diff options
| -rw-r--r-- | libswscale/colorspace-test.c | 22 | ||||
| -rw-r--r-- | libswscale/ppc/yuv2rgb_altivec.c | 36 | ||||
| -rw-r--r-- | libswscale/rgb2rgb.c | 16 | ||||
| -rw-r--r-- | libswscale/rgb2rgb.h | 16 | ||||
| -rw-r--r-- | libswscale/rgb2rgb_template.c | 2736 | ||||
| -rw-r--r-- | libswscale/sparc/yuv2rgb_vis.c | 188 | ||||
| -rw-r--r-- | libswscale/swscale.c | 270 | ||||
| -rw-r--r-- | libswscale/swscale_template.c | 228 |
8 files changed, 1756 insertions, 1756 deletions
diff --git a/libswscale/colorspace-test.c b/libswscale/colorspace-test.c index 2223ee3a31..a1ec1be1b4 100644 --- a/libswscale/colorspace-test.c +++ b/libswscale/colorspace-test.c @@ -41,17 +41,17 @@ static char *args_parse(int argc, char *argv[]) while ((o = getopt(argc, argv, "m23")) != -1) { switch (o) { - case 'm': - cpu_caps |= SWS_CPU_CAPS_MMX; - break; - case '2': - cpu_caps |= SWS_CPU_CAPS_MMX2; - break; - case '3': - cpu_caps |= SWS_CPU_CAPS_3DNOW; - break; - default: - av_log(NULL, AV_LOG_ERROR, "Unknown option %c\n", o); + case 'm': + cpu_caps |= SWS_CPU_CAPS_MMX; + break; + case '2': + cpu_caps |= SWS_CPU_CAPS_MMX2; + break; + case '3': + cpu_caps |= SWS_CPU_CAPS_3DNOW; + break; + default: + av_log(NULL, AV_LOG_ERROR, "Unknown option %c\n", o); } } diff --git a/libswscale/ppc/yuv2rgb_altivec.c b/libswscale/ppc/yuv2rgb_altivec.c index 37fbcb5dd5..57acc6bc3e 100644 --- a/libswscale/ppc/yuv2rgb_altivec.c +++ b/libswscale/ppc/yuv2rgb_altivec.c @@ -855,13 +855,13 @@ ff_yuv2packedX_altivec(SwsContext *c, B = vec_packclp (B0,B1); switch(c->dstFormat) { - case PIX_FMT_ABGR: out_abgr (R,G,B,out); break; - case PIX_FMT_BGRA: out_bgra (R,G,B,out); break; - case PIX_FMT_RGBA: out_rgba (R,G,B,out); break; - case PIX_FMT_ARGB: out_argb (R,G,B,out); break; - case PIX_FMT_RGB24: out_rgb24 (R,G,B,out); break; - case PIX_FMT_BGR24: out_bgr24 (R,G,B,out); break; - default: + case PIX_FMT_ABGR: out_abgr (R,G,B,out); break; + case PIX_FMT_BGRA: out_bgra (R,G,B,out); break; + case PIX_FMT_RGBA: out_rgba (R,G,B,out); break; + case PIX_FMT_ARGB: out_argb (R,G,B,out); break; + case PIX_FMT_RGB24: out_rgb24 (R,G,B,out); break; + case PIX_FMT_BGR24: out_bgr24 (R,G,B,out); break; + default: { /* If this is reached, the caller should have called yuv2packedXinC instead. */ @@ -934,17 +934,17 @@ ff_yuv2packedX_altivec(SwsContext *c, nout = (vector unsigned char *)scratch; switch(c->dstFormat) { - case PIX_FMT_ABGR: out_abgr (R,G,B,nout); break; - case PIX_FMT_BGRA: out_bgra (R,G,B,nout); break; - case PIX_FMT_RGBA: out_rgba (R,G,B,nout); break; - case PIX_FMT_ARGB: out_argb (R,G,B,nout); break; - case PIX_FMT_RGB24: out_rgb24 (R,G,B,nout); break; - case PIX_FMT_BGR24: out_bgr24 (R,G,B,nout); break; - default: - /* Unreachable, I think. */ - av_log(c, AV_LOG_ERROR, "altivec_yuv2packedX doesn't support %s output\n", - sws_format_name(c->dstFormat)); - return; + case PIX_FMT_ABGR: out_abgr (R,G,B,nout); break; + case PIX_FMT_BGRA: out_bgra (R,G,B,nout); break; + case PIX_FMT_RGBA: out_rgba (R,G,B,nout); break; + case PIX_FMT_ARGB: out_argb (R,G,B,nout); break; + case PIX_FMT_RGB24: out_rgb24 (R,G,B,nout); break; + case PIX_FMT_BGR24: out_bgr24 (R,G,B,nout); break; + default: + /* Unreachable, I think. */ + av_log(c, AV_LOG_ERROR, "altivec_yuv2packedX doesn't support %s output\n", + sws_format_name(c->dstFormat)); + return; } memcpy (&((uint32_t*)dest)[i], scratch, (dstW-i)/4); diff --git a/libswscale/rgb2rgb.c b/libswscale/rgb2rgb.c index 7aeb2a3ac2..1b1d4db77c 100644 --- a/libswscale/rgb2rgb.c +++ b/libswscale/rgb2rgb.c @@ -88,17 +88,17 @@ void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *sr long srcStride1, long srcStride2, long srcStride3, long dstStride); void (*uyvytoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src, - long width, long height, - long lumStride, long chromStride, long srcStride); + long width, long height, + long lumStride, long chromStride, long srcStride); void (*uyvytoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src, - long width, long height, - long lumStride, long chromStride, long srcStride); + long width, long height, + long lumStride, long chromStride, long srcStride); void (*yuyvtoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src, - long width, long height, - long lumStride, long chromStride, long srcStride); + long width, long height, + long lumStride, long chromStride, long srcStride); void (*yuyvtoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src, - long width, long height, - long lumStride, long chromStride, long srcStride); + long width, long height, + long lumStride, long chromStride, long srcStride); #if ARCH_X86 && CONFIG_GPL diff --git a/libswscale/rgb2rgb.h b/libswscale/rgb2rgb.h index 3850ef291d..9d6e6df415 100644 --- a/libswscale/rgb2rgb.h +++ b/libswscale/rgb2rgb.h @@ -144,17 +144,17 @@ extern void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint extern void (*uyvytoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src, - long width, long height, - long lumStride, long chromStride, long srcStride); + long width, long height, + long lumStride, long chromStride, long srcStride); extern void (*uyvytoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src, - long width, long height, - long lumStride, long chromStride, long srcStride); + long width, long height, + long lumStride, long chromStride, long srcStride); extern void (*yuyvtoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src, - long width, long height, - long lumStride, long chromStride, long srcStride); + long width, long height, + long lumStride, long chromStride, long srcStride); extern void (*yuyvtoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src, - long width, long height, - long lumStride, long chromStride, long srcStride); + long width, long height, + long lumStride, long chromStride, long srcStride); void sws_rgb2rgb_init(int flags); diff --git a/libswscale/rgb2rgb_template.c b/libswscale/rgb2rgb_template.c index 3d1fb8aea6..475a810f0b 100644 --- a/libswscale/rgb2rgb_template.c +++ b/libswscale/rgb2rgb_template.c @@ -77,16 +77,16 @@ static inline void RENAME(rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long s const uint8_t *s = src; const uint8_t *end; #if HAVE_MMX - const uint8_t *mm_end; + const uint8_t *mm_end; #endif end = s + src_size; #if HAVE_MMX - __asm__ volatile(PREFETCH" %0"::"m"(*s):"memory"); - mm_end = end - 23; - __asm__ volatile("movq %0, %%mm7"::"m"(mask32a):"memory"); - while (s < mm_end) - { - __asm__ volatile( + __asm__ volatile(PREFETCH" %0"::"m"(*s):"memory"); + mm_end = end - 23; + __asm__ volatile("movq %0, %%mm7"::"m"(mask32a):"memory"); + while (s < mm_end) + { + __asm__ volatile( PREFETCH" 32%1 \n\t" "movd %1, %%mm0 \n\t" "punpckldq 3%1, %%mm0 \n\t" @@ -107,11 +107,11 @@ static inline void RENAME(rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long s :"=m"(*dest) :"m"(*s) :"memory"); - dest += 32; - s += 24; - } - __asm__ volatile(SFENCE:::"memory"); - __asm__ volatile(EMMS:::"memory"); + dest += 32; + s += 24; + } + __asm__ volatile(SFENCE:::"memory"); + __asm__ volatile(EMMS:::"memory"); #endif while (s < end) { @@ -146,53 +146,53 @@ static inline void RENAME(rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long s while (s < mm_end) { __asm__ volatile( - PREFETCH" 32%1 \n\t" - "movq %1, %%mm0 \n\t" - "movq 8%1, %%mm1 \n\t" - "movq 16%1, %%mm4 \n\t" - "movq 24%1, %%mm5 \n\t" - "movq %%mm0, %%mm2 \n\t" - "movq %%mm1, %%mm3 \n\t" - "movq %%mm4, %%mm6 \n\t" - "movq %%mm5, %%mm7 \n\t" - "psrlq $8, %%mm2 \n\t" - "psrlq $8, %%mm3 \n\t" - "psrlq $8, %%mm6 \n\t" - "psrlq $8, %%mm7 \n\t" - "pand %2, %%mm0 \n\t" - "pand %2, %%mm1 \n\t" - "pand %2, %%mm4 \n\t" - "pand %2, %%mm5 \n\t" - "pand %3, %%mm2 \n\t" - "pand %3, %%mm3 \n\t" - "pand %3, %%mm6 \n\t" - "pand %3, %%mm7 \n\t" - "por %%mm2, %%mm0 \n\t" - "por %%mm3, %%mm1 \n\t" - "por %%mm6, %%mm4 \n\t" - "por %%mm7, %%mm5 \n\t" - - "movq %%mm1, %%mm2 \n\t" - "movq %%mm4, %%mm3 \n\t" - "psllq $48, %%mm2 \n\t" - "psllq $32, %%mm3 \n\t" - "pand %4, %%mm2 \n\t" - "pand %5, %%mm3 \n\t" - "por %%mm2, %%mm0 \n\t" - "psrlq $16, %%mm1 \n\t" - "psrlq $32, %%mm4 \n\t" - "psllq $16, %%mm5 \n\t" - "por %%mm3, %%mm1 \n\t" - "pand %6, %%mm5 \n\t" - "por %%mm5, %%mm4 \n\t" - - MOVNTQ" %%mm0, %0 \n\t" - MOVNTQ" %%mm1, 8%0 \n\t" - MOVNTQ" %%mm4, 16%0" - :"=m"(*dest) - :"m"(*s),"m"(mask24l), - "m"(mask24h),"m"(mask24hh),"m"(mask24hhh),"m"(mask24hhhh) - :"memory"); + PREFETCH" 32%1 \n\t" + "movq %1, %%mm0 \n\t" + "movq 8%1, %%mm1 \n\t" + "movq 16%1, %%mm4 \n\t" + "movq 24%1, %%mm5 \n\t" + "movq %%mm0, %%mm2 \n\t" + "movq %%mm1, %%mm3 \n\t" + "movq %%mm4, %%mm6 \n\t" + "movq %%mm5, %%mm7 \n\t" + "psrlq $8, %%mm2 \n\t" + "psrlq $8, %%mm3 \n\t" + "psrlq $8, %%mm6 \n\t" + "psrlq $8, %%mm7 \n\t" + "pand %2, %%mm0 \n\t" + "pand %2, %%mm1 \n\t" + "pand %2, %%mm4 \n\t" + "pand %2, %%mm5 \n\t" + "pand %3, %%mm2 \n\t" + "pand %3, %%mm3 \n\t" + "pand %3, %%mm6 \n\t" + "pand %3, %%mm7 \n\t" + "por %%mm2, %%mm0 \n\t" + "por %%mm3, %%mm1 \n\t" + "por %%mm6, %%mm4 \n\t" + "por %%mm7, %%mm5 \n\t" + + "movq %%mm1, %%mm2 \n\t" + "movq %%mm4, %%mm3 \n\t" + "psllq $48, %%mm2 \n\t" + "psllq $32, %%mm3 \n\t" + "pand %4, %%mm2 \n\t" + "pand %5, %%mm3 \n\t" + "por %%mm2, %%mm0 \n\t" + "psrlq $16, %%mm1 \n\t" + "psrlq $32, %%mm4 \n\t" + "psllq $16, %%mm5 \n\t" + "por %%mm3, %%mm1 \n\t" + "pand %6, %%mm5 \n\t" + "por %%mm5, %%mm4 \n\t" + + MOVNTQ" %%mm0, %0 \n\t" + MOVNTQ" %%mm1, 8%0 \n\t" + MOVNTQ" %%mm4, 16%0" + :"=m"(*dest) + :"m"(*s),"m"(mask24l), + "m"(mask24h),"m"(mask24hh),"m"(mask24hhh),"m"(mask24hhhh) + :"memory"); dest += 24; s += 32; } @@ -237,19 +237,19 @@ static inline void RENAME(rgb15to16)(const uint8_t *src, uint8_t *dst, long src_ while (s<mm_end) { __asm__ volatile( - PREFETCH" 32%1 \n\t" - "movq %1, %%mm0 \n\t" - "movq 8%1, %%mm2 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm2, %%mm3 \n\t" - "pand %%mm4, %%mm0 \n\t" - "pand %%mm4, %%mm2 \n\t" - "paddw %%mm1, %%mm0 \n\t" - "paddw %%mm3, %%mm2 \n\t" - MOVNTQ" %%mm0, %0 \n\t" - MOVNTQ" %%mm2, 8%0" - :"=m"(*d) - :"m"(*s) + PREFETCH" 32%1 \n\t" + "movq %1, %%mm0 \n\t" + "movq 8%1, %%mm2 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm2, %%mm3 \n\t" + "pand %%mm4, %%mm0 \n\t" + "pand %%mm4, %%mm2 \n\t" + "paddw %%mm1, %%mm0 \n\t" + "paddw %%mm3, %%mm2 \n\t" + MOVNTQ" %%mm0, %0 \n\t" + MOVNTQ" %%mm2, 8%0" + :"=m"(*d) + :"m"(*s) ); d+=16; s+=16; @@ -287,23 +287,23 @@ static inline void RENAME(rgb16to15)(const uint8_t *src, uint8_t *dst, long src_ while (s<mm_end) { __asm__ volatile( - PREFETCH" 32%1 \n\t" - "movq %1, %%mm0 \n\t" - "movq 8%1, %%mm2 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm2, %%mm3 \n\t" - "psrlq $1, %%mm0 \n\t" - "psrlq $1, %%mm2 \n\t" - "pand %%mm7, %%mm0 \n\t" - "pand %%mm7, %%mm2 \n\t" - "pand %%mm6, %%mm1 \n\t" - "pand %%mm6, %%mm3 \n\t" - "por %%mm1, %%mm0 \n\t" - "por %%mm3, %%mm2 \n\t" - MOVNTQ" %%mm0, %0 \n\t" - MOVNTQ" %%mm2, 8%0" - :"=m"(*d) - :"m"(*s) + PREFETCH" 32%1 \n\t" + "movq %1, %%mm0 \n\t" + "movq 8%1, %%mm2 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm2, %%mm3 \n\t" + "psrlq $1, %%mm0 \n\t" + "psrlq $1, %%mm2 \n\t" + "pand %%mm7, %%mm0 \n\t" + "pand %%mm7, %%mm2 \n\t" + "pand %%mm6, %%mm1 \n\t" + "pand %%mm6, %%mm3 \n\t" + "por %%mm1, %%mm0 \n\t" + "por %%mm3, %%mm2 \n\t" + MOVNTQ" %%mm0, %0 \n\t" + MOVNTQ" %%mm2, 8%0" + :"=m"(*d) + :"m"(*s) ); d+=16; s+=16; @@ -339,38 +339,38 @@ static inline void RENAME(rgb32to16)(const uint8_t *src, uint8_t *dst, long src_ mm_end = end - 15; #if 1 //is faster only if multiplies are reasonably fast (FIXME figure out on which CPUs this is faster, on Athlon it is slightly faster) __asm__ volatile( - "movq %3, %%mm5 \n\t" - "movq %4, %%mm6 \n\t" - "movq %5, %%mm7 \n\t" - "jmp 2f \n\t" - ASMALIGN(4) - "1: \n\t" - 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 %%mm3, %%mm4 \n\t" - "pand %%mm6, %%mm0 \n\t" - "pand %%mm6, %%mm3 \n\t" - "pmaddwd %%mm7, %%mm0 \n\t" - "pmaddwd %%mm7, %%mm3 \n\t" - "pand %%mm5, %%mm1 \n\t" - "pand %%mm5, %%mm4 \n\t" - "por %%mm1, %%mm0 \n\t" - "por %%mm4, %%mm3 \n\t" - "psrld $5, %%mm0 \n\t" - "pslld $11, %%mm3 \n\t" - "por %%mm3, %%mm0 \n\t" - MOVNTQ" %%mm0, (%0) \n\t" - "add $16, %1 \n\t" - "add $8, %0 \n\t" - "2: \n\t" - "cmp %2, %1 \n\t" - " jb 1b \n\t" - : "+r" (d), "+r"(s) - : "r" (mm_end), "m" (mask3216g), "m" (mask3216br), "m" (mul3216) + "movq %3, %%mm5 \n\t" + "movq %4, %%mm6 \n\t" + "movq %5, %%mm7 \n\t" + "jmp 2f \n\t" + ASMALIGN(4) + "1: \n\t" + 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 %%mm3, %%mm4 \n\t" + "pand %%mm6, %%mm0 \n\t" + "pand %%mm6, %%mm3 \n\t" + "pmaddwd %%mm7, %%mm0 \n\t" + "pmaddwd %%mm7, %%mm3 \n\t" + "pand %%mm5, %%mm1 \n\t" + "pand %%mm5, %%mm4 \n\t" + "por %%mm1, %%mm0 \n\t" + "por %%mm4, %%mm3 \n\t" + "psrld $5, %%mm0 \n\t" + "pslld $11, %%mm3 \n\t" + "por %%mm3, %%mm0 \n\t" + MOVNTQ" %%mm0, (%0) \n\t" + "add $16, %1 \n\t" + "add $8, %0 \n\t" + "2: \n\t" + "cmp %2, %1 \n\t" + " jb 1b \n\t" + : "+r" (d), "+r"(s) + : "r" (mm_end), "m" (mask3216g), "m" (mask3216br), "m" (mul3216) ); #else __asm__ volatile(PREFETCH" %0"::"m"(*src):"memory"); @@ -381,35 +381,35 @@ static inline void RENAME(rgb32to16)(const uint8_t *src, uint8_t *dst, long src_ 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"); + 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; } @@ -443,35 +443,35 @@ static inline void RENAME(rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long s 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" - "psllq $8, %%mm0 \n\t" - "psllq $8, %%mm3 \n\t" - "pand %%mm7, %%mm0 \n\t" - "pand %%mm7, %%mm3 \n\t" - "psrlq $5, %%mm1 \n\t" - "psrlq $5, %%mm4 \n\t" - "pand %%mm6, %%mm1 \n\t" - "pand %%mm6, %%mm4 \n\t" - "psrlq $19, %%mm2 \n\t" - "psrlq $19, %%mm5 \n\t" - "pand %2, %%mm2 \n\t" - "pand %2, %%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"); + 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" + "psllq $8, %%mm0 \n\t" + "psllq $8, %%mm3 \n\t" + "pand %%mm7, %%mm0 \n\t" + "pand %%mm7, %%mm3 \n\t" + "psrlq $5, %%mm1 \n\t" + "psrlq $5, %%mm4 \n\t" + "pand %%mm6, %%mm1 \n\t" + "pand %%mm6, %%mm4 \n\t" + "psrlq $19, %%mm2 \n\t" + "psrlq $19, %%mm5 \n\t" + "pand %2, %%mm2 \n\t" + "pand %2, %%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; } @@ -498,38 +498,38 @@ static inline void RENAME(rgb32to15)(const uint8_t *src, uint8_t *dst, long src_ mm_end = end - 15; #if 1 //is faster only if multiplies are reasonably fast (FIXME figure out on which CPUs this is faster, on Athlon it is slightly faster) __asm__ volatile( - "movq %3, %%mm5 \n\t" - "movq %4, %%mm6 \n\t" - "movq %5, %%mm7 \n\t" - "jmp 2f \n\t" - ASMALIGN(4) - "1: \n\t" - 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 %%mm3, %%mm4 \n\t" - "pand %%mm6, %%mm0 \n\t" - "pand %%mm6, %%mm3 \n\t" - "pmaddwd %%mm7, %%mm0 \n\t" - "pmaddwd %%mm7, %%mm3 \n\t" - "pand %%mm5, %%mm1 \n\t" - "pand %%mm5, %%mm4 \n\t" - "por %%mm1, %%mm0 \n\t" - "por %%mm4, %%mm3 \n\t" - "psrld $6, %%mm0 \n\t" - "pslld $10, %%mm3 \n\t" - "por %%mm3, %%mm0 \n\t" - MOVNTQ" %%mm0, (%0) \n\t" - "add $16, %1 \n\t" - "add $8, %0 \n\t" - "2: \n\t" - "cmp %2, %1 \n\t" - " jb 1b \n\t" - : "+r" (d), "+r"(s) - : "r" (mm_end), "m" (mask3215g), "m" (mask3216br), "m" (mul3215) + "movq %3, %%mm5 \n\t" + "movq %4, %%mm6 \n\t" + "movq %5, %%mm7 \n\t" + "jmp 2f \n\t" + ASMALIGN(4) + "1: \n\t" + 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 %%mm3, %%mm4 \n\t" + "pand %%mm6, %%mm0 \n\t" + "pand %%mm6, %%mm3 \n\t" + "pmaddwd %%mm7, %%mm0 \n\t" + "pmaddwd %%mm7, %%mm3 \n\t" + "pand %%mm5, %%mm1 \n\t" + "pand %%mm5, %%mm4 \n\t" + "por %%mm1, %%mm0 \n\t" + "por %%mm4, %%mm3 \n\t" + "psrld $6, %%mm0 \n\t" + "pslld $10, %%mm3 \n\t" + "por %%mm3, %%mm0 \n\t" + MOVNTQ" %%mm0, (%0) \n\t" + "add $16, %1 \n\t" + "add $8, %0 \n\t" + "2: \n\t" + "cmp %2, %1 \n\t" + " jb 1b \n\t" + : "+r" (d), "+r"(s) + : "r" (mm_end), "m" (mask3215g), "m" (mask3216br), "m" (mul3215) ); #else __asm__ volatile(PREFETCH" %0"::"m"(*src):"memory"); @@ -540,35 +540,35 @@ static inline void RENAME(rgb32to15)(const uint8_t *src, uint8_t *dst, long src_ 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"); + 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; } @@ -602,35 +602,35 @@ static inline void RENAME(rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long s 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" - "psllq $7, %%mm0 \n\t" - "psllq $7, %%mm3 \n\t" - "pand %%mm7, %%mm0 \n\t" - "pand %%mm7, %%mm3 \n\t" - "psrlq $6, %%mm1 \n\t" - "psrlq $6, %%mm4 \n\t" - "pand %%mm6, %%mm1 \n\t" - "pand %%mm6, %%mm4 \n\t" - "psrlq $19, %%mm2 \n\t" - "psrlq $19, %%mm5 \n\t" - "pand %2, %%mm2 \n\t" - "pand %2, %%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"); + 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" + "psllq $7, %%mm0 \n\t" + "psllq $7, %%mm3 \n\t" + "pand %%mm7, %%mm0 \n\t" + "pand %%mm7, %%mm3 \n\t" + "psrlq $6, %%mm1 \n\t" + "psrlq $6, %%mm4 \n\t" + "pand %%mm6, %%mm1 \n\t" + "pand %%mm6, %%mm4 \n\t" + "psrlq $19, %%mm2 \n\t" + "psrlq $19, %%mm5 \n\t" + "pand %2, %%mm2 \n\t" + "pand %2, %%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; } @@ -663,35 +663,35 @@ static inline void RENAME(rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long s 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 $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"); + 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 $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 += 12; } @@ -726,35 +726,35 @@ static inline void RENAME(rgb24to16)(const uint8_t *src, uint8_t *dst, long src_ 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" - "psllq $8, %%mm0 \n\t" - "psllq $8, %%mm3 \n\t" - "pand %%mm7, %%mm0 \n\t" - "pand %%mm7, %%mm3 \n\t" - "psrlq $5, %%mm1 \n\t" - "psrlq $5, %%mm4 \n\t" - "pand %%mm6, %%mm1 \n\t" - "pand %%mm6, %%mm4 \n\t" - "psrlq $19, %%mm2 \n\t" - "psrlq $19, %%mm5 \n\t" - "pand %2, %%mm2 \n\t" - "pand %2, %%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"); + 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" + "psllq $8, %%mm0 \n\t" + "psllq $8, %%mm3 \n\t" + "pand %%mm7, %%mm0 \n\t" + "pand %%mm7, %%mm3 \n\t" + "psrlq $5, %%mm1 \n\t" + "psrlq $5, %%mm4 \n\t" + "pand %%mm6, %%mm1 \n\t" + "pand %%mm6, %%mm4 \n\t" + "psrlq $19, %%mm2 \n\t" + "psrlq $19, %%mm5 \n\t" + "pand %2, %%mm2 \n\t" + "pand %2, %%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 += 12; } @@ -789,35 +789,35 @@ static inline void RENAME(rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long s 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"); + 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; } @@ -852,35 +852,35 @@ static inline void RENAME(rgb24to15)(const uint8_t *src, uint8_t *dst, long src_ 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" - "psllq $7, %%mm0 \n\t" - "psllq $7, %%mm3 \n\t" - "pand %%mm7, %%mm0 \n\t" - "pand %%mm7, %%mm3 \n\t" - "psrlq $6, %%mm1 \n\t" - "psrlq $6, %%mm4 \n\t" - "pand %%mm6, %%mm1 \n\t" - "pand %%mm6, %%mm4 \n\t" - "psrlq $19, %%mm2 \n\t" - "psrlq $19, %%mm5 \n\t" - "pand %2, %%mm2 \n\t" - "pand %2, %%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"); + 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" + "psllq $7, %%mm0 \n\t" + "psllq $7, %%mm3 \n\t" + "pand %%mm7, %%mm0 \n\t" + "pand %%mm7, %%mm3 \n\t" + "psrlq $6, %%mm1 \n\t" + "psrlq $6, %%mm4 \n\t" + "pand %%mm6, %%mm1 \n\t" + "pand %%mm6, %%mm4 \n\t" + "psrlq $19, %%mm2 \n\t" + "psrlq $19, %%mm5 \n\t" + "pand %2, %%mm2 \n\t" + "pand %2, %%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; } @@ -932,117 +932,117 @@ static inline void RENAME(rgb15tobgr24)(const uint8_t *src, uint8_t *dst, long s while (s < mm_end) { __asm__ volatile( - PREFETCH" 32%1 \n\t" - "movq %1, %%mm0 \n\t" - "movq %1, %%mm1 \n\t" - "movq %1, %%mm2 \n\t" - "pand %2, %%mm0 \n\t" - "pand %3, %%mm1 \n\t" - "pand %4, %%mm2 \n\t" - "psllq $3, %%mm0 \n\t" - "psrlq $2, %%mm1 \n\t" - "psrlq $7, %%mm2 \n\t" - "movq %%mm0, %%mm3 \n\t" - "movq %%mm1, %%mm4 \n\t" - "movq %%mm2, %%mm5 \n\t" - "punpcklwd %5, %%mm0 \n\t" - "punpcklwd %5, %%mm1 \n\t" - "punpcklwd %5, %%mm2 \n\t" - "punpckhwd %5, %%mm3 \n\t" - "punpckhwd %5, %%mm4 \n\t" - "punpckhwd %5, %%mm5 \n\t" - "psllq $8, %%mm1 \n\t" - "psllq $16, %%mm2 \n\t" - "por %%mm1, %%mm0 \n\t" - "por %%mm2, %%mm0 \n\t" - "psllq $8, %%mm4 \n\t" - "psllq $16, %%mm5 \n\t" - "por %%mm4, %%mm3 \n\t" - "por %%mm5, %%mm3 \n\t" - - "movq %%mm0, %%mm6 \n\t" - "movq %%mm3, %%mm7 \n\t" - - "movq 8%1, %%mm0 \n\t" - "movq 8%1, %%mm1 \n\t" - "movq 8%1, %%mm2 \n\t" - "pand %2, %%mm0 \n\t" - "pand %3, %%mm1 \n\t" - "pand %4, %%mm2 \n\t" - "psllq $3, %%mm0 \n\t" - "psrlq $2, %%mm1 \n\t" - "psrlq $7, %%mm2 \n\t" - "movq %%mm0, %%mm3 \n\t" - "movq %%mm1, %%mm4 \n\t" - "movq %%mm2, %%mm5 \n\t" - "punpcklwd %5, %%mm0 \n\t" - "punpcklwd %5, %%mm1 \n\t" - "punpcklwd %5, %%mm2 \n\t" - "punpckhwd %5, %%mm3 \n\t" - "punpckhwd %5, %%mm4 \n\t" - "punpckhwd %5, %%mm5 \n\t" - "psllq $8, %%mm1 \n\t" - "psllq $16, %%mm2 \n\t" - "por %%mm1, %%mm0 \n\t" - "por %%mm2, %%mm0 \n\t" - "psllq $8, %%mm4 \n\t" - "psllq $16, %%mm5 \n\t" - "por %%mm4, %%mm3 \n\t" - "por %%mm5, %%mm3 \n\t" - - :"=m"(*d) - :"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r), "m"(mmx_null) - :"memory"); + PREFETCH" 32%1 \n\t" + "movq %1, %%mm0 \n\t" + "movq %1, %%mm1 \n\t" + "movq %1, %%mm2 \n\t" + "pand %2, %%mm0 \n\t" + "pand %3, %%mm1 \n\t" + "pand %4, %%mm2 \n\t" + "psllq $3, %%mm0 \n\t" + "psrlq $2, %%mm1 \n\t" + "psrlq $7, %%mm2 \n\t" + "movq %%mm0, %%mm3 \n\t" + "movq %%mm1, %%mm4 \n\t" + "movq %%mm2, %%mm5 \n\t" + "punpcklwd %5, %%mm0 \n\t" + "punpcklwd %5, %%mm1 \n\t" + "punpcklwd %5, %%mm2 \n\t" + "punpckhwd %5, %%mm3 \n\t" + "punpckhwd %5, %%mm4 \n\t" + "punpckhwd %5, %%mm5 \n\t" + "psllq $8, %%mm1 \n\t" + "psllq $16, %%mm2 \n\t" + "por %%mm1, %%mm0 \n\t" + "por %%mm2, %%mm0 \n\t" + "psllq $8, %%mm4 \n\t" + "psllq $16, %%mm5 \n\t" + "por %%mm4, %%mm3 \n\t" + "por %%mm5, %%mm3 \n\t" + + "movq %%mm0, %%mm6 \n\t" + "movq %%mm3, %%mm7 \n\t" + + "movq 8%1, %%mm0 \n\t" + "movq 8%1, %%mm1 \n\t" + "movq 8%1, %%mm2 \n\t" + "pand %2, %%mm0 \n\t" + "pand %3, %%mm1 \n\t" + "pand %4, %%mm2 \n\t" + "psllq $3, %%mm0 \n\t" + "psrlq $2, %%mm1 \n\t" + "psrlq $7, %%mm2 \n\t" + "movq %%mm0, %%mm3 \n\t" + "movq %%mm1, %%mm4 \n\t" + "movq %%mm2, %%mm5 \n\t" + "punpcklwd %5, %%mm0 \n\t" + "punpcklwd %5, %%mm1 \n\t" + "punpcklwd %5, %%mm2 \n\t" + "punpckhwd %5, %%mm3 \n\t" + "punpckhwd %5, %%mm4 \n\t" + "punpckhwd %5, %%mm5 \n\t" + "psllq $8, %%mm1 \n\t" + "psllq $16, %%mm2 \n\t" + "por %%mm1, %%mm0 \n\t" + "por %%mm2, %%mm0 \n\t" + "psllq $8, %%mm4 \n\t" + "psllq $16, %%mm5 \n\t" + "por %%mm4, %%mm3 \n\t" + "por %%mm5, %%mm3 \n\t" + + :"=m"(*d) + :"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r), "m"(mmx_null) + :"memory"); /* borrowed 32 to 24 */ __asm__ volatile( - "movq %%mm0, %%mm4 \n\t" - "movq %%mm3, %%mm5 \n\t" - "movq %%mm6, %%mm0 \n\t" - "movq %%mm7, %%mm1 \n\t" - - "movq %%mm4, %%mm6 \n\t" - "movq %%mm5, %%mm7 \n\t" - "movq %%mm0, %%mm2 \n\t" - "movq %%mm1, %%mm3 \n\t" - - "psrlq $8, %%mm2 \n\t" - "psrlq $8, %%mm3 \n\t" - "psrlq $8, %%mm6 \n\t" - "psrlq $8, %%mm7 \n\t" - "pand %2, %%mm0 \n\t" - "pand %2, %%mm1 \n\t" - "pand %2, %%mm4 \n\t" - "pand %2, %%mm5 \n\t" - "pand %3, %%mm2 \n\t" - "pand %3, %%mm3 \n\t" - "pand %3, %%mm6 \n\t" - "pand %3, %%mm7 \n\t" - "por %%mm2, %%mm0 \n\t" - "por %%mm3, %%mm1 \n\t" - "por %%mm6, %%mm4 \n\t" - "por %%mm7, %%mm5 \n\t" - - "movq %%mm1, %%mm2 \n\t" - "movq %%mm4, %%mm3 \n\t" - "psllq $48, %%mm2 \n\t" - "psllq $32, %%mm3 \n\t" - "pand %4, %%mm2 \n\t" - "pand %5, %%mm3 \n\t" - "por %%mm2, %%mm0 \n\t" - "psrlq $16, %%mm1 \n\t" - "psrlq $32, %%mm4 \n\t" - "psllq $16, %%mm5 \n\t" - "por %%mm3, %%mm1 \n\t" - "pand %6, %%mm5 \n\t" - "por %%mm5, %%mm4 \n\t" - - MOVNTQ" %%mm0, %0 \n\t" - MOVNTQ" %%mm1, 8%0 \n\t" - MOVNTQ" %%mm4, 16%0" - - :"=m"(*d) - :"m"(*s),"m"(mask24l),"m"(mask24h),"m"(mask24hh),"m"(mask24hhh),"m"(mask24hhhh) - :"memory"); + "movq %%mm0, %%mm4 \n\t" + "movq %%mm3, %%mm5 \n\t" + "movq %%mm6, %%mm0 \n\t" + "movq %%mm7, %%mm1 \n\t" + + "movq %%mm4, %%mm6 \n\t" + "movq %%mm5, %%mm7 \n\t" + "movq %%mm0, %%mm2 \n\t" + "movq %%mm1, %%mm3 \n\t" + + "psrlq $8, %%mm2 \n\t" + "psrlq $8, %%mm3 \n\t" + "psrlq $8, %%mm6 \n\t" + "psrlq $8, %%mm7 \n\t" + "pand %2, %%mm0 \n\t" + "pand %2, %%mm1 \n\t" + "pand %2, %%mm4 \n\t" + "pand %2, %%mm5 \n\t" + "pand %3, %%mm2 \n\t" + "pand %3, %%mm3 \n\t" + "pand %3, %%mm6 \n\t" + "pand %3, %%mm7 \n\t" + "por %%mm2, %%mm0 \n\t" + "por %%mm3, %%mm1 \n\t" + "por %%mm6, %%mm4 \n\t" + "por %%mm7, %%mm5 \n\t" + + "movq %%mm1, %%mm2 \n\t" + "movq %%mm4, %%mm3 \n\t" + "psllq $48, %%mm2 \n\t" + "psllq $32, %%mm3 \n\t" + "pand %4, %%mm2 \n\t" + "pand %5, %%mm3 \n\t" + "por %%mm2, %%mm0 \n\t" + "psrlq $16, %%mm1 \n\t" + "psrlq $32, %%mm4 \n\t" + "psllq $16, %%mm5 \n\t" + "por %%mm3, %%mm1 \n\t" + "pand %6, %%mm5 \n\t" + "por %%mm5, %%mm4 \n\t" + + MOVNTQ" %%mm0, %0 \n\t" + MOVNTQ" %%mm1, 8%0 \n\t" + MOVNTQ" %%mm4, 16%0" + + :"=m"(*d) + :"m"(*s),"m"(mask24l),"m"(mask24h),"m"(mask24hh),"m"(mask24hhh),"m"(mask24hhhh) + :"memory"); d += 24; s += 8; } @@ -1074,116 +1074,116 @@ static inline void RENAME(rgb16tobgr24)(const uint8_t *src, uint8_t *dst, long s while (s < mm_end) { __asm__ volatile( - PREFETCH" 32%1 \n\t" - "movq %1, %%mm0 \n\t" - "movq %1, %%mm1 \n\t" - "movq %1, %%mm2 \n\t" - "pand %2, %%mm0 \n\t" - "pand %3, %%mm1 \n\t" - "pand %4, %%mm2 \n\t" - "psllq $3, %%mm0 \n\t" - "psrlq $3, %%mm1 \n\t" - "psrlq $8, %%mm2 \n\t" - "movq %%mm0, %%mm3 \n\t" - "movq %%mm1, %%mm4 \n\t" - "movq %%mm2, %%mm5 \n\t" - "punpcklwd %5, %%mm0 \n\t" - "punpcklwd %5, %%mm1 \n\t" - "punpcklwd %5, %%mm2 \n\t" - "punpckhwd %5, %%mm3 \n\t" - "punpckhwd %5, %%mm4 \n\t" - "punpckhwd %5, %%mm5 \n\t" - "psllq $8, %%mm1 \n\t" - "psllq $16, %%mm2 \n\t" - "por %%mm1, %%mm0 \n\t" - "por %%mm2, %%mm0 \n\t" - "psllq $8, %%mm4 \n\t" - "psllq $16, %%mm5 \n\t" - "por %%mm4, %%mm3 \n\t" - "por %%mm5, %%mm3 \n\t" - - "movq %%mm0, %%mm6 \n\t" - "movq %%mm3, %%mm7 \n\t" - - "movq 8%1, %%mm0 \n\t" - "movq 8%1, %%mm1 \n\t" - "movq 8%1, %%mm2 \n\t" - "pand %2, %%mm0 \n\t" - "pand %3, %%mm1 \n\t" - "pand %4, %%mm2 \n\t" - "psllq $3, %%mm0 \n\t" - "psrlq $3, %%mm1 \n\t" - "psrlq $8, %%mm2 \n\t" - "movq %%mm0, %%mm3 \n\t" - "movq %%mm1, %%mm4 \n\t" - "movq %%mm2, %%mm5 \n\t" - "punpcklwd %5, %%mm0 \n\t" - "punpcklwd %5, %%mm1 \n\t" - "punpcklwd %5, %%mm2 \n\t" - "punpckhwd %5, %%mm3 \n\t" - "punpckhwd %5, %%mm4 \n\t" - "punpckhwd %5, %%mm5 \n\t" - "psllq $8, %%mm1 \n\t" - "psllq $16, %%mm2 \n\t" - "por %%mm1, %%mm0 \n\t" - "por %%mm2, %%mm0 \n\t" - "psllq $8, %%mm4 \n\t" - "psllq $16, %%mm5 \n\t" - "por %%mm4, %%mm3 \n\t" - "por %%mm5, %%mm3 \n\t" - :"=m"(*d) - :"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r),"m"(mmx_null) - :"memory"); + PREFETCH" 32%1 \n\t" + "movq %1, %%mm0 \n\t" + "movq %1, %%mm1 \n\t" + "movq %1, %%mm2 \n\t" + "pand %2, %%mm0 \n\t" + "pand %3, %%mm1 \n\t" + "pand %4, %%mm2 \n\t" + "psllq $3, %%mm0 \n\t" + "psrlq $3, %%mm1 \n\t" + "psrlq $8, %%mm2 \n\t" + "movq %%mm0, %%mm3 \n\t" + "movq %%mm1, %%mm4 \n\t" + "movq %%mm2, %%mm5 \n\t" + "punpcklwd %5, %%mm0 \n\t" + "punpcklwd %5, %%mm1 \n\t" + "punpcklwd %5, %%mm2 \n\t" + "punpckhwd %5, %%mm3 \n\t" + "punpckhwd %5, %%mm4 \n\t" + "punpckhwd %5, %%mm5 \n\t" + "psllq $8, %%mm1 \n\t" + "psllq $16, %%mm2 \n\t" + "por %%mm1, %%mm0 \n\t" + "por %%mm2, %%mm0 \n\t" + "psllq $8, %%mm4 \n\t" + "psllq $16, %%mm5 \n\t" + "por %%mm4, %%mm3 \n\t" + "por %%mm5, %%mm3 \n\t" + + "movq %%mm0, %%mm6 \n\t" + "movq %%mm3, %%mm7 \n\t" + + "movq 8%1, %%mm0 \n\t" + "movq 8%1, %%mm1 \n\t" + "movq 8%1, %%mm2 \n\t" + "pand %2, %%mm0 \n\t" + "pand %3, %%mm1 \n\t" + "pand %4, %%mm2 \n\t" + "psllq $3, %%mm0 \n\t" + "psrlq $3, %%mm1 \n\t" + "psrlq $8, %%mm2 \n\t" + "movq %%mm0, %%mm3 \n\t" + "movq %%mm1, %%mm4 \n\t" + "movq %%mm2, %%mm5 \n\t" + "punpcklwd %5, %%mm0 \n\t" + "punpcklwd %5, %%mm1 \n\t" + "punpcklwd %5, %%mm2 \n\t" + "punpckhwd %5, %%mm3 \n\t" + "punpckhwd %5, %%mm4 \n\t" + "punpckhwd %5, %%mm5 \n\t" + "psllq $8, %%mm1 \n\t" + "psllq $16, %%mm2 \n\t" + "por %%mm1, %%mm0 \n\t" + "por %%mm2, %%mm0 \n\t" + "psllq $8, %%mm4 \n\t" + "psllq $16, %%mm5 \n\t" + "por %%mm4, %%mm3 \n\t" + "por %%mm5, %%mm3 \n\t" + :"=m"(*d) + :"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r),"m"(mmx_null) + :"memory"); /* borrowed 32 to 24 */ __asm__ volatile( - "movq %%mm0, %%mm4 \n\t" - "movq %%mm3, %%mm5 \n\t" - "movq %%mm6, %%mm0 \n\t" - "movq %%mm7, %%mm1 \n\t" - - "movq %%mm4, %%mm6 \n\t" - "movq %%mm5, %%mm7 \n\t" - "movq %%mm0, %%mm2 \n\t" - "movq %%mm1, %%mm3 \n\t" - - "psrlq $8, %%mm2 \n\t" - "psrlq $8, %%mm3 \n\t" - "psrlq $8, %%mm6 \n\t" - "psrlq $8, %%mm7 \n\t" - "pand %2, %%mm0 \n\t" - "pand %2, %%mm1 \n\t" - "pand %2, %%mm4 \n\t" - "pand %2, %%mm5 \n\t" - "pand %3, %%mm2 \n\t" - "pand %3, %%mm3 \n\t" - "pand %3, %%mm6 \n\t" - "pand %3, %%mm7 \n\t" - "por %%mm2, %%mm0 \n\t" - "por %%mm3, %%mm1 \n\t" - "por %%mm6, %%mm4 \n\t" - "por %%mm7, %%mm5 \n\t" - - "movq %%mm1, %%mm2 \n\t" - "movq %%mm4, %%mm3 \n\t" - "psllq $48, %%mm2 \n\t" - "psllq $32, %%mm3 \n\t" - "pand %4, %%mm2 \n\t" - "pand %5, %%mm3 \n\t" - "por %%mm2, %%mm0 \n\t" - "psrlq $16, %%mm1 \n\t" - "psrlq $32, %%mm4 \n\t" - "psllq $16, %%mm5 \n\t" - "por %%mm3, %%mm1 \n\t" - "pand %6, %%mm5 \n\t" - "por %%mm5, %%mm4 \n\t" - - MOVNTQ" %%mm0, %0 \n\t" - MOVNTQ" %%mm1, 8%0 \n\t" - MOVNTQ" %%mm4, 16%0" - - :"=m"(*d) - :"m"(*s),"m"(mask24l),"m"(mask24h),"m"(mask24hh),"m"(mask24hhh),"m"(mask24hhhh) - :"memory"); + "movq %%mm0, %%mm4 \n\t" + "movq %%mm3, %%mm5 \n\t" + "movq %%mm6, %%mm0 \n\t" + "movq %%mm7, %%mm1 \n\t" + + "movq %%mm4, %%mm6 \n\t" + "movq %%mm5, %%mm7 \n\t" + "movq %%mm0, %%mm2 \n\t" + "movq %%mm1, %%mm3 \n\t" + + "psrlq $8, %%mm2 \n\t" + "psrlq $8, %%mm3 \n\t" + "psrlq $8, %%mm6 \n\t" + "psrlq $8, %%mm7 \n\t" + "pand %2, %%mm0 \n\t" + "pand %2, %%mm1 \n\t" + "pand %2, %%mm4 \n\t" + "pand %2, %%mm5 \n\t" + "pand %3, %%mm2 \n\t" + "pand %3, %%mm3 \n\t" + "pand %3, %%mm6 \n\t" + "pand %3, %%mm7 \n\t" + "por %%mm2, %%mm0 \n\t" + "por %%mm3, %%mm1 \n\t" + "por %%mm6, %%mm4 \n\t" + "por %%mm7, %%mm5 \n\t" + + "movq %%mm1, %%mm2 \n\t" + "movq %%mm4, %%mm3 \n\t" + "psllq $48, %%mm2 \n\t" + "psllq $32, %%mm3 \n\t" + "pand %4, %%mm2 \n\t" + "pand %5, %%mm3 \n\t" + "por %%mm2, %%mm0 \n\t" + "psrlq $16, %%mm1 \n\t" + "psrlq $32, %%mm4 \n\t" + "psllq $16, %%mm5 \n\t" + "por %%mm3, %%mm1 \n\t" + "pand %6, %%mm5 \n\t" + "por %%mm5, %%mm4 \n\t" + + MOVNTQ" %%mm0, %0 \n\t" + MOVNTQ" %%mm1, 8%0 \n\t" + MOVNTQ" %%mm4, 16%0" + + :"=m"(*d) + :"m"(*s),"m"(mask24l),"m"(mask24h),"m"(mask24hh),"m"(mask24hhh),"m"(mask24hhhh) + :"memory"); d += 24; s += 8; } @@ -1236,20 +1236,20 @@ static inline void RENAME(rgb15to32)(const uint8_t *src, uint8_t *dst, long src_ while (s < mm_end) { __asm__ volatile( - PREFETCH" 32%1 \n\t" - "movq %1, %%mm0 \n\t" - "movq %1, %%mm1 \n\t" - "movq %1, %%mm2 \n\t" - "pand %2, %%mm0 \n\t" - "pand %3, %%mm1 \n\t" - "pand %4, %%mm2 \n\t" - "psllq $3, %%mm0 \n\t" - "psrlq $2, %%mm1 \n\t" - "psrlq $7, %%mm2 \n\t" - PACK_RGB32 - :"=m"(*d) - :"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r) - :"memory"); + PREFETCH" 32%1 \n\t" + "movq %1, %%mm0 \n\t" + "movq %1, %%mm1 \n\t" + "movq %1, %%mm2 \n\t" + "pand %2, %%mm0 \n\t" + "pand %3, %%mm1 \n\t" + "pand %4, %%mm2 \n\t" + "psllq $3, %%mm0 \n\t" + "psrlq $2, %%mm1 \n\t" + "psrlq $7, %%mm2 \n\t" + PACK_RGB32 + :"=m"(*d) + :"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r) + :"memory"); d += 16; s += 4; } @@ -1291,20 +1291,20 @@ static inline void RENAME(rgb16to32)(const uint8_t *src, uint8_t *dst, long src_ while (s < mm_end) { __asm__ volatile( - PREFETCH" 32%1 \n\t" - "movq %1, %%mm0 \n\t" - "movq %1, %%mm1 \n\t" - "movq %1, %%mm2 \n\t" - "pand %2, %%mm0 \n\t" - "pand %3, %%mm1 \n\t" - "pand %4, %%mm2 \n\t" - "psllq $3, %%mm0 \n\t" - "psrlq $3, %%mm1 \n\t" - "psrlq $8, %%mm2 \n\t" - PACK_RGB32 - :"=m"(*d) - :"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r) - :"memory"); + PREFETCH" 32%1 \n\t" + "movq %1, %%mm0 \n\t" + "movq %1, %%mm1 \n\t" + "movq %1, %%mm2 \n\t" + "pand %2, %%mm0 \n\t" + "pand %3, %%mm1 \n\t" + "pand %4, %%mm2 \n\t" + "psllq $3, %%mm0 \n\t" + "psrlq $3, %%mm1 \n\t" + "psrlq $8, %%mm2 \n\t" + PACK_RGB32 + :"=m"(*d) + :"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r) + :"memory"); d += 16; s += 4; } @@ -1336,55 +1336,55 @@ static inline void RENAME(rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long s uint8_t *d = dst-idx; #if HAVE_MMX __asm__ volatile( - "test %0, %0 \n\t" - "jns 2f \n\t" - PREFETCH" (%1, %0) \n\t" - "movq %3, %%mm7 \n\t" - "pxor %4, %%mm7 \n\t" - "movq %%mm7, %%mm6 \n\t" - "pxor %5, %%mm7 \n\t" - ASMALIGN(4) - "1: \n\t" - PREFETCH" 32(%1, %0) \n\t" - "movq (%1, %0), %%mm0 \n\t" - "movq 8(%1, %0), %%mm1 \n\t" + "test %0, %0 \n\t" + "jns 2f \n\t" + PREFETCH" (%1, %0) \n\t" + "movq %3, %%mm7 \n\t" + "pxor %4, %%mm7 \n\t" + "movq %%mm7, %%mm6 \n\t" + "pxor %5, %%mm7 \n\t" + ASMALIGN(4) + "1: \n\t" + PREFETCH" 32(%1, %0) \n\t" + "movq (%1, %0), %%mm0 \n\t" + "movq 8(%1, %0), %%mm1 \n\t" # if HAVE_MMX2 - "pshufw $177, %%mm0, %%mm3 \n\t" - "pshufw $177, %%mm1, %%mm5 \n\t" - "pand %%mm7, %%mm0 \n\t" - "pand %%mm6, %%mm3 \n\t" - "pand %%mm7, %%mm1 \n\t" - "pand %%mm6, %%mm5 \n\t" - "por %%mm3, %%mm0 \n\t" - "por %%mm5, %%mm1 \n\t" + "pshufw $177, %%mm0, %%mm3 \n\t" + "pshufw $177, %%mm1, %%mm5 \n\t" + "pand %%mm7, %%mm0 \n\t" + "pand %%mm6, %%mm3 \n\t" + "pand %%mm7, %%mm1 \n\t" + "pand %%mm6, %%mm5 \n\t" + "por %%mm3, %%mm0 \n\t" + "por %%mm5, %%mm1 \n\t" # else - "movq %%mm0, %%mm2 \n\t" - "movq %%mm1, %%mm4 \n\t" - "pand %%mm7, %%mm0 \n\t" - "pand %%mm6, %%mm2 \n\t" - "pand %%mm7, %%mm1 \n\t" - "pand %%mm6, %%mm4 \n\t" - "movq %%mm2, %%mm3 \n\t" - "movq %%mm4, %%mm5 \n\t" - "pslld $16, %%mm2 \n\t" - "psrld $16, %%mm3 \n\t" - "pslld $16, %%mm4 \n\t" - "psrld $16, %%mm5 \n\t" - "por %%mm2, %%mm0 \n\t" - "por %%mm4, %%mm1 \n\t" - "por %%mm3, %%mm0 \n\t" - "por %%mm5, %%mm1 \n\t" + "movq %%mm0, %%mm2 \n\t" + "movq %%mm1, %%mm4 \n\t" + "pand %%mm7, %%mm0 \n\t" + "pand %%mm6, %%mm2 \n\t" + "pand %%mm7, %%mm1 \n\t" + "pand %%mm6, %%mm4 \n\t" + "movq %%mm2, %%mm3 \n\t" + "movq %%mm4, %%mm5 \n\t" + "pslld $16, %%mm2 \n\t" + "psrld $16, %%mm3 \n\t" + "pslld $16, %%mm4 \n\t" + "psrld $16, %%mm5 \n\t" + "por %%mm2, %%mm0 \n\t" + "por %%mm4, %%mm1 \n\t" + "por %%mm3, %%mm0 \n\t" + "por %%mm5, %%mm1 \n\t" # endif - MOVNTQ" %%mm0, (%2, %0) \n\t" - MOVNTQ" %%mm1, 8(%2, %0) \n\t" - "add $16, %0 \n\t" - "js 1b \n\t" - SFENCE" \n\t" - EMMS" \n\t" - "2: \n\t" - : "+&r"(idx) - : "r" (s), "r" (d), "m" (mask32b), "m" (mask32r), "m" (mmx_one) - : "memory"); + MOVNTQ" %%mm0, (%2, %0) \n\t" + MOVNTQ" %%mm1, 8(%2, %0) \n\t" + "add $16, %0 \n\t" + "js 1b \n\t" + SFENCE" \n\t" + EMMS" \n\t" + "2: \n\t" + : "+&r"(idx) + : "r" (s), "r" (d), "m" (mask32b), "m" (mask32r), "m" (mmx_one) + : "memory"); #endif for (; idx<15; idx+=4) { register int v = *(const uint32_t *)&s[idx], g = v & 0xff00ff00; @@ -1399,47 +1399,47 @@ static inline void RENAME(rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long s #if HAVE_MMX x86_reg mmx_size= 23 - src_size; __asm__ volatile ( - "test %%"REG_a", %%"REG_a" \n\t" - "jns 2f \n\t" - "movq "MANGLE(mask24r)", %%mm5 \n\t" - "movq "MANGLE(mask24g)", %%mm6 \n\t" - "movq "MANGLE(mask24b)", %%mm7 \n\t" - ASMALIGN(4) - "1: \n\t" - PREFETCH" 32(%1, %%"REG_a") \n\t" - "movq (%1, %%"REG_a"), %%mm0 \n\t" // BGR BGR BG - "movq (%1, %%"REG_a"), %%mm1 \n\t" // BGR BGR BG - "movq 2(%1, %%"REG_a"), %%mm2 \n\t" // R BGR BGR B - "psllq $16, %%mm0 \n\t" // 00 BGR BGR - "pand %%mm5, %%mm0 \n\t" - "pand %%mm6, %%mm1 \n\t" - "pand %%mm7, %%mm2 \n\t" - "por %%mm0, %%mm1 \n\t" - "por %%mm2, %%mm1 \n\t" - "movq 6(%1, %%"REG_a"), %%mm0 \n\t" // BGR BGR BG - MOVNTQ" %%mm1, (%2, %%"REG_a") \n\t" // RGB RGB RG - "movq 8(%1, %%"REG_a"), %%mm1 \n\t" // R BGR BGR B - "movq 10(%1, %%"REG_a"), %%mm2 \n\t" // GR BGR BGR - "pand %%mm7, %%mm0 \n\t" - "pand %%mm5, %%mm1 \n\t" - "pand %%mm6, %%mm2 \n\t" - "por %%mm0, %%mm1 \n\t" - "por %%mm2, %%mm1 \n\t" - "movq 14(%1, %%"REG_a"), %%mm0 \n\t" // R BGR BGR B - MOVNTQ" %%mm1, 8(%2, %%"REG_a") \n\t" // B RGB RGB R - "movq 16(%1, %%"REG_a"), %%mm1 \n\t" // GR BGR BGR - "movq 18(%1, %%"REG_a"), %%mm2 \n\t" // BGR BGR BG - "pand %%mm6, %%mm0 \n\t" - "pand %%mm7, %%mm1 \n\t" - "pand %%mm5, %%mm2 \n\t" - "por %%mm0, %%mm1 \n\t" - "por %%mm2, %%mm1 \n\t" - MOVNTQ" %%mm1, 16(%2, %%"REG_a") \n\t" - "add $24, %%"REG_a" \n\t" - " js 1b \n\t" - "2: \n\t" - : "+a" (mmx_size) - : "r" (src-mmx_size), "r"(dst-mmx_size) + "test %%"REG_a", %%"REG_a" \n\t" + "jns 2f \n\t" + "movq "MANGLE(mask24r)", %%mm5 \n\t" + "movq "MANGLE(mask24g)", %%mm6 \n\t" + "movq "MANGLE(mask24b)", %%mm7 \n\t" + ASMALIGN(4) + "1: \n\t" + PREFETCH" 32(%1, %%"REG_a") \n\t" + "movq (%1, %%"REG_a"), %%mm0 \n\t" // BGR BGR BG + "movq (%1, %%"REG_a"), %%mm1 \n\t" // BGR BGR BG + "movq 2(%1, %%"REG_a"), %%mm2 \n\t" // R BGR BGR B + "psllq $16, %%mm0 \n\t" // 00 BGR BGR + "pand %%mm5, %%mm0 \n\t" + "pand %%mm6, %%mm1 \n\t" + "pand %%mm7, %%mm2 \n\t" + "por %%mm0, %%mm1 \n\t" + "por %%mm2, %%mm1 \n\t" + "movq 6(%1, %%"REG_a"), %%mm0 \n\t" // BGR BGR BG + MOVNTQ" %%mm1, (%2, %%"REG_a") \n\t" // RGB RGB RG + "movq 8(%1, %%"REG_a"), %%mm1 \n\t" // R BGR BGR B + "movq 10(%1, %%"REG_a"), %%mm2 \n\t" // GR BGR BGR + "pand %%mm7, %%mm0 \n\t" + "pand %%mm5, %%mm1 \n\t" + "pand %%mm6, %%mm2 \n\t" + "por %%mm0, %%mm1 \n\t" + "por %%mm2, %%mm1 \n\t" + "movq 14(%1, %%"REG_a"), %%mm0 \n\t" // R BGR BGR B + MOVNTQ" %%mm1, 8(%2, %%"REG_a") \n\t" // B RGB RGB R + "movq 16(%1, %%"REG_a"), %%mm1 \n\t" // GR BGR BGR + "movq 18(%1, %%"REG_a"), %%mm2 \n\t" // BGR BGR BG + "pand %%mm6, %%mm0 \n\t" + "pand %%mm7, %%mm1 \n\t" + "pand %%mm5, %%mm2 \n\t" + "por %%mm0, %%mm1 \n\t" + "por %%mm2, %%mm1 \n\t" + MOVNTQ" %%mm1, 16(%2, %%"REG_a") \n\t" + "add $24, %%"REG_a" \n\t" + " js 1b \n\t" + "2: \n\t" + : "+a" (mmx_size) + : "r" (src-mmx_size), "r"(dst-mmx_size) ); __asm__ volatile(SFENCE:::"memory"); @@ -1474,37 +1474,37 @@ static inline void RENAME(yuvPlanartoyuy2)(const uint8_t *ysrc, const uint8_t *u #if HAVE_MMX //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" - PREFETCH" 32(%1, %%"REG_a", 2) \n\t" - PREFETCH" 32(%2, %%"REG_a") \n\t" - PREFETCH" 32(%3, %%"REG_a") \n\t" - "movq (%2, %%"REG_a"), %%mm0 \n\t" // U(0) - "movq %%mm0, %%mm2 \n\t" // U(0) - "movq (%3, %%"REG_a"), %%mm1 \n\t" // V(0) - "punpcklbw %%mm1, %%mm0 \n\t" // UVUV UVUV(0) - "punpckhbw %%mm1, %%mm2 \n\t" // UVUV UVUV(8) - - "movq (%1, %%"REG_a",2), %%mm3 \n\t" // Y(0) - "movq 8(%1, %%"REG_a",2), %%mm5 \n\t" // Y(8) - "movq %%mm3, %%mm4 \n\t" // Y(0) - "movq %%mm5, %%mm6 \n\t" // Y(8) - "punpcklbw %%mm0, %%mm3 \n\t" // YUYV YUYV(0) - "punpckhbw %%mm0, %%mm4 \n\t" // YUYV YUYV(4) - "punpcklbw %%mm2, %%mm5 \n\t" // YUYV YUYV(8) - "punpckhbw %%mm2, %%mm6 \n\t" // YUYV YUYV(12) - - MOVNTQ" %%mm3, (%0, %%"REG_a", 4) \n\t" - MOVNTQ" %%mm4, 8(%0, %%"REG_a", 4) \n\t" - MOVNTQ" %%mm5, 16(%0, %%"REG_a", 4) \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" (chromWidth) - : "%"REG_a + "xor %%"REG_a", %%"REG_a" \n\t" + ASMALIGN(4) + "1: \n\t" + PREFETCH" 32(%1, %%"REG_a", 2) \n\t" + PREFETCH" 32(%2, %%"REG_a") \n\t" + PREFETCH" 32(%3, %%"REG_a") \n\t" + "movq (%2, %%"REG_a"), %%mm0 \n\t" // U(0) + "movq %%mm0, %%mm2 \n\t" // U(0) + "movq (%3, %%"REG_a"), %%mm1 \n\t" // V(0) + "punpcklbw %%mm1, %%mm0 \n\t" // UVUV UVUV(0) + "punpckhbw %%mm1, %%mm2 \n\t" // UVUV UVUV(8) + + "movq (%1, %%"REG_a",2), %%mm3 \n\t" // Y(0) + "movq 8(%1, %%"REG_a",2), %%mm5 \n\t" // Y(8) + "movq %%mm3, %%mm4 \n\t" // Y(0) + "movq %%mm5, %%mm6 \n\t" // Y(8) + "punpcklbw %%mm0, %%mm3 \n\t" // YUYV YUYV(0) + "punpckhbw %%mm0, %%mm4 \n\t" // YUYV YUYV(4) + "punpcklbw %%mm2, %%mm5 \n\t" // YUYV YUYV(8) + "punpckhbw %%mm2, %%mm6 \n\t" // YUYV YUYV(12) + + MOVNTQ" %%mm3, (%0, %%"REG_a", 4) \n\t" + MOVNTQ" %%mm4, 8(%0, %%"REG_a", 4) \n\t" + MOVNTQ" %%mm5, 16(%0, %%"REG_a", 4) \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" (chromWidth) + : "%"REG_a ); #else @@ -1597,9 +1597,9 @@ static inline void RENAME(yuvPlanartoyuy2)(const uint8_t *ysrc, const uint8_t *u dst += dstStride; } #if HAVE_MMX -__asm__( EMMS" \n\t" - SFENCE" \n\t" - :::"memory"); + __asm__(EMMS" \n\t" + SFENCE" \n\t" + :::"memory"); #endif } @@ -1626,37 +1626,37 @@ static inline void RENAME(yuvPlanartouyvy)(const uint8_t *ysrc, const uint8_t *u #if HAVE_MMX //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" - PREFETCH" 32(%1, %%"REG_a", 2) \n\t" - PREFETCH" 32(%2, %%"REG_a") \n\t" - PREFETCH" 32(%3, %%"REG_a") \n\t" - "movq (%2, %%"REG_a"), %%mm0 \n\t" // U(0) - "movq %%mm0, %%mm2 \n\t" // U(0) - "movq (%3, %%"REG_a"), %%mm1 \n\t" // V(0) - "punpcklbw %%mm1, %%mm0 \n\t" // UVUV UVUV(0) - "punpckhbw %%mm1, %%mm2 \n\t" // UVUV UVUV(8) - - "movq (%1, %%"REG_a",2), %%mm3 \n\t" // Y(0) - "movq 8(%1, %%"REG_a",2), %%mm5 \n\t" // Y(8) - "movq %%mm0, %%mm4 \n\t" // Y(0) - "movq %%mm2, %%mm6 \n\t" // Y(8) - "punpcklbw %%mm3, %%mm0 \n\t" // YUYV YUYV(0) - "punpckhbw %%mm3, %%mm4 \n\t" // YUYV YUYV(4) - "punpcklbw %%mm5, %%mm2 \n\t" // YUYV YUYV(8) - "punpckhbw %%mm5, %%mm6 \n\t" // YUYV YUYV(12) - - MOVNTQ" %%mm0, (%0, %%"REG_a", 4) \n\t" - MOVNTQ" %%mm4, 8(%0, %%"REG_a", 4) \n\t" - MOVNTQ" %%mm2, 16(%0, %%"REG_a", 4) \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" (chromWidth) - : "%"REG_a + "xor %%"REG_a", %%"REG_a" \n\t" + ASMALIGN(4) + "1: \n\t" + PREFETCH" 32(%1, %%"REG_a", 2) \n\t" + PREFETCH" 32(%2, %%"REG_a") \n\t" + PREFETCH" 32(%3, %%"REG_a") \n\t" + "movq (%2, %%"REG_a"), %%mm0 \n\t" // U(0) + "movq %%mm0, %%mm2 \n\t" // U(0) + "movq (%3, %%"REG_a"), %%mm1 \n\t" // V(0) + "punpcklbw %%mm1, %%mm0 \n\t" // UVUV UVUV(0) + "punpckhbw %%mm1, %%mm2 \n\t" // UVUV UVUV(8) + + "movq (%1, %%"REG_a",2), %%mm3 \n\t" // Y(0) + "movq 8(%1, %%"REG_a",2), %%mm5 \n\t" // Y(8) + "movq %%mm0, %%mm4 \n\t" // Y(0) + "movq %%mm2, %%mm6 \n\t" // Y(8) + "punpcklbw %%mm3, %%mm0 \n\t" // YUYV YUYV(0) + "punpckhbw %%mm3, %%mm4 \n\t" // YUYV YUYV(4) + "punpcklbw %%mm5, %%mm2 \n\t" // YUYV YUYV(8) + "punpckhbw %%mm5, %%mm6 \n\t" // YUYV YUYV(12) + + MOVNTQ" %%mm0, (%0, %%"REG_a", 4) \n\t" + MOVNTQ" %%mm4, 8(%0, %%"REG_a", 4) \n\t" + MOVNTQ" %%mm2, 16(%0, %%"REG_a", 4) \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" (chromWidth) + : "%"REG_a ); #else //FIXME adapt the Alpha ASM code from yv12->yuy2 @@ -1703,9 +1703,9 @@ static inline void RENAME(yuvPlanartouyvy)(const uint8_t *ysrc, const uint8_t *u dst += dstStride; } #if HAVE_MMX -__asm__( EMMS" \n\t" - SFENCE" \n\t" - :::"memory"); + __asm__(EMMS" \n\t" + SFENCE" \n\t" + :::"memory"); #endif } @@ -1755,85 +1755,85 @@ static inline void RENAME(yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t { #if HAVE_MMX __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) - "1: \n\t" - PREFETCH" 64(%0, %%"REG_a", 4) \n\t" - "movq (%0, %%"REG_a", 4), %%mm0 \n\t" // YUYV YUYV(0) - "movq 8(%0, %%"REG_a", 4), %%mm1 \n\t" // YUYV YUYV(4) - "movq %%mm0, %%mm2 \n\t" // YUYV YUYV(0) - "movq %%mm1, %%mm3 \n\t" // YUYV YUYV(4) - "psrlw $8, %%mm0 \n\t" // U0V0 U0V0(0) - "psrlw $8, %%mm1 \n\t" // U0V0 U0V0(4) - "pand %%mm7, %%mm2 \n\t" // Y0Y0 Y0Y0(0) - "pand %%mm7, %%mm3 \n\t" // Y0Y0 Y0Y0(4) - "packuswb %%mm1, %%mm0 \n\t" // UVUV UVUV(0) - "packuswb %%mm3, %%mm2 \n\t" // YYYY YYYY(0) - - MOVNTQ" %%mm2, (%1, %%"REG_a", 2) \n\t" - - "movq 16(%0, %%"REG_a", 4), %%mm1 \n\t" // YUYV YUYV(8) - "movq 24(%0, %%"REG_a", 4), %%mm2 \n\t" // YUYV YUYV(12) - "movq %%mm1, %%mm3 \n\t" // YUYV YUYV(8) - "movq %%mm2, %%mm4 \n\t" // YUYV YUYV(12) - "psrlw $8, %%mm1 \n\t" // U0V0 U0V0(8) - "psrlw $8, %%mm2 \n\t" // U0V0 U0V0(12) - "pand %%mm7, %%mm3 \n\t" // Y0Y0 Y0Y0(8) - "pand %%mm7, %%mm4 \n\t" // Y0Y0 Y0Y0(12) - "packuswb %%mm2, %%mm1 \n\t" // UVUV UVUV(8) - "packuswb %%mm4, %%mm3 \n\t" // YYYY YYYY(8) - - MOVNTQ" %%mm3, 8(%1, %%"REG_a", 2) \n\t" - - "movq %%mm0, %%mm2 \n\t" // UVUV UVUV(0) - "movq %%mm1, %%mm3 \n\t" // UVUV UVUV(8) - "psrlw $8, %%mm0 \n\t" // V0V0 V0V0(0) - "psrlw $8, %%mm1 \n\t" // V0V0 V0V0(8) - "pand %%mm7, %%mm2 \n\t" // U0U0 U0U0(0) - "pand %%mm7, %%mm3 \n\t" // U0U0 U0U0(8) - "packuswb %%mm1, %%mm0 \n\t" // VVVV VVVV(0) - "packuswb %%mm3, %%mm2 \n\t" // UUUU UUUU(0) - - MOVNTQ" %%mm0, (%3, %%"REG_a") \n\t" - 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" (chromWidth) - : "memory", "%"REG_a + "xor %%"REG_a", %%"REG_a" \n\t" + "pcmpeqw %%mm7, %%mm7 \n\t" + "psrlw $8, %%mm7 \n\t" // FF,00,FF,00... + ASMALIGN(4) + "1: \n\t" + PREFETCH" 64(%0, %%"REG_a", 4) \n\t" + "movq (%0, %%"REG_a", 4), %%mm0 \n\t" // YUYV YUYV(0) + "movq 8(%0, %%"REG_a", 4), %%mm1 \n\t" // YUYV YUYV(4) + "movq %%mm0, %%mm2 \n\t" // YUYV YUYV(0) + "movq %%mm1, %%mm3 \n\t" // YUYV YUYV(4) + "psrlw $8, %%mm0 \n\t" // U0V0 U0V0(0) + "psrlw $8, %%mm1 \n\t" // U0V0 U0V0(4) + "pand %%mm7, %%mm2 \n\t" // Y0Y0 Y0Y0(0) + "pand %%mm7, %%mm3 \n\t" // Y0Y0 Y0Y0(4) + "packuswb %%mm1, %%mm0 \n\t" // UVUV UVUV(0) + "packuswb %%mm3, %%mm2 \n\t" // YYYY YYYY(0) + + MOVNTQ" %%mm2, (%1, %%"REG_a", 2) \n\t" + + "movq 16(%0, %%"REG_a", 4), %%mm1 \n\t" // YUYV YUYV(8) + "movq 24(%0, %%"REG_a", 4), %%mm2 \n\t" // YUYV YUYV(12) + "movq %%mm1, %%mm3 \n\t" // YUYV YUYV(8) + "movq %%mm2, %%mm4 \n\t" // YUYV YUYV(12) + "psrlw $8, %%mm1 \n\t" // U0V0 U0V0(8) + "psrlw $8, %%mm2 \n\t" // U0V0 U0V0(12) + "pand %%mm7, %%mm3 \n\t" // Y0Y0 Y0Y0(8) + "pand %%mm7, %%mm4 \n\t" // Y0Y0 Y0Y0(12) + "packuswb %%mm2, %%mm1 \n\t" // UVUV UVUV(8) + "packuswb %%mm4, %%mm3 \n\t" // YYYY YYYY(8) + + MOVNTQ" %%mm3, 8(%1, %%"REG_a", 2) \n\t" + + "movq %%mm0, %%mm2 \n\t" // UVUV UVUV(0) + "movq %%mm1, %%mm3 \n\t" // UVUV UVUV(8) + "psrlw $8, %%mm0 \n\t" // V0V0 V0V0(0) + "psrlw $8, %%mm1 \n\t" // V0V0 V0V0(8) + "pand %%mm7, %%mm2 \n\t" // U0U0 U0U0(0) + "pand %%mm7, %%mm3 \n\t" // U0U0 U0U0(8) + "packuswb %%mm1, %%mm0 \n\t" // VVVV VVVV(0) + "packuswb %%mm3, %%mm2 \n\t" // UUUU UUUU(0) + + MOVNTQ" %%mm0, (%3, %%"REG_a") \n\t" + 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" (chromWidth) + : "memory", "%"REG_a ); ydst += lumStride; src += srcStride; __asm__ volatile( - "xor %%"REG_a", %%"REG_a" \n\t" - ASMALIGN(4) - "1: \n\t" - PREFETCH" 64(%0, %%"REG_a", 4) \n\t" - "movq (%0, %%"REG_a", 4), %%mm0 \n\t" // YUYV YUYV(0) - "movq 8(%0, %%"REG_a", 4), %%mm1 \n\t" // YUYV YUYV(4) - "movq 16(%0, %%"REG_a", 4), %%mm2 \n\t" // YUYV YUYV(8) - "movq 24(%0, %%"REG_a", 4), %%mm3 \n\t" // YUYV YUYV(12) - "pand %%mm7, %%mm0 \n\t" // Y0Y0 Y0Y0(0) - "pand %%mm7, %%mm1 \n\t" // Y0Y0 Y0Y0(4) - "pand %%mm7, %%mm2 \n\t" // Y0Y0 Y0Y0(8) - "pand %%mm7, %%mm3 \n\t" // Y0Y0 Y0Y0(12) - "packuswb %%mm1, %%mm0 \n\t" // YYYY YYYY(0) - "packuswb %%mm3, %%mm2 \n\t" // YYYY YYYY(8) - - MOVNTQ" %%mm0, (%1, %%"REG_a", 2) \n\t" - MOVNTQ" %%mm2, 8(%1, %%"REG_a", 2) \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" (chromWidth) - : "memory", "%"REG_a + "xor %%"REG_a", %%"REG_a" \n\t" + ASMALIGN(4) + "1: \n\t" + PREFETCH" 64(%0, %%"REG_a", 4) \n\t" + "movq (%0, %%"REG_a", 4), %%mm0 \n\t" // YUYV YUYV(0) + "movq 8(%0, %%"REG_a", 4), %%mm1 \n\t" // YUYV YUYV(4) + "movq 16(%0, %%"REG_a", 4), %%mm2 \n\t" // YUYV YUYV(8) + "movq 24(%0, %%"REG_a", 4), %%mm3 \n\t" // YUYV YUYV(12) + "pand %%mm7, %%mm0 \n\t" // Y0Y0 Y0Y0(0) + "pand %%mm7, %%mm1 \n\t" // Y0Y0 Y0Y0(4) + "pand %%mm7, %%mm2 \n\t" // Y0Y0 Y0Y0(8) + "pand %%mm7, %%mm3 \n\t" // Y0Y0 Y0Y0(12) + "packuswb %%mm1, %%mm0 \n\t" // YYYY YYYY(0) + "packuswb %%mm3, %%mm2 \n\t" // YYYY YYYY(8) + + MOVNTQ" %%mm0, (%1, %%"REG_a", 2) \n\t" + MOVNTQ" %%mm2, 8(%1, %%"REG_a", 2) \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" (chromWidth) + : "memory", "%"REG_a ); #else long i; @@ -1859,9 +1859,9 @@ static inline void RENAME(yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t src += srcStride; } #if HAVE_MMX -__asm__ volatile( EMMS" \n\t" - SFENCE" \n\t" - :::"memory"); + __asm__ volatile(EMMS" \n\t" + SFENCE" \n\t" + :::"memory"); #endif } @@ -1888,51 +1888,51 @@ static inline void RENAME(planar2x)(const uint8_t *src, uint8_t *dst, long srcWi } dst[2*srcWidth-1]= src[srcWidth-1]; - dst+= dstStride; + dst+= dstStride; for (y=1; y<srcHeight; y++){ #if HAVE_MMX2 || HAVE_AMD3DNOW const x86_reg mmxSize= srcWidth&~15; __asm__ volatile( - "mov %4, %%"REG_a" \n\t" - "1: \n\t" - "movq (%0, %%"REG_a"), %%mm0 \n\t" - "movq (%1, %%"REG_a"), %%mm1 \n\t" - "movq 1(%0, %%"REG_a"), %%mm2 \n\t" - "movq 1(%1, %%"REG_a"), %%mm3 \n\t" - "movq -1(%0, %%"REG_a"), %%mm4 \n\t" - "movq -1(%1, %%"REG_a"), %%mm5 \n\t" - PAVGB" %%mm0, %%mm5 \n\t" - PAVGB" %%mm0, %%mm3 \n\t" - PAVGB" %%mm0, %%mm5 \n\t" - PAVGB" %%mm0, %%mm3 \n\t" - PAVGB" %%mm1, %%mm4 \n\t" - PAVGB" %%mm1, %%mm2 \n\t" - PAVGB" %%mm1, %%mm4 \n\t" - PAVGB" %%mm1, %%mm2 \n\t" - "movq %%mm5, %%mm7 \n\t" - "movq %%mm4, %%mm6 \n\t" - "punpcklbw %%mm3, %%mm5 \n\t" - "punpckhbw %%mm3, %%mm7 \n\t" - "punpcklbw %%mm2, %%mm4 \n\t" - "punpckhbw %%mm2, %%mm6 \n\t" + "mov %4, %%"REG_a" \n\t" + "1: \n\t" + "movq (%0, %%"REG_a"), %%mm0 \n\t" + "movq (%1, %%"REG_a"), %%mm1 \n\t" + "movq 1(%0, %%"REG_a"), %%mm2 \n\t" + "movq 1(%1, %%"REG_a"), %%mm3 \n\t" + "movq -1(%0, %%"REG_a"), %%mm4 \n\t" + "movq -1(%1, %%"REG_a"), %%mm5 \n\t" + PAVGB" %%mm0, %%mm5 \n\t" + PAVGB" %%mm0, %%mm3 \n\t" + PAVGB" %%mm0, %%mm5 \n\t" + PAVGB" %%mm0, %%mm3 \n\t" + PAVGB" %%mm1, %%mm4 \n\t" + PAVGB" %%mm1, %%mm2 \n\t" + PAVGB" %%mm1, %%mm4 \n\t" + PAVGB" %%mm1, %%mm2 \n\t" + "movq %%mm5, %%mm7 \n\t" + "movq %%mm4, %%mm6 \n\t" + "punpcklbw %%mm3, %%mm5 \n\t" + "punpckhbw %%mm3, %%mm7 \n\t" + "punpcklbw %%mm2, %%mm4 \n\t" + "punpckhbw %%mm2, %%mm6 \n\t" #if 1 - MOVNTQ" %%mm5, (%2, %%"REG_a", 2) \n\t" - MOVNTQ" %%mm7, 8(%2, %%"REG_a", 2) \n\t" - MOVNTQ" %%mm4, (%3, %%"REG_a", 2) \n\t" - MOVNTQ" %%mm6, 8(%3, %%"REG_a", 2) \n\t" + MOVNTQ" %%mm5, (%2, %%"REG_a", 2) \n\t" + MOVNTQ" %%mm7, 8(%2, %%"REG_a", 2) \n\t" + MOVNTQ" %%mm4, (%3, %%"REG_a", 2) \n\t" + MOVNTQ" %%mm6, 8(%3, %%"REG_a", 2) \n\t" #else - "movq %%mm5, (%2, %%"REG_a", 2) \n\t" - "movq %%mm7, 8(%2, %%"REG_a", 2) \n\t" - "movq %%mm4, (%3, %%"REG_a", 2) \n\t" - "movq %%mm6, 8(%3, %%"REG_a", 2) \n\t" -#endif - "add $8, %%"REG_a" \n\t" - " js 1b \n\t" - :: "r" (src + mmxSize ), "r" (src + srcStride + mmxSize ), - "r" (dst + mmxSize*2), "r" (dst + dstStride + mmxSize*2), - "g" (-mmxSize) - : "%"REG_a + "movq %%mm5, (%2, %%"REG_a", 2) \n\t" + "movq %%mm7, 8(%2, %%"REG_a", 2) \n\t" + "movq %%mm4, (%3, %%"REG_a", 2) \n\t" + "movq %%mm6, 8(%3, %%"REG_a", 2) \n\t" +#endif + "add $8, %%"REG_a" \n\t" + " js 1b \n\t" + :: "r" (src + mmxSize ), "r" (src + srcStride + mmxSize ), + "r" (dst + mmxSize*2), "r" (dst + dstStride + mmxSize*2), + "g" (-mmxSize) + : "%"REG_a ); #else @@ -1971,9 +1971,9 @@ static inline void RENAME(planar2x)(const uint8_t *src, uint8_t *dst, long srcWi #endif #if HAVE_MMX -__asm__ volatile( EMMS" \n\t" - SFENCE" \n\t" - :::"memory"); + __asm__ volatile(EMMS" \n\t" + SFENCE" \n\t" + :::"memory"); #endif } @@ -1993,85 +1993,85 @@ static inline void RENAME(uyvytoyv12)(const uint8_t *src, uint8_t *ydst, uint8_t { #if HAVE_MMX __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) - "1: \n\t" - PREFETCH" 64(%0, %%"REG_a", 4) \n\t" - "movq (%0, %%"REG_a", 4), %%mm0 \n\t" // UYVY UYVY(0) - "movq 8(%0, %%"REG_a", 4), %%mm1 \n\t" // UYVY UYVY(4) - "movq %%mm0, %%mm2 \n\t" // UYVY UYVY(0) - "movq %%mm1, %%mm3 \n\t" // UYVY UYVY(4) - "pand %%mm7, %%mm0 \n\t" // U0V0 U0V0(0) - "pand %%mm7, %%mm1 \n\t" // U0V0 U0V0(4) - "psrlw $8, %%mm2 \n\t" // Y0Y0 Y0Y0(0) - "psrlw $8, %%mm3 \n\t" // Y0Y0 Y0Y0(4) - "packuswb %%mm1, %%mm0 \n\t" // UVUV UVUV(0) - "packuswb %%mm3, %%mm2 \n\t" // YYYY YYYY(0) - - MOVNTQ" %%mm2, (%1, %%"REG_a", 2) \n\t" - - "movq 16(%0, %%"REG_a", 4), %%mm1 \n\t" // UYVY UYVY(8) - "movq 24(%0, %%"REG_a", 4), %%mm2 \n\t" // UYVY UYVY(12) - "movq %%mm1, %%mm3 \n\t" // UYVY UYVY(8) - "movq %%mm2, %%mm4 \n\t" // UYVY UYVY(12) - "pand %%mm7, %%mm1 \n\t" // U0V0 U0V0(8) - "pand %%mm7, %%mm2 \n\t" // U0V0 U0V0(12) - "psrlw $8, %%mm3 \n\t" // Y0Y0 Y0Y0(8) - "psrlw $8, %%mm4 \n\t" // Y0Y0 Y0Y0(12) - "packuswb %%mm2, %%mm1 \n\t" // UVUV UVUV(8) - "packuswb %%mm4, %%mm3 \n\t" // YYYY YYYY(8) - - MOVNTQ" %%mm3, 8(%1, %%"REG_a", 2) \n\t" - - "movq %%mm0, %%mm2 \n\t" // UVUV UVUV(0) - "movq %%mm1, %%mm3 \n\t" // UVUV UVUV(8) - "psrlw $8, %%mm0 \n\t" // V0V0 V0V0(0) - "psrlw $8, %%mm1 \n\t" // V0V0 V0V0(8) - "pand %%mm7, %%mm2 \n\t" // U0U0 U0U0(0) - "pand %%mm7, %%mm3 \n\t" // U0U0 U0U0(8) - "packuswb %%mm1, %%mm0 \n\t" // VVVV VVVV(0) - "packuswb %%mm3, %%mm2 \n\t" // UUUU UUUU(0) - - MOVNTQ" %%mm0, (%3, %%"REG_a") \n\t" - 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" (chromWidth) - : "memory", "%"REG_a + "xor %%"REG_a", %%"REG_a" \n\t" + "pcmpeqw %%mm7, %%mm7 \n\t" + "psrlw $8, %%mm7 \n\t" // FF,00,FF,00... + ASMALIGN(4) + "1: \n\t" + PREFETCH" 64(%0, %%"REG_a", 4) \n\t" + "movq (%0, %%"REG_a", 4), %%mm0 \n\t" // UYVY UYVY(0) + "movq 8(%0, %%"REG_a", 4), %%mm1 \n\t" // UYVY UYVY(4) + "movq %%mm0, %%mm2 \n\t" // UYVY UYVY(0) + "movq %%mm1, %%mm3 \n\t" // UYVY UYVY(4) + "pand %%mm7, %%mm0 \n\t" // U0V0 U0V0(0) + "pand %%mm7, %%mm1 \n\t" // U0V0 U0V0(4) + "psrlw $8, %%mm2 \n\t" // Y0Y0 Y0Y0(0) + "psrlw $8, %%mm3 \n\t" // Y0Y0 Y0Y0(4) + "packuswb %%mm1, %%mm0 \n\t" // UVUV UVUV(0) + "packuswb %%mm3, %%mm2 \n\t" // YYYY YYYY(0) + + MOVNTQ" %%mm2, (%1, %%"REG_a", 2) \n\t" + + "movq 16(%0, %%"REG_a", 4), %%mm1 \n\t" // UYVY UYVY(8) + "movq 24(%0, %%"REG_a", 4), %%mm2 \n\t" // UYVY UYVY(12) + "movq %%mm1, %%mm3 \n\t" // UYVY UYVY(8) + "movq %%mm2, %%mm4 \n\t" // UYVY UYVY(12) + "pand %%mm7, %%mm1 \n\t" // U0V0 U0V0(8) + "pand %%mm7, %%mm2 \n\t" // U0V0 U0V0(12) + "psrlw $8, %%mm3 \n\t" // Y0Y0 Y0Y0(8) + "psrlw $8, %%mm4 \n\t" // Y0Y0 Y0Y0(12) + "packuswb %%mm2, %%mm1 \n\t" // UVUV UVUV(8) + "packuswb %%mm4, %%mm3 \n\t" // YYYY YYYY(8) + + MOVNTQ" %%mm3, 8(%1, %%"REG_a", 2) \n\t" + + "movq %%mm0, %%mm2 \n\t" // UVUV UVUV(0) + "movq %%mm1, %%mm3 \n\t" // UVUV UVUV(8) + "psrlw $8, %%mm0 \n\t" // V0V0 V0V0(0) + "psrlw $8, %%mm1 \n\t" // V0V0 V0V0(8) + "pand %%mm7, %%mm2 \n\t" // U0U0 U0U0(0) + "pand %%mm7, %%mm3 \n\t" // U0U0 U0U0(8) + "packuswb %%mm1, %%mm0 \n\t" // VVVV VVVV(0) + "packuswb %%mm3, %%mm2 \n\t" // UUUU UUUU(0) + + MOVNTQ" %%mm0, (%3, %%"REG_a") \n\t" + 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" (chromWidth) + : "memory", "%"REG_a ); ydst += lumStride; src += srcStride; __asm__ volatile( - "xor %%"REG_a", %%"REG_a" \n\t" - ASMALIGN(4) - "1: \n\t" - PREFETCH" 64(%0, %%"REG_a", 4) \n\t" - "movq (%0, %%"REG_a", 4), %%mm0 \n\t" // YUYV YUYV(0) - "movq 8(%0, %%"REG_a", 4), %%mm1 \n\t" // YUYV YUYV(4) - "movq 16(%0, %%"REG_a", 4), %%mm2 \n\t" // YUYV YUYV(8) - "movq 24(%0, %%"REG_a", 4), %%mm3 \n\t" // YUYV YUYV(12) - "psrlw $8, %%mm0 \n\t" // Y0Y0 Y0Y0(0) - "psrlw $8, %%mm1 \n\t" // Y0Y0 Y0Y0(4) - "psrlw $8, %%mm2 \n\t" // Y0Y0 Y0Y0(8) - "psrlw $8, %%mm3 \n\t" // Y0Y0 Y0Y0(12) - "packuswb %%mm1, %%mm0 \n\t" // YYYY YYYY(0) - "packuswb %%mm3, %%mm2 \n\t" // YYYY YYYY(8) - - MOVNTQ" %%mm0, (%1, %%"REG_a", 2) \n\t" - MOVNTQ" %%mm2, 8(%1, %%"REG_a", 2) \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" (chromWidth) - : "memory", "%"REG_a + "xor %%"REG_a", %%"REG_a" \n\t" + ASMALIGN(4) + "1: \n\t" + PREFETCH" 64(%0, %%"REG_a", 4) \n\t" + "movq (%0, %%"REG_a", 4), %%mm0 \n\t" // YUYV YUYV(0) + "movq 8(%0, %%"REG_a", 4), %%mm1 \n\t" // YUYV YUYV(4) + "movq 16(%0, %%"REG_a", 4), %%mm2 \n\t" // YUYV YUYV(8) + "movq 24(%0, %%"REG_a", 4), %%mm3 \n\t" // YUYV YUYV(12) + "psrlw $8, %%mm0 \n\t" // Y0Y0 Y0Y0(0) + "psrlw $8, %%mm1 \n\t" // Y0Y0 Y0Y0(4) + "psrlw $8, %%mm2 \n\t" // Y0Y0 Y0Y0(8) + "psrlw $8, %%mm3 \n\t" // Y0Y0 Y0Y0(12) + "packuswb %%mm1, %%mm0 \n\t" // YYYY YYYY(0) + "packuswb %%mm3, %%mm2 \n\t" // YYYY YYYY(8) + + MOVNTQ" %%mm0, (%1, %%"REG_a", 2) \n\t" + MOVNTQ" %%mm2, 8(%1, %%"REG_a", 2) \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" (chromWidth) + : "memory", "%"REG_a ); #else long i; @@ -2097,9 +2097,9 @@ static inline void RENAME(uyvytoyv12)(const uint8_t *src, uint8_t *ydst, uint8_t src += srcStride; } #if HAVE_MMX -__asm__ volatile( EMMS" \n\t" - SFENCE" \n\t" - :::"memory"); + __asm__ volatile(EMMS" \n\t" + SFENCE" \n\t" + :::"memory"); #endif } @@ -2123,225 +2123,225 @@ static inline void RENAME(rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_ for (i=0; i<2; i++) { __asm__ volatile( - "mov %2, %%"REG_a" \n\t" - "movq "MANGLE(ff_bgr2YCoeff)", %%mm6 \n\t" + "mov %2, %%"REG_a" \n\t" + "movq "MANGLE(ff_bgr2YCoeff)", %%mm6 \n\t" + "movq "MANGLE(ff_w1111)", %%mm5 \n\t" + "pxor %%mm7, %%mm7 \n\t" + "lea (%%"REG_a", %%"REG_a", 2), %%"REG_d" \n\t" + ASMALIGN(4) + "1: \n\t" + PREFETCH" 64(%0, %%"REG_d") \n\t" + "movd (%0, %%"REG_d"), %%mm0 \n\t" + "movd 3(%0, %%"REG_d"), %%mm1 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpcklbw %%mm7, %%mm1 \n\t" + "movd 6(%0, %%"REG_d"), %%mm2 \n\t" + "movd 9(%0, %%"REG_d"), %%mm3 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" + "punpcklbw %%mm7, %%mm3 \n\t" + "pmaddwd %%mm6, %%mm0 \n\t" + "pmaddwd %%mm6, %%mm1 \n\t" + "pmaddwd %%mm6, %%mm2 \n\t" + "pmaddwd %%mm6, %%mm3 \n\t" +#ifndef FAST_BGR2YV12 + "psrad $8, %%mm0 \n\t" + "psrad $8, %%mm1 \n\t" + "psrad $8, %%mm2 \n\t" + "psrad $8, %%mm3 \n\t" +#endif + "packssdw %%mm1, %%mm0 \n\t" + "packssdw %%mm3, %%mm2 \n\t" + "pmaddwd %%mm5, %%mm0 \n\t" + "pmaddwd %%mm5, %%mm2 \n\t" + "packssdw %%mm2, %%mm0 \n\t" + "psraw $7, %%mm0 \n\t" + + "movd 12(%0, %%"REG_d"), %%mm4 \n\t" + "movd 15(%0, %%"REG_d"), %%mm1 \n\t" + "punpcklbw %%mm7, %%mm4 \n\t" + "punpcklbw %%mm7, %%mm1 \n\t" + "movd 18(%0, %%"REG_d"), %%mm2 \n\t" + "movd 21(%0, %%"REG_d"), %%mm3 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" + "punpcklbw %%mm7, %%mm3 \n\t" + "pmaddwd %%mm6, %%mm4 \n\t" + "pmaddwd %%mm6, %%mm1 \n\t" + "pmaddwd %%mm6, %%mm2 \n\t" + "pmaddwd %%mm6, %%mm3 \n\t" +#ifndef FAST_BGR2YV12 + "psrad $8, %%mm4 \n\t" + "psrad $8, %%mm1 \n\t" + "psrad $8, %%mm2 \n\t" + "psrad $8, %%mm3 \n\t" +#endif + "packssdw %%mm1, %%mm4 \n\t" + "packssdw %%mm3, %%mm2 \n\t" + "pmaddwd %%mm5, %%mm4 \n\t" + "pmaddwd %%mm5, %%mm2 \n\t" + "add $24, %%"REG_d" \n\t" + "packssdw %%mm2, %%mm4 \n\t" + "psraw $7, %%mm4 \n\t" + + "packuswb %%mm4, %%mm0 \n\t" + "paddusb "MANGLE(ff_bgr2YOffset)", %%mm0 \n\t" + + MOVNTQ" %%mm0, (%1, %%"REG_a") \n\t" + "add $8, %%"REG_a" \n\t" + " js 1b \n\t" + : : "r" (src+width*3), "r" (ydst+width), "g" ((x86_reg)-width) + : "%"REG_a, "%"REG_d + ); + ydst += lumStride; + src += srcStride; + } + src -= srcStride*2; + __asm__ volatile( + "mov %4, %%"REG_a" \n\t" "movq "MANGLE(ff_w1111)", %%mm5 \n\t" + "movq "MANGLE(ff_bgr2UCoeff)", %%mm6 \n\t" "pxor %%mm7, %%mm7 \n\t" "lea (%%"REG_a", %%"REG_a", 2), %%"REG_d" \n\t" + "add %%"REG_d", %%"REG_d" \n\t" ASMALIGN(4) "1: \n\t" PREFETCH" 64(%0, %%"REG_d") \n\t" + PREFETCH" 64(%1, %%"REG_d") \n\t" +#if HAVE_MMX2 || HAVE_AMD3DNOW + "movq (%0, %%"REG_d"), %%mm0 \n\t" + "movq (%1, %%"REG_d"), %%mm1 \n\t" + "movq 6(%0, %%"REG_d"), %%mm2 \n\t" + "movq 6(%1, %%"REG_d"), %%mm3 \n\t" + PAVGB" %%mm1, %%mm0 \n\t" + PAVGB" %%mm3, %%mm2 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm2, %%mm3 \n\t" + "psrlq $24, %%mm0 \n\t" + "psrlq $24, %%mm2 \n\t" + PAVGB" %%mm1, %%mm0 \n\t" + PAVGB" %%mm3, %%mm2 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" +#else "movd (%0, %%"REG_d"), %%mm0 \n\t" - "movd 3(%0, %%"REG_d"), %%mm1 \n\t" + "movd (%1, %%"REG_d"), %%mm1 \n\t" + "movd 3(%0, %%"REG_d"), %%mm2 \n\t" + "movd 3(%1, %%"REG_d"), %%mm3 \n\t" "punpcklbw %%mm7, %%mm0 \n\t" "punpcklbw %%mm7, %%mm1 \n\t" - "movd 6(%0, %%"REG_d"), %%mm2 \n\t" - "movd 9(%0, %%"REG_d"), %%mm3 \n\t" "punpcklbw %%mm7, %%mm2 \n\t" "punpcklbw %%mm7, %%mm3 \n\t" + "paddw %%mm1, %%mm0 \n\t" + "paddw %%mm3, %%mm2 \n\t" + "paddw %%mm2, %%mm0 \n\t" + "movd 6(%0, %%"REG_d"), %%mm4 \n\t" + "movd 6(%1, %%"REG_d"), %%mm1 \n\t" + "movd 9(%0, %%"REG_d"), %%mm2 \n\t" + "movd 9(%1, %%"REG_d"), %%mm3 \n\t" + "punpcklbw %%mm7, %%mm4 \n\t" + "punpcklbw %%mm7, %%mm1 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" + "punpcklbw %%mm7, %%mm3 \n\t" + "paddw %%mm1, %%mm4 \n\t" + "paddw %%mm3, %%mm2 \n\t" + "paddw %%mm4, %%mm2 \n\t" + "psrlw $2, %%mm0 \n\t" + "psrlw $2, %%mm2 \n\t" +#endif + "movq "MANGLE(ff_bgr2VCoeff)", %%mm1 \n\t" + "movq "MANGLE(ff_bgr2VCoeff)", %%mm3 \n\t" + + "pmaddwd %%mm0, %%mm1 \n\t" + "pmaddwd %%mm2, %%mm3 \n\t" "pmaddwd %%mm6, %%mm0 \n\t" - "pmaddwd %%mm6, %%mm1 \n\t" "pmaddwd %%mm6, %%mm2 \n\t" - "pmaddwd %%mm6, %%mm3 \n\t" #ifndef FAST_BGR2YV12 "psrad $8, %%mm0 \n\t" "psrad $8, %%mm1 \n\t" "psrad $8, %%mm2 \n\t" "psrad $8, %%mm3 \n\t" #endif - "packssdw %%mm1, %%mm0 \n\t" - "packssdw %%mm3, %%mm2 \n\t" - "pmaddwd %%mm5, %%mm0 \n\t" - "pmaddwd %%mm5, %%mm2 \n\t" "packssdw %%mm2, %%mm0 \n\t" + "packssdw %%mm3, %%mm1 \n\t" + "pmaddwd %%mm5, %%mm0 \n\t" + "pmaddwd %%mm5, %%mm1 \n\t" + "packssdw %%mm1, %%mm0 \n\t" // V1 V0 U1 U0 "psraw $7, %%mm0 \n\t" +#if HAVE_MMX2 || HAVE_AMD3DNOW + "movq 12(%0, %%"REG_d"), %%mm4 \n\t" + "movq 12(%1, %%"REG_d"), %%mm1 \n\t" + "movq 18(%0, %%"REG_d"), %%mm2 \n\t" + "movq 18(%1, %%"REG_d"), %%mm3 \n\t" + PAVGB" %%mm1, %%mm4 \n\t" + PAVGB" %%mm3, %%mm2 \n\t" + "movq %%mm4, %%mm1 \n\t" + "movq %%mm2, %%mm3 \n\t" + "psrlq $24, %%mm4 \n\t" + "psrlq $24, %%mm2 \n\t" + PAVGB" %%mm1, %%mm4 \n\t" + PAVGB" %%mm3, %%mm2 \n\t" + "punpcklbw %%mm7, %%mm4 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" +#else "movd 12(%0, %%"REG_d"), %%mm4 \n\t" - "movd 15(%0, %%"REG_d"), %%mm1 \n\t" + "movd 12(%1, %%"REG_d"), %%mm1 \n\t" + "movd 15(%0, %%"REG_d"), %%mm2 \n\t" + "movd 15(%1, %%"REG_d"), %%mm3 \n\t" "punpcklbw %%mm7, %%mm4 \n\t" "punpcklbw %%mm7, %%mm1 \n\t" - "movd 18(%0, %%"REG_d"), %%mm2 \n\t" - "movd 21(%0, %%"REG_d"), %%mm3 \n\t" "punpcklbw %%mm7, %%mm2 \n\t" "punpcklbw %%mm7, %%mm3 \n\t" + "paddw %%mm1, %%mm4 \n\t" + "paddw %%mm3, %%mm2 \n\t" + "paddw %%mm2, %%mm4 \n\t" + "movd 18(%0, %%"REG_d"), %%mm5 \n\t" + "movd 18(%1, %%"REG_d"), %%mm1 \n\t" + "movd 21(%0, %%"REG_d"), %%mm2 \n\t" + "movd 21(%1, %%"REG_d"), %%mm3 \n\t" + "punpcklbw %%mm7, %%mm5 \n\t" + "punpcklbw %%mm7, %%mm1 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" + "punpcklbw %%mm7, %%mm3 \n\t" + "paddw %%mm1, %%mm5 \n\t" + "paddw %%mm3, %%mm2 \n\t" + "paddw %%mm5, %%mm2 \n\t" + "movq "MANGLE(ff_w1111)", %%mm5 \n\t" + "psrlw $2, %%mm4 \n\t" + "psrlw $2, %%mm2 \n\t" +#endif + "movq "MANGLE(ff_bgr2VCoeff)", %%mm1 \n\t" + "movq "MANGLE(ff_bgr2VCoeff)", %%mm3 \n\t" + + "pmaddwd %%mm4, %%mm1 \n\t" + "pmaddwd %%mm2, %%mm3 \n\t" "pmaddwd %%mm6, %%mm4 \n\t" - "pmaddwd %%mm6, %%mm1 \n\t" "pmaddwd %%mm6, %%mm2 \n\t" - "pmaddwd %%mm6, %%mm3 \n\t" #ifndef FAST_BGR2YV12 "psrad $8, %%mm4 \n\t" "psrad $8, %%mm1 \n\t" "psrad $8, %%mm2 \n\t" "psrad $8, %%mm3 \n\t" #endif - "packssdw %%mm1, %%mm4 \n\t" - "packssdw %%mm3, %%mm2 \n\t" + "packssdw %%mm2, %%mm4 \n\t" + "packssdw %%mm3, %%mm1 \n\t" "pmaddwd %%mm5, %%mm4 \n\t" - "pmaddwd %%mm5, %%mm2 \n\t" + "pmaddwd %%mm5, %%mm1 \n\t" "add $24, %%"REG_d" \n\t" - "packssdw %%mm2, %%mm4 \n\t" + "packssdw %%mm1, %%mm4 \n\t" // V3 V2 U3 U2 "psraw $7, %%mm4 \n\t" - "packuswb %%mm4, %%mm0 \n\t" - "paddusb "MANGLE(ff_bgr2YOffset)", %%mm0 \n\t" - - MOVNTQ" %%mm0, (%1, %%"REG_a") \n\t" - "add $8, %%"REG_a" \n\t" + "movq %%mm0, %%mm1 \n\t" + "punpckldq %%mm4, %%mm0 \n\t" + "punpckhdq %%mm4, %%mm1 \n\t" + "packsswb %%mm1, %%mm0 \n\t" + "paddb "MANGLE(ff_bgr2UVOffset)", %%mm0 \n\t" + "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+width*3), "r" (ydst+width), "g" ((x86_reg)-width) + : : "r" (src+chromWidth*6), "r" (src+srcStride+chromWidth*6), "r" (udst+chromWidth), "r" (vdst+chromWidth), "g" (-chromWidth) : "%"REG_a, "%"REG_d - ); - ydst += lumStride; - src += srcStride; - } - src -= srcStride*2; - __asm__ volatile( - "mov %4, %%"REG_a" \n\t" - "movq "MANGLE(ff_w1111)", %%mm5 \n\t" - "movq "MANGLE(ff_bgr2UCoeff)", %%mm6 \n\t" - "pxor %%mm7, %%mm7 \n\t" - "lea (%%"REG_a", %%"REG_a", 2), %%"REG_d" \n\t" - "add %%"REG_d", %%"REG_d" \n\t" - ASMALIGN(4) - "1: \n\t" - PREFETCH" 64(%0, %%"REG_d") \n\t" - PREFETCH" 64(%1, %%"REG_d") \n\t" -#if HAVE_MMX2 || HAVE_AMD3DNOW - "movq (%0, %%"REG_d"), %%mm0 \n\t" - "movq (%1, %%"REG_d"), %%mm1 \n\t" - "movq 6(%0, %%"REG_d"), %%mm2 \n\t" - "movq 6(%1, %%"REG_d"), %%mm3 \n\t" - PAVGB" %%mm1, %%mm0 \n\t" - PAVGB" %%mm3, %%mm2 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm2, %%mm3 \n\t" - "psrlq $24, %%mm0 \n\t" - "psrlq $24, %%mm2 \n\t" - PAVGB" %%mm1, %%mm0 \n\t" - PAVGB" %%mm3, %%mm2 \n\t" - "punpcklbw %%mm7, %%mm0 \n\t" - "punpcklbw %%mm7, %%mm2 \n\t" -#else - "movd (%0, %%"REG_d"), %%mm0 \n\t" - "movd (%1, %%"REG_d"), %%mm1 \n\t" - "movd 3(%0, %%"REG_d"), %%mm2 \n\t" - "movd 3(%1, %%"REG_d"), %%mm3 \n\t" - "punpcklbw %%mm7, %%mm0 \n\t" - "punpcklbw %%mm7, %%mm1 \n\t" - "punpcklbw %%mm7, %%mm2 \n\t" - "punpcklbw %%mm7, %%mm3 \n\t" - "paddw %%mm1, %%mm0 \n\t" - "paddw %%mm3, %%mm2 \n\t" - "paddw %%mm2, %%mm0 \n\t" - "movd 6(%0, %%"REG_d"), %%mm4 \n\t" - "movd 6(%1, %%"REG_d"), %%mm1 \n\t" - "movd 9(%0, %%"REG_d"), %%mm2 \n\t" - "movd 9(%1, %%"REG_d"), %%mm3 \n\t" - "punpcklbw %%mm7, %%mm4 \n\t" - "punpcklbw %%mm7, %%mm1 \n\t" - "punpcklbw %%mm7, %%mm2 \n\t" - "punpcklbw %%mm7, %%mm3 \n\t" - "paddw %%mm1, %%mm4 \n\t" - "paddw %%mm3, %%mm2 \n\t" - "paddw %%mm4, %%mm2 \n\t" - "psrlw $2, %%mm0 \n\t" - "psrlw $2, %%mm2 \n\t" -#endif - "movq "MANGLE(ff_bgr2VCoeff)", %%mm1 \n\t" - "movq "MANGLE(ff_bgr2VCoeff)", %%mm3 \n\t" - - "pmaddwd %%mm0, %%mm1 \n\t" - "pmaddwd %%mm2, %%mm3 \n\t" - "pmaddwd %%mm6, %%mm0 \n\t" - "pmaddwd %%mm6, %%mm2 \n\t" -#ifndef FAST_BGR2YV12 - "psrad $8, %%mm0 \n\t" - "psrad $8, %%mm1 \n\t" - "psrad $8, %%mm2 \n\t" - "psrad $8, %%mm3 \n\t" -#endif - "packssdw %%mm2, %%mm0 \n\t" - "packssdw %%mm3, %%mm1 \n\t" - "pmaddwd %%mm5, %%mm0 \n\t" - "pmaddwd %%mm5, %%mm1 \n\t" - "packssdw %%mm1, %%mm0 \n\t" // V1 V0 U1 U0 - "psraw $7, %%mm0 \n\t" - -#if HAVE_MMX2 || HAVE_AMD3DNOW - "movq 12(%0, %%"REG_d"), %%mm4 \n\t" - "movq 12(%1, %%"REG_d"), %%mm1 \n\t" - "movq 18(%0, %%"REG_d"), %%mm2 \n\t" - "movq 18(%1, %%"REG_d"), %%mm3 \n\t" - PAVGB" %%mm1, %%mm4 \n\t" - PAVGB" %%mm3, %%mm2 \n\t" - "movq %%mm4, %%mm1 \n\t" - "movq %%mm2, %%mm3 \n\t" - "psrlq $24, %%mm4 \n\t" - "psrlq $24, %%mm2 \n\t" - PAVGB" %%mm1, %%mm4 \n\t" - PAVGB" %%mm3, %%mm2 \n\t" - "punpcklbw %%mm7, %%mm4 \n\t" - "punpcklbw %%mm7, %%mm2 \n\t" -#else - "movd 12(%0, %%"REG_d"), %%mm4 \n\t" - "movd 12(%1, %%"REG_d"), %%mm1 \n\t" - "movd 15(%0, %%"REG_d"), %%mm2 \n\t" - "movd 15(%1, %%"REG_d"), %%mm3 \n\t" - "punpcklbw %%mm7, %%mm4 \n\t" - "punpcklbw %%mm7, %%mm1 \n\t" - "punpcklbw %%mm7, %%mm2 \n\t" - "punpcklbw %%mm7, %%mm3 \n\t" - "paddw %%mm1, %%mm4 \n\t" - "paddw %%mm3, %%mm2 \n\t" - "paddw %%mm2, %%mm4 \n\t" - "movd 18(%0, %%"REG_d"), %%mm5 \n\t" - "movd 18(%1, %%"REG_d"), %%mm1 \n\t" - "movd 21(%0, %%"REG_d"), %%mm2 \n\t" - "movd 21(%1, %%"REG_d"), %%mm3 \n\t" - "punpcklbw %%mm7, %%mm5 \n\t" - "punpcklbw %%mm7, %%mm1 \n\t" - "punpcklbw %%mm7, %%mm2 \n\t" - "punpcklbw %%mm7, %%mm3 \n\t" - "paddw %%mm1, %%mm5 \n\t" - "paddw %%mm3, %%mm2 \n\t" - "paddw %%mm5, %%mm2 \n\t" - "movq "MANGLE(ff_w1111)", %%mm5 \n\t" - "psrlw $2, %%mm4 \n\t" - "psrlw $2, %%mm2 \n\t" -#endif - "movq "MANGLE(ff_bgr2VCoeff)", %%mm1 \n\t" - "movq "MANGLE(ff_bgr2VCoeff)", %%mm3 \n\t" - - "pmaddwd %%mm4, %%mm1 \n\t" - "pmaddwd %%mm2, %%mm3 \n\t" - "pmaddwd %%mm6, %%mm4 \n\t" - "pmaddwd %%mm6, %%mm2 \n\t" -#ifndef FAST_BGR2YV12 - "psrad $8, %%mm4 \n\t" - "psrad $8, %%mm1 \n\t" - "psrad $8, %%mm2 \n\t" - "psrad $8, %%mm3 \n\t" -#endif - "packssdw %%mm2, %%mm4 \n\t" - "packssdw %%mm3, %%mm1 \n\t" - "pmaddwd %%mm5, %%mm4 \n\t" - "pmaddwd %%mm5, %%mm1 \n\t" - "add $24, %%"REG_d" \n\t" - "packssdw %%mm1, %%mm4 \n\t" // V3 V2 U3 U2 - "psraw $7, %%mm4 \n\t" - - "movq %%mm0, %%mm1 \n\t" - "punpckldq %%mm4, %%mm0 \n\t" - "punpckhdq %%mm4, %%mm1 \n\t" - "packsswb %%mm1, %%mm0 \n\t" - "paddb "MANGLE(ff_bgr2UVOffset)", %%mm0 \n\t" - "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+chromWidth*6), "r" (src+srcStride+chromWidth*6), "r" (udst+chromWidth), "r" (vdst+chromWidth), "g" (-chromWidth) - : "%"REG_a, "%"REG_d ); udst += chromStride; @@ -2349,9 +2349,9 @@ static inline void RENAME(rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_ src += srcStride*2; } - __asm__ volatile( EMMS" \n\t" - SFENCE" \n\t" - :::"memory"); + __asm__ volatile(EMMS" \n\t" + SFENCE" \n\t" + :::"memory"); #else y=0; #endif @@ -2418,48 +2418,48 @@ static void RENAME(interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dest, #if HAVE_MMX #if HAVE_SSE2 __asm__( - "xor %%"REG_a", %%"REG_a" \n\t" - "1: \n\t" - PREFETCH" 64(%1, %%"REG_a") \n\t" - PREFETCH" 64(%2, %%"REG_a") \n\t" - "movdqa (%1, %%"REG_a"), %%xmm0 \n\t" - "movdqa (%1, %%"REG_a"), %%xmm1 \n\t" - "movdqa (%2, %%"REG_a"), %%xmm2 \n\t" - "punpcklbw %%xmm2, %%xmm0 \n\t" - "punpckhbw %%xmm2, %%xmm1 \n\t" - "movntdq %%xmm0, (%0, %%"REG_a", 2) \n\t" - "movntdq %%xmm1, 16(%0, %%"REG_a", 2) \n\t" - "add $16, %%"REG_a" \n\t" - "cmp %3, %%"REG_a" \n\t" - " jb 1b \n\t" - ::"r"(dest), "r"(src1), "r"(src2), "r" ((x86_reg)width-15) - : "memory", "%"REG_a"" + "xor %%"REG_a", %%"REG_a" \n\t" + "1: \n\t" + PREFETCH" 64(%1, %%"REG_a") \n\t" + PREFETCH" 64(%2, %%"REG_a") \n\t" + "movdqa (%1, %%"REG_a"), %%xmm0 \n\t" + "movdqa (%1, %%"REG_a"), %%xmm1 \n\t" + "movdqa (%2, %%"REG_a"), %%xmm2 \n\t" + "punpcklbw %%xmm2, %%xmm0 \n\t" + "punpckhbw %%xmm2, %%xmm1 \n\t" + "movntdq %%xmm0, (%0, %%"REG_a", 2) \n\t" + "movntdq %%xmm1, 16(%0, %%"REG_a", 2) \n\t" + "add $16, %%"REG_a" \n\t" + "cmp %3, %%"REG_a" \n\t" + " jb 1b \n\t" + ::"r"(dest), "r"(src1), "r"(src2), "r" ((x86_reg)width-15) + : "memory", "%"REG_a"" ); #else __asm__( - "xor %%"REG_a", %%"REG_a" \n\t" - "1: \n\t" - PREFETCH" 64(%1, %%"REG_a") \n\t" - PREFETCH" 64(%2, %%"REG_a") \n\t" - "movq (%1, %%"REG_a"), %%mm0 \n\t" - "movq 8(%1, %%"REG_a"), %%mm2 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm2, %%mm3 \n\t" - "movq (%2, %%"REG_a"), %%mm4 \n\t" - "movq 8(%2, %%"REG_a"), %%mm5 \n\t" - "punpcklbw %%mm4, %%mm0 \n\t" - "punpckhbw %%mm4, %%mm1 \n\t" - "punpcklbw %%mm5, %%mm2 \n\t" - "punpckhbw %%mm5, %%mm3 \n\t" - MOVNTQ" %%mm0, (%0, %%"REG_a", 2) \n\t" - MOVNTQ" %%mm1, 8(%0, %%"REG_a", 2) \n\t" - MOVNTQ" %%mm2, 16(%0, %%"REG_a", 2) \n\t" - MOVNTQ" %%mm3, 24(%0, %%"REG_a", 2) \n\t" - "add $16, %%"REG_a" \n\t" - "cmp %3, %%"REG_a" \n\t" - " jb 1b \n\t" - ::"r"(dest), "r"(src1), "r"(src2), "r" ((x86_reg)width-15) - : "memory", "%"REG_a + "xor %%"REG_a", %%"REG_a" \n\t" + "1: \n\t" + PREFETCH" 64(%1, %%"REG_a") \n\t" + PREFETCH" 64(%2, %%"REG_a") \n\t" + "movq (%1, %%"REG_a"), %%mm0 \n\t" + "movq 8(%1, %%"REG_a"), %%mm2 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm2, %%mm3 \n\t" + "movq (%2, %%"REG_a"), %%mm4 \n\t" + "movq 8(%2, %%"REG_a"), %%mm5 \n\t" + "punpcklbw %%mm4, %%mm0 \n\t" + "punpckhbw %%mm4, %%mm1 \n\t" + "punpcklbw %%mm5, %%mm2 \n\t" + "punpckhbw %%mm5, %%mm3 \n\t" + MOVNTQ" %%mm0, (%0, %%"REG_a", 2) \n\t" + MOVNTQ" %%mm1, 8(%0, %%"REG_a", 2) \n\t" + MOVNTQ" %%mm2, 16(%0, %%"REG_a", 2) \n\t" + MOVNTQ" %%mm3, 24(%0, %%"REG_a", 2) \n\t" + "add $16, %%"REG_a" \n\t" + "cmp %3, %%"REG_a" \n\t" + " jb 1b \n\t" + ::"r"(dest), "r"(src1), "r"(src2), "r" ((x86_reg)width-15) + : "memory", "%"REG_a ); #endif for (w= (width&(~15)); w < width; w++) @@ -2480,10 +2480,10 @@ static void RENAME(interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dest, } #if HAVE_MMX __asm__( - EMMS" \n\t" - SFENCE" \n\t" - ::: "memory" - ); + EMMS" \n\t" + SFENCE" \n\t" + ::: "memory" + ); #endif } @@ -2498,95 +2498,95 @@ static inline void RENAME(vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2, w=width/2; h=height/2; #if HAVE_MMX __asm__ volatile( - PREFETCH" %0 \n\t" - PREFETCH" %1 \n\t" - ::"m"(*(src1+srcStride1)),"m"(*(src2+srcStride2)):"memory"); + PREFETCH" %0 \n\t" + PREFETCH" %1 \n\t" + ::"m"(*(src1+srcStride1)),"m"(*(src2+srcStride2)):"memory"); #endif for (y=0;y<h;y++){ - const uint8_t* s1=src1+srcStride1*(y>>1); - uint8_t* d=dst1+dstStride1*y; - x=0; + const uint8_t* s1=src1+srcStride1*(y>>1); + uint8_t* d=dst1+dstStride1*y; + x=0; #if HAVE_MMX - for (;x<w-31;x+=32) - { - __asm__ volatile( - PREFETCH" 32%1 \n\t" - "movq %1, %%mm0 \n\t" - "movq 8%1, %%mm2 \n\t" - "movq 16%1, %%mm4 \n\t" - "movq 24%1, %%mm6 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm2, %%mm3 \n\t" - "movq %%mm4, %%mm5 \n\t" - "movq %%mm6, %%mm7 \n\t" - "punpcklbw %%mm0, %%mm0 \n\t" - "punpckhbw %%mm1, %%mm1 \n\t" - "punpcklbw %%mm2, %%mm2 \n\t" - "punpckhbw %%mm3, %%mm3 \n\t" - "punpcklbw %%mm4, %%mm4 \n\t" - "punpckhbw %%mm5, %%mm5 \n\t" - "punpcklbw %%mm6, %%mm6 \n\t" - "punpckhbw %%mm7, %%mm7 \n\t" - MOVNTQ" %%mm0, %0 \n\t" - MOVNTQ" %%mm1, 8%0 \n\t" - MOVNTQ" %%mm2, 16%0 \n\t" - MOVNTQ" %%mm3, 24%0 \n\t" - MOVNTQ" %%mm4, 32%0 \n\t" - MOVNTQ" %%mm5, 40%0 \n\t" - MOVNTQ" %%mm6, 48%0 \n\t" - MOVNTQ" %%mm7, 56%0" - :"=m"(d[2*x]) - :"m"(s1[x]) - :"memory"); - } -#endif - for (;x<w;x++) d[2*x]=d[2*x+1]=s1[x]; + for (;x<w-31;x+=32) + { + __asm__ volatile( + PREFETCH" 32%1 \n\t" + "movq %1, %%mm0 \n\t" + "movq 8%1, %%mm2 \n\t" + "movq 16%1, %%mm4 \n\t" + "movq 24%1, %%mm6 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm2, %%mm3 \n\t" + "movq %%mm4, %%mm5 \n\t" + "movq %%mm6, %%mm7 \n\t" + "punpcklbw %%mm0, %%mm0 \n\t" + "punpckhbw %%mm1, %%mm1 \n\t" + "punpcklbw %%mm2, %%mm2 \n\t" + "punpckhbw %%mm3, %%mm3 \n\t" + "punpcklbw %%mm4, %%mm4 \n\t" + "punpckhbw %%mm5, %%mm5 \n\t" + "punpcklbw %%mm6, %%mm6 \n\t" + "punpckhbw %%mm7, %%mm7 \n\t" + MOVNTQ" %%mm0, %0 \n\t" + MOVNTQ" %%mm1, 8%0 \n\t" + MOVNTQ" %%mm2, 16%0 \n\t" + MOVNTQ" %%mm3, 24%0 \n\t" + MOVNTQ" %%mm4, 32%0 \n\t" + MOVNTQ" %%mm5, 40%0 \n\t" + MOVNTQ" %%mm6, 48%0 \n\t" + MOVNTQ" %%mm7, 56%0" + :"=m"(d[2*x]) + :"m"(s1[x]) + :"memory"); + } +#endif + for (;x<w;x++) d[2*x]=d[2*x+1]=s1[x]; } for (y=0;y<h;y++){ - const uint8_t* s2=src2+srcStride2*(y>>1); - uint8_t* d=dst2+dstStride2*y; - x=0; + const uint8_t* s2=src2+srcStride2*(y>>1); + uint8_t* d=dst2+dstStride2*y; + x=0; #if HAVE_MMX - for (;x<w-31;x+=32) - { - __asm__ volatile( - PREFETCH" 32%1 \n\t" - "movq %1, %%mm0 \n\t" - "movq 8%1, %%mm2 \n\t" - "movq 16%1, %%mm4 \n\t" - "movq 24%1, %%mm6 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm2, %%mm3 \n\t" - "movq %%mm4, %%mm5 \n\t" - "movq %%mm6, %%mm7 \n\t" - "punpcklbw %%mm0, %%mm0 \n\t" - "punpckhbw %%mm1, %%mm1 \n\t" - "punpcklbw %%mm2, %%mm2 \n\t" - "punpckhbw %%mm3, %%mm3 \n\t" - "punpcklbw %%mm4, %%mm4 \n\t" - "punpckhbw %%mm5, %%mm5 \n\t" - "punpcklbw %%mm6, %%mm6 \n\t" - "punpckhbw %%mm7, %%mm7 \n\t" - MOVNTQ" %%mm0, %0 \n\t" - MOVNTQ" %%mm1, 8%0 \n\t" - MOVNTQ" %%mm2, 16%0 \n\t" - MOVNTQ" %%mm3, 24%0 \n\t" - MOVNTQ" %%mm4, 32%0 \n\t" - MOVNTQ" %%mm5, 40%0 \n\t" - MOVNTQ" %%mm6, 48%0 \n\t" - MOVNTQ" %%mm7, 56%0" - :"=m"(d[2*x]) - :"m"(s2[x]) - :"memory"); - } -#endif - for (;x<w;x++) d[2*x]=d[2*x+1]=s2[x]; + for (;x<w-31;x+=32) + { + __asm__ volatile( + PREFETCH" 32%1 \n\t" + "movq %1, %%mm0 \n\t" + "movq 8%1, %%mm2 \n\t" + "movq 16%1, %%mm4 \n\t" + "movq 24%1, %%mm6 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm2, %%mm3 \n\t" + "movq %%mm4, %%mm5 \n\t" + "movq %%mm6, %%mm7 \n\t" + "punpcklbw %%mm0, %%mm0 \n\t" + "punpckhbw %%mm1, %%mm1 \n\t" + "punpcklbw %%mm2, %%mm2 \n\t" + "punpckhbw %%mm3, %%mm3 \n\t" + "punpcklbw %%mm4, %%mm4 \n\t" + "punpckhbw %%mm5, %%mm5 \n\t" + "punpcklbw %%mm6, %%mm6 \n\t" + "punpckhbw %%mm7, %%mm7 \n\t" + MOVNTQ" %%mm0, %0 \n\t" + MOVNTQ" %%mm1, 8%0 \n\t" + MOVNTQ" %%mm2, 16%0 \n\t" + MOVNTQ" %%mm3, 24%0 \n\t" + MOVNTQ" %%mm4, 32%0 \n\t" + MOVNTQ" %%mm5, 40%0 \n\t" + MOVNTQ" %%mm6, 48%0 \n\t" + MOVNTQ" %%mm7, 56%0" + :"=m"(d[2*x]) + :"m"(s2[x]) + :"memory"); + } +#endif + for (;x<w;x++) d[2*x]=d[2*x+1]=s2[x]; } #if HAVE_MMX __asm__( - EMMS" \n\t" - SFENCE" \n\t" - ::: "memory" + EMMS" \n\t" + SFENCE" \n\t" + ::: "memory" ); #endif } @@ -2601,84 +2601,84 @@ static inline void RENAME(yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2 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; - x=0; + 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; + 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" - PREFETCH" 32(%3, %0) \n\t" - "movq (%1, %0, 4), %%mm0 \n\t" /* Y0Y1Y2Y3Y4Y5Y6Y7 */ - "movq (%2, %0), %%mm1 \n\t" /* U0U1U2U3U4U5U6U7 */ - "movq (%3, %0), %%mm2 \n\t" /* V0V1V2V3V4V5V6V7 */ - "movq %%mm0, %%mm3 \n\t" /* Y0Y1Y2Y3Y4Y5Y6Y7 */ - "movq %%mm1, %%mm4 \n\t" /* U0U1U2U3U4U5U6U7 */ - "movq %%mm2, %%mm5 \n\t" /* V0V1V2V3V4V5V6V7 */ - "punpcklbw %%mm1, %%mm1 \n\t" /* U0U0 U1U1 U2U2 U3U3 */ - "punpcklbw %%mm2, %%mm2 \n\t" /* V0V0 V1V1 V2V2 V3V3 */ - "punpckhbw %%mm4, %%mm4 \n\t" /* U4U4 U5U5 U6U6 U7U7 */ - "punpckhbw %%mm5, %%mm5 \n\t" /* V4V4 V5V5 V6V6 V7V7 */ - - "movq %%mm1, %%mm6 \n\t" - "punpcklbw %%mm2, %%mm1 \n\t" /* U0V0 U0V0 U1V1 U1V1*/ - "punpcklbw %%mm1, %%mm0 \n\t" /* Y0U0 Y1V0 Y2U0 Y3V0*/ - "punpckhbw %%mm1, %%mm3 \n\t" /* Y4U1 Y5V1 Y6U1 Y7V1*/ - MOVNTQ" %%mm0, (%4, %0, 8) \n\t" - MOVNTQ" %%mm3, 8(%4, %0, 8) \n\t" - - "punpckhbw %%mm2, %%mm6 \n\t" /* U2V2 U2V2 U3V3 U3V3*/ - "movq 8(%1, %0, 4), %%mm0 \n\t" - "movq %%mm0, %%mm3 \n\t" - "punpcklbw %%mm6, %%mm0 \n\t" /* Y U2 Y V2 Y U2 Y V2*/ - "punpckhbw %%mm6, %%mm3 \n\t" /* Y U3 Y V3 Y U3 Y V3*/ - MOVNTQ" %%mm0, 16(%4, %0, 8) \n\t" - MOVNTQ" %%mm3, 24(%4, %0, 8) \n\t" - - "movq %%mm4, %%mm6 \n\t" - "movq 16(%1, %0, 4), %%mm0 \n\t" - "movq %%mm0, %%mm3 \n\t" - "punpcklbw %%mm5, %%mm4 \n\t" - "punpcklbw %%mm4, %%mm0 \n\t" /* Y U4 Y V4 Y U4 Y V4*/ - "punpckhbw %%mm4, %%mm3 \n\t" /* Y U5 Y V5 Y U5 Y V5*/ - MOVNTQ" %%mm0, 32(%4, %0, 8) \n\t" - MOVNTQ" %%mm3, 40(%4, %0, 8) \n\t" - - "punpckhbw %%mm5, %%mm6 \n\t" - "movq 24(%1, %0, 4), %%mm0 \n\t" - "movq %%mm0, %%mm3 \n\t" - "punpcklbw %%mm6, %%mm0 \n\t" /* Y U6 Y V6 Y U6 Y V6*/ - "punpckhbw %%mm6, %%mm3 \n\t" /* Y U7 Y V7 Y U7 Y V7*/ - MOVNTQ" %%mm0, 48(%4, %0, 8) \n\t" - MOVNTQ" %%mm3, 56(%4, %0, 8) \n\t" - - : "+r" (x) - : "r"(yp), "r" (up), "r"(vp), "r"(d) - :"memory"); - } -#endif - for (; x<w; x++) - { - const long x2 = x<<2; - d[8*x+0] = yp[x2]; - d[8*x+1] = up[x]; - d[8*x+2] = yp[x2+1]; - d[8*x+3] = vp[x]; - d[8*x+4] = yp[x2+2]; - d[8*x+5] = up[x]; - d[8*x+6] = yp[x2+3]; - d[8*x+7] = vp[x]; - } + for (;x<w-7;x+=8) + { + __asm__ volatile( + PREFETCH" 32(%1, %0) \n\t" + PREFETCH" 32(%2, %0) \n\t" + PREFETCH" 32(%3, %0) \n\t" + "movq (%1, %0, 4), %%mm0 \n\t" /* Y0Y1Y2Y3Y4Y5Y6Y7 */ + "movq (%2, %0), %%mm1 \n\t" /* U0U1U2U3U4U5U6U7 */ + "movq (%3, %0), %%mm2 \n\t" /* V0V1V2V3V4V5V6V7 */ + "movq %%mm0, %%mm3 \n\t" /* Y0Y1Y2Y3Y4Y5Y6Y7 */ + "movq %%mm1, %%mm4 \n\t" /* U0U1U2U3U4U5U6U7 */ + "movq %%mm2, %%mm5 \n\t" /* V0V1V2V3V4V5V6V7 */ + "punpcklbw %%mm1, %%mm1 \n\t" /* U0U0 U1U1 U2U2 U3U3 */ + "punpcklbw %%mm2, %%mm2 \n\t" /* V0V0 V1V1 V2V2 V3V3 */ + "punpckhbw %%mm4, %%mm4 \n\t" /* U4U4 U5U5 U6U6 U7U7 */ + "punpckhbw %%mm5, %%mm5 \n\t" /* V4V4 V5V5 V6V6 V7V7 */ + + "movq %%mm1, %%mm6 \n\t" + "punpcklbw %%mm2, %%mm1 \n\t" /* U0V0 U0V0 U1V1 U1V1*/ + "punpcklbw %%mm1, %%mm0 \n\t" /* Y0U0 Y1V0 Y2U0 Y3V0*/ + "punpckhbw %%mm1, %%mm3 \n\t" /* Y4U1 Y5V1 Y6U1 Y7V1*/ + MOVNTQ" %%mm0, (%4, %0, 8) \n\t" + MOVNTQ" %%mm3, 8(%4, %0, 8) \n\t" + + "punpckhbw %%mm2, %%mm6 \n\t" /* U2V2 U2V2 U3V3 U3V3*/ + "movq 8(%1, %0, 4), %%mm0 \n\t" + "movq %%mm0, %%mm3 \n\t" + "punpcklbw %%mm6, %%mm0 \n\t" /* Y U2 Y V2 Y U2 Y V2*/ + "punpckhbw %%mm6, %%mm3 \n\t" /* Y U3 Y V3 Y U3 Y V3*/ + MOVNTQ" %%mm0, 16(%4, %0, 8) \n\t" + MOVNTQ" %%mm3, 24(%4, %0, 8) \n\t" + + "movq %%mm4, %%mm6 \n\t" + "movq 16(%1, %0, 4), %%mm0 \n\t" + "movq %%mm0, %%mm3 \n\t" + "punpcklbw %%mm5, %%mm4 \n\t" + "punpcklbw %%mm4, %%mm0 \n\t" /* Y U4 Y V4 Y U4 Y V4*/ + "punpckhbw %%mm4, %%mm3 \n\t" /* Y U5 Y V5 Y U5 Y V5*/ + MOVNTQ" %%mm0, 32(%4, %0, 8) \n\t" + MOVNTQ" %%mm3, 40(%4, %0, 8) \n\t" + + "punpckhbw %%mm5, %%mm6 \n\t" + "movq 24(%1, %0, 4), %%mm0 \n\t" + "movq %%mm0, %%mm3 \n\t" + "punpcklbw %%mm6, %%mm0 \n\t" /* Y U6 Y V6 Y U6 Y V6*/ + "punpckhbw %%mm6, %%mm3 \n\t" /* Y U7 Y V7 Y U7 Y V7*/ + MOVNTQ" %%mm0, 48(%4, %0, 8) \n\t" + MOVNTQ" %%mm3, 56(%4, %0, 8) \n\t" + + : "+r" (x) + : "r"(yp), "r" (up), "r"(vp), "r"(d) + :"memory"); + } +#endif + for (; x<w; x++) + { + const long x2 = x<<2; + d[8*x+0] = yp[x2]; + d[8*x+1] = up[x]; + d[8*x+2] = yp[x2+1]; + d[8*x+3] = vp[x]; + d[8*x+4] = yp[x2+2]; + d[8*x+5] = up[x]; + d[8*x+6] = yp[x2+3]; + d[8*x+7] = vp[x]; + } } #if HAVE_MMX __asm__( - EMMS" \n\t" - SFENCE" \n\t" - ::: "memory" + EMMS" \n\t" + SFENCE" \n\t" + ::: "memory" ); #endif } @@ -2947,9 +2947,9 @@ static void RENAME(yuyvtoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, co } #if HAVE_MMX __asm__( - EMMS" \n\t" - SFENCE" \n\t" - ::: "memory" + EMMS" \n\t" + SFENCE" \n\t" + ::: "memory" ); #endif } @@ -2972,9 +2972,9 @@ static void RENAME(yuyvtoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, co } #if HAVE_MMX __asm__( - EMMS" \n\t" - SFENCE" \n\t" - ::: "memory" + EMMS" \n\t" + SFENCE" \n\t" + ::: "memory" ); #endif } @@ -2999,9 +2999,9 @@ static void RENAME(uyvytoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, co } #if HAVE_MMX __asm__( - EMMS" \n\t" - SFENCE" \n\t" - ::: "memory" + EMMS" \n\t" + SFENCE" \n\t" + ::: "memory" ); #endif } @@ -3024,9 +3024,9 @@ static void RENAME(uyvytoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, co } #if HAVE_MMX __asm__( - EMMS" \n\t" - SFENCE" \n\t" - ::: "memory" + EMMS" \n\t" + SFENCE" \n\t" + ::: "memory" ); #endif } diff --git a/libswscale/sparc/yuv2rgb_vis.c b/libswscale/sparc/yuv2rgb_vis.c index cdbe140b20..3abe315e55 100644 --- a/libswscale/sparc/yuv2rgb_vis.c +++ b/libswscale/sparc/yuv2rgb_vis.c @@ -83,105 +83,105 @@ // FIXME: must be changed to set alpha to 255 instead of 0 static int vis_420P_ARGB32(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, int srcSliceH, uint8_t* dst[], int dstStride[]){ - int y, out1, out2, out3, out4, out5, out6; - - for(y=0;y < srcSliceH;++y) { - __asm__ volatile ( - YUV2RGB_INIT - "wr %%g0, 0xd2, %%asi \n\t" /* ASI_FL16_P */ - "1: \n\t" - "ldda [%1] %%asi, %%f2 \n\t" - "ldda [%1+2] %%asi, %%f18 \n\t" - "ldda [%2] %%asi, %%f4 \n\t" - "ldda [%2+2] %%asi, %%f20 \n\t" - "ld [%0], %%f0 \n\t" - "ld [%0+4], %%f16 \n\t" - "fpmerge %%f3, %%f3, %%f2 \n\t" - "fpmerge %%f19, %%f19, %%f18 \n\t" - "fpmerge %%f5, %%f5, %%f4 \n\t" - "fpmerge %%f21, %%f21, %%f20 \n\t" - YUV2RGB_KERNEL - "fzero %%f0 \n\t" - "fpmerge %%f4, %%f6, %%f8 \n\t" // r,b,t1 - "fpmerge %%f20, %%f22, %%f24 \n\t" // r,b,t1 - "fpmerge %%f0, %%f2, %%f10 \n\t" // 0,g,t2 - "fpmerge %%f0, %%f18, %%f26 \n\t" // 0,g,t2 - "fpmerge %%f10, %%f8, %%f4 \n\t" // t2,t1,msb - "fpmerge %%f26, %%f24, %%f20 \n\t" // t2,t1,msb - "fpmerge %%f11, %%f9, %%f6 \n\t" // t2,t1,lsb - "fpmerge %%f27, %%f25, %%f22 \n\t" // t2,t1,lsb - "std %%f4, [%3] \n\t" - "std %%f20, [%3+16] \n\t" - "std %%f6, [%3+8] \n\t" - "std %%f22, [%3+24] \n\t" - - "add %0, 8, %0 \n\t" - "add %1, 4, %1 \n\t" - "add %2, 4, %2 \n\t" - "subcc %4, 8, %4 \n\t" - "bne 1b \n\t" - "add %3, 32, %3 \n\t" //delay slot - : "=r" (out1), "=r" (out2), "=r" (out3), "=r" (out4), "=r" (out5), "=r" (out6) - : "0" (src[0]+(y+srcSliceY)*srcStride[0]), "1" (src[1]+((y+srcSliceY)>>1)*srcStride[1]), - "2" (src[2]+((y+srcSliceY)>>1)*srcStride[2]), "3" (dst[0]+(y+srcSliceY)*dstStride[0]), - "4" (c->dstW), - "5" (c->sparc_coeffs) - ); - } - - return srcSliceH; + int y, out1, out2, out3, out4, out5, out6; + + for(y=0;y < srcSliceH;++y) { + __asm__ volatile ( + YUV2RGB_INIT + "wr %%g0, 0xd2, %%asi \n\t" /* ASI_FL16_P */ + "1: \n\t" + "ldda [%1] %%asi, %%f2 \n\t" + "ldda [%1+2] %%asi, %%f18 \n\t" + "ldda [%2] %%asi, %%f4 \n\t" + "ldda [%2+2] %%asi, %%f20 \n\t" + "ld [%0], %%f0 \n\t" + "ld [%0+4], %%f16 \n\t" + "fpmerge %%f3, %%f3, %%f2 \n\t" + "fpmerge %%f19, %%f19, %%f18 \n\t" + "fpmerge %%f5, %%f5, %%f4 \n\t" + "fpmerge %%f21, %%f21, %%f20 \n\t" + YUV2RGB_KERNEL + "fzero %%f0 \n\t" + "fpmerge %%f4, %%f6, %%f8 \n\t" // r,b,t1 + "fpmerge %%f20, %%f22, %%f24 \n\t" // r,b,t1 + "fpmerge %%f0, %%f2, %%f10 \n\t" // 0,g,t2 + "fpmerge %%f0, %%f18, %%f26 \n\t" // 0,g,t2 + "fpmerge %%f10, %%f8, %%f4 \n\t" // t2,t1,msb + "fpmerge %%f26, %%f24, %%f20 \n\t" // t2,t1,msb + "fpmerge %%f11, %%f9, %%f6 \n\t" // t2,t1,lsb + "fpmerge %%f27, %%f25, %%f22 \n\t" // t2,t1,lsb + "std %%f4, [%3] \n\t" + "std %%f20, [%3+16] \n\t" + "std %%f6, [%3+8] \n\t" + "std %%f22, [%3+24] \n\t" + + "add %0, 8, %0 \n\t" + "add %1, 4, %1 \n\t" + "add %2, 4, %2 \n\t" + "subcc %4, 8, %4 \n\t" + "bne 1b \n\t" + "add %3, 32, %3 \n\t" //delay slot + : "=r" (out1), "=r" (out2), "=r" (out3), "=r" (out4), "=r" (out5), "=r" (out6) + : "0" (src[0]+(y+srcSliceY)*srcStride[0]), "1" (src[1]+((y+srcSliceY)>>1)*srcStride[1]), + "2" (src[2]+((y+srcSliceY)>>1)*srcStride[2]), "3" (dst[0]+(y+srcSliceY)*dstStride[0]), + "4" (c->dstW), + "5" (c->sparc_coeffs) + ); + } + + return srcSliceH; } // FIXME: must be changed to set alpha to 255 instead of 0 static int vis_422P_ARGB32(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, int srcSliceH, uint8_t* dst[], int dstStride[]){ - int y, out1, out2, out3, out4, out5, out6; - - for(y=0;y < srcSliceH;++y) { - __asm__ volatile ( - YUV2RGB_INIT - "wr %%g0, 0xd2, %%asi \n\t" /* ASI_FL16_P */ - "1: \n\t" - "ldda [%1] %%asi, %%f2 \n\t" - "ldda [%1+2] %%asi, %%f18 \n\t" - "ldda [%2] %%asi, %%f4 \n\t" - "ldda [%2+2] %%asi, %%f20 \n\t" - "ld [%0], %%f0 \n\t" - "ld [%0+4], %%f16 \n\t" - "fpmerge %%f3, %%f3, %%f2 \n\t" - "fpmerge %%f19, %%f19, %%f18 \n\t" - "fpmerge %%f5, %%f5, %%f4 \n\t" - "fpmerge %%f21, %%f21, %%f20 \n\t" - YUV2RGB_KERNEL - "fzero %%f0 \n\t" - "fpmerge %%f4, %%f6, %%f8 \n\t" // r,b,t1 - "fpmerge %%f20, %%f22, %%f24 \n\t" // r,b,t1 - "fpmerge %%f0, %%f2, %%f10 \n\t" // 0,g,t2 - "fpmerge %%f0, %%f18, %%f26 \n\t" // 0,g,t2 - "fpmerge %%f10, %%f8, %%f4 \n\t" // t2,t1,msb - "fpmerge %%f26, %%f24, %%f20 \n\t" // t2,t1,msb - "fpmerge %%f11, %%f9, %%f6 \n\t" // t2,t1,lsb - "fpmerge %%f27, %%f25, %%f22 \n\t" // t2,t1,lsb - "std %%f4, [%3] \n\t" - "std %%f20, [%3+16] \n\t" - "std %%f6, [%3+8] \n\t" - "std %%f22, [%3+24] \n\t" - - "add %0, 8, %0 \n\t" - "add %1, 4, %1 \n\t" - "add %2, 4, %2 \n\t" - "subcc %4, 8, %4 \n\t" - "bne 1b \n\t" - "add %3, 32, %3 \n\t" //delay slot - : "=r" (out1), "=r" (out2), "=r" (out3), "=r" (out4), "=r" (out5), "=r" (out6) - : "0" (src[0]+(y+srcSliceY)*srcStride[0]), "1" (src[1]+(y+srcSliceY)*srcStride[1]), - "2" (src[2]+(y+srcSliceY)*srcStride[2]), "3" (dst[0]+(y+srcSliceY)*dstStride[0]), - "4" (c->dstW), - "5" (c->sparc_coeffs) - ); - } - - return srcSliceH; + int y, out1, out2, out3, out4, out5, out6; + + for(y=0;y < srcSliceH;++y) { + __asm__ volatile ( + YUV2RGB_INIT + "wr %%g0, 0xd2, %%asi \n\t" /* ASI_FL16_P */ + "1: \n\t" + "ldda [%1] %%asi, %%f2 \n\t" + "ldda [%1+2] %%asi, %%f18 \n\t" + "ldda [%2] %%asi, %%f4 \n\t" + "ldda [%2+2] %%asi, %%f20 \n\t" + "ld [%0], %%f0 \n\t" + "ld [%0+4], %%f16 \n\t" + "fpmerge %%f3, %%f3, %%f2 \n\t" + "fpmerge %%f19, %%f19, %%f18 \n\t" + "fpmerge %%f5, %%f5, %%f4 \n\t" + "fpmerge %%f21, %%f21, %%f20 \n\t" + YUV2RGB_KERNEL + "fzero %%f0 \n\t" + "fpmerge %%f4, %%f6, %%f8 \n\t" // r,b,t1 + "fpmerge %%f20, %%f22, %%f24 \n\t" // r,b,t1 + "fpmerge %%f0, %%f2, %%f10 \n\t" // 0,g,t2 + "fpmerge %%f0, %%f18, %%f26 \n\t" // 0,g,t2 + "fpmerge %%f10, %%f8, %%f4 \n\t" // t2,t1,msb + "fpmerge %%f26, %%f24, %%f20 \n\t" // t2,t1,msb + "fpmerge %%f11, %%f9, %%f6 \n\t" // t2,t1,lsb + "fpmerge %%f27, %%f25, %%f22 \n\t" // t2,t1,lsb + "std %%f4, [%3] \n\t" + "std %%f20, [%3+16] \n\t" + "std %%f6, [%3+8] \n\t" + "std %%f22, [%3+24] \n\t" + + "add %0, 8, %0 \n\t" + "add %1, 4, %1 \n\t" + "add %2, 4, %2 \n\t" + "subcc %4, 8, %4 \n\t" + "bne 1b \n\t" + "add %3, 32, %3 \n\t" //delay slot + : "=r" (out1), "=r" (out2), "=r" (out3), "=r" (out4), "=r" (out5), "=r" (out6) + : "0" (src[0]+(y+srcSliceY)*srcStride[0]), "1" (src[1]+(y+srcSliceY)*srcStride[1]), + "2" (src[2]+(y+srcSliceY)*srcStride[2]), "3" (dst[0]+(y+srcSliceY)*dstStride[0]), + "4" (c->dstW), + "5" (c->sparc_coeffs) + ); + } + + return srcSliceH; } SwsFunc ff_yuv2rgb_init_vis(SwsContext *c){ diff --git a/libswscale/swscale.c b/libswscale/swscale.c index c513066b25..4742f2fc04 100644 --- a/libswscale/swscale.c +++ b/libswscale/swscale.c @@ -366,112 +366,112 @@ DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={ const char *sws_format_name(enum PixelFormat format) { switch (format) { - case PIX_FMT_YUV420P: - return "yuv420p"; - case PIX_FMT_YUVA420P: - return "yuva420p"; - case PIX_FMT_YUYV422: - return "yuyv422"; - case PIX_FMT_RGB24: - return "rgb24"; - case PIX_FMT_BGR24: - return "bgr24"; - case PIX_FMT_YUV422P: - return "yuv422p"; - case PIX_FMT_YUV444P: - return "yuv444p"; - case PIX_FMT_RGB32: - return "rgb32"; - case PIX_FMT_YUV410P: - return "yuv410p"; - case PIX_FMT_YUV411P: - return "yuv411p"; - case PIX_FMT_RGB565: - return "rgb565"; - case PIX_FMT_RGB555: - return "rgb555"; - case PIX_FMT_GRAY16BE: - return "gray16be"; - case PIX_FMT_GRAY16LE: - return "gray16le"; - case PIX_FMT_GRAY8: - return "gray8"; - case PIX_FMT_MONOWHITE: - return "mono white"; - case PIX_FMT_MONOBLACK: - return "mono black"; - case PIX_FMT_PAL8: - return "Palette"; - case PIX_FMT_YUVJ420P: - return "yuvj420p"; - case PIX_FMT_YUVJ422P: - return "yuvj422p"; - case PIX_FMT_YUVJ444P: - return "yuvj444p"; - case PIX_FMT_XVMC_MPEG2_MC: - return "xvmc_mpeg2_mc"; - case PIX_FMT_XVMC_MPEG2_IDCT: - return "xvmc_mpeg2_idct"; - case PIX_FMT_UYVY422: - return "uyvy422"; - case PIX_FMT_UYYVYY411: - return "uyyvyy411"; - case PIX_FMT_RGB32_1: - return "rgb32x"; - case PIX_FMT_BGR32_1: - return "bgr32x"; - case PIX_FMT_BGR32: - return "bgr32"; - case PIX_FMT_BGR565: - return "bgr565"; - case PIX_FMT_BGR555: - return "bgr555"; - case PIX_FMT_BGR8: - return "bgr8"; - case PIX_FMT_BGR4: - return "bgr4"; - case PIX_FMT_BGR4_BYTE: - return "bgr4 byte"; - case PIX_FMT_RGB8: - return "rgb8"; - case PIX_FMT_RGB4: - return "rgb4"; - case PIX_FMT_RGB4_BYTE: - return "rgb4 byte"; - case PIX_FMT_RGB48BE: - return "rgb48be"; - case PIX_FMT_RGB48LE: - return "rgb48le"; - case PIX_FMT_NV12: - return "nv12"; - case PIX_FMT_NV21: - return "nv21"; - case PIX_FMT_YUV440P: - return "yuv440p"; - case PIX_FMT_VDPAU_H264: - return "vdpau_h264"; - case PIX_FMT_VDPAU_MPEG1: - return "vdpau_mpeg1"; - case PIX_FMT_VDPAU_MPEG2: - return "vdpau_mpeg2"; - case PIX_FMT_VDPAU_WMV3: - return "vdpau_wmv3"; - case PIX_FMT_VDPAU_VC1: - return "vdpau_vc1"; - case PIX_FMT_YUV420PLE: - return "yuv420ple"; - case PIX_FMT_YUV422PLE: - return "yuv422ple"; - case PIX_FMT_YUV444PLE: - return "yuv444ple"; - case PIX_FMT_YUV420PBE: - return "yuv420pbe"; - case PIX_FMT_YUV422PBE: - return "yuv422pbe"; - case PIX_FMT_YUV444PBE: - return "yuv444pbe"; - default: - return "Unknown format"; + case PIX_FMT_YUV420P: + return "yuv420p"; + case PIX_FMT_YUVA420P: + return "yuva420p"; + case PIX_FMT_YUYV422: + return "yuyv422"; + case PIX_FMT_RGB24: + return "rgb24"; + case PIX_FMT_BGR24: + return "bgr24"; + case PIX_FMT_YUV422P: + return "yuv422p"; + case PIX_FMT_YUV444P: + return "yuv444p"; + case PIX_FMT_RGB32: + return "rgb32"; + case PIX_FMT_YUV410P: + return "yuv410p"; + case PIX_FMT_YUV411P: + return "yuv411p"; + case PIX_FMT_RGB565: + return "rgb565"; + case PIX_FMT_RGB555: + return "rgb555"; + case PIX_FMT_GRAY16BE: + return "gray16be"; + case PIX_FMT_GRAY16LE: + return "gray16le"; + case PIX_FMT_GRAY8: + return "gray8"; + case PIX_FMT_MONOWHITE: + return "mono white"; + case PIX_FMT_MONOBLACK: + return "mono black"; + case PIX_FMT_PAL8: + return "Palette"; + case PIX_FMT_YUVJ420P: + return "yuvj420p"; + case PIX_FMT_YUVJ422P: + return "yuvj422p"; + case PIX_FMT_YUVJ444P: + return "yuvj444p"; + case PIX_FMT_XVMC_MPEG2_MC: + return "xvmc_mpeg2_mc"; + case PIX_FMT_XVMC_MPEG2_IDCT: + return "xvmc_mpeg2_idct"; + case PIX_FMT_UYVY422: + return "uyvy422"; + case PIX_FMT_UYYVYY411: + return "uyyvyy411"; + case PIX_FMT_RGB32_1: + return "rgb32x"; + case PIX_FMT_BGR32_1: + return "bgr32x"; + case PIX_FMT_BGR32: + return "bgr32"; + case PIX_FMT_BGR565: + return "bgr565"; + case PIX_FMT_BGR555: + return "bgr555"; + case PIX_FMT_BGR8: + return "bgr8"; + case PIX_FMT_BGR4: + return "bgr4"; + case PIX_FMT_BGR4_BYTE: + return "bgr4 byte"; + case PIX_FMT_RGB8: + return "rgb8"; + case PIX_FMT_RGB4: + return "rgb4"; + case PIX_FMT_RGB4_BYTE: + return "rgb4 byte"; + case PIX_FMT_RGB48BE: + return "rgb48be"; + case PIX_FMT_RGB48LE: + return "rgb48le"; + case PIX_FMT_NV12: + return "nv12"; + case PIX_FMT_NV21: + return "nv21"; + case PIX_FMT_YUV440P: + return "yuv440p"; + case PIX_FMT_VDPAU_H264: + return "vdpau_h264"; + case PIX_FMT_VDPAU_MPEG1: + return "vdpau_mpeg1"; + case PIX_FMT_VDPAU_MPEG2: + return "vdpau_mpeg2"; + case PIX_FMT_VDPAU_WMV3: + return "vdpau_wmv3"; + case PIX_FMT_VDPAU_VC1: + return "vdpau_vc1"; + case PIX_FMT_YUV420PLE: + return "yuv420ple"; + case PIX_FMT_YUV422PLE: + return "yuv422ple"; + case PIX_FMT_YUV444PLE: + return "yuv444ple"; + case PIX_FMT_YUV420PBE: + return "yuv420pbe"; + case PIX_FMT_YUV422PBE: + return "yuv422pbe"; + case PIX_FMT_YUV444PBE: + return "yuv444pbe"; + default: + return "Unknown format"; } } @@ -1443,12 +1443,12 @@ static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uin static double getSplineCoeff(double a, double b, double c, double d, double dist) { // printf("%f %f %f %f %f\n", a,b,c,d,dist); - if (dist<=1.0) return ((d*dist + c)*dist + b)*dist +a; - else return getSplineCoeff( 0.0, - b+ 2.0*c + 3.0*d, - c + 3.0*d, - -b- 3.0*c - 6.0*d, - dist-1.0); + if (dist<=1.0) return ((d*dist + c)*dist + b)*dist +a; + else return getSplineCoeff( 0.0, + b+ 2.0*c + 3.0*d, + c + 3.0*d, + -b- 3.0*c - 6.0*d, + dist-1.0); } static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc, @@ -1515,14 +1515,14 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF int j; (*filterPos)[i]= xx; - //bilinear upscale / linear interpolate / area averaging - for (j=0; j<filterSize; j++) - { - int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16); - if (coeff<0) coeff=0; - filter[i*filterSize + j]= coeff; - xx++; - } + //bilinear upscale / linear interpolate / area averaging + for (j=0; j<filterSize; j++) + { + int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16); + if (coeff<0) coeff=0; + filter[i*filterSize + j]= coeff; + xx++; + } xDstInSrc+= xInc; } } @@ -2562,20 +2562,20 @@ int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int static int handle_jpeg(enum PixelFormat *format) { switch (*format) { - case PIX_FMT_YUVJ420P: - *format = PIX_FMT_YUV420P; - return 1; - case PIX_FMT_YUVJ422P: - *format = PIX_FMT_YUV422P; - return 1; - case PIX_FMT_YUVJ444P: - *format = PIX_FMT_YUV444P; - return 1; - case PIX_FMT_YUVJ440P: - *format = PIX_FMT_YUV440P; - return 1; - default: - return 0; + case PIX_FMT_YUVJ420P: + *format = PIX_FMT_YUV420P; + return 1; + case PIX_FMT_YUVJ422P: + *format = PIX_FMT_YUV422P; + return 1; + case PIX_FMT_YUVJ444P: + *format = PIX_FMT_YUV444P; + return 1; + case PIX_FMT_YUVJ440P: + *format = PIX_FMT_YUV440P; + return 1; + default: + return 0; } } @@ -2837,7 +2837,7 @@ SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int d if (c->swScale){ if (flags&SWS_PRINT_INFO) av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n", - sws_format_name(srcFormat), sws_format_name(dstFormat)); + sws_format_name(srcFormat), sws_format_name(dstFormat)); return c; } } diff --git a/libswscale/swscale_template.c b/libswscale/swscale_template.c index 4681d42995..5b73069d6d 100644 --- a/libswscale/swscale_template.c +++ b/libswscale/swscale_template.c @@ -932,13 +932,13 @@ static inline void RENAME(yuv2yuvX)(SwsContext *c, const int16_t *lumFilter, con } #endif #if COMPILE_TEMPLATE_ALTIVEC -yuv2yuvX_altivec_real(lumFilter, lumSrc, lumFilterSize, - chrFilter, chrSrc, chrFilterSize, - dest, uDest, vDest, dstW, chrDstW); + yuv2yuvX_altivec_real(lumFilter, lumSrc, lumFilterSize, + chrFilter, chrSrc, chrFilterSize, + dest, uDest, vDest, dstW, chrDstW); #else //COMPILE_TEMPLATE_ALTIVEC -yuv2yuvXinC(lumFilter, lumSrc, lumFilterSize, - chrFilter, chrSrc, chrFilterSize, - alpSrc, dest, uDest, vDest, aDest, dstW, chrDstW); + yuv2yuvXinC(lumFilter, lumSrc, lumFilterSize, + chrFilter, chrSrc, chrFilterSize, + alpSrc, dest, uDest, vDest, aDest, dstW, chrDstW); #endif //!COMPILE_TEMPLATE_ALTIVEC } @@ -946,9 +946,9 @@ static inline void RENAME(yuv2nv12X)(SwsContext *c, const int16_t *lumFilter, co const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize, uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat) { -yuv2nv12XinC(lumFilter, lumSrc, lumFilterSize, - chrFilter, chrSrc, chrFilterSize, - dest, uDest, dstW, chrDstW, dstFormat); + yuv2nv12XinC(lumFilter, lumSrc, lumFilterSize, + chrFilter, chrSrc, chrFilterSize, + dest, uDest, dstW, chrDstW, dstFormat); } static inline void RENAME(yuv2yuv1)(SwsContext *c, const int16_t *lumSrc, const int16_t *chrSrc, const int16_t *alpSrc, @@ -1198,9 +1198,9 @@ static inline void RENAME(yuv2packedX)(SwsContext *c, const int16_t *lumFilter, /* The following list of supported dstFormat values should match what's found in the body of ff_yuv2packedX_altivec() */ if (!(c->flags & SWS_BITEXACT) && !c->alpPixBuf && - (c->dstFormat==PIX_FMT_ABGR || c->dstFormat==PIX_FMT_BGRA || - c->dstFormat==PIX_FMT_BGR24 || c->dstFormat==PIX_FMT_RGB24 || - c->dstFormat==PIX_FMT_RGBA || c->dstFormat==PIX_FMT_ARGB)) + (c->dstFormat==PIX_FMT_ABGR || c->dstFormat==PIX_FMT_BGRA || + c->dstFormat==PIX_FMT_BGR24 || c->dstFormat==PIX_FMT_RGB24 || + c->dstFormat==PIX_FMT_RGBA || c->dstFormat==PIX_FMT_ARGB)) ff_yuv2packedX_altivec(c, lumFilter, lumSrc, lumFilterSize, chrFilter, chrSrc, chrFilterSize, dest, dstW, dstY); @@ -1358,7 +1358,7 @@ static inline void RENAME(yuv2packed2)(SwsContext *c, const uint16_t *buf0, cons } } #endif //COMPILE_TEMPLATE_MMX -YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB2_C, YSCALE_YUV_2_PACKED2_C(void,0), YSCALE_YUV_2_GRAY16_2_C, YSCALE_YUV_2_MONO2_C) + YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB2_C, YSCALE_YUV_2_PACKED2_C(void,0), YSCALE_YUV_2_GRAY16_2_C, YSCALE_YUV_2_MONO2_C) } /** @@ -2305,14 +2305,14 @@ static inline void RENAME(hyscale)(SwsContext *c, uint16_t *dst, long dstWidth, #endif /* ARCH_X86_64 */ -CALL_MMX2_FILTER_CODE -CALL_MMX2_FILTER_CODE -CALL_MMX2_FILTER_CODE -CALL_MMX2_FILTER_CODE -CALL_MMX2_FILTER_CODE -CALL_MMX2_FILTER_CODE -CALL_MMX2_FILTER_CODE -CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE #if defined(PIC) "mov %5, %%"REG_b" \n\t" @@ -2470,10 +2470,10 @@ inline static void RENAME(hcscale)(SwsContext *c, uint16_t *dst, long dstWidth, PREFETCH" 32(%%"REG_c") \n\t" PREFETCH" 64(%%"REG_c") \n\t" -CALL_MMX2_FILTER_CODE -CALL_MMX2_FILTER_CODE -CALL_MMX2_FILTER_CODE -CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE "xor %%"REG_a", %%"REG_a" \n\t" // i "mov %5, %%"REG_c" \n\t" // src "mov %1, %%"REG_D" \n\t" // buf1 @@ -2482,10 +2482,10 @@ CALL_MMX2_FILTER_CODE PREFETCH" 32(%%"REG_c") \n\t" PREFETCH" 64(%%"REG_c") \n\t" -CALL_MMX2_FILTER_CODE -CALL_MMX2_FILTER_CODE -CALL_MMX2_FILTER_CODE -CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE + CALL_MMX2_FILTER_CODE #if defined(PIC) "mov %6, %%"REG_b" \n\t" @@ -2754,60 +2754,60 @@ static int RENAME(swScale)(SwsContext *c, uint8_t* src[], int srcStride[], int s const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? (const int16_t **) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL; #if COMPILE_TEMPLATE_MMX int i; - if (flags & SWS_ACCURATE_RND){ - int s= APCK_SIZE / 8; - for (i=0; i<vLumFilterSize; i+=2){ - *(void**)&lumMmxFilter[s*i ]= lumSrcPtr[i ]; - *(void**)&lumMmxFilter[s*i+APCK_PTR2/4 ]= lumSrcPtr[i+(vLumFilterSize>1)]; - lumMmxFilter[s*i+APCK_COEF/4 ]= - lumMmxFilter[s*i+APCK_COEF/4+1]= vLumFilter[dstY*vLumFilterSize + i ] - + (vLumFilterSize>1 ? vLumFilter[dstY*vLumFilterSize + i + 1]<<16 : 0); - if (CONFIG_SWSCALE_ALPHA && alpPixBuf){ - *(void**)&alpMmxFilter[s*i ]= alpSrcPtr[i ]; - *(void**)&alpMmxFilter[s*i+APCK_PTR2/4 ]= alpSrcPtr[i+(vLumFilterSize>1)]; - alpMmxFilter[s*i+APCK_COEF/4 ]= - alpMmxFilter[s*i+APCK_COEF/4+1]= lumMmxFilter[s*i+APCK_COEF/4 ]; + if (flags & SWS_ACCURATE_RND){ + int s= APCK_SIZE / 8; + for (i=0; i<vLumFilterSize; i+=2){ + *(void**)&lumMmxFilter[s*i ]= lumSrcPtr[i ]; + *(void**)&lumMmxFilter[s*i+APCK_PTR2/4 ]= lumSrcPtr[i+(vLumFilterSize>1)]; + lumMmxFilter[s*i+APCK_COEF/4 ]= + lumMmxFilter[s*i+APCK_COEF/4+1]= vLumFilter[dstY*vLumFilterSize + i ] + + (vLumFilterSize>1 ? vLumFilter[dstY*vLumFilterSize + i + 1]<<16 : 0); + if (CONFIG_SWSCALE_ALPHA && alpPixBuf){ + *(void**)&alpMmxFilter[s*i ]= alpSrcPtr[i ]; + *(void**)&alpMmxFilter[s*i+APCK_PTR2/4 ]= alpSrcPtr[i+(vLumFilterSize>1)]; + alpMmxFilter[s*i+APCK_COEF/4 ]= + alpMmxFilter[s*i+APCK_COEF/4+1]= lumMmxFilter[s*i+APCK_COEF/4 ]; + } } - } - for (i=0; i<vChrFilterSize; i+=2){ - *(void**)&chrMmxFilter[s*i ]= chrSrcPtr[i ]; - *(void**)&chrMmxFilter[s*i+APCK_PTR2/4 ]= chrSrcPtr[i+(vChrFilterSize>1)]; - chrMmxFilter[s*i+APCK_COEF/4 ]= - chrMmxFilter[s*i+APCK_COEF/4+1]= vChrFilter[chrDstY*vChrFilterSize + i ] - + (vChrFilterSize>1 ? vChrFilter[chrDstY*vChrFilterSize + i + 1]<<16 : 0); - } - }else{ - for (i=0; i<vLumFilterSize; i++) - { - lumMmxFilter[4*i+0]= (int32_t)lumSrcPtr[i]; - lumMmxFilter[4*i+1]= (uint64_t)lumSrcPtr[i] >> 32; - lumMmxFilter[4*i+2]= - lumMmxFilter[4*i+3]= - ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001; - if (CONFIG_SWSCALE_ALPHA && alpPixBuf){ - alpMmxFilter[4*i+0]= (int32_t)alpSrcPtr[i]; - alpMmxFilter[4*i+1]= (uint64_t)alpSrcPtr[i] >> 32; - alpMmxFilter[4*i+2]= - alpMmxFilter[4*i+3]= lumMmxFilter[4*i+2]; + for (i=0; i<vChrFilterSize; i+=2){ + *(void**)&chrMmxFilter[s*i ]= chrSrcPtr[i ]; + *(void**)&chrMmxFilter[s*i+APCK_PTR2/4 ]= chrSrcPtr[i+(vChrFilterSize>1)]; + chrMmxFilter[s*i+APCK_COEF/4 ]= + chrMmxFilter[s*i+APCK_COEF/4+1]= vChrFilter[chrDstY*vChrFilterSize + i ] + + (vChrFilterSize>1 ? vChrFilter[chrDstY*vChrFilterSize + i + 1]<<16 : 0); + } + }else{ + for (i=0; i<vLumFilterSize; i++) + { + lumMmxFilter[4*i+0]= (int32_t)lumSrcPtr[i]; + lumMmxFilter[4*i+1]= (uint64_t)lumSrcPtr[i] >> 32; + lumMmxFilter[4*i+2]= + lumMmxFilter[4*i+3]= + ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001; + if (CONFIG_SWSCALE_ALPHA && alpPixBuf){ + alpMmxFilter[4*i+0]= (int32_t)alpSrcPtr[i]; + alpMmxFilter[4*i+1]= (uint64_t)alpSrcPtr[i] >> 32; + alpMmxFilter[4*i+2]= + alpMmxFilter[4*i+3]= lumMmxFilter[4*i+2]; + } + } + for (i=0; i<vChrFilterSize; i++) + { + chrMmxFilter[4*i+0]= (int32_t)chrSrcPtr[i]; + chrMmxFilter[4*i+1]= (uint64_t)chrSrcPtr[i] >> 32; + chrMmxFilter[4*i+2]= + chrMmxFilter[4*i+3]= + ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001; } } - for (i=0; i<vChrFilterSize; i++) - { - chrMmxFilter[4*i+0]= (int32_t)chrSrcPtr[i]; - chrMmxFilter[4*i+1]= (uint64_t)chrSrcPtr[i] >> 32; - chrMmxFilter[4*i+2]= - chrMmxFilter[4*i+3]= - ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001; - } - } #endif if (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21){ const int chrSkipMask= (1<<c->chrDstVSubSample)-1; if (dstY&chrSkipMask) uDest= NULL; //FIXME split functions in lumi / chromi c->yuv2nv12X(c, - vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, - vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, - dest, uDest, dstW, chrDstW, dstFormat); + vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, + vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, + dest, uDest, dstW, chrDstW, dstFormat); } else if (isPlanarYUV(dstFormat) || dstFormat==PIX_FMT_GRAY8) //YV12 like { @@ -2816,10 +2816,10 @@ static int RENAME(swScale)(SwsContext *c, uint8_t* src[], int srcStride[], int s if (is16BPS(dstFormat)) { yuv2yuvX16inC( - vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, - vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, - alpSrcPtr, (uint16_t *) dest, (uint16_t *) uDest, (uint16_t *) vDest, (uint16_t *) aDest, dstW, chrDstW, - dstFormat); + vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, + vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, + alpSrcPtr, (uint16_t *) dest, (uint16_t *) uDest, (uint16_t *) vDest, (uint16_t *) aDest, dstW, chrDstW, + dstFormat); } else if (vLumFilterSize == 1 && vChrFilterSize == 1) // unscaled YV12 @@ -2832,9 +2832,9 @@ static int RENAME(swScale)(SwsContext *c, uint8_t* src[], int srcStride[], int s else //General YV12 { c->yuv2yuvX(c, - vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, - vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, - alpSrcPtr, dest, uDest, vDest, aDest, dstW, chrDstW); + vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, + vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, + alpSrcPtr, dest, uDest, vDest, aDest, dstW, chrDstW); } } else @@ -2846,13 +2846,13 @@ static int RENAME(swScale)(SwsContext *c, uint8_t* src[], int srcStride[], int s int chrAlpha= vChrFilter[2*dstY+1]; if(flags & SWS_FULL_CHR_H_INT){ yuv2rgbXinC_full(c, //FIXME write a packed1_full function - vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, - vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, - alpSrcPtr, dest, dstW, dstY); + vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, + vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, + alpSrcPtr, dest, dstW, dstY); }else{ c->yuv2packed1(c, *lumSrcPtr, *chrSrcPtr, *(chrSrcPtr+1), - alpPixBuf ? *alpSrcPtr : NULL, - dest, dstW, chrAlpha, dstFormat, flags, dstY); + alpPixBuf ? *alpSrcPtr : NULL, + dest, dstW, chrAlpha, dstFormat, flags, dstY); } } else if (vLumFilterSize == 2 && vChrFilterSize == 2) //bilinear upscale RGB @@ -2865,27 +2865,27 @@ static int RENAME(swScale)(SwsContext *c, uint8_t* src[], int srcStride[], int s chrMmxFilter[3]= vChrFilter[2*chrDstY]*0x10001; if(flags & SWS_FULL_CHR_H_INT){ yuv2rgbXinC_full(c, //FIXME write a packed2_full function - vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, - vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, - alpSrcPtr, dest, dstW, dstY); + vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, + vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, + alpSrcPtr, dest, dstW, dstY); }else{ c->yuv2packed2(c, *lumSrcPtr, *(lumSrcPtr+1), *chrSrcPtr, *(chrSrcPtr+1), - alpPixBuf ? *alpSrcPtr : NULL, alpPixBuf ? *(alpSrcPtr+1) : NULL, - dest, dstW, lumAlpha, chrAlpha, dstY); + alpPixBuf ? *alpSrcPtr : NULL, alpPixBuf ? *(alpSrcPtr+1) : NULL, + dest, dstW, lumAlpha, chrAlpha, dstY); } } else //general RGB { if(flags & SWS_FULL_CHR_H_INT){ yuv2rgbXinC_full(c, - vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, - vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, - alpSrcPtr, dest, dstW, dstY); + vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, + vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, + alpSrcPtr, dest, dstW, dstY); }else{ c->yuv2packedX(c, - vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, - vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, - alpSrcPtr, dest, dstW, dstY); + vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, + vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, + alpSrcPtr, dest, dstW, dstY); } } } @@ -2899,9 +2899,9 @@ static int RENAME(swScale)(SwsContext *c, uint8_t* src[], int srcStride[], int s const int chrSkipMask= (1<<c->chrDstVSubSample)-1; if (dstY&chrSkipMask) uDest= NULL; //FIXME split functions in lumi / chromi yuv2nv12XinC( - vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, - vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, - dest, uDest, dstW, chrDstW, dstFormat); + vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, + vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, + dest, uDest, dstW, chrDstW, dstFormat); } else if (isPlanarYUV(dstFormat) || dstFormat==PIX_FMT_GRAY8) //YV12 { @@ -2910,17 +2910,17 @@ static int RENAME(swScale)(SwsContext *c, uint8_t* src[], int srcStride[], int s if (is16BPS(dstFormat)) { yuv2yuvX16inC( - vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, - vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, - alpSrcPtr, (uint16_t *) dest, (uint16_t *) uDest, (uint16_t *) vDest, (uint16_t *) aDest, dstW, chrDstW, - dstFormat); + vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, + vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, + alpSrcPtr, (uint16_t *) dest, (uint16_t *) uDest, (uint16_t *) vDest, (uint16_t *) aDest, dstW, chrDstW, + dstFormat); } else { yuv2yuvXinC( - vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, - vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, - alpSrcPtr, dest, uDest, vDest, aDest, dstW, chrDstW); + vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, + vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, + alpSrcPtr, dest, uDest, vDest, aDest, dstW, chrDstW); } } else @@ -2929,14 +2929,14 @@ static int RENAME(swScale)(SwsContext *c, uint8_t* src[], int srcStride[], int s assert(chrSrcPtr + vChrFilterSize - 1 < chrPixBuf + vChrBufSize*2); if(flags & SWS_FULL_CHR_H_INT){ yuv2rgbXinC_full(c, - vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, - vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, - alpSrcPtr, dest, dstW, dstY); + vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, + vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, + alpSrcPtr, dest, dstW, dstY); }else{ yuv2packedXinC(c, - vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, - vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, - alpSrcPtr, dest, dstW, dstY); + vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, + vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, + alpSrcPtr, dest, dstW, dstY); } } } |