mplayer.hg: libswscale/rgb2rgb_template.c comparison

comparison libswscale/rgb2rgb_template.c @ 28968:e67354af521e

Revert Date: Wed Mar 18 23:11:50 2009 New Revision: 28996 Log: Fix libswscale compilation on non-x86, hopefully without breaking MinGW64 again. This change was non optimal, correct would have been to revert the offending commits if no time was available to find a clean fix.

author	michael
date	Thu, 19 Mar 2009 15:02:44 +0000
parents	a7e795e068ad
children	a03804d10dbf

comparison

equal deleted inserted replaced

-:128847ca9780
+:e67354af521e
 }
 }
 static inline void RENAME(rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size)
 {
-#if HAVE_MMX
+x86_reg idx = 15 - src_size;
-x86_reg
-#else
-long
-#endif
-idx = 15 - src_size;
 const uint8_t *s = src-idx;
 uint8_t *d = dst-idx;
 #if HAVE_MMX
 __asm__ volatile(
 "test          %0, %0           \n\t"
 static inline void RENAME(yuvPlanartoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
 long width, long height,
 long lumStride, long chromStride, long dstStride, long vertLumPerChroma)
 {
 long y;
-const long chromWidth= width>>1;
+const x86_reg chromWidth= width>>1;
 for (y=0; y<height; y++)
 {
 #if HAVE_MMX
-x86_reg cw = chromWidth;
 //FIXME handle 2 lines at once (fewer prefetches, reuse some chroma, but very likely memory-limited anyway)
 __asm__ volatile(
 "xor                 %%"REG_a", %%"REG_a"   \n\t"
 ASMALIGN(4)
 "1:                                         \n\t"
 MOVNTQ"                  %%mm6, 24(%0, %%"REG_a", 4)    \n\t"
 "add                        $8, %%"REG_a"   \n\t"
 "cmp                        %4, %%"REG_a"   \n\t"
 " jb                        1b              \n\t"
-::"r"(dst), "r"(ysrc), "r"(usrc), "r"(vsrc), "g" (cw)
+::"r"(dst), "r"(ysrc), "r"(usrc), "r"(vsrc), "g" (chromWidth)
 : "%"REG_a
 );
 #else
 #if ARCH_ALPHA && HAVE_MVI
 static inline void RENAME(yuvPlanartouyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
 long width, long height,
 long lumStride, long chromStride, long dstStride, long vertLumPerChroma)
 {
 long y;
-const long chromWidth= width>>1;
+const x86_reg chromWidth= width>>1;
 for (y=0; y<height; y++)
 {
 #if HAVE_MMX
-x86_reg cw = chromWidth;
 //FIXME handle 2 lines at once (fewer prefetches, reuse some chroma, but very likely memory-limited anyway)
 __asm__ volatile(
 "xor                %%"REG_a", %%"REG_a"    \n\t"
 ASMALIGN(4)
 "1:                                         \n\t"
 MOVNTQ"                 %%mm6, 24(%0, %%"REG_a", 4)     \n\t"
 "add                       $8, %%"REG_a"    \n\t"
 "cmp                       %4, %%"REG_a"    \n\t"
 " jb                       1b               \n\t"
-::"r"(dst), "r"(ysrc), "r"(usrc), "r"(vsrc), "g" (cw)
+::"r"(dst), "r"(ysrc), "r"(usrc), "r"(vsrc), "g" (chromWidth)
 : "%"REG_a
 );
 #else
 //FIXME adapt the Alpha ASM code from yv12->yuy2
 static inline void RENAME(yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
 long width, long height,
 long lumStride, long chromStride, long srcStride)
 {
 long y;
-const long chromWidth= width>>1;
+const x86_reg chromWidth= width>>1;
 for (y=0; y<height; y+=2)
 {
 #if HAVE_MMX
-x86_reg cw = chromWidth;
 __asm__ volatile(
 "xor                 %%"REG_a", %%"REG_a"   \n\t"
 "pcmpeqw                 %%mm7, %%mm7       \n\t"
 "psrlw                      $8, %%mm7       \n\t" // FF,00,FF,00...
 ASMALIGN(4)
 MOVNTQ"                  %%mm2, (%2, %%"REG_a")     \n\t"
 "add                        $8, %%"REG_a"   \n\t"
 "cmp                        %4, %%"REG_a"   \n\t"
 " jb                        1b              \n\t"
-::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (cw)
+::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (chromWidth)
 : "memory", "%"REG_a
 );
 ydst += lumStride;
 src  += srcStride;
 "add                        $8, %%"REG_a"   \n\t"
 "cmp                        %4, %%"REG_a"   \n\t"
 " jb                        1b              \n\t"
-::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (cw)
+::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (chromWidth)
 : "memory", "%"REG_a
 );
 #else
 long i;
 for (i=0; i<chromWidth; i++)
 "g" (-mmxSize)
 : "%"REG_a
 );
 #else
-const long mmxSize=1;
+const x86_reg mmxSize=1;
 #endif
 dst[0        ]= (3*src[0] +   src[srcStride])>>2;
 dst[dstStride]= (  src[0] + 3*src[srcStride])>>2;
 for (x=mmxSize-1; x<srcWidth-1; x++){
 static inline void RENAME(uyvytoyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
 long width, long height,
 long lumStride, long chromStride, long srcStride)
 {
 long y;
-const long chromWidth= width>>1;
+const x86_reg chromWidth= width>>1;
 for (y=0; y<height; y+=2)
 {
 #if HAVE_MMX
-x86_reg cw = chromWidth;
 __asm__ volatile(
 "xor                 %%"REG_a", %%"REG_a"   \n\t"
 "pcmpeqw             %%mm7, %%mm7   \n\t"
 "psrlw                  $8, %%mm7   \n\t" // FF,00,FF,00...
 ASMALIGN(4)
 MOVNTQ"              %%mm2, (%2, %%"REG_a") \n\t"
 "add                    $8, %%"REG_a"   \n\t"
 "cmp                    %4, %%"REG_a"   \n\t"
 " jb                    1b          \n\t"
-::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (cw)
+::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (chromWidth)
 : "memory", "%"REG_a
 );
 ydst += lumStride;
 src  += srcStride;
 "add                    $8, %%"REG_a"   \n\t"
 "cmp                    %4, %%"REG_a"   \n\t"
 " jb                    1b          \n\t"
-::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (cw)
+::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (chromWidth)
 : "memory", "%"REG_a
 );
 #else
 long i;
 for (i=0; i<chromWidth; i++)
 static inline void RENAME(rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
 long width, long height,
 long lumStride, long chromStride, long srcStride)
 {
 long y;
-const long chromWidth= width>>1;
+const x86_reg chromWidth= width>>1;
 #if HAVE_MMX
-x86_reg cw = chromWidth;
 for (y=0; y<height-2; y+=2)
 {
 long i;
 for (i=0; i<2; i++)
 {
 "movd                    %%mm0, (%2, %%"REG_a") \n\t"
 "punpckhdq               %%mm0, %%mm0           \n\t"
 "movd                    %%mm0, (%3, %%"REG_a") \n\t"
 "add                        $4, %%"REG_a"       \n\t"
 " js                        1b                  \n\t"
-: : "r" (src+cw*6), "r" (src+srcStride+cw*6), "r" (udst+cw), "r" (vdst+cw), "g" (-cw)
+: : "r" (src+chromWidth*6), "r" (src+srcStride+chromWidth*6), "r" (udst+chromWidth), "r" (vdst+chromWidth), "g" (-chromWidth)
 : "%"REG_a, "%"REG_d
 );
 udst += chromStride;
 vdst += chromStride;
 uint8_t *dst1, uint8_t *dst2,
 long width, long height,
 long srcStride1, long srcStride2,
 long dstStride1, long dstStride2)
 {
-long x,y,w,h;
+x86_reg y;
+long x,w,h;
 w=width/2; h=height/2;
 #if HAVE_MMX
 __asm__ volatile(
 PREFETCH" %0    \n\t"
 PREFETCH" %1    \n\t"
 uint8_t *dst,
 long width, long height,
 long srcStride1, long srcStride2,
 long srcStride3, long dstStride)
 {
+x86_reg x;
 long y,w,h;
 w=width/2; h=height;
 for (y=0;y<h;y++){
 const uint8_t* yp=src1+srcStride1*y;
 const uint8_t* up=src2+srcStride2*(y>>2);
 const uint8_t* vp=src3+srcStride3*(y>>2);
 uint8_t* d=dst+dstStride*y;
-#if HAVE_MMX
+x=0;
-x86_reg x = 0;
+#if HAVE_MMX
 for (;x<w-7;x+=8)
 {
 __asm__ volatile(
 PREFETCH"   32(%1, %0)          \n\t"
 PREFETCH"   32(%2, %0)          \n\t"
 : "+r" (x)
 : "r"(yp), "r" (up), "r"(vp), "r"(d)
 :"memory");
 }
-#else
-long x = 0;
 #endif
 for (; x<w; x++)
 {
 const long x2 = x<<2;
 d[8*x+0] = yp[x2];

Mercurial > mplayer.hg

comparison libswscale/rgb2rgb_template.c @ 28968:e67354af521e