 1, ABGR, 4, 4, 1)
TESTQPLANARTOB(H422Alpha, 2, 1, ARGB, 4, 4, 1)
TESTQPLANARTOB(H422Alpha, 2, 1, ABGR, 4, 4, 1)
TESTQPLANARTOB(U422Alpha, 2, 1, ARGB, 4, 4, 1)
TESTQPLANARTOB(U422Alpha, 2, 1, ABGR, 4, 4, 1)
TESTQPLANARTOB(I444Alpha, 1, 1, ARGB, 4, 4, 1)
TESTQPLANARTOB(I444Alpha, 1, 1, ABGR, 4, 4, 1)
TESTQPLANARTOB(J444Alpha, 1, 1, ARGB, 4, 4, 1)
TESTQPLANARTOB(J444Alpha, 1, 1, ABGR, 4, 4, 1)
TESTQPLANARTOB(H444Alpha, 1, 1, ARGB, 4, 4, 1)
TESTQPLANARTOB(H444Alpha, 1, 1, ABGR, 4, 4, 1)
TESTQPLANARTOB(U444Alpha, 1, 1, ARGB, 4, 4, 1)
TESTQPLANARTOB(U444Alpha, 1, 1, ABGR, 4, 4, 1)
#define TESTBIPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, FMT_C, \ #define TESTBIPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, FMT_C, \
BPP_B, W1280, N, NEG, OFF) \ BPP_B, W1280, N, NEG, OFF) \
@ -2762,6 +2814,22 @@ TESTQPLANARTOE(H420Alpha, 2, 2, ARGB, 1, 4, ABGR, 4)
TESTQPLANARTOE(H420Alpha, 2, 2, ABGR, 1, 4, ARGB, 4) TESTQPLANARTOE(H420Alpha, 2, 2, ABGR, 1, 4, ARGB, 4)
TESTQPLANARTOE(U420Alpha, 2, 2, ARGB, 1, 4, ABGR, 4) TESTQPLANARTOE(U420Alpha, 2, 2, ARGB, 1, 4, ABGR, 4)
TESTQPLANARTOE(U420Alpha, 2, 2, ABGR, 1, 4, ARGB, 4) TESTQPLANARTOE(U420Alpha, 2, 2, ABGR, 1, 4, ARGB, 4)
TESTQPLANARTOE(I422Alpha, 2, 1, ARGB, 1, 4, ABGR, 4)
TESTQPLANARTOE(I422Alpha, 2, 1, ABGR, 1, 4, ARGB, 4)
TESTQPLANARTOE(J422Alpha, 2, 1, ARGB, 1, 4, ABGR, 4)
TESTQPLANARTOE(J422Alpha, 2, 1, ABGR, 1, 4, ARGB, 4)
TESTQPLANARTOE(H422Alpha, 2, 1, ARGB, 1, 4, ABGR, 4)
TESTQPLANARTOE(H422Alpha, 2, 1, ABGR, 1, 4, ARGB, 4)
TESTQPLANARTOE(U422Alpha, 2, 1, ARGB, 1, 4, ABGR, 4)
TESTQPLANARTOE(U422Alpha, 2, 1, ABGR, 1, 4, ARGB, 4)
TESTQPLANARTOE(I444Alpha, 1, 1, ARGB, 1, 4, ABGR, 4)
TESTQPLANARTOE(I444Alpha, 1, 1, ABGR, 1, 4, ARGB, 4)
TESTQPLANARTOE(J444Alpha, 1, 1, ARGB, 1, 4, ABGR, 4)
TESTQPLANARTOE(J444Alpha, 1, 1, ABGR, 1, 4, ARGB, 4)
TESTQPLANARTOE(H444Alpha, 1, 1, ARGB, 1, 4, ABGR, 4)
TESTQPLANARTOE(H444Alpha, 1, 1, ABGR, 1, 4, ARGB, 4)
TESTQPLANARTOE(U444Alpha, 1, 1, ARGB, 1, 4, ABGR, 4)
TESTQPLANARTOE(U444Alpha, 1, 1, ABGR, 1, 4, ARGB, 4)
#define TESTPLANETOEI(FMT_A, SUB_A, BPP_A, FMT_B, SUB_B, BPP_B, W1280, N, NEG, \ #define TESTPLANETOEI(FMT_A, SUB_A, BPP_A, FMT_B, SUB_B, BPP_B, W1280, N, NEG, \
OFF, FMT_C, BPP_C) \ OFF, FMT_C, BPP_C) \