#include <inttypes.h>
#include "../config.h"
#include "rgb2rgb.h"
+#include "swscale.h"
#include "../cpudetect.h"
#include "../mangle.h"
-
-#ifdef ARCH_X86
-#define CAN_COMPILE_X86_ASM
-#endif
+#include "../bswap.h"
+#include "../libvo/fastmemcpy.h"
#define FAST_BGR2YV12 // use 7 bit coeffs instead of 15bit
-#ifdef CAN_COMPILE_X86_ASM
-static const uint64_t mask32b __attribute__((aligned(8))) = 0x000000FF000000FFULL;
-static const uint64_t mask32g __attribute__((aligned(8))) = 0x0000FF000000FF00ULL;
-static const uint64_t mask32r __attribute__((aligned(8))) = 0x00FF000000FF0000ULL;
+void (*rgb24to32)(const uint8_t *src,uint8_t *dst,unsigned src_size);
+void (*rgb24to16)(const uint8_t *src,uint8_t *dst,unsigned src_size);
+void (*rgb24to15)(const uint8_t *src,uint8_t *dst,unsigned src_size);
+void (*rgb32to24)(const uint8_t *src,uint8_t *dst,unsigned src_size);
+void (*rgb32to16)(const uint8_t *src,uint8_t *dst,unsigned src_size);
+void (*rgb32to15)(const uint8_t *src,uint8_t *dst,unsigned src_size);
+void (*rgb15to16)(const uint8_t *src,uint8_t *dst,unsigned src_size);
+void (*rgb15to24)(const uint8_t *src,uint8_t *dst,unsigned src_size);
+void (*rgb15to32)(const uint8_t *src,uint8_t *dst,unsigned src_size);
+void (*rgb16to15)(const uint8_t *src,uint8_t *dst,unsigned src_size);
+void (*rgb16to24)(const uint8_t *src,uint8_t *dst,unsigned src_size);
+void (*rgb16to32)(const uint8_t *src,uint8_t *dst,unsigned src_size);
+//void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, unsigned src_size);
+void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, unsigned src_size);
+void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, unsigned src_size);
+void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, unsigned src_size);
+void (*rgb32tobgr32)(const uint8_t *src, uint8_t *dst, unsigned src_size);
+//void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, unsigned src_size);
+void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, unsigned src_size);
+void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, unsigned src_size);
+
+void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
+ unsigned int width, unsigned int height,
+ int lumStride, int chromStride, int dstStride);
+void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
+ unsigned int width, unsigned int height,
+ int lumStride, int chromStride, int dstStride);
+void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
+ unsigned int width, unsigned int height,
+ int lumStride, int chromStride, int dstStride);
+void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
+ unsigned int width, unsigned int height,
+ int lumStride, int chromStride, int srcStride);
+void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
+ unsigned int width, unsigned int height,
+ int lumStride, int chromStride, int srcStride);
+void (*planar2x)(const uint8_t *src, uint8_t *dst, int width, int height, int srcStride, int dstStride);
+void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst,
+ unsigned width, unsigned height, int src1Stride,
+ int src2Stride, int dstStride);
+void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
+ uint8_t *dst1, uint8_t *dst2,
+ unsigned width, unsigned height,
+ int srcStride1, int srcStride2,
+ int dstStride1, int dstStride2);
+void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
+ uint8_t *dst,
+ unsigned width, unsigned height,
+ int srcStride1, int srcStride2,
+ int srcStride3, int dstStride);
+
+#if defined(ARCH_X86) || defined(ARCH_X86_64)
+static const uint64_t mmx_null __attribute__((aligned(8))) = 0x0000000000000000ULL;
+static const uint64_t mmx_one __attribute__((aligned(8))) = 0xFFFFFFFFFFFFFFFFULL;
+static const uint64_t mask32b attribute_used __attribute__((aligned(8))) = 0x000000FF000000FFULL;
+static const uint64_t mask32g attribute_used __attribute__((aligned(8))) = 0x0000FF000000FF00ULL;
+static const uint64_t mask32r attribute_used __attribute__((aligned(8))) = 0x00FF000000FF0000ULL;
static const uint64_t mask32 __attribute__((aligned(8))) = 0x00FFFFFF00FFFFFFULL;
+static const uint64_t mask3216br __attribute__((aligned(8)))=0x00F800F800F800F8ULL;
+static const uint64_t mask3216g __attribute__((aligned(8)))=0x0000FC000000FC00ULL;
+static const uint64_t mask3215g __attribute__((aligned(8)))=0x0000F8000000F800ULL;
+static const uint64_t mul3216 __attribute__((aligned(8))) = 0x2000000420000004ULL;
+static const uint64_t mul3215 __attribute__((aligned(8))) = 0x2000000820000008ULL;
+static const uint64_t mask24b attribute_used __attribute__((aligned(8))) = 0x00FF0000FF0000FFULL;
+static const uint64_t mask24g attribute_used __attribute__((aligned(8))) = 0xFF0000FF0000FF00ULL;
+static const uint64_t mask24r attribute_used __attribute__((aligned(8))) = 0x0000FF0000FF0000ULL;
static const uint64_t mask24l __attribute__((aligned(8))) = 0x0000000000FFFFFFULL;
static const uint64_t mask24h __attribute__((aligned(8))) = 0x0000FFFFFF000000ULL;
static const uint64_t mask24hh __attribute__((aligned(8))) = 0xffff000000000000ULL;
static const uint64_t mask15b __attribute__((aligned(8))) = 0x001F001F001F001FULL; /* 00000000 00011111 xxB */
static const uint64_t mask15rg __attribute__((aligned(8))) = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000 RGx */
static const uint64_t mask15s __attribute__((aligned(8))) = 0xFFE0FFE0FFE0FFE0ULL;
+static const uint64_t mask15g __attribute__((aligned(8))) = 0x03E003E003E003E0ULL;
+static const uint64_t mask15r __attribute__((aligned(8))) = 0x7C007C007C007C00ULL;
+#define mask16b mask15b
+static const uint64_t mask16g __attribute__((aligned(8))) = 0x07E007E007E007E0ULL;
+static const uint64_t mask16r __attribute__((aligned(8))) = 0xF800F800F800F800ULL;
static const uint64_t red_16mask __attribute__((aligned(8))) = 0x0000f8000000f800ULL;
static const uint64_t green_16mask __attribute__((aligned(8)))= 0x000007e0000007e0ULL;
static const uint64_t blue_16mask __attribute__((aligned(8))) = 0x0000001f0000001fULL;
static const uint64_t blue_15mask __attribute__((aligned(8))) = 0x0000001f0000001fULL;
#ifdef FAST_BGR2YV12
-static const uint64_t bgr2YCoeff __attribute__((aligned(8))) = 0x000000210041000DULL;
-static const uint64_t bgr2UCoeff __attribute__((aligned(8))) = 0x0000FFEEFFDC0038ULL;
-static const uint64_t bgr2VCoeff __attribute__((aligned(8))) = 0x00000038FFD2FFF8ULL;
+static const uint64_t bgr2YCoeff attribute_used __attribute__((aligned(8))) = 0x000000210041000DULL;
+static const uint64_t bgr2UCoeff attribute_used __attribute__((aligned(8))) = 0x0000FFEEFFDC0038ULL;
+static const uint64_t bgr2VCoeff attribute_used __attribute__((aligned(8))) = 0x00000038FFD2FFF8ULL;
#else
-static const uint64_t bgr2YCoeff __attribute__((aligned(8))) = 0x000020E540830C8BULL;
-static const uint64_t bgr2UCoeff __attribute__((aligned(8))) = 0x0000ED0FDAC23831ULL;
-static const uint64_t bgr2VCoeff __attribute__((aligned(8))) = 0x00003831D0E6F6EAULL;
+static const uint64_t bgr2YCoeff attribute_used __attribute__((aligned(8))) = 0x000020E540830C8BULL;
+static const uint64_t bgr2UCoeff attribute_used __attribute__((aligned(8))) = 0x0000ED0FDAC23831ULL;
+static const uint64_t bgr2VCoeff attribute_used __attribute__((aligned(8))) = 0x00003831D0E6F6EAULL;
#endif
-static const uint64_t bgr2YOffset __attribute__((aligned(8))) = 0x1010101010101010ULL;
-static const uint64_t bgr2UVOffset __attribute__((aligned(8)))= 0x8080808080808080ULL;
-static const uint64_t w1111 __attribute__((aligned(8))) = 0x0001000100010001ULL;
+static const uint64_t bgr2YOffset attribute_used __attribute__((aligned(8))) = 0x1010101010101010ULL;
+static const uint64_t bgr2UVOffset attribute_used __attribute__((aligned(8)))= 0x8080808080808080ULL;
+static const uint64_t w1111 attribute_used __attribute__((aligned(8))) = 0x0001000100010001ULL;
#if 0
static volatile uint64_t __attribute__((aligned(8))) b5Dither;
#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
-#undef ARCH_X86
+#undef HAVE_SSE2
#define RENAME(a) a ## _C
#include "rgb2rgb_template.c"
-#ifdef CAN_COMPILE_X86_ASM
+#if defined(ARCH_X86) || defined(ARCH_X86_64)
//MMX versions
#undef RENAME
#define HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
-#define ARCH_X86
+#undef HAVE_SSE2
#define RENAME(a) a ## _MMX
#include "rgb2rgb_template.c"
#define HAVE_MMX
#define HAVE_MMX2
#undef HAVE_3DNOW
-#define ARCH_X86
+#undef HAVE_SSE2
#define RENAME(a) a ## _MMX2
#include "rgb2rgb_template.c"
#define HAVE_MMX
#undef HAVE_MMX2
#define HAVE_3DNOW
-#define ARCH_X86
-#define RENAME(a) a ## _3DNow
+#undef HAVE_SSE2
+#define RENAME(a) a ## _3DNOW
#include "rgb2rgb_template.c"
-#endif //CAN_COMPILE_X86_ASM
-
-void rgb24to32(const uint8_t *src,uint8_t *dst,unsigned src_size)
-{
-#ifdef CAN_COMPILE_X86_ASM
- // ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- rgb24to32_MMX2(src, dst, src_size);
- else if(gCpuCaps.has3DNow)
- rgb24to32_3DNow(src, dst, src_size);
- else if(gCpuCaps.hasMMX)
- rgb24to32_MMX(src, dst, src_size);
- else
- rgb24to32_C(src, dst, src_size);
-#else
- rgb24to32_C(src, dst, src_size);
-#endif
-}
-
-void rgb32to24(const uint8_t *src,uint8_t *dst,unsigned src_size)
-{
-#ifdef CAN_COMPILE_X86_ASM
- // ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- rgb32to24_MMX2(src, dst, src_size);
- else if(gCpuCaps.has3DNow)
- rgb32to24_3DNow(src, dst, src_size);
- else if(gCpuCaps.hasMMX)
- rgb32to24_MMX(src, dst, src_size);
- else
- rgb32to24_C(src, dst, src_size);
-#else
- rgb32to24_C(src, dst, src_size);
-#endif
-}
+#endif //ARCH_X86 || ARCH_X86_64
/*
- Original by Strepto/Astral
+ rgb15->rgb16 Original by Strepto/Astral
ported to gcc & bugfixed : A'rpi
MMX2, 3DNOW optimization by Nick Kurshev
32bit c version, and and&add trick by Michael Niedermayer
*/
-void rgb15to16(const uint8_t *src,uint8_t *dst,unsigned src_size)
-{
-#ifdef CAN_COMPILE_X86_ASM
- // ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- rgb15to16_MMX2(src, dst, src_size);
- else if(gCpuCaps.has3DNow)
- rgb15to16_3DNow(src, dst, src_size);
- else if(gCpuCaps.hasMMX)
- rgb15to16_MMX(src, dst, src_size);
- else
- rgb15to16_C(src, dst, src_size);
-#else
- rgb15to16_C(src, dst, src_size);
+
+void sws_rgb2rgb_init(int flags){
+#if defined(ARCH_X86) || defined(ARCH_X86_64)
+ if(flags & SWS_CPU_CAPS_MMX2){
+ rgb15to16= rgb15to16_MMX2;
+ rgb15to24= rgb15to24_MMX2;
+ rgb15to32= rgb15to32_MMX2;
+ rgb16to24= rgb16to24_MMX2;
+ rgb16to32= rgb16to32_MMX2;
+ rgb16to15= rgb16to15_MMX2;
+ rgb24to16= rgb24to16_MMX2;
+ rgb24to15= rgb24to15_MMX2;
+ rgb24to32= rgb24to32_MMX2;
+ rgb32to16= rgb32to16_MMX2;
+ rgb32to15= rgb32to15_MMX2;
+ rgb32to24= rgb32to24_MMX2;
+ rgb24tobgr15= rgb24tobgr15_MMX2;
+ rgb24tobgr16= rgb24tobgr16_MMX2;
+ rgb24tobgr24= rgb24tobgr24_MMX2;
+ rgb32tobgr32= rgb32tobgr32_MMX2;
+ rgb32tobgr16= rgb32tobgr16_MMX2;
+ rgb32tobgr15= rgb32tobgr15_MMX2;
+ yv12toyuy2= yv12toyuy2_MMX2;
+ yv12touyvy= yv12touyvy_MMX2;
+ yuv422ptoyuy2= yuv422ptoyuy2_MMX2;
+ yuy2toyv12= yuy2toyv12_MMX2;
+// uyvytoyv12= uyvytoyv12_MMX2;
+// yvu9toyv12= yvu9toyv12_MMX2;
+ planar2x= planar2x_MMX2;
+ rgb24toyv12= rgb24toyv12_MMX2;
+ interleaveBytes= interleaveBytes_MMX2;
+ vu9_to_vu12= vu9_to_vu12_MMX2;
+ yvu9_to_yuy2= yvu9_to_yuy2_MMX2;
+ }else if(flags & SWS_CPU_CAPS_3DNOW){
+ rgb15to16= rgb15to16_3DNOW;
+ rgb15to24= rgb15to24_3DNOW;
+ rgb15to32= rgb15to32_3DNOW;
+ rgb16to24= rgb16to24_3DNOW;
+ rgb16to32= rgb16to32_3DNOW;
+ rgb16to15= rgb16to15_3DNOW;
+ rgb24to16= rgb24to16_3DNOW;
+ rgb24to15= rgb24to15_3DNOW;
+ rgb24to32= rgb24to32_3DNOW;
+ rgb32to16= rgb32to16_3DNOW;
+ rgb32to15= rgb32to15_3DNOW;
+ rgb32to24= rgb32to24_3DNOW;
+ rgb24tobgr15= rgb24tobgr15_3DNOW;
+ rgb24tobgr16= rgb24tobgr16_3DNOW;
+ rgb24tobgr24= rgb24tobgr24_3DNOW;
+ rgb32tobgr32= rgb32tobgr32_3DNOW;
+ rgb32tobgr16= rgb32tobgr16_3DNOW;
+ rgb32tobgr15= rgb32tobgr15_3DNOW;
+ yv12toyuy2= yv12toyuy2_3DNOW;
+ yv12touyvy= yv12touyvy_3DNOW;
+ yuv422ptoyuy2= yuv422ptoyuy2_3DNOW;
+ yuy2toyv12= yuy2toyv12_3DNOW;
+// uyvytoyv12= uyvytoyv12_3DNOW;
+// yvu9toyv12= yvu9toyv12_3DNOW;
+ planar2x= planar2x_3DNOW;
+ rgb24toyv12= rgb24toyv12_3DNOW;
+ interleaveBytes= interleaveBytes_3DNOW;
+ vu9_to_vu12= vu9_to_vu12_3DNOW;
+ yvu9_to_yuy2= yvu9_to_yuy2_3DNOW;
+ }else if(flags & SWS_CPU_CAPS_MMX){
+ rgb15to16= rgb15to16_MMX;
+ rgb15to24= rgb15to24_MMX;
+ rgb15to32= rgb15to32_MMX;
+ rgb16to24= rgb16to24_MMX;
+ rgb16to32= rgb16to32_MMX;
+ rgb16to15= rgb16to15_MMX;
+ rgb24to16= rgb24to16_MMX;
+ rgb24to15= rgb24to15_MMX;
+ rgb24to32= rgb24to32_MMX;
+ rgb32to16= rgb32to16_MMX;
+ rgb32to15= rgb32to15_MMX;
+ rgb32to24= rgb32to24_MMX;
+ rgb24tobgr15= rgb24tobgr15_MMX;
+ rgb24tobgr16= rgb24tobgr16_MMX;
+ rgb24tobgr24= rgb24tobgr24_MMX;
+ rgb32tobgr32= rgb32tobgr32_MMX;
+ rgb32tobgr16= rgb32tobgr16_MMX;
+ rgb32tobgr15= rgb32tobgr15_MMX;
+ yv12toyuy2= yv12toyuy2_MMX;
+ yv12touyvy= yv12touyvy_MMX;
+ yuv422ptoyuy2= yuv422ptoyuy2_MMX;
+ yuy2toyv12= yuy2toyv12_MMX;
+// uyvytoyv12= uyvytoyv12_MMX;
+// yvu9toyv12= yvu9toyv12_MMX;
+ planar2x= planar2x_MMX;
+ rgb24toyv12= rgb24toyv12_MMX;
+ interleaveBytes= interleaveBytes_MMX;
+ vu9_to_vu12= vu9_to_vu12_MMX;
+ yvu9_to_yuy2= yvu9_to_yuy2_MMX;
+ }else
#endif
+ {
+ rgb15to16= rgb15to16_C;
+ rgb15to24= rgb15to24_C;
+ rgb15to32= rgb15to32_C;
+ rgb16to24= rgb16to24_C;
+ rgb16to32= rgb16to32_C;
+ rgb16to15= rgb16to15_C;
+ rgb24to16= rgb24to16_C;
+ rgb24to15= rgb24to15_C;
+ rgb24to32= rgb24to32_C;
+ rgb32to16= rgb32to16_C;
+ rgb32to15= rgb32to15_C;
+ rgb32to24= rgb32to24_C;
+ rgb24tobgr15= rgb24tobgr15_C;
+ rgb24tobgr16= rgb24tobgr16_C;
+ rgb24tobgr24= rgb24tobgr24_C;
+ rgb32tobgr32= rgb32tobgr32_C;
+ rgb32tobgr16= rgb32tobgr16_C;
+ rgb32tobgr15= rgb32tobgr15_C;
+ yv12toyuy2= yv12toyuy2_C;
+ yv12touyvy= yv12touyvy_C;
+ yuv422ptoyuy2= yuv422ptoyuy2_C;
+ yuy2toyv12= yuy2toyv12_C;
+// uyvytoyv12= uyvytoyv12_C;
+// yvu9toyv12= yvu9toyv12_C;
+ planar2x= planar2x_C;
+ rgb24toyv12= rgb24toyv12_C;
+ interleaveBytes= interleaveBytes_C;
+ vu9_to_vu12= vu9_to_vu12_C;
+ yvu9_to_yuy2= yvu9_to_yuy2_C;
+ }
}
/**
void palette8torgb32(const uint8_t *src, uint8_t *dst, unsigned num_pixels, const uint8_t *palette)
{
unsigned i;
+
+/*
for(i=0; i<num_pixels; i++)
((unsigned *)dst)[i] = ((unsigned *)palette)[ src[i] ];
+*/
+
+ for(i=0; i<num_pixels; i++)
+ {
+ #ifdef WORDS_BIGENDIAN
+ dst[3]= palette[ src[i]*4+2 ];
+ dst[2]= palette[ src[i]*4+1 ];
+ dst[1]= palette[ src[i]*4+0 ];
+ #else
+ //FIXME slow?
+ dst[0]= palette[ src[i]*4+2 ];
+ dst[1]= palette[ src[i]*4+1 ];
+ dst[2]= palette[ src[i]*4+0 ];
+// dst[3]= 0; /* do we need this cleansing? */
+ #endif
+ dst+= 4;
+ }
+}
+
+void palette8tobgr32(const uint8_t *src, uint8_t *dst, unsigned num_pixels, const uint8_t *palette)
+{
+ unsigned i;
+ for(i=0; i<num_pixels; i++)
+ {
+ #ifdef WORDS_BIGENDIAN
+ dst[3]= palette[ src[i]*4+0 ];
+ dst[2]= palette[ src[i]*4+1 ];
+ dst[1]= palette[ src[i]*4+2 ];
+ #else
+ //FIXME slow?
+ dst[0]= palette[ src[i]*4+0 ];
+ dst[1]= palette[ src[i]*4+1 ];
+ dst[2]= palette[ src[i]*4+2 ];
+// dst[3]= 0; /* do we need this cleansing? */
+ #endif
+
+ dst+= 4;
+ }
}
/**
for(i=0; i<num_pixels; i++)
{
//FIXME slow?
- dst[0]= palette[ src[i]*4+0 ];
+ dst[0]= palette[ src[i]*4+2 ];
dst[1]= palette[ src[i]*4+1 ];
- dst[2]= palette[ src[i]*4+2 ];
+ dst[2]= palette[ src[i]*4+0 ];
dst+= 3;
}
}
-void rgb32to16(const uint8_t *src, uint8_t *dst, unsigned src_size)
+void palette8tobgr24(const uint8_t *src, uint8_t *dst, unsigned num_pixels, const uint8_t *palette)
{
-#ifdef CAN_COMPILE_X86_ASM
- // ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- rgb32to16_MMX2(src, dst, src_size);
- else if(gCpuCaps.has3DNow)
- rgb32to16_3DNow(src, dst, src_size);
- else if(gCpuCaps.hasMMX)
- rgb32to16_MMX(src, dst, src_size);
- else
- rgb32to16_C(src, dst, src_size);
-#else
- rgb32to16_C(src, dst, src_size);
-#endif
-}
-
-void rgb32to15(const uint8_t *src, uint8_t *dst, unsigned src_size)
-{
-#ifdef CAN_COMPILE_X86_ASM
- // ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- rgb32to15_MMX2(src, dst, src_size);
- else if(gCpuCaps.has3DNow)
- rgb32to15_3DNow(src, dst, src_size);
- else if(gCpuCaps.hasMMX)
- rgb32to15_MMX(src, dst, src_size);
- else
- rgb32to15_C(src, dst, src_size);
-#else
- rgb32to15_C(src, dst, src_size);
-#endif
-}
-
-void rgb24to16(const uint8_t *src, uint8_t *dst, unsigned src_size)
-{
-#ifdef CAN_COMPILE_X86_ASM
- // ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- rgb24to16_MMX2(src, dst, src_size);
- else if(gCpuCaps.has3DNow)
- rgb24to16_3DNow(src, dst, src_size);
- else if(gCpuCaps.hasMMX)
- rgb24to16_MMX(src, dst, src_size);
- else
- rgb24to16_C(src, dst, src_size);
-#else
- rgb24to16_C(src, dst, src_size);
-#endif
-}
-
-void rgb24to15(const uint8_t *src, uint8_t *dst, unsigned src_size)
-{
-#ifdef CAN_COMPILE_X86_ASM
- // ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- rgb24to15_MMX2(src, dst, src_size);
- else if(gCpuCaps.has3DNow)
- rgb24to15_3DNow(src, dst, src_size);
- else if(gCpuCaps.hasMMX)
- rgb24to15_MMX(src, dst, src_size);
- else
- rgb24to15_C(src, dst, src_size);
-#else
- rgb24to15_C(src, dst, src_size);
-#endif
+ unsigned i;
+/*
+ writes 1 byte o much and might cause alignment issues on some architectures?
+ for(i=0; i<num_pixels; i++)
+ ((unsigned *)(&dst[i*3])) = ((unsigned *)palette)[ src[i] ];
+*/
+ for(i=0; i<num_pixels; i++)
+ {
+ //FIXME slow?
+ dst[0]= palette[ src[i]*4+0 ];
+ dst[1]= palette[ src[i]*4+1 ];
+ dst[2]= palette[ src[i]*4+2 ];
+ dst+= 3;
+ }
}
/**
for(i=0; i<num_pixels; i++)
((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ];
}
+void palette8tobgr16(const uint8_t *src, uint8_t *dst, unsigned num_pixels, const uint8_t *palette)
+{
+ unsigned i;
+ for(i=0; i<num_pixels; i++)
+ ((uint16_t *)dst)[i] = bswap_16(((uint16_t *)palette)[ src[i] ]);
+}
/**
* Pallete is assumed to contain bgr15, see rgb32to15 to convert the palette
for(i=0; i<num_pixels; i++)
((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ];
}
+void palette8tobgr15(const uint8_t *src, uint8_t *dst, unsigned num_pixels, const uint8_t *palette)
+{
+ unsigned i;
+ for(i=0; i<num_pixels; i++)
+ ((uint16_t *)dst)[i] = bswap_16(((uint16_t *)palette)[ src[i] ]);
+}
-void rgb32tobgr32(const uint8_t *src, uint8_t *dst, unsigned int src_size)
+void rgb32tobgr24(const uint8_t *src, uint8_t *dst, unsigned int src_size)
{
-#ifdef CAN_COMPILE_X86_ASM
- // ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- rgb32tobgr32_MMX2(src, dst, src_size);
- else if(gCpuCaps.has3DNow)
- rgb32tobgr32_3DNow(src, dst, src_size);
- else if(gCpuCaps.hasMMX)
- rgb32tobgr32_MMX(src, dst, src_size);
- else
- rgb32tobgr32_C(src, dst, src_size);
-#else
- rgb32tobgr32_C(src, dst, src_size);
-#endif
+ unsigned i;
+ unsigned num_pixels = src_size >> 2;
+ for(i=0; i<num_pixels; i++)
+ {
+ dst[3*i + 0] = src[4*i + 2];
+ dst[3*i + 1] = src[4*i + 1];
+ dst[3*i + 2] = src[4*i + 0];
+ }
}
-/**
- *
- * height should be a multiple of 2 and width should be a multiple of 16 (if this is a
- * problem for anyone then tell me, and ill fix it)
- */
-void yv12toyuy2(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
- unsigned int width, unsigned int height,
- unsigned int lumStride, unsigned int chromStride, unsigned int dstStride)
+void rgb24tobgr32(const uint8_t *src, uint8_t *dst, unsigned int src_size)
{
-#ifdef CAN_COMPILE_X86_ASM
- // ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- yv12toyuy2_MMX2(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride);
- else if(gCpuCaps.has3DNow)
- yv12toyuy2_3DNow(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride);
- else if(gCpuCaps.hasMMX)
- yv12toyuy2_MMX(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride);
- else
- yv12toyuy2_C(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride);
-#else
- yv12toyuy2_C(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride);
-#endif
+ unsigned i;
+ for(i=0; 3*i<src_size; i++)
+ {
+ dst[4*i + 0] = src[3*i + 2];
+ dst[4*i + 1] = src[3*i + 1];
+ dst[4*i + 2] = src[3*i + 0];
+ dst[4*i + 3] = 0;
+ }
}
-/**
- *
- * height should be a multiple of 2 and width should be a multiple of 16 (if this is a
- * problem for anyone then tell me, and ill fix it)
- */
-void yuy2toyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
- unsigned int width, unsigned int height,
- unsigned int lumStride, unsigned int chromStride, unsigned int srcStride)
+void rgb16tobgr32(const uint8_t *src, uint8_t *dst, unsigned int src_size)
{
-#ifdef CAN_COMPILE_X86_ASM
- // ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- yuy2toyv12_MMX2(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
- else if(gCpuCaps.has3DNow)
- yuy2toyv12_3DNow(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
- else if(gCpuCaps.hasMMX)
- yuy2toyv12_MMX(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
- else
- yuy2toyv12_C(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
-#else
- yuy2toyv12_C(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
-#endif
+ const uint16_t *end;
+ uint8_t *d = (uint8_t *)dst;
+ const uint16_t *s = (uint16_t *)src;
+ end = s + src_size/2;
+ while(s < end)
+ {
+ register uint16_t bgr;
+ bgr = *s++;
+ *d++ = (bgr&0xF800)>>8;
+ *d++ = (bgr&0x7E0)>>3;
+ *d++ = (bgr&0x1F)<<3;
+ *d++ = 0;
+ }
}
-/**
- *
- * height should be a multiple of 2 and width should be a multiple of 16 (if this is a
- * problem for anyone then tell me, and ill fix it)
- * chrominance data is only taken from every secound line others are ignored FIXME write HQ version
- */
-void uyvytoyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
- unsigned int width, unsigned int height,
- unsigned int lumStride, unsigned int chromStride, unsigned int srcStride)
+void rgb16tobgr24(const uint8_t *src, uint8_t *dst, unsigned int src_size)
{
-#ifdef CAN_COMPILE_X86_ASM
- // ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- uyvytoyv12_MMX2(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
- else if(gCpuCaps.has3DNow)
- uyvytoyv12_3DNow(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
- else if(gCpuCaps.hasMMX)
- uyvytoyv12_MMX(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
- else
- uyvytoyv12_C(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
-#else
- uyvytoyv12_C(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
-#endif
+ const uint16_t *end;
+ uint8_t *d = (uint8_t *)dst;
+ const uint16_t *s = (const uint16_t *)src;
+ end = s + src_size/2;
+ while(s < end)
+ {
+ register uint16_t bgr;
+ bgr = *s++;
+ *d++ = (bgr&0xF800)>>8;
+ *d++ = (bgr&0x7E0)>>3;
+ *d++ = (bgr&0x1F)<<3;
+ }
}
-/**
- *
- * height should be a multiple of 2 and width should be a multiple of 2 (if this is a
- * problem for anyone then tell me, and ill fix it)
- * chrominance data is only taken from every secound line others are ignored FIXME write HQ version
- */
-void rgb24toyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
- unsigned int width, unsigned int height,
- unsigned int lumStride, unsigned int chromStride, unsigned int srcStride)
+void rgb16tobgr16(const uint8_t *src, uint8_t *dst, unsigned int src_size)
{
-#ifdef CAN_COMPILE_X86_ASM
- // ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- rgb24toyv12_MMX2(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
- else if(gCpuCaps.has3DNow)
- rgb24toyv12_3DNow(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
- else if(gCpuCaps.hasMMX)
- rgb24toyv12_MMX(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
- else
- rgb24toyv12_C(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
-#else
- rgb24toyv12_C(src, ydst, udst, vdst, width, height, lumStride, chromStride, srcStride);
-#endif
+ unsigned i;
+ unsigned num_pixels = src_size >> 1;
+
+ for(i=0; i<num_pixels; i++)
+ {
+ unsigned b,g,r;
+ register uint16_t rgb;
+ rgb = src[2*i];
+ r = rgb&0x1F;
+ g = (rgb&0x7E0)>>5;
+ b = (rgb&0xF800)>>11;
+ dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11);
+ }
}
-void interleaveBytes(uint8_t *src1, uint8_t *src2, uint8_t *dst,
- int width, int height, int src1Stride, int src2Stride, int dstStride)
+void rgb16tobgr15(const uint8_t *src, uint8_t *dst, unsigned int src_size)
{
-#ifdef CAN_COMPILE_X86_ASM
- // ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- interleaveBytes_MMX2(src1, src2, dst, width, height, src1Stride, src2Stride, dstStride);
- else if(gCpuCaps.has3DNow)
- interleaveBytes_3DNow(src1, src2, dst, width, height, src1Stride, src2Stride, dstStride);
- else if(gCpuCaps.hasMMX)
- interleaveBytes_MMX(src1, src2, dst, width, height, src1Stride, src2Stride, dstStride);
- else
- interleaveBytes_C(src1, src2, dst, width, height, src1Stride, src2Stride, dstStride);
-#else
- interleaveBytes_C(src1, src2, dst, width, height, src1Stride, src2Stride, dstStride);
-#endif
+ unsigned i;
+ unsigned num_pixels = src_size >> 1;
+
+ for(i=0; i<num_pixels; i++)
+ {
+ unsigned b,g,r;
+ register uint16_t rgb;
+ rgb = src[2*i];
+ r = rgb&0x1F;
+ g = (rgb&0x7E0)>>5;
+ b = (rgb&0xF800)>>11;
+ dst[2*i] = (b&0x1F) | ((g&0x1F)<<5) | ((r&0x1F)<<10);
+ }
+}
+
+void rgb15tobgr32(const uint8_t *src, uint8_t *dst, unsigned int src_size)
+{
+ const uint16_t *end;
+ uint8_t *d = (uint8_t *)dst;
+ const uint16_t *s = (const uint16_t *)src;
+ end = s + src_size/2;
+ while(s < end)
+ {
+ register uint16_t bgr;
+ bgr = *s++;
+ *d++ = (bgr&0x7C00)>>7;
+ *d++ = (bgr&0x3E0)>>2;
+ *d++ = (bgr&0x1F)<<3;
+ *d++ = 0;
+ }
+}
+
+void rgb15tobgr24(const uint8_t *src, uint8_t *dst, unsigned int src_size)
+{
+ const uint16_t *end;
+ uint8_t *d = (uint8_t *)dst;
+ const uint16_t *s = (uint16_t *)src;
+ end = s + src_size/2;
+ while(s < end)
+ {
+ register uint16_t bgr;
+ bgr = *s++;
+ *d++ = (bgr&0x7C00)>>7;
+ *d++ = (bgr&0x3E0)>>2;
+ *d++ = (bgr&0x1F)<<3;
+ }
+}
+
+void rgb15tobgr16(const uint8_t *src, uint8_t *dst, unsigned int src_size)
+{
+ unsigned i;
+ unsigned num_pixels = src_size >> 1;
+
+ for(i=0; i<num_pixels; i++)
+ {
+ unsigned b,g,r;
+ register uint16_t rgb;
+ rgb = src[2*i];
+ r = rgb&0x1F;
+ g = (rgb&0x3E0)>>5;
+ b = (rgb&0x7C00)>>10;
+ dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11);
+ }
+}
+
+void rgb15tobgr15(const uint8_t *src, uint8_t *dst, unsigned int src_size)
+{
+ unsigned i;
+ unsigned num_pixels = src_size >> 1;
+
+ for(i=0; i<num_pixels; i++)
+ {
+ unsigned b,g,r;
+ register uint16_t rgb;
+ rgb = src[2*i];
+ r = rgb&0x1F;
+ g = (rgb&0x3E0)>>5;
+ b = (rgb&0x7C00)>>10;
+ dst[2*i] = (b&0x1F) | ((g&0x1F)<<5) | ((r&0x1F)<<10);
+ }
+}
+
+void rgb8tobgr8(const uint8_t *src, uint8_t *dst, unsigned int src_size)
+{
+ unsigned i;
+ unsigned num_pixels = src_size;
+ for(i=0; i<num_pixels; i++)
+ {
+ unsigned b,g,r;
+ register uint8_t rgb;
+ rgb = src[i];
+ r = (rgb&0x07);
+ g = (rgb&0x38)>>3;
+ b = (rgb&0xC0)>>6;
+ dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6);
+ }
}
projects / ffmpeg / blobdiff