Skip to content

L
libyuv
Commit ce4c2fad authored by Frank Barchard's avatar Frank Barchard
Browse files
Options

Raw 24 bit RGB to RGB24 (bgr) 

Add unittests that do 1 step conversion vs 2 step conversion.

Tests end swapping versions match direct conversions.

R=harryjin@google.com
BUG=libyuv:518

Review URL: https://codereview.chromium.org/1419103007 .
parent 87926cec
Hide whitespace changes
Inline Side-by-side
Showing with 220 additions and 7 deletions
include/libyuv/planar_functions.h
View file @ ce4c2fad
......@@ -170,6 +170,14 @@ int I422ToRGBA(const uint8* src_y, int src_stride_y,
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,
......
include/libyuv/row.h
View file @ ce4c2fad
......@@ -130,6 +130,7 @@ extern "C" {
#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
......@@ -301,6 +302,7 @@ extern "C" {
#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
......@@ -908,6 +910,7 @@ void ARGBShuffleRow_Any_NEON(const uint8* src_argb, uint8* dst_argb,
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);
......@@ -921,6 +924,7 @@ void ARGB4444ToARGBRow_AVX2(const uint8* src_argb4444, uint8* dst_argb,
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);
......@@ -928,12 +932,14 @@ 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);
......@@ -951,6 +957,7 @@ void ARGB4444ToARGBRow_Any_AVX2(const uint8* src_argb4444, uint8* dst_argb,
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,
......
source/planar_functions.cc
View file @ ce4c2fad
......@@ -920,6 +920,56 @@ int NV12ToRGB565(const uint8* src_y, int src_stride_y,
return 0;
}
// Convert RAW to RGB24.
LIBYUV_API
int RAWToRGB24(const uint8* src_raw, int src_stride_raw,
uint8* dst_rgb24, int dst_stride_rgb24,
int width, int height) {
int y;
void (*RAWToRGB24Row)(const uint8* src_rgb, uint8* dst_rgb24, int width) =
RAWToRGB24Row_C;
if (!src_raw || !dst_rgb24 ||
width <= 0 || height == 0) {
return -1;
}
// Negative height means invert the image.
if (height < 0) {
height = -height;
src_raw = src_raw + (height - 1) * src_stride_raw;
src_stride_raw = -src_stride_raw;
}
// Coalesce rows.
if (src_stride_raw == width * 3 &&
dst_stride_rgb24 == width * 3) {
width *= height;
height = 1;
src_stride_raw = dst_stride_rgb24 = 0;
}
#if defined(HAS_RAWTORGB24ROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
RAWToRGB24Row = RAWToRGB24Row_Any_SSSE3;
if (IS_ALIGNED(width, 8)) {
RAWToRGB24Row = RAWToRGB24Row_SSSE3;
}
}
#endif
#if defined(HAS_RAWTORGB24ROW_NEON)
if (TestCpuFlag(kCpuHasNEON)) {
RAWToRGB24Row = RAWToRGB24Row_Any_NEON;
if (IS_ALIGNED(width, 8)) {
RAWToRGB24Row = RAWToRGB24Row_NEON;
}
}
#endif
for (y = 0; y < height; ++y) {
RAWToRGB24Row(src_raw, dst_rgb24, width);
src_raw += src_stride_raw;
dst_rgb24 += dst_stride_rgb24;
}
return 0;
}
LIBYUV_API
void SetPlane(uint8* dst_y, int dst_stride_y,
int width, int height,
......
source/row_any.cc
View file @ ce4c2fad
......@@ -338,6 +338,9 @@ ANY11(RGB565ToARGBRow_Any_SSE2, RGB565ToARGBRow_SSE2, 0, 2, 4, 7)
ANY11(ARGB1555ToARGBRow_Any_SSE2, ARGB1555ToARGBRow_SSE2, 0, 2, 4, 7)
ANY11(ARGB4444ToARGBRow_Any_SSE2, ARGB4444ToARGBRow_SSE2, 0, 2, 4, 7)
#endif
#if defined(HAS_RAWTORGB24ROW_SSSE3)
ANY11(RAWToRGB24Row_Any_SSSE3, RAWToRGB24Row_SSSE3, 0, 3, 3, 7)
#endif
#if defined(HAS_RGB565TOARGBROW_AVX2)
ANY11(RGB565ToARGBRow_Any_AVX2, RGB565ToARGBRow_AVX2, 0, 2, 4, 15)
#endif
......@@ -356,6 +359,9 @@ ANY11(ARGBToARGB4444Row_Any_NEON, ARGBToARGB4444Row_NEON, 0, 4, 2, 7)
ANY11(J400ToARGBRow_Any_NEON, J400ToARGBRow_NEON, 0, 1, 4, 7)
ANY11(I400ToARGBRow_Any_NEON, I400ToARGBRow_NEON, 0, 1, 4, 7)
#endif
#if defined(HAS_RAWTORGB24ROW_NEON)
ANY11(RAWToRGB24Row_Any_NEON, RAWToRGB24Row_NEON, 0, 3, 3, 7)
#endif
#ifdef HAS_ARGBTOYROW_AVX2
ANY11(ARGBToYRow_Any_AVX2, ARGBToYRow_AVX2, 0, 4, 1, 31)
#endif
......
source/row_common.cc
View file @ ce4c2fad
......@@ -100,6 +100,20 @@ void RAWToARGBRow_C(const uint8* src_raw, uint8* dst_argb, int width) {
}
}
void RAWToRGB24Row_C(const uint8* src_raw, uint8* dst_rgb24, int width) {
int x;
for (x = 0; x < width; ++x) {
uint8 r = src_raw[0];
uint8 g = src_raw[1];
uint8 b = src_raw[2];
dst_rgb24[0] = b;
dst_rgb24[1] = g;
dst_rgb24[2] = r;
dst_rgb24 += 3;
src_raw += 3;
}
}
void RGB565ToARGBRow_C(const uint8* src_rgb565, uint8* dst_argb, int width) {
int x;
for (x = 0; x < width; ++x) {
......
source/row_gcc.cc
View file @ ce4c2fad
......@@ -121,6 +121,24 @@ static uvec8 kShuffleMaskRAWToARGB = {
2u, 1u, 0u, 12u, 5u, 4u, 3u, 13u, 8u, 7u, 6u, 14u, 11u, 10u, 9u, 15u
};
// Shuffle table for converting RAW to RGB24. First 8.
static const uvec8 kShuffleMaskRAWToRGB24_0 = {
2u, 1u, 0u, 5u, 4u, 3u, 8u, 7u,
128u, 128u, 128u, 128u, 128u, 128u, 128u, 128u
};
// Shuffle table for converting RAW to RGB24. Middle 8.
static const uvec8 kShuffleMaskRAWToRGB24_1 = {
2u, 7u, 6u, 5u, 10u, 9u, 8u, 13u,
128u, 128u, 128u, 128u, 128u, 128u, 128u, 128u
};
// Shuffle table for converting RAW to RGB24. Last 8.
static const uvec8 kShuffleMaskRAWToRGB24_2 = {
8u, 7u, 12u, 11u, 10u, 15u, 14u, 13u,
128u, 128u, 128u, 128u, 128u, 128u, 128u, 128u
};
// Shuffle table for converting ARGB to RGB24.
static uvec8 kShuffleMaskARGBToRGB24 = {
0u, 1u, 2u, 4u, 5u, 6u, 8u, 9u, 10u, 12u, 13u, 14u, 128u, 128u, 128u, 128u
......@@ -272,6 +290,36 @@ void RAWToARGBRow_SSSE3(const uint8* src_raw, uint8* dst_argb, int width) {
);
}
void RAWToRGB24Row_SSSE3(const uint8* src_raw, uint8* dst_rgb24, int width) {
asm volatile (
"movdqa %3,%%xmm3 \n"
"movdqa %4,%%xmm4 \n"
"movdqa %5,%%xmm5 \n"
LABELALIGN
"1: \n"
"movdqu " MEMACCESS(0) ",%%xmm0 \n"
"movdqu " MEMACCESS2(0x4,0) ",%%xmm1 \n"
"movdqu " MEMACCESS2(0x8,0) ",%%xmm2 \n"
"lea " MEMLEA(0x18,0) ",%0 \n"
"pshufb %%xmm3,%%xmm0 \n"
"pshufb %%xmm4,%%xmm1 \n"
"pshufb %%xmm5,%%xmm2 \n"
"movq %%xmm0," MEMACCESS(1) " \n"
"movq %%xmm1," MEMACCESS2(0x8,1) " \n"
"movq %%xmm2," MEMACCESS2(0x10,1) " \n"
"lea " MEMLEA(0x18,1) ",%1 \n"
"sub $0x8,%2 \n"
"jg 1b \n"
: "+r"(src_raw), // %0
"+r"(dst_rgb24), // %1
"+r"(width) // %2
: "m"(kShuffleMaskRAWToRGB24_0), // %3
"m"(kShuffleMaskRAWToRGB24_1), // %4
"m"(kShuffleMaskRAWToRGB24_2) // %5
: "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
);
}
void RGB565ToARGBRow_SSE2(const uint8* src, uint8* dst, int width) {
asm volatile (
"mov $0x1080108,%%eax \n"
......
source/row_neon.cc
View file @ ce4c2fad
......@@ -798,12 +798,30 @@ void RAWToARGBRow_NEON(const uint8* src_raw, uint8* dst_argb, int width) {
"bgt 1b \n"
: "+r"(src_raw), // %0
"+r"(dst_argb), // %1
"+r"(width) // %2
"+r"(width) // %2
:
: "cc", "memory", "d1", "d2", "d3", "d4" // Clobber List
);
}
void RAWToRGB24Row_NEON(const uint8* src_raw, uint8* dst_rgb24, int width) {
asm volatile (
"1: \n"
MEMACCESS(0)
"vld3.8 {d1, d2, d3}, [%0]! \n" // load 8 pixels of RAW.
"subs %2, %2, #8 \n" // 8 processed per loop.
"vswp.u8 d1, d3 \n" // swap R, B
MEMACCESS(1)
"vst3.8 {d1, d2, d3}, [%1]! \n" // store 8 pixels of RGB24.
"bgt 1b \n"
: "+r"(src_raw), // %0
"+r"(dst_rgb24), // %1
"+r"(width) // %2
:
: "cc", "memory", "d1", "d2", "d3" // Clobber List
);
}
#define RGB565TOARGB \
"vshrn.u16 d6, q0, #5 \n" /* G xxGGGGGG */ \
"vuzp.u8 d0, d1 \n" /* d0 xxxBBBBB RRRRRxxx */ \
......
source/row_neon64.cc
View file @ ce4c2fad
......@@ -847,6 +847,25 @@ void RAWToARGBRow_NEON(const uint8* src_raw, uint8* dst_argb, int width) {
}
#endif // HAS_RAWTOARGBROW_NEON
void RAWToRGB24Row_NEON(const uint8* src_raw, uint8* dst_rgb24, int width) {
asm volatile (
"1: \n"
MEMACCESS(0)
"ld3 {v0.8b,v1.8b,v2.8b}, [%0], #24 \n" // read r g b
"subs %w2, %w2, #8 \n" // 8 processed per loop.
"orr v3.8b, v1.8b, v1.8b \n" // move g
"orr v4.8b, v0.8b, v0.8b \n" // move r
MEMACCESS(1)
"st3 {v2.8b,v3.8b,v4.8b}, [%1], #24 \n" // store b g r
"b.gt 1b \n"
: "+r"(src_raw), // %0
"+r"(dst_rgb24), // %1
"+r"(width) // %2
:
: "cc", "memory", "v0", "v1", "v2", "v3", "v4" // Clobber List
);
}
#define RGB565TOARGB \
"shrn v6.8b, v0.8h, #5 \n" /* G xxGGGGGG */ \
"shl v6.8b, v6.8b, #2 \n" /* G GGGGGG00 upper 6 */ \
......
source/row_win.cc
View file @ ce4c2fad
......@@ -226,6 +226,24 @@ static const uvec8 kShuffleMaskRAWToARGB = {
2u, 1u, 0u, 12u, 5u, 4u, 3u, 13u, 8u, 7u, 6u, 14u, 11u, 10u, 9u, 15u
};
// Shuffle table for converting RAW to RGB24. First 8.
static const uvec8 kShuffleMaskRAWToRGB24_0 = {
2u, 1u, 0u, 5u, 4u, 3u, 8u, 7u,
128u, 128u, 128u, 128u, 128u, 128u, 128u, 128u
};
// Shuffle table for converting RAW to RGB24. Middle 8.
static const uvec8 kShuffleMaskRAWToRGB24_1 = {
2u, 7u, 6u, 5u, 10u, 9u, 8u, 13u,
128u, 128u, 128u, 128u, 128u, 128u, 128u, 128u
};
// Shuffle table for converting RAW to RGB24. Last 8.
static const uvec8 kShuffleMaskRAWToRGB24_2 = {
8u, 7u, 12u, 11u, 10u, 15u, 14u, 13u,
128u, 128u, 128u, 128u, 128u, 128u, 128u, 128u
};
// Shuffle table for converting ARGB to RGB24.
static const uvec8 kShuffleMaskARGBToRGB24 = {
0u, 1u, 2u, 4u, 5u, 6u, 8u, 9u, 10u, 12u, 13u, 14u, 128u, 128u, 128u, 128u
......@@ -241,11 +259,6 @@ static const uvec8 kShuffleMaskARGBToRGB24_0 = {
0u, 1u, 2u, 4u, 5u, 6u, 8u, 9u, 128u, 128u, 128u, 128u, 10u, 12u, 13u, 14u
};
// Shuffle table for converting ARGB to RAW.
static const uvec8 kShuffleMaskARGBToRAW_0 = {
2u, 1u, 0u, 6u, 5u, 4u, 10u, 9u, 128u, 128u, 128u, 128u, 8u, 14u, 13u, 12u
};
// YUY2 shuf 16 Y to 32 Y.
static const lvec8 kShuffleYUY2Y = {
0, 0, 2, 2, 4, 4, 6, 6, 8, 8, 10, 10, 12, 12, 14, 14,
......@@ -413,6 +426,34 @@ void RAWToARGBRow_SSSE3(const uint8* src_raw, uint8* dst_argb,
}
}
__declspec(naked)
void RAWToRGB24Row_SSSE3(const uint8* src_raw, uint8* dst_rgb24, int width) {
__asm {
mov eax, [esp + 4] // src_raw
mov edx, [esp + 8] // dst_rgb24
mov ecx, [esp + 12] // width
movdqa xmm3, xmmword ptr kShuffleMaskRAWToRGB24_0
movdqa xmm4, xmmword ptr kShuffleMaskRAWToRGB24_1
movdqa xmm5, xmmword ptr kShuffleMaskRAWToRGB24_2
convertloop:
movdqu xmm0, [eax]
movdqu xmm1, [eax + 4]
movdqu xmm2, [eax + 8]
lea eax, [eax + 24]
pshufb xmm0, xmm3
pshufb xmm1, xmm4
pshufb xmm2, xmm5
movq qword ptr [edx], xmm0
movq qword ptr [edx + 8], xmm1
movq qword ptr [edx + 16], xmm2
lea edx, [edx + 24]
sub ecx, 8
jg convertloop
ret
}
}
// pmul method to replicate bits.
// Math to replicate bits:
// (v << 8) | (v << 3)
......
unit_test/convert_test.cc
View file @ ce4c2fad
......@@ -992,6 +992,7 @@ TESTATOB(BGRA, 4, 4, 1, ARGB, 4, 4, 1, 0)
TESTATOB(ABGR, 4, 4, 1, ARGB, 4, 4, 1, 0)
TESTATOB(RGBA, 4, 4, 1, ARGB, 4, 4, 1, 0)
TESTATOB(RAW, 3, 3, 1, ARGB, 4, 4, 1, 0)
TESTATOB(RAW, 3, 3, 1, RGB24, 3, 3, 1, 0)
TESTATOB(RGB24, 3, 3, 1, ARGB, 4, 4, 1, 0)
TESTATOB(RGB565, 2, 2, 1, ARGB, 4, 4, 1, 0)
TESTATOB(ARGB1555, 2, 2, 1, ARGB, 4, 4, 1, 0)
......@@ -1765,7 +1766,8 @@ TESTPLANARTOE(I420, 2, 2, BGRA, 4, ARGB, 4)
TESTPLANARTOE(I420, 2, 2, ABGR, 4, ARGB, 4)
TESTPLANARTOE(I420, 2, 2, RGBA, 4, ARGB, 4)
TESTPLANARTOE(I420, 2, 2, RGB24, 3, ARGB, 4)
// TESTPLANARTOE(I420, 2, 2, RGB24, 3, RAW, 3)
TESTPLANARTOE(I420, 2, 2, RAW, 3, RGB24, 3)
TESTPLANARTOE(I420, 2, 2, RGB24, 3, RAW, 3)
TESTPLANARTOE(I420, 2, 2, ARGB, 4, RAW, 3)
TESTPLANARTOE(I420, 2, 2, RAW, 3, ARGB, 4)
TESTPLANARTOE(I420, 2, 2, ARGB, 4, RGB565, 2)
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment