]> git.sesse.net Git - ffmpeg/blobdiff - postproc/rgb2rgb_template.c
projects / ffmpeg / blobdiff
? search:
summary | shortlog | log | commit | commitdiff | tree
raw | inline | side by side
MMX2, 3DNOW, MMX optimized rgb32(24)to16(15) stuff
[ffmpeg] / postproc / rgb2rgb_template.c
diff --git a/postproc/rgb2rgb_template.c b/postproc/rgb2rgb_template.c
index bb74bbda0c1783ecc254d479ba041795ba078d60..9109fcc3031ee851b6bfa15c9b9ba6dd89332e30 100644 (file)
--- a/postproc/rgb2rgb_template.c
+++ b/postproc/rgb2rgb_template.c
@@ -19,9 +19,12 @@ static const uint64_t mask24h  __attribute__((aligned(8))) = 0x0000FFFFFF000000U
 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 red_mask  __attribute__((aligned(8))) = 0x0000f8000000f800ULL;
-static const uint64_t green_mask __attribute__((aligned(8)))= 0x000007e0000007e0ULL;
-static const uint64_t blue_mask __attribute__((aligned(8))) = 0x0000001f0000001fULL;
+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 red_15mask  __attribute__((aligned(8))) = 0x00007c000000f800ULL;
+static const uint64_t green_15mask __attribute__((aligned(8)))= 0x000003e0000007e0ULL;
+static const uint64_t blue_15mask __attribute__((aligned(8))) = 0x0000001f0000001fULL;
 #endif
 
 void rgb24to32(const uint8_t *src,uint8_t *dst,unsigned src_size)
@@ -92,7 +95,6 @@ void rgb32to24(const uint8_t *src,uint8_t *dst,unsigned src_size)
        "movq   %0, %%mm7\n\t"
        "movq   %1, %%mm6"
        ::"m"(mask24l),"m"(mask24h):"memory");
-  if(mm_end == end) mm_end -= MMREG_SIZE*2;
   while(s < mm_end)
   {
     __asm __volatile(
@@ -228,6 +230,62 @@ void palette8torgb24(const uint8_t *src, uint8_t *dst, unsigned num_pixels, cons
 
 void rgb32to16(const uint8_t *src, uint8_t *dst, unsigned src_size)
 {
+#ifdef HAVE_MMX
+       const uint8_t *s = src;
+       const uint8_t *end,*mm_end;
+       uint16_t *d = (uint16_t *)dst;
+       end = s + src_size;
+       mm_end = (uint8_t*)((((unsigned long)end)/(MMREG_SIZE*2))*(MMREG_SIZE*2));
+       __asm __volatile(PREFETCH"      %0"::"m"(*src):"memory");
+       __asm __volatile(
+           "movq       %0, %%mm7\n\t"
+           "movq       %1, %%mm6\n\t"
+           ::"m"(red_16mask),"m"(green_16mask));
+       while(s < mm_end)
+       {
+           __asm __volatile(
+               PREFETCH" 32%1\n\t"
+               "movd   %1, %%mm0\n\t"
+               "movd   4%1, %%mm3\n\t"
+               "punpckldq 8%1, %%mm0\n\t"
+               "punpckldq 12%1, %%mm3\n\t"
+               "movq   %%mm0, %%mm1\n\t"
+               "movq   %%mm0, %%mm2\n\t"
+               "movq   %%mm3, %%mm4\n\t"
+               "movq   %%mm3, %%mm5\n\t"
+               "psrlq  $3, %%mm0\n\t"
+               "psrlq  $3, %%mm3\n\t"
+               "pand   %2, %%mm0\n\t"
+               "pand   %2, %%mm3\n\t"
+               "psrlq  $5, %%mm1\n\t"
+               "psrlq  $5, %%mm4\n\t"
+               "pand   %%mm6, %%mm1\n\t"
+               "pand   %%mm6, %%mm4\n\t"
+               "psrlq  $8, %%mm2\n\t"
+               "psrlq  $8, %%mm5\n\t"
+               "pand   %%mm7, %%mm2\n\t"
+               "pand   %%mm7, %%mm5\n\t"
+               "por    %%mm1, %%mm0\n\t"
+               "por    %%mm4, %%mm3\n\t"
+               "por    %%mm2, %%mm0\n\t"
+               "por    %%mm5, %%mm3\n\t"
+               "psllq  $16, %%mm3\n\t"
+               "por    %%mm3, %%mm0\n\t"
+               MOVNTQ" %%mm0, %0\n\t"
+               :"=m"(*d):"m"(*s),"m"(blue_16mask):"memory");
+               d += 4;
+               s += 16;
+       }
+       while(s < end)
+       {
+               const int b= *s++;
+               const int g= *s++;
+               const int r= *s++;
+               *d++ = (b>>3) | ((g&0xFC)<<3) | ((r&0xF8)<<8);
+       }
+       __asm __volatile(SFENCE:::"memory");
+       __asm __volatile(EMMS:::"memory");
+#else
        unsigned j,i,num_pixels=src_size/4;
        uint16_t *d = (uint16_t *)dst;
        for(i=0,j=0; j<num_pixels; i+=4,j++)
@@ -238,10 +296,67 @@ void rgb32to16(const uint8_t *src, uint8_t *dst, unsigned src_size)
 
                d[j]= (b>>3) | ((g&0xFC)<<3) | ((r&0xF8)<<8);
        }
+#endif
 }
 
 void rgb32to15(const uint8_t *src, uint8_t *dst, unsigned src_size)
 {
+#ifdef HAVE_MMX
+       const uint8_t *s = src;
+       const uint8_t *end,*mm_end;
+       uint16_t *d = (uint16_t *)dst;
+       end = s + src_size;
+       mm_end = (uint8_t*)((((unsigned long)end)/(MMREG_SIZE*2))*(MMREG_SIZE*2));
+       __asm __volatile(PREFETCH"      %0"::"m"(*src):"memory");
+       __asm __volatile(
+           "movq       %0, %%mm7\n\t"
+           "movq       %1, %%mm6\n\t"
+           ::"m"(red_15mask),"m"(green_15mask));
+       while(s < mm_end)
+       {
+           __asm __volatile(
+               PREFETCH" 32%1\n\t"
+               "movd   %1, %%mm0\n\t"
+               "movd   4%1, %%mm3\n\t"
+               "punpckldq 8%1, %%mm0\n\t"
+               "punpckldq 12%1, %%mm3\n\t"
+               "movq   %%mm0, %%mm1\n\t"
+               "movq   %%mm0, %%mm2\n\t"
+               "movq   %%mm3, %%mm4\n\t"
+               "movq   %%mm3, %%mm5\n\t"
+               "psrlq  $3, %%mm0\n\t"
+               "psrlq  $3, %%mm3\n\t"
+               "pand   %2, %%mm0\n\t"
+               "pand   %2, %%mm3\n\t"
+               "psrlq  $6, %%mm1\n\t"
+               "psrlq  $6, %%mm4\n\t"
+               "pand   %%mm6, %%mm1\n\t"
+               "pand   %%mm6, %%mm4\n\t"
+               "psrlq  $9, %%mm2\n\t"
+               "psrlq  $9, %%mm5\n\t"
+               "pand   %%mm7, %%mm2\n\t"
+               "pand   %%mm7, %%mm5\n\t"
+               "por    %%mm1, %%mm0\n\t"
+               "por    %%mm4, %%mm3\n\t"
+               "por    %%mm2, %%mm0\n\t"
+               "por    %%mm5, %%mm3\n\t"
+               "psllq  $16, %%mm3\n\t"
+               "por    %%mm3, %%mm0\n\t"
+               MOVNTQ" %%mm0, %0\n\t"
+               :"=m"(*d):"m"(*s),"m"(blue_15mask):"memory");
+               d += 4;
+               s += 16;
+       }
+       while(s < end)
+       {
+               const int b= *s++;
+               const int g= *s++;
+               const int r= *s++;
+               *d++ = (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7);
+       }
+       __asm __volatile(SFENCE:::"memory");
+       __asm __volatile(EMMS:::"memory");
+#else
        unsigned j,i,num_pixels=src_size/4;
        uint16_t *d = (uint16_t *)dst;
        for(i=0,j=0; j<num_pixels; i+=4,j++)
@@ -252,6 +367,7 @@ void rgb32to15(const uint8_t *src, uint8_t *dst, unsigned src_size)
 
                d[j]= (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7);
        }
+#endif
 }
 
 void rgb24to16(const uint8_t *src, uint8_t *dst, unsigned src_size)
@@ -266,7 +382,8 @@ void rgb24to16(const uint8_t *src, uint8_t *dst, unsigned src_size)
        __asm __volatile(
            "movq       %0, %%mm7\n\t"
            "movq       %1, %%mm6\n\t"
-           ::"m"(red_mask),"m"(green_mask));
+           ::"m"(red_16mask),"m"(green_16mask));
+       if(mm_end == end) mm_end -= MMREG_SIZE*2;
        while(s < mm_end)
        {
            __asm __volatile(
@@ -298,7 +415,7 @@ void rgb24to16(const uint8_t *src, uint8_t *dst, unsigned src_size)
                "psllq  $16, %%mm3\n\t"
                "por    %%mm3, %%mm0\n\t"
                MOVNTQ" %%mm0, %0\n\t"
-               :"=m"(*d):"m"(*s),"m"(blue_mask):"memory");
+               :"=m"(*d):"m"(*s),"m"(blue_16mask):"memory");
                d += 4;
                s += 12;
        }
@@ -327,6 +444,63 @@ void rgb24to16(const uint8_t *src, uint8_t *dst, unsigned src_size)
 
 void rgb24to15(const uint8_t *src, uint8_t *dst, unsigned src_size)
 {
+#ifdef HAVE_MMX
+       const uint8_t *s = src;
+       const uint8_t *end,*mm_end;
+       uint16_t *d = (uint16_t *)dst;
+       end = s + src_size;
+       mm_end = (uint8_t*)((((unsigned long)end)/(MMREG_SIZE*2))*(MMREG_SIZE*2));
+       __asm __volatile(PREFETCH"      %0"::"m"(*src):"memory");
+       __asm __volatile(
+           "movq       %0, %%mm7\n\t"
+           "movq       %1, %%mm6\n\t"
+           ::"m"(red_15mask),"m"(green_15mask));
+       if(mm_end == end) mm_end -= MMREG_SIZE*2;
+       while(s < mm_end)
+       {
+           __asm __volatile(
+               PREFETCH" 32%1\n\t"
+               "movd   %1, %%mm0\n\t"
+               "movd   3%1, %%mm3\n\t"
+               "punpckldq 6%1, %%mm0\n\t"
+               "punpckldq 9%1, %%mm3\n\t"
+               "movq   %%mm0, %%mm1\n\t"
+               "movq   %%mm0, %%mm2\n\t"
+               "movq   %%mm3, %%mm4\n\t"
+               "movq   %%mm3, %%mm5\n\t"
+               "psrlq  $3, %%mm0\n\t"
+               "psrlq  $3, %%mm3\n\t"
+               "pand   %2, %%mm0\n\t"
+               "pand   %2, %%mm3\n\t"
+               "psrlq  $6, %%mm1\n\t"
+               "psrlq  $6, %%mm4\n\t"
+               "pand   %%mm6, %%mm1\n\t"
+               "pand   %%mm6, %%mm4\n\t"
+               "psrlq  $9, %%mm2\n\t"
+               "psrlq  $9, %%mm5\n\t"
+               "pand   %%mm7, %%mm2\n\t"
+               "pand   %%mm7, %%mm5\n\t"
+               "por    %%mm1, %%mm0\n\t"
+               "por    %%mm4, %%mm3\n\t"
+               "por    %%mm2, %%mm0\n\t"
+               "por    %%mm5, %%mm3\n\t"
+               "psllq  $16, %%mm3\n\t"
+               "por    %%mm3, %%mm0\n\t"
+               MOVNTQ" %%mm0, %0\n\t"
+               :"=m"(*d):"m"(*s),"m"(blue_15mask):"memory");
+               d += 4;
+               s += 12;
+       }
+       while(s < end)
+       {
+               const int b= *s++;
+               const int g= *s++;
+               const int r= *s++;
+               *d++ = (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7);
+       }
+       __asm __volatile(SFENCE:::"memory");
+       __asm __volatile(EMMS:::"memory");
+#else
        unsigned j,i,num_pixels=src_size/3;
        uint16_t *d = (uint16_t *)dst;
        for(i=0,j=0; j<num_pixels; i+=3,j++)
@@ -337,6 +511,7 @@ void rgb24to15(const uint8_t *src, uint8_t *dst, unsigned src_size)
 
                d[j]= (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7);
        }
+#endif
 }
 
 /**
Unnamed repository; edit this file 'description' to name the repository.