Commit ffaea7ee authored by fbarchard@google.com's avatar fbarchard@google.com

ARGBGray function for convertion a rectangle of ARGB to gray scale

BUG=none
TEST=none
Review URL: https://webrtc-codereview.appspot.com/582007

git-svn-id: http://libyuv.googlecode.com/svn/trunk@269 16f28f9a-4ce2-e073-06de-1de4eb20be90
parent 9c02ba53
Name: libyuv Name: libyuv
URL: http://code.google.com/p/libyuv/ URL: http://code.google.com/p/libyuv/
Version: 268 Version: 269
License: BSD License: BSD
License File: LICENSE License File: LICENSE
......
...@@ -133,6 +133,11 @@ int ARGBRect(uint8* dst_argb, int dst_stride_argb, ...@@ -133,6 +133,11 @@ int ARGBRect(uint8* dst_argb, int dst_stride_argb,
int width, int height, int width, int height,
uint32 value); uint32 value);
// Make a rectangle of ARGB gray scale.
int ARGBGray(uint8* dst_argb, int dst_stride_argb,
int x, int y,
int width, int height);
// Copy ARGB to ARGB. // Copy ARGB to ARGB.
int ARGBCopy(const uint8* src_argb, int src_stride_argb, int ARGBCopy(const uint8* src_argb, int src_stride_argb,
uint8* dst_argb, int dst_stride_argb, uint8* dst_argb, int dst_stride_argb,
......
...@@ -12,19 +12,13 @@ ...@@ -12,19 +12,13 @@
#define INCLUDE_LIBYUV_SCALE_ARGB_H_ #define INCLUDE_LIBYUV_SCALE_ARGB_H_
#include "libyuv/basic_types.h" #include "libyuv/basic_types.h"
#include "libyuv/scale.h" // For FilterMode
#ifdef __cplusplus #ifdef __cplusplus
namespace libyuv { namespace libyuv {
extern "C" { extern "C" {
#endif #endif
// Supported filtering
enum FilterMode {
kFilterNone = 0, // Point sample; Fastest
kFilterBilinear = 1, // Faster than box, but lower quality scaling down.
kFilterBox = 2 // Highest quality (not supported for ARGB)
};
int ARGBScale(const uint8* src_argb, int src_stride_argb, int ARGBScale(const uint8* src_argb, int src_stride_argb,
int src_width, int src_height, int src_width, int src_height,
uint8* dst_argb, int dst_stride_argb, uint8* dst_argb, int dst_stride_argb,
......
...@@ -11,7 +11,7 @@ ...@@ -11,7 +11,7 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ #ifndef INCLUDE_LIBYUV_VERSION_H_
#define INCLUDE_LIBYUV_VERSION_H_ #define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 268 #define LIBYUV_VERSION 269
#endif // INCLUDE_LIBYUV_VERSION_H_ #endif // INCLUDE_LIBYUV_VERSION_H_
...@@ -1037,6 +1037,28 @@ AddRow GetSubRow(uint16* dst, int width) { ...@@ -1037,6 +1037,28 @@ AddRow GetSubRow(uint16* dst, int width) {
return SubRowF; return SubRowF;
} }
// Make a rectangle of ARGB gray scale.
int ARGBGray(uint8* dst_argb, int dst_stride_argb,
int dst_x, int dst_y,
int width, int height) {
if (!dst_argb || width <= 0 || height <= 0 || dst_x < 0 || dst_y < 0) {
return -1;
}
void (*ARGBGrayRow)(uint8* dst_argb, int width) = ARGBGrayRow_C;
#if defined(HAS_ARGBGRAYROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3) && IS_ALIGNED(width, 8) &&
IS_ALIGNED(dst_argb, 16) && IS_ALIGNED(dst_stride_argb, 16)) {
ARGBGrayRow = ARGBGrayRow_SSSE3;
}
#endif
uint8* dst = dst_argb + dst_y * dst_stride_argb + dst_x * 4;
for (int y = 0; y < height; ++y) {
ARGBGrayRow(dst, width);
dst += dst_stride_argb;
}
return 0;
}
#ifdef __cplusplus #ifdef __cplusplus
} // extern "C" } // extern "C"
} // namespace libyuv } // namespace libyuv
......
...@@ -71,6 +71,7 @@ extern "C" { ...@@ -71,6 +71,7 @@ extern "C" {
#define HAS_YTOARGBROW_SSE2 #define HAS_YTOARGBROW_SSE2
#define HAS_YUY2TOUVROW_SSE2 #define HAS_YUY2TOUVROW_SSE2
#define HAS_YUY2TOYROW_SSE2 #define HAS_YUY2TOYROW_SSE2
#define HAS_ARGBGRAYROW_SSSE3
#endif #endif
// The following are available only useful when SSSE3 is unavailable. // The following are available only useful when SSSE3 is unavailable.
...@@ -399,6 +400,9 @@ extern uint32 fixed_invtbl8[256]; ...@@ -399,6 +400,9 @@ extern uint32 fixed_invtbl8[256];
void ARGBUnattenuateRow_C(const uint8* src_argb, uint8* dst_argb, int width); void ARGBUnattenuateRow_C(const uint8* src_argb, uint8* dst_argb, int width);
void ARGBUnattenuateRow_SSE2(const uint8* src_argb, uint8* dst_argb, int width); void ARGBUnattenuateRow_SSE2(const uint8* src_argb, uint8* dst_argb, int width);
void ARGBGrayRow_C(uint8* dst_argb, int width);
void ARGBGrayRow_SSSE3(uint8* dst_argb, int width);
#ifdef __cplusplus #ifdef __cplusplus
} // extern "C" } // extern "C"
} // namespace libyuv } // namespace libyuv
......
...@@ -234,6 +234,21 @@ MAKEROWY(ARGB, 2, 1, 0) ...@@ -234,6 +234,21 @@ MAKEROWY(ARGB, 2, 1, 0)
MAKEROWY(BGRA, 1, 2, 3) MAKEROWY(BGRA, 1, 2, 3)
MAKEROWY(ABGR, 0, 1, 2) MAKEROWY(ABGR, 0, 1, 2)
// http://en.wikipedia.org/wiki/Grayscale.
// 0.11 * B + 0.59 * G + 0.30 * R
// Coefficients rounded to multiple of 2 for consistency with SSSE3 version.
static __inline int RGBToGray(uint8 r, uint8 g, uint8 b) {
return (( 76 * r + 152 * g + 28 * b) >> 8);
}
void ARGBGrayRow_C(uint8* dst_argb, int width) {
for (int x = 0; x < width; ++x) {
uint8 y = RGBToGray(dst_argb[2], dst_argb[1], dst_argb[0]);
dst_argb[2] = dst_argb[1] = dst_argb[0] = y;
dst_argb += 4;
}
}
void I400ToARGBRow_C(const uint8* src_y, uint8* dst_argb, int width) { void I400ToARGBRow_C(const uint8* src_y, uint8* dst_argb, int width) {
// Copy a Y to RGB. // Copy a Y to RGB.
for (int x = 0; x < width; ++x) { for (int x = 0; x < width; ++x) {
......
...@@ -108,6 +108,11 @@ CONST uvec8 kShuffleMaskARGBToRAW = { ...@@ -108,6 +108,11 @@ CONST uvec8 kShuffleMaskARGBToRAW = {
2u, 1u, 0u, 6u, 5u, 4u, 10u, 9u, 8u, 14u, 13u, 12u, 128u, 128u, 128u, 128u 2u, 1u, 0u, 6u, 5u, 4u, 10u, 9u, 8u, 14u, 13u, 12u, 128u, 128u, 128u, 128u
}; };
// Constant for ARGB color to gray scale. 0.11 * B + 0.59 * G + 0.30 * R
CONST vec8 kARGBToGray = {
14, 76, 38, 0, 14, 76, 38, 0, 14, 76, 38, 0, 14, 76, 38, 0
};
void I400ToARGBRow_SSE2(const uint8* src_y, uint8* dst_argb, int pix) { void I400ToARGBRow_SSE2(const uint8* src_y, uint8* dst_argb, int pix) {
asm volatile ( asm volatile (
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
...@@ -2527,6 +2532,53 @@ void ARGBUnattenuateRow_SSE2(const uint8* src_argb, uint8* dst_argb, ...@@ -2527,6 +2532,53 @@ void ARGBUnattenuateRow_SSE2(const uint8* src_argb, uint8* dst_argb,
} }
#endif // HAS_ARGBUNATTENUATE_SSE2 #endif // HAS_ARGBUNATTENUATE_SSE2
#ifdef HAS_ARGBGRAYROW_SSSE3
// Convert 8 ARGB pixels (64 bytes) to 8 Gray ARGB pixels
void ARGBGrayRow_SSSE3(uint8* dst_argb, int width) {
asm volatile (
"movdqa %2,%%xmm4 \n"
"pcmpeqb %%xmm5,%%xmm5 \n"
"pslld $0x18,%%xmm5 \n"
"pcmpeqb %%xmm3,%%xmm3 \n"
"psrld $0x8,%%xmm3 \n"
// 8 pixel loop \n"
".p2align 4 \n"
"1: \n"
"movdqa (%0),%%xmm0 \n"
"movdqa 0x10(%0),%%xmm1 \n"
"pmaddubsw %%xmm4,%%xmm0 \n"
"pmaddubsw %%xmm4,%%xmm1 \n"
"movdqa (%0),%%xmm6 \n"
"movdqa 0x10(%0),%%xmm7 \n"
"pand %%xmm5,%%xmm6 \n"
"pand %%xmm5,%%xmm7 \n"
"phaddw %%xmm1,%%xmm0 \n"
"psrlw $0x7,%%xmm0 \n"
"packuswb %%xmm0,%%xmm0 \n"
"punpcklbw %%xmm0,%%xmm0 \n"
"movdqa %%xmm0,%%xmm1 \n"
"punpcklwd %%xmm0,%%xmm0 \n"
"punpckhwd %%xmm1,%%xmm1 \n"
"pand %%xmm3,%%xmm0 \n"
"pand %%xmm3,%%xmm1 \n"
"por %%xmm6,%%xmm0 \n"
"por %%xmm7,%%xmm1 \n"
"sub $0x8,%1 \n"
"movdqa %%xmm0,(%0) \n"
"movdqa %%xmm1,0x10(%0) \n"
"lea 0x20(%0),%0 \n"
"jg 1b \n"
: "+r"(dst_argb), // %0
"+r"(width) // %1
: "m"(kARGBToGray) // %2
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7"
#endif
);
}
#endif // HAS_ARGBGRAYROW_SSSE3
#endif // defined(__x86_64__) || defined(__i386__) #endif // defined(__x86_64__) || defined(__i386__)
#ifdef __cplusplus #ifdef __cplusplus
......
...@@ -93,12 +93,16 @@ static const uvec8 kShuffleMaskARGBToRGB24 = { ...@@ -93,12 +93,16 @@ static const uvec8 kShuffleMaskARGBToRGB24 = {
0u, 1u, 2u, 4u, 5u, 6u, 8u, 9u, 10u, 12u, 13u, 14u, 128u, 128u, 128u, 128u 0u, 1u, 2u, 4u, 5u, 6u, 8u, 9u, 10u, 12u, 13u, 14u, 128u, 128u, 128u, 128u
}; };
// Shuffle table for converting ARGB to RAW. // Shuffle table for converting ARGB to RAW.
static const uvec8 kShuffleMaskARGBToRAW = { static const uvec8 kShuffleMaskARGBToRAW = {
2u, 1u, 0u, 6u, 5u, 4u, 10u, 9u, 8u, 14u, 13u, 12u, 128u, 128u, 128u, 128u 2u, 1u, 0u, 6u, 5u, 4u, 10u, 9u, 8u, 14u, 13u, 12u, 128u, 128u, 128u, 128u
}; };
// Constant for ARGB color to gray scale. 0.11 * B + 0.59 * G + 0.30 * R
static const vec8 kARGBToGray = {
14, 76, 38, 0, 14, 76, 38, 0, 14, 76, 38, 0, 14, 76, 38, 0
};
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
void I400ToARGBRow_SSE2(const uint8* src_y, uint8* dst_argb, int pix) { void I400ToARGBRow_SSE2(const uint8* src_y, uint8* dst_argb, int pix) {
__asm { __asm {
...@@ -2553,6 +2557,50 @@ void ARGBUnattenuateRow_SSE2(const uint8* src_argb, uint8* dst_argb, ...@@ -2553,6 +2557,50 @@ void ARGBUnattenuateRow_SSE2(const uint8* src_argb, uint8* dst_argb,
} }
#endif // HAS_ARGBUNATTENUATE_SSE2 #endif // HAS_ARGBUNATTENUATE_SSE2
#ifdef HAS_ARGBGRAYROW_SSSE3
// Convert 8 ARGB pixels (64 bytes) to 8 Gray ARGB pixels
__declspec(naked) __declspec(align(16))
void ARGBGrayRow_SSSE3(uint8* dst_argb, int width) {
__asm {
mov eax, [esp + 4] /* dst_argb */
mov ecx, [esp + 8] /* width */
movdqa xmm4, kARGBToGray
pcmpeqb xmm5, xmm5 // generate mask 0xff000000
pslld xmm5, 24
pcmpeqb xmm3, xmm3 // generate mask 0x00ffffff
psrld xmm3, 8
align 16
convertloop:
movdqa xmm0, [eax]
movdqa xmm1, [eax + 16]
pmaddubsw xmm0, xmm4
pmaddubsw xmm1, xmm4
movdqa xmm6, [eax] // preserve alpha
movdqa xmm7, [eax + 16]
pand xmm6, xmm5
pand xmm7, xmm5
phaddw xmm0, xmm1
psrlw xmm0, 7
packuswb xmm0, xmm0 // 8 Y values
punpcklbw xmm0, xmm0
movdqa xmm1, xmm0
punpcklwd xmm0, xmm0
punpckhwd xmm1, xmm1
pand xmm0, xmm3 // mask in alpha
pand xmm1, xmm3
por xmm0, xmm6
por xmm1, xmm7
sub ecx, 8
movdqa [eax], xmm0
movdqa [eax + 16], xmm1
lea eax, [eax + 32]
jg convertloop
ret
}
}
#endif // HAS_ARGBGRAYROW_SSSE3
#endif // _M_IX86 #endif // _M_IX86
#ifdef __cplusplus #ifdef __cplusplus
......
...@@ -220,4 +220,57 @@ TEST_F(libyuvTest, TestAddRow) { ...@@ -220,4 +220,57 @@ TEST_F(libyuvTest, TestAddRow) {
AddRow(orig_pixels, added_pixels, 256); AddRow(orig_pixels, added_pixels, 256);
} }
} }
TEST_F(libyuvTest, TestARGBGray) {
SIMD_ALIGNED(uint8 orig_pixels[256][4]);
// Test blue
orig_pixels[0][0] = 255u;
orig_pixels[0][1] = 0u;
orig_pixels[0][2] = 0u;
orig_pixels[0][3] = 128u;
// Test green
orig_pixels[1][0] = 0u;
orig_pixels[1][1] = 255u;
orig_pixels[1][2] = 0u;
orig_pixels[1][3] = 0u;
// Test red
orig_pixels[2][0] = 0u;
orig_pixels[2][1] = 0u;
orig_pixels[2][2] = 255u;
orig_pixels[2][3] = 255u;
// Test color
orig_pixels[3][0] = 16u;
orig_pixels[3][1] = 64u;
orig_pixels[3][2] = 192u;
orig_pixels[3][3] = 224u;
ARGBGray(&orig_pixels[0][0], 0, 0, 0, 4, 1);
EXPECT_EQ(27u, orig_pixels[0][0]);
EXPECT_EQ(27u, orig_pixels[0][1]);
EXPECT_EQ(27u, orig_pixels[0][2]);
EXPECT_EQ(128u, orig_pixels[0][3]);
EXPECT_EQ(151u, orig_pixels[1][0]);
EXPECT_EQ(151u, orig_pixels[1][1]);
EXPECT_EQ(151u, orig_pixels[1][2]);
EXPECT_EQ(0u, orig_pixels[1][3]);
EXPECT_EQ(75u, orig_pixels[2][0]);
EXPECT_EQ(75u, orig_pixels[2][1]);
EXPECT_EQ(75u, orig_pixels[2][2]);
EXPECT_EQ(255u, orig_pixels[2][3]);
EXPECT_EQ(96u, orig_pixels[3][0]);
EXPECT_EQ(96u, orig_pixels[3][1]);
EXPECT_EQ(96u, orig_pixels[3][2]);
EXPECT_EQ(224u, orig_pixels[3][3]);
for (int i = 0; i < 256; ++i) {
orig_pixels[i][0] = i;
orig_pixels[i][1] = i / 2;
orig_pixels[i][2] = i / 3;
orig_pixels[i][3] = i;
}
for (int i = 0; i < 1000 * 1280 * 720 / 256; ++i) {
ARGBGray(&orig_pixels[0][0], 0, 0, 0, 256, 1);
}
} }
} // namespace libyuv
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