+
+#define RGB2YUV_SHIFT 8
+#define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
+#define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
+#define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
+#define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
+#define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
+#define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
+#define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
+#define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
+#define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
+
+//Note: we have C, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one
+//Plain C versions
+#undef HAVE_MMX
+#undef HAVE_MMX2
+#undef HAVE_3DNOW
+#undef HAVE_SSE2
+#define RENAME(a) a ## _C
+#include "rgb2rgb_template.c"
+
+#if defined(ARCH_X86) || defined(ARCH_X86_64)
+
+//MMX versions
+#undef RENAME
+#define HAVE_MMX
+#undef HAVE_MMX2
+#undef HAVE_3DNOW
+#undef HAVE_SSE2
+#define RENAME(a) a ## _MMX
+#include "rgb2rgb_template.c"
+
+//MMX2 versions
+#undef RENAME
+#define HAVE_MMX
+#define HAVE_MMX2
+#undef HAVE_3DNOW
+#undef HAVE_SSE2
+#define RENAME(a) a ## _MMX2
+#include "rgb2rgb_template.c"
+
+//3DNOW versions
+#undef RENAME
+#define HAVE_MMX
+#undef HAVE_MMX2
+#define HAVE_3DNOW
+#undef HAVE_SSE2
+#define RENAME(a) a ## _3DNOW
+#include "rgb2rgb_template.c"
+
+#endif //ARCH_X86 || ARCH_X86_64
+
+/*
+ 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 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;
+ }
+}
+
+/**
+ * Pallete is assumed to contain bgr32
+ */
+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;
+ }
+}
+
+/**
+ * Pallete is assumed to contain bgr32
+ */
+void palette8torgb24(const uint8_t *src, uint8_t *dst, unsigned num_pixels, const uint8_t *palette)
+{
+ 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+2 ];
+ dst[1]= palette[ src[i]*4+1 ];
+ dst[2]= palette[ src[i]*4+0 ];
+ dst+= 3;
+ }
+}
+
+void palette8tobgr24(const uint8_t *src, uint8_t *dst, unsigned num_pixels, const uint8_t *palette)
+{
+ 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;
+ }
+}
+
+/**
+ * Palette is assumed to contain bgr16, see rgb32to16 to convert the palette
+ */
+void palette8torgb16(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] = ((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
+ */
+void palette8torgb15(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] = ((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 rgb32tobgr24(const uint8_t *src, uint8_t *dst, unsigned int src_size)
+{
+ 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];
+ }
+}
+
+void rgb24tobgr32(const uint8_t *src, uint8_t *dst, unsigned int src_size)
+{
+ 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;
+ }
+}
+
+void rgb16tobgr32(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&0xF800)>>8;
+ *d++ = (bgr&0x7E0)>>3;
+ *d++ = (bgr&0x1F)<<3;
+ *d++ = 0;
+ }
+}
+
+void rgb16tobgr24(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&0xF800)>>8;
+ *d++ = (bgr&0x7E0)>>3;
+ *d++ = (bgr&0x1F)<<3;
+ }
+}
+
+void rgb16tobgr16(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&0x7E0)>>5;
+ b = (rgb&0xF800)>>11;
+ dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11);
+ }
+}
+
+void rgb16tobgr15(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&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);
+ }