#undef PREFETCHW
#undef PAVGB
-#ifdef HAVE_SSE2
+#if HAVE_SSE2
#define MMREG_SIZE 16
#else
#define MMREG_SIZE 8
#endif
-#ifdef HAVE_3DNOW
+#if HAVE_AMD3DNOW
#define PREFETCH "prefetch"
#define PREFETCHW "prefetchw"
#define PAVGB "pavgusb"
-#elif defined (HAVE_MMX2)
+#elif HAVE_MMX2
#define PREFETCH "prefetchnta"
#define PREFETCHW "prefetcht0"
#define PAVGB "pavgb"
#define PREFETCHW " # nop"
#endif
-#ifdef HAVE_3DNOW
-/* On K6 femms is faster than emms. On K7 femms is directly mapped on emms. */
+#if HAVE_AMD3DNOW
+/* On K6 femms is faster than emms. On K7 femms is directly mapped to emms. */
#define EMMS "femms"
#else
#define EMMS "emms"
#endif
-#ifdef HAVE_MMX2
+#if HAVE_MMX2
#define MOVNTQ "movntq"
#define SFENCE "sfence"
#else
uint8_t *dest = dst;
const uint8_t *s = src;
const uint8_t *end;
- #ifdef HAVE_MMX
+ #if HAVE_MMX
const uint8_t *mm_end;
#endif
end = s + src_size;
- #ifdef HAVE_MMX
+ #if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*s):"memory");
mm_end = end - 23;
- __asm__ volatile("movq %0, %%mm7"::"m"(mask32):"memory");
+ __asm__ volatile("movq %0, %%mm7"::"m"(mask32a):"memory");
while (s < mm_end)
{
__asm__ volatile(
"punpckldq 15%1, %%mm2 \n\t"
"movd 18%1, %%mm3 \n\t"
"punpckldq 21%1, %%mm3 \n\t"
- "pand %%mm7, %%mm0 \n\t"
- "pand %%mm7, %%mm1 \n\t"
- "pand %%mm7, %%mm2 \n\t"
- "pand %%mm7, %%mm3 \n\t"
+ "por %%mm7, %%mm0 \n\t"
+ "por %%mm7, %%mm1 \n\t"
+ "por %%mm7, %%mm2 \n\t"
+ "por %%mm7, %%mm3 \n\t"
MOVNTQ" %%mm0, %0 \n\t"
MOVNTQ" %%mm1, 8%0 \n\t"
MOVNTQ" %%mm2, 16%0 \n\t"
{
#ifdef WORDS_BIGENDIAN
/* RGB24 (= R,G,B) -> RGB32 (= A,B,G,R) */
- *dest++ = 0;
+ *dest++ = 255;
*dest++ = s[2];
*dest++ = s[1];
*dest++ = s[0];
*dest++ = *s++;
*dest++ = *s++;
*dest++ = *s++;
- *dest++ = 0;
+ *dest++ = 255;
#endif
}
}
uint8_t *dest = dst;
const uint8_t *s = src;
const uint8_t *end;
-#ifdef HAVE_MMX
+#if HAVE_MMX
const uint8_t *mm_end;
#endif
end = s + src_size;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*s):"memory");
mm_end = end - 31;
while (s < mm_end)
register const uint8_t *end;
const uint8_t *mm_end;
end = s + src_size;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*s));
__asm__ volatile("movq %0, %%mm4"::"m"(mask15s));
mm_end = end - 15;
register const uint8_t *end;
const uint8_t *mm_end;
end = s + src_size;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*s));
__asm__ volatile("movq %0, %%mm7"::"m"(mask15rg));
__asm__ volatile("movq %0, %%mm6"::"m"(mask15b));
{
const uint8_t *s = src;
const uint8_t *end;
-#ifdef HAVE_MMX
+#if HAVE_MMX
const uint8_t *mm_end;
#endif
uint16_t *d = (uint16_t *)dst;
end = s + src_size;
-#ifdef HAVE_MMX
+#if HAVE_MMX
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(
{
const uint8_t *s = src;
const uint8_t *end;
-#ifdef HAVE_MMX
+#if HAVE_MMX
const uint8_t *mm_end;
#endif
uint16_t *d = (uint16_t *)dst;
end = s + src_size;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*src):"memory");
__asm__ volatile(
"movq %0, %%mm7 \n\t"
{
const uint8_t *s = src;
const uint8_t *end;
-#ifdef HAVE_MMX
+#if HAVE_MMX
const uint8_t *mm_end;
#endif
uint16_t *d = (uint16_t *)dst;
end = s + src_size;
-#ifdef HAVE_MMX
+#if HAVE_MMX
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(
{
const uint8_t *s = src;
const uint8_t *end;
-#ifdef HAVE_MMX
+#if HAVE_MMX
const uint8_t *mm_end;
#endif
uint16_t *d = (uint16_t *)dst;
end = s + src_size;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*src):"memory");
__asm__ volatile(
"movq %0, %%mm7 \n\t"
{
const uint8_t *s = src;
const uint8_t *end;
-#ifdef HAVE_MMX
+#if HAVE_MMX
const uint8_t *mm_end;
#endif
uint16_t *d = (uint16_t *)dst;
end = s + src_size;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*src):"memory");
__asm__ volatile(
"movq %0, %%mm7 \n\t"
{
const uint8_t *s = src;
const uint8_t *end;
-#ifdef HAVE_MMX
+#if HAVE_MMX
const uint8_t *mm_end;
#endif
uint16_t *d = (uint16_t *)dst;
end = s + src_size;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*src):"memory");
__asm__ volatile(
"movq %0, %%mm7 \n\t"
{
const uint8_t *s = src;
const uint8_t *end;
-#ifdef HAVE_MMX
+#if HAVE_MMX
const uint8_t *mm_end;
#endif
uint16_t *d = (uint16_t *)dst;
end = s + src_size;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*src):"memory");
__asm__ volatile(
"movq %0, %%mm7 \n\t"
{
const uint8_t *s = src;
const uint8_t *end;
-#ifdef HAVE_MMX
+#if HAVE_MMX
const uint8_t *mm_end;
#endif
uint16_t *d = (uint16_t *)dst;
end = s + src_size;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*src):"memory");
__asm__ volatile(
"movq %0, %%mm7 \n\t"
static inline void RENAME(rgb15tobgr24)(const uint8_t *src, uint8_t *dst, long src_size)
{
const uint16_t *end;
-#ifdef HAVE_MMX
+#if HAVE_MMX
const uint16_t *mm_end;
#endif
uint8_t *d = dst;
const uint16_t *s = (const uint16_t*)src;
end = s + src_size/2;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*s):"memory");
mm_end = end - 7;
while (s < mm_end)
static inline void RENAME(rgb16tobgr24)(const uint8_t *src, uint8_t *dst, long src_size)
{
const uint16_t *end;
-#ifdef HAVE_MMX
+#if HAVE_MMX
const uint16_t *mm_end;
#endif
uint8_t *d = (uint8_t *)dst;
const uint16_t *s = (const uint16_t *)src;
end = s + src_size/2;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*s):"memory");
mm_end = end - 7;
while (s < mm_end)
}
}
+/*
+ * mm0 = 00 B3 00 B2 00 B1 00 B0
+ * mm1 = 00 G3 00 G2 00 G1 00 G0
+ * mm2 = 00 R3 00 R2 00 R1 00 R0
+ * mm6 = FF FF FF FF FF FF FF FF
+ * mm7 = 00 00 00 00 00 00 00 00
+ */
+#define PACK_RGB32 \
+ "packuswb %%mm7, %%mm0 \n\t" /* 00 00 00 00 B3 B2 B1 B0 */ \
+ "packuswb %%mm7, %%mm1 \n\t" /* 00 00 00 00 G3 G2 G1 G0 */ \
+ "packuswb %%mm7, %%mm2 \n\t" /* 00 00 00 00 R3 R2 R1 R0 */ \
+ "punpcklbw %%mm1, %%mm0 \n\t" /* G3 B3 G2 B2 G1 B1 G0 B0 */ \
+ "punpcklbw %%mm6, %%mm2 \n\t" /* FF R3 FF R2 FF R1 FF R0 */ \
+ "movq %%mm0, %%mm3 \n\t" \
+ "punpcklwd %%mm2, %%mm0 \n\t" /* FF R1 G1 B1 FF R0 G0 B0 */ \
+ "punpckhwd %%mm2, %%mm3 \n\t" /* FF R3 G3 B3 FF R2 G2 B2 */ \
+ MOVNTQ" %%mm0, %0 \n\t" \
+ MOVNTQ" %%mm3, 8%0 \n\t" \
+
static inline void RENAME(rgb15to32)(const uint8_t *src, uint8_t *dst, long src_size)
{
const uint16_t *end;
-#ifdef HAVE_MMX
+#if HAVE_MMX
const uint16_t *mm_end;
#endif
uint8_t *d = dst;
const uint16_t *s = (const uint16_t *)src;
end = s + src_size/2;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*s):"memory");
__asm__ volatile("pxor %%mm7,%%mm7 \n\t":::"memory");
+ __asm__ volatile("pcmpeqd %%mm6,%%mm6 \n\t":::"memory");
mm_end = end - 3;
while (s < mm_end)
{
"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 %%mm7, %%mm0 \n\t"
- "punpcklwd %%mm7, %%mm1 \n\t"
- "punpcklwd %%mm7, %%mm2 \n\t"
- "punpckhwd %%mm7, %%mm3 \n\t"
- "punpckhwd %%mm7, %%mm4 \n\t"
- "punpckhwd %%mm7, %%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"
- MOVNTQ" %%mm0, %0 \n\t"
- MOVNTQ" %%mm3, 8%0 \n\t"
+ PACK_RGB32
:"=m"(*d)
:"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r)
:"memory");
register uint16_t bgr;
bgr = *s++;
#ifdef WORDS_BIGENDIAN
- *d++ = 0;
+ *d++ = 255;
*d++ = (bgr&0x7C00)>>7;
*d++ = (bgr&0x3E0)>>2;
*d++ = (bgr&0x1F)<<3;
*d++ = (bgr&0x1F)<<3;
*d++ = (bgr&0x3E0)>>2;
*d++ = (bgr&0x7C00)>>7;
- *d++ = 0;
+ *d++ = 255;
#endif
#endif
static inline void RENAME(rgb16to32)(const uint8_t *src, uint8_t *dst, long src_size)
{
const uint16_t *end;
-#ifdef HAVE_MMX
+#if HAVE_MMX
const uint16_t *mm_end;
#endif
uint8_t *d = dst;
const uint16_t *s = (const uint16_t*)src;
end = s + src_size/2;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(PREFETCH" %0"::"m"(*s):"memory");
__asm__ volatile("pxor %%mm7,%%mm7 \n\t":::"memory");
+ __asm__ volatile("pcmpeqd %%mm6,%%mm6 \n\t":::"memory");
mm_end = end - 3;
while (s < mm_end)
{
"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 %%mm7, %%mm0 \n\t"
- "punpcklwd %%mm7, %%mm1 \n\t"
- "punpcklwd %%mm7, %%mm2 \n\t"
- "punpckhwd %%mm7, %%mm3 \n\t"
- "punpckhwd %%mm7, %%mm4 \n\t"
- "punpckhwd %%mm7, %%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"
- MOVNTQ" %%mm0, %0 \n\t"
- MOVNTQ" %%mm3, 8%0 \n\t"
+ PACK_RGB32
:"=m"(*d)
:"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r)
:"memory");
register uint16_t bgr;
bgr = *s++;
#ifdef WORDS_BIGENDIAN
- *d++ = 0;
+ *d++ = 255;
*d++ = (bgr&0xF800)>>8;
*d++ = (bgr&0x7E0)>>3;
*d++ = (bgr&0x1F)<<3;
*d++ = (bgr&0x1F)<<3;
*d++ = (bgr&0x7E0)>>3;
*d++ = (bgr&0xF800)>>8;
- *d++ = 0;
+ *d++ = 255;
#endif
}
}
static inline void RENAME(rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size)
{
- long idx = 15 - src_size;
+ x86_reg idx = 15 - src_size;
const uint8_t *s = src-idx;
uint8_t *d = dst-idx;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(
"test %0, %0 \n\t"
"jns 2f \n\t"
PREFETCH" 32(%1, %0) \n\t"
"movq (%1, %0), %%mm0 \n\t"
"movq 8(%1, %0), %%mm1 \n\t"
-# ifdef HAVE_MMX2
+# if HAVE_MMX2
"pshufw $177, %%mm0, %%mm3 \n\t"
"pshufw $177, %%mm1, %%mm5 \n\t"
"pand %%mm7, %%mm0 \n\t"
static inline void RENAME(rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size)
{
unsigned i;
-#ifdef HAVE_MMX
- long mmx_size= 23 - src_size;
+#if HAVE_MMX
+ x86_reg mmx_size= 23 - src_size;
__asm__ volatile (
"test %%"REG_a", %%"REG_a" \n\t"
"jns 2f \n\t"
long lumStride, long chromStride, long dstStride, long vertLumPerChroma)
{
long y;
- const long chromWidth= width>>1;
+ const x86_reg chromWidth= width>>1;
for (y=0; y<height; y++)
{
-#ifdef HAVE_MMX
+#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"
);
#else
-#if defined ARCH_ALPHA && defined HAVE_MVI
+#if ARCH_ALPHA && HAVE_MVI
#define pl2yuy2(n) \
y1 = yc[n]; \
y2 = yc2[n]; \
ysrc += lumStride;
dst += dstStride;
}
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__( EMMS" \n\t"
SFENCE" \n\t"
:::"memory");
long lumStride, long chromStride, long dstStride, long vertLumPerChroma)
{
long y;
- const long chromWidth= width>>1;
+ const x86_reg chromWidth= width>>1;
for (y=0; y<height; y++)
{
-#ifdef HAVE_MMX
+#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"
ysrc += lumStride;
dst += dstStride;
}
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__( EMMS" \n\t"
SFENCE" \n\t"
:::"memory");
long lumStride, long chromStride, long srcStride)
{
long y;
- const long chromWidth= width>>1;
+ const x86_reg chromWidth= width>>1;
for (y=0; y<height; y+=2)
{
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(
"xor %%"REG_a", %%"REG_a" \n\t"
"pcmpeqw %%mm7, %%mm7 \n\t"
ydst += lumStride;
src += srcStride;
}
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile( EMMS" \n\t"
SFENCE" \n\t"
:::"memory");
dst+= dstStride;
for (y=1; y<srcHeight; y++){
-#if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
- const long mmxSize= srcWidth&~15;
+#if HAVE_MMX2 || HAVE_AMD3DNOW
+ const x86_reg mmxSize= srcWidth&~15;
__asm__ volatile(
"mov %4, %%"REG_a" \n\t"
"1: \n\t"
);
#else
- const long mmxSize=1;
+ const x86_reg mmxSize=1;
#endif
dst[0 ]= (3*src[0] + src[srcStride])>>2;
dst[dstStride]= ( src[0] + 3*src[srcStride])>>2;
}
#endif
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile( EMMS" \n\t"
SFENCE" \n\t"
:::"memory");
long lumStride, long chromStride, long srcStride)
{
long y;
- const long chromWidth= width>>1;
+ const x86_reg chromWidth= width>>1;
for (y=0; y<height; y+=2)
{
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(
- "xorl %%eax, %%eax \n\t"
+ "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, %%eax, 4) \n\t"
- "movq (%0, %%eax, 4), %%mm0 \n\t" // UYVY UYVY(0)
- "movq 8(%0, %%eax, 4), %%mm1 \n\t" // UYVY UYVY(4)
+ 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)
"packuswb %%mm1, %%mm0 \n\t" // UVUV UVUV(0)
"packuswb %%mm3, %%mm2 \n\t" // YYYY YYYY(0)
- MOVNTQ" %%mm2, (%1, %%eax, 2) \n\t"
+ MOVNTQ" %%mm2, (%1, %%"REG_a", 2) \n\t"
- "movq 16(%0, %%eax, 4), %%mm1 \n\t" // UYVY UYVY(8)
- "movq 24(%0, %%eax, 4), %%mm2 \n\t" // UYVY UYVY(12)
+ "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)
"packuswb %%mm2, %%mm1 \n\t" // UVUV UVUV(8)
"packuswb %%mm4, %%mm3 \n\t" // YYYY YYYY(8)
- MOVNTQ" %%mm3, 8(%1, %%eax, 2) \n\t"
+ 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)
"packuswb %%mm1, %%mm0 \n\t" // VVVV VVVV(0)
"packuswb %%mm3, %%mm2 \n\t" // UUUU UUUU(0)
- MOVNTQ" %%mm0, (%3, %%eax) \n\t"
- MOVNTQ" %%mm2, (%2, %%eax) \n\t"
+ MOVNTQ" %%mm0, (%3, %%"REG_a") \n\t"
+ MOVNTQ" %%mm2, (%2, %%"REG_a") \n\t"
- "addl $8, %%eax \n\t"
- "cmpl %4, %%eax \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", "%eax"
+ : "memory", "%"REG_a
);
ydst += lumStride;
src += srcStride;
__asm__ volatile(
- "xorl %%eax, %%eax \n\t"
+ "xor %%"REG_a", %%"REG_a" \n\t"
ASMALIGN(4)
"1: \n\t"
- PREFETCH" 64(%0, %%eax, 4) \n\t"
- "movq (%0, %%eax, 4), %%mm0 \n\t" // YUYV YUYV(0)
- "movq 8(%0, %%eax, 4), %%mm1 \n\t" // YUYV YUYV(4)
- "movq 16(%0, %%eax, 4), %%mm2 \n\t" // YUYV YUYV(8)
- "movq 24(%0, %%eax, 4), %%mm3 \n\t" // YUYV YUYV(12)
+ 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)
"packuswb %%mm1, %%mm0 \n\t" // YYYY YYYY(0)
"packuswb %%mm3, %%mm2 \n\t" // YYYY YYYY(8)
- MOVNTQ" %%mm0, (%1, %%eax, 2) \n\t"
- MOVNTQ" %%mm2, 8(%1, %%eax, 2) \n\t"
+ MOVNTQ" %%mm0, (%1, %%"REG_a", 2) \n\t"
+ MOVNTQ" %%mm2, 8(%1, %%"REG_a", 2) \n\t"
- "addl $8, %%eax \n\t"
- "cmpl %4, %%eax \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", "%eax"
+ : "memory", "%"REG_a
);
#else
long i;
ydst += lumStride;
src += srcStride;
}
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile( EMMS" \n\t"
SFENCE" \n\t"
:::"memory");
long lumStride, long chromStride, long srcStride)
{
long y;
- const long chromWidth= width>>1;
-#ifdef HAVE_MMX
+ const x86_reg chromWidth= width>>1;
+#if HAVE_MMX
for (y=0; y<height-2; y+=2)
{
long i;
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" (-width)
+ : : "r" (src+width*3), "r" (ydst+width), "g" ((x86_reg)-width)
: "%"REG_a, "%"REG_d
);
ydst += lumStride;
"1: \n\t"
PREFETCH" 64(%0, %%"REG_d") \n\t"
PREFETCH" 64(%1, %%"REG_d") \n\t"
-#if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
+#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"
"packssdw %%mm1, %%mm0 \n\t" // V1 V0 U1 U0
"psraw $7, %%mm0 \n\t"
-#if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
+#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"
{
long w;
-#ifdef HAVE_MMX
-#ifdef HAVE_SSE2
+#if HAVE_MMX
+#if HAVE_SSE2
__asm__(
"xor %%"REG_a", %%"REG_a" \n\t"
"1: \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" (width-15)
+ ::"r"(dest), "r"(src1), "r"(src2), "r" ((x86_reg)width-15)
: "memory", "%"REG_a""
);
#else
"add $16, %%"REG_a" \n\t"
"cmp %3, %%"REG_a" \n\t"
" jb 1b \n\t"
- ::"r"(dest), "r"(src1), "r"(src2), "r" (width-15)
+ ::"r"(dest), "r"(src1), "r"(src2), "r" ((x86_reg)width-15)
: "memory", "%"REG_a
);
#endif
src1 += src1Stride;
src2 += src2Stride;
}
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__(
EMMS" \n\t"
SFENCE" \n\t"
long srcStride1, long srcStride2,
long dstStride1, long dstStride2)
{
- long y,x,w,h;
+ x86_reg y;
+ long x,w,h;
w=width/2; h=height/2;
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__ volatile(
PREFETCH" %0 \n\t"
PREFETCH" %1 \n\t"
const uint8_t* s1=src1+srcStride1*(y>>1);
uint8_t* d=dst1+dstStride1*y;
x=0;
-#ifdef HAVE_MMX
+#if HAVE_MMX
for (;x<w-31;x+=32)
{
__asm__ volatile(
const uint8_t* s2=src2+srcStride2*(y>>1);
uint8_t* d=dst2+dstStride2*y;
x=0;
-#ifdef HAVE_MMX
+#if HAVE_MMX
for (;x<w-31;x+=32)
{
__asm__ volatile(
#endif
for (;x<w;x++) d[2*x]=d[2*x+1]=s2[x];
}
-#ifdef HAVE_MMX
+#if HAVE_MMX
__asm__(
EMMS" \n\t"
SFENCE" \n\t"
long srcStride1, long srcStride2,
long srcStride3, long dstStride)
{
- long y,x,w,h;
+ x86_reg x;
+ long y,w,h;
w=width/2; h=height;
for (y=0;y<h;y++){
const uint8_t* yp=src1+srcStride1*y;
const uint8_t* vp=src3+srcStride3*(y>>2);
uint8_t* d=dst+dstStride*y;
x=0;
-#ifdef HAVE_MMX
+#if HAVE_MMX
for (;x<w-7;x+=8)
{
__asm__ volatile(
d[8*x+7] = vp[x];
}
}
-#ifdef HAVE_MMX
+#if HAVE_MMX
+ __asm__(
+ EMMS" \n\t"
+ SFENCE" \n\t"
+ ::: "memory"
+ );
+#endif
+}
+
+static void RENAME(extract_even)(const uint8_t *src, uint8_t *dst, x86_reg count)
+{
+ dst += count;
+ src += 2*count;
+ count= - count;
+
+#if HAVE_MMX
+ if(count <= -16){
+ count += 15;
+ __asm__ volatile(
+ "pcmpeqw %%mm7, %%mm7 \n\t"
+ "psrlw $8, %%mm7 \n\t"
+ "1: \n\t"
+ "movq -30(%1, %0, 2), %%mm0 \n\t"
+ "movq -22(%1, %0, 2), %%mm1 \n\t"
+ "movq -14(%1, %0, 2), %%mm2 \n\t"
+ "movq -6(%1, %0, 2), %%mm3 \n\t"
+ "pand %%mm7, %%mm0 \n\t"
+ "pand %%mm7, %%mm1 \n\t"
+ "pand %%mm7, %%mm2 \n\t"
+ "pand %%mm7, %%mm3 \n\t"
+ "packuswb %%mm1, %%mm0 \n\t"
+ "packuswb %%mm3, %%mm2 \n\t"
+ MOVNTQ" %%mm0,-15(%2, %0) \n\t"
+ MOVNTQ" %%mm2,- 7(%2, %0) \n\t"
+ "add $16, %0 \n\t"
+ " js 1b \n\t"
+ : "+r"(count)
+ : "r"(src), "r"(dst)
+ );
+ count -= 15;
+ }
+#endif
+ while(count<0){
+ dst[count]= src[2*count];
+ count++;
+ }
+}
+
+static void RENAME(extract_even2)(const uint8_t *src, uint8_t *dst0, uint8_t *dst1, x86_reg count)
+{
+ dst0+= count;
+ dst1+= count;
+ src += 4*count;
+ count= - count;
+#if HAVE_MMX
+ if(count <= -8){
+ count += 7;
+ __asm__ volatile(
+ "pcmpeqw %%mm7, %%mm7 \n\t"
+ "psrlw $8, %%mm7 \n\t"
+ "1: \n\t"
+ "movq -28(%1, %0, 4), %%mm0 \n\t"
+ "movq -20(%1, %0, 4), %%mm1 \n\t"
+ "movq -12(%1, %0, 4), %%mm2 \n\t"
+ "movq -4(%1, %0, 4), %%mm3 \n\t"
+ "pand %%mm7, %%mm0 \n\t"
+ "pand %%mm7, %%mm1 \n\t"
+ "pand %%mm7, %%mm2 \n\t"
+ "pand %%mm7, %%mm3 \n\t"
+ "packuswb %%mm1, %%mm0 \n\t"
+ "packuswb %%mm3, %%mm2 \n\t"
+ "movq %%mm0, %%mm1 \n\t"
+ "movq %%mm2, %%mm3 \n\t"
+ "psrlw $8, %%mm0 \n\t"
+ "psrlw $8, %%mm2 \n\t"
+ "pand %%mm7, %%mm1 \n\t"
+ "pand %%mm7, %%mm3 \n\t"
+ "packuswb %%mm2, %%mm0 \n\t"
+ "packuswb %%mm3, %%mm1 \n\t"
+ MOVNTQ" %%mm0,- 7(%3, %0) \n\t"
+ MOVNTQ" %%mm1,- 7(%2, %0) \n\t"
+ "add $8, %0 \n\t"
+ " js 1b \n\t"
+ : "+r"(count)
+ : "r"(src), "r"(dst0), "r"(dst1)
+ );
+ count -= 7;
+ }
+#endif
+ while(count<0){
+ dst0[count]= src[4*count+0];
+ dst1[count]= src[4*count+2];
+ count++;
+ }
+}
+
+static void RENAME(extract_even2avg)(const uint8_t *src0, const uint8_t *src1, uint8_t *dst0, uint8_t *dst1, x86_reg count)
+{
+ dst0 += count;
+ dst1 += count;
+ src0 += 4*count;
+ src1 += 4*count;
+ count= - count;
+#ifdef PAVGB
+ if(count <= -8){
+ count += 7;
+ __asm__ volatile(
+ "pcmpeqw %%mm7, %%mm7 \n\t"
+ "psrlw $8, %%mm7 \n\t"
+ "1: \n\t"
+ "movq -28(%1, %0, 4), %%mm0 \n\t"
+ "movq -20(%1, %0, 4), %%mm1 \n\t"
+ "movq -12(%1, %0, 4), %%mm2 \n\t"
+ "movq -4(%1, %0, 4), %%mm3 \n\t"
+ PAVGB" -28(%2, %0, 4), %%mm0 \n\t"
+ PAVGB" -20(%2, %0, 4), %%mm1 \n\t"
+ PAVGB" -12(%2, %0, 4), %%mm2 \n\t"
+ PAVGB" - 4(%2, %0, 4), %%mm3 \n\t"
+ "pand %%mm7, %%mm0 \n\t"
+ "pand %%mm7, %%mm1 \n\t"
+ "pand %%mm7, %%mm2 \n\t"
+ "pand %%mm7, %%mm3 \n\t"
+ "packuswb %%mm1, %%mm0 \n\t"
+ "packuswb %%mm3, %%mm2 \n\t"
+ "movq %%mm0, %%mm1 \n\t"
+ "movq %%mm2, %%mm3 \n\t"
+ "psrlw $8, %%mm0 \n\t"
+ "psrlw $8, %%mm2 \n\t"
+ "pand %%mm7, %%mm1 \n\t"
+ "pand %%mm7, %%mm3 \n\t"
+ "packuswb %%mm2, %%mm0 \n\t"
+ "packuswb %%mm3, %%mm1 \n\t"
+ MOVNTQ" %%mm0,- 7(%4, %0) \n\t"
+ MOVNTQ" %%mm1,- 7(%3, %0) \n\t"
+ "add $8, %0 \n\t"
+ " js 1b \n\t"
+ : "+r"(count)
+ : "r"(src0), "r"(src1), "r"(dst0), "r"(dst1)
+ );
+ count -= 7;
+ }
+#endif
+ while(count<0){
+ dst0[count]= (src0[4*count+0]+src1[4*count+0])>>1;
+ dst1[count]= (src0[4*count+2]+src1[4*count+2])>>1;
+ count++;
+ }
+}
+
+static void RENAME(extract_odd2)(const uint8_t *src, uint8_t *dst0, uint8_t *dst1, x86_reg count)
+{
+ dst0+= count;
+ dst1+= count;
+ src += 4*count;
+ count= - count;
+#if HAVE_MMX
+ if(count <= -8){
+ count += 7;
+ __asm__ volatile(
+ "pcmpeqw %%mm7, %%mm7 \n\t"
+ "psrlw $8, %%mm7 \n\t"
+ "1: \n\t"
+ "movq -28(%1, %0, 4), %%mm0 \n\t"
+ "movq -20(%1, %0, 4), %%mm1 \n\t"
+ "movq -12(%1, %0, 4), %%mm2 \n\t"
+ "movq -4(%1, %0, 4), %%mm3 \n\t"
+ "psrlw $8, %%mm0 \n\t"
+ "psrlw $8, %%mm1 \n\t"
+ "psrlw $8, %%mm2 \n\t"
+ "psrlw $8, %%mm3 \n\t"
+ "packuswb %%mm1, %%mm0 \n\t"
+ "packuswb %%mm3, %%mm2 \n\t"
+ "movq %%mm0, %%mm1 \n\t"
+ "movq %%mm2, %%mm3 \n\t"
+ "psrlw $8, %%mm0 \n\t"
+ "psrlw $8, %%mm2 \n\t"
+ "pand %%mm7, %%mm1 \n\t"
+ "pand %%mm7, %%mm3 \n\t"
+ "packuswb %%mm2, %%mm0 \n\t"
+ "packuswb %%mm3, %%mm1 \n\t"
+ MOVNTQ" %%mm0,- 7(%3, %0) \n\t"
+ MOVNTQ" %%mm1,- 7(%2, %0) \n\t"
+ "add $8, %0 \n\t"
+ " js 1b \n\t"
+ : "+r"(count)
+ : "r"(src), "r"(dst0), "r"(dst1)
+ );
+ count -= 7;
+ }
+#endif
+ src++;
+ while(count<0){
+ dst0[count]= src[4*count+0];
+ dst1[count]= src[4*count+2];
+ count++;
+ }
+}
+
+static void RENAME(extract_odd2avg)(const uint8_t *src0, const uint8_t *src1, uint8_t *dst0, uint8_t *dst1, x86_reg count)
+{
+ dst0 += count;
+ dst1 += count;
+ src0 += 4*count;
+ src1 += 4*count;
+ count= - count;
+#ifdef PAVGB
+ if(count <= -8){
+ count += 7;
+ __asm__ volatile(
+ "pcmpeqw %%mm7, %%mm7 \n\t"
+ "psrlw $8, %%mm7 \n\t"
+ "1: \n\t"
+ "movq -28(%1, %0, 4), %%mm0 \n\t"
+ "movq -20(%1, %0, 4), %%mm1 \n\t"
+ "movq -12(%1, %0, 4), %%mm2 \n\t"
+ "movq -4(%1, %0, 4), %%mm3 \n\t"
+ PAVGB" -28(%2, %0, 4), %%mm0 \n\t"
+ PAVGB" -20(%2, %0, 4), %%mm1 \n\t"
+ PAVGB" -12(%2, %0, 4), %%mm2 \n\t"
+ PAVGB" - 4(%2, %0, 4), %%mm3 \n\t"
+ "psrlw $8, %%mm0 \n\t"
+ "psrlw $8, %%mm1 \n\t"
+ "psrlw $8, %%mm2 \n\t"
+ "psrlw $8, %%mm3 \n\t"
+ "packuswb %%mm1, %%mm0 \n\t"
+ "packuswb %%mm3, %%mm2 \n\t"
+ "movq %%mm0, %%mm1 \n\t"
+ "movq %%mm2, %%mm3 \n\t"
+ "psrlw $8, %%mm0 \n\t"
+ "psrlw $8, %%mm2 \n\t"
+ "pand %%mm7, %%mm1 \n\t"
+ "pand %%mm7, %%mm3 \n\t"
+ "packuswb %%mm2, %%mm0 \n\t"
+ "packuswb %%mm3, %%mm1 \n\t"
+ MOVNTQ" %%mm0,- 7(%4, %0) \n\t"
+ MOVNTQ" %%mm1,- 7(%3, %0) \n\t"
+ "add $8, %0 \n\t"
+ " js 1b \n\t"
+ : "+r"(count)
+ : "r"(src0), "r"(src1), "r"(dst0), "r"(dst1)
+ );
+ count -= 7;
+ }
+#endif
+ src0++;
+ src1++;
+ while(count<0){
+ dst0[count]= (src0[4*count+0]+src1[4*count+0])>>1;
+ dst1[count]= (src0[4*count+2]+src1[4*count+2])>>1;
+ count++;
+ }
+}
+
+static void RENAME(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 y;
+ const long chromWidth= -((-width)>>1);
+
+ for (y=0; y<height; y++){
+ RENAME(extract_even)(src, ydst, width);
+ if(y&1){
+ RENAME(extract_odd2avg)(src-srcStride, src, udst, vdst, chromWidth);
+ udst+= chromStride;
+ vdst+= chromStride;
+ }
+
+ src += srcStride;
+ ydst+= lumStride;
+ }
+#if HAVE_MMX
+ __asm__(
+ EMMS" \n\t"
+ SFENCE" \n\t"
+ ::: "memory"
+ );
+#endif
+}
+
+static void RENAME(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 y;
+ const long chromWidth= -((-width)>>1);
+
+ for (y=0; y<height; y++){
+ RENAME(extract_even)(src, ydst, width);
+ RENAME(extract_odd2)(src, udst, vdst, chromWidth);
+
+ src += srcStride;
+ ydst+= lumStride;
+ udst+= chromStride;
+ vdst+= chromStride;
+ }
+#if HAVE_MMX
+ __asm__(
+ EMMS" \n\t"
+ SFENCE" \n\t"
+ ::: "memory"
+ );
+#endif
+}
+
+static void RENAME(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 y;
+ const long chromWidth= -((-width)>>1);
+
+ for (y=0; y<height; y++){
+ RENAME(extract_even)(src+1, ydst, width);
+ if(y&1){
+ RENAME(extract_even2avg)(src-srcStride, src, udst, vdst, chromWidth);
+ udst+= chromStride;
+ vdst+= chromStride;
+ }
+
+ src += srcStride;
+ ydst+= lumStride;
+ }
+#if HAVE_MMX
+ __asm__(
+ EMMS" \n\t"
+ SFENCE" \n\t"
+ ::: "memory"
+ );
+#endif
+}
+
+static void RENAME(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 y;
+ const long chromWidth= -((-width)>>1);
+
+ for (y=0; y<height; y++){
+ RENAME(extract_even)(src+1, ydst, width);
+ RENAME(extract_even2)(src, udst, vdst, chromWidth);
+
+ src += srcStride;
+ ydst+= lumStride;
+ udst+= chromStride;
+ vdst+= chromStride;
+ }
+#if HAVE_MMX
__asm__(
EMMS" \n\t"
SFENCE" \n\t"
yuv422ptoyuy2 = RENAME(yuv422ptoyuy2);
yuv422ptouyvy = RENAME(yuv422ptouyvy);
yuy2toyv12 = RENAME(yuy2toyv12);
-// uyvytoyv12 = RENAME(uyvytoyv12);
// yvu9toyv12 = RENAME(yvu9toyv12);
planar2x = RENAME(planar2x);
rgb24toyv12 = RENAME(rgb24toyv12);
interleaveBytes = RENAME(interleaveBytes);
vu9_to_vu12 = RENAME(vu9_to_vu12);
yvu9_to_yuy2 = RENAME(yvu9_to_yuy2);
+
+ uyvytoyuv420 = RENAME(uyvytoyuv420);
+ uyvytoyuv422 = RENAME(uyvytoyuv422);
+ yuyvtoyuv420 = RENAME(yuyvtoyuv420);
+ yuyvtoyuv422 = RENAME(yuyvtoyuv422);
}
projects / ffmpeg / blobdiff