#if COMPILE_TEMPLATE_AMD3DNOW
#define PREFETCH "prefetch"
#define PAVGB "pavgusb"
-#elif COMPILE_TEMPLATE_MMX2
+#elif COMPILE_TEMPLATE_MMXEXT
#define PREFETCH "prefetchnta"
#define PAVGB "pavgb"
#else
#define EMMS "emms"
#endif
-#if COMPILE_TEMPLATE_MMX2
+#if COMPILE_TEMPLATE_MMXEXT
#define MOVNTQ "movntq"
#define SFENCE "sfence"
#else
#if !COMPILE_TEMPLATE_AMD3DNOW
-static inline void RENAME(rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb24tobgr32)(const uint8_t *src, uint8_t *dst, int src_size)
{
uint8_t *dest = dst;
const uint8_t *s = src;
__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"
- "movd 6%1, %%mm1 \n\t"
- "punpckldq 9%1, %%mm1 \n\t"
- "movd 12%1, %%mm2 \n\t"
- "punpckldq 15%1, %%mm2 \n\t"
- "movd 18%1, %%mm3 \n\t"
- "punpckldq 21%1, %%mm3 \n\t"
+ PREFETCH" 32(%1) \n\t"
+ "movd (%1), %%mm0 \n\t"
+ "punpckldq 3(%1), %%mm0 \n\t"
+ "movd 6(%1), %%mm1 \n\t"
+ "punpckldq 9(%1), %%mm1 \n\t"
+ "movd 12(%1), %%mm2 \n\t"
+ "punpckldq 15(%1), %%mm2 \n\t"
+ "movd 18(%1), %%mm3 \n\t"
+ "punpckldq 21(%1), %%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"
- MOVNTQ" %%mm3, 24%0"
- :"=m"(*dest)
- :"m"(*s)
+ MOVNTQ" %%mm0, (%0) \n\t"
+ MOVNTQ" %%mm1, 8(%0) \n\t"
+ MOVNTQ" %%mm2, 16(%0) \n\t"
+ MOVNTQ" %%mm3, 24(%0)"
+ :: "r"(dest), "r"(s)
:"memory");
dest += 32;
s += 24;
"pand "MANGLE(mask24hhhh)", %%mm5\n\t" \
"por %%mm5, %%mm4 \n\t" \
\
- MOVNTQ" %%mm0, %0 \n\t" \
- MOVNTQ" %%mm1, 8%0 \n\t" \
- MOVNTQ" %%mm4, 16%0"
+ MOVNTQ" %%mm0, (%0) \n\t" \
+ MOVNTQ" %%mm1, 8(%0) \n\t" \
+ MOVNTQ" %%mm4, 16(%0)"
-static inline void RENAME(rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb32tobgr24)(const uint8_t *src, uint8_t *dst, int src_size)
{
uint8_t *dest = dst;
const uint8_t *s = src;
mm_end = end - 31;
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"
+ 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"
STORE_BGR24_MMX
- :"=m"(*dest)
- :"m"(*s)
+ :: "r"(dest), "r"(s)
:"memory");
dest += 24;
s += 32;
MMX2, 3DNOW optimization by Nick Kurshev
32-bit C version, and and&add trick by Michael Niedermayer
*/
-static inline void RENAME(rgb15to16)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb15to16)(const uint8_t *src, uint8_t *dst, int src_size)
{
register const uint8_t* s=src;
register uint8_t* d=dst;
mm_end = end - 15;
while (s<mm_end) {
__asm__ volatile(
- PREFETCH" 32%1 \n\t"
- "movq %1, %%mm0 \n\t"
- "movq 8%1, %%mm2 \n\t"
+ 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)
+ MOVNTQ" %%mm0, (%0) \n\t"
+ MOVNTQ" %%mm2, 8(%0)"
+ :: "r"(d), "r"(s)
);
d+=16;
s+=16;
}
}
-static inline void RENAME(rgb16to15)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb16to15)(const uint8_t *src, uint8_t *dst, int src_size)
{
register const uint8_t* s=src;
register uint8_t* d=dst;
mm_end = end - 15;
while (s<mm_end) {
__asm__ volatile(
- PREFETCH" 32%1 \n\t"
- "movq %1, %%mm0 \n\t"
- "movq 8%1, %%mm2 \n\t"
+ 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"
"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)
+ MOVNTQ" %%mm0, (%0) \n\t"
+ MOVNTQ" %%mm2, 8(%0)"
+ :: "r"(d), "r"(s)
);
d+=16;
s+=16;
}
}
-static inline void RENAME(rgb32to16)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb32to16)(const uint8_t *src, uint8_t *dst, int src_size)
{
const uint8_t *s = src;
const uint8_t *end;
uint16_t *d = (uint16_t *)dst;
end = s + src_size;
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"
: "+r" (d), "+r"(s)
: "r" (mm_end), "m" (mask3216g), "m" (mask3216br), "m" (mul3216)
);
-#else
- __asm__ volatile(PREFETCH" %0"::"m"(*src):"memory");
- __asm__ volatile(
- "movq %0, %%mm7 \n\t"
- "movq %1, %%mm6 \n\t"
- ::"m"(red_16mask),"m"(green_16mask));
- while (s < mm_end) {
- __asm__ volatile(
- PREFETCH" 32%1 \n\t"
- "movd %1, %%mm0 \n\t"
- "movd 4%1, %%mm3 \n\t"
- "punpckldq 8%1, %%mm0 \n\t"
- "punpckldq 12%1, %%mm3 \n\t"
- "movq %%mm0, %%mm1 \n\t"
- "movq %%mm0, %%mm2 \n\t"
- "movq %%mm3, %%mm4 \n\t"
- "movq %%mm3, %%mm5 \n\t"
- "psrlq $3, %%mm0 \n\t"
- "psrlq $3, %%mm3 \n\t"
- "pand %2, %%mm0 \n\t"
- "pand %2, %%mm3 \n\t"
- "psrlq $5, %%mm1 \n\t"
- "psrlq $5, %%mm4 \n\t"
- "pand %%mm6, %%mm1 \n\t"
- "pand %%mm6, %%mm4 \n\t"
- "psrlq $8, %%mm2 \n\t"
- "psrlq $8, %%mm5 \n\t"
- "pand %%mm7, %%mm2 \n\t"
- "pand %%mm7, %%mm5 \n\t"
- "por %%mm1, %%mm0 \n\t"
- "por %%mm4, %%mm3 \n\t"
- "por %%mm2, %%mm0 \n\t"
- "por %%mm5, %%mm3 \n\t"
- "psllq $16, %%mm3 \n\t"
- "por %%mm3, %%mm0 \n\t"
- MOVNTQ" %%mm0, %0 \n\t"
- :"=m"(*d):"m"(*s),"m"(blue_16mask):"memory");
- d += 4;
- s += 16;
- }
-#endif
__asm__ volatile(SFENCE:::"memory");
__asm__ volatile(EMMS:::"memory");
while (s < end) {
}
}
-static inline void RENAME(rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb32tobgr16)(const uint8_t *src, uint8_t *dst, int src_size)
{
const uint8_t *s = src;
const uint8_t *end;
mm_end = end - 15;
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"
+ 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"
"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");
+ MOVNTQ" %%mm0, (%0) \n\t"
+ :: "r"(d),"r"(s),"m"(blue_16mask):"memory");
d += 4;
s += 16;
}
}
}
-static inline void RENAME(rgb32to15)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb32to15)(const uint8_t *src, uint8_t *dst, int src_size)
{
const uint8_t *s = src;
const uint8_t *end;
uint16_t *d = (uint16_t *)dst;
end = s + src_size;
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"
: "+r" (d), "+r"(s)
: "r" (mm_end), "m" (mask3215g), "m" (mask3216br), "m" (mul3215)
);
-#else
- __asm__ volatile(PREFETCH" %0"::"m"(*src):"memory");
- __asm__ volatile(
- "movq %0, %%mm7 \n\t"
- "movq %1, %%mm6 \n\t"
- ::"m"(red_15mask),"m"(green_15mask));
- while (s < mm_end) {
- __asm__ volatile(
- PREFETCH" 32%1 \n\t"
- "movd %1, %%mm0 \n\t"
- "movd 4%1, %%mm3 \n\t"
- "punpckldq 8%1, %%mm0 \n\t"
- "punpckldq 12%1, %%mm3 \n\t"
- "movq %%mm0, %%mm1 \n\t"
- "movq %%mm0, %%mm2 \n\t"
- "movq %%mm3, %%mm4 \n\t"
- "movq %%mm3, %%mm5 \n\t"
- "psrlq $3, %%mm0 \n\t"
- "psrlq $3, %%mm3 \n\t"
- "pand %2, %%mm0 \n\t"
- "pand %2, %%mm3 \n\t"
- "psrlq $6, %%mm1 \n\t"
- "psrlq $6, %%mm4 \n\t"
- "pand %%mm6, %%mm1 \n\t"
- "pand %%mm6, %%mm4 \n\t"
- "psrlq $9, %%mm2 \n\t"
- "psrlq $9, %%mm5 \n\t"
- "pand %%mm7, %%mm2 \n\t"
- "pand %%mm7, %%mm5 \n\t"
- "por %%mm1, %%mm0 \n\t"
- "por %%mm4, %%mm3 \n\t"
- "por %%mm2, %%mm0 \n\t"
- "por %%mm5, %%mm3 \n\t"
- "psllq $16, %%mm3 \n\t"
- "por %%mm3, %%mm0 \n\t"
- MOVNTQ" %%mm0, %0 \n\t"
- :"=m"(*d):"m"(*s),"m"(blue_15mask):"memory");
- d += 4;
- s += 16;
- }
-#endif
__asm__ volatile(SFENCE:::"memory");
__asm__ volatile(EMMS:::"memory");
while (s < end) {
}
}
-static inline void RENAME(rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb32tobgr15)(const uint8_t *src, uint8_t *dst, int src_size)
{
const uint8_t *s = src;
const uint8_t *end;
mm_end = end - 15;
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"
+ 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"
"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");
+ MOVNTQ" %%mm0, (%0) \n\t"
+ ::"r"(d),"r"(s),"m"(blue_15mask):"memory");
d += 4;
s += 16;
}
}
}
-static inline void RENAME(rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb24tobgr16)(const uint8_t *src, uint8_t *dst, int src_size)
{
const uint8_t *s = src;
const uint8_t *end;
mm_end = end - 11;
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"
+ 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"
"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");
+ MOVNTQ" %%mm0, (%0) \n\t"
+ ::"r"(d),"r"(s),"m"(blue_16mask):"memory");
d += 4;
s += 12;
}
}
}
-static inline void RENAME(rgb24to16)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb24to16)(const uint8_t *src, uint8_t *dst, int src_size)
{
const uint8_t *s = src;
const uint8_t *end;
mm_end = end - 15;
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"
+ 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"
"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");
+ MOVNTQ" %%mm0, (%0) \n\t"
+ ::"r"(d),"r"(s),"m"(blue_16mask):"memory");
d += 4;
s += 12;
}
}
}
-static inline void RENAME(rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb24tobgr15)(const uint8_t *src, uint8_t *dst, int src_size)
{
const uint8_t *s = src;
const uint8_t *end;
mm_end = end - 11;
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"
+ 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"
"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");
+ MOVNTQ" %%mm0, (%0) \n\t"
+ ::"r"(d),"r"(s),"m"(blue_15mask):"memory");
d += 4;
s += 12;
}
}
}
-static inline void RENAME(rgb24to15)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb24to15)(const uint8_t *src, uint8_t *dst, int src_size)
{
const uint8_t *s = src;
const uint8_t *end;
mm_end = end - 15;
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"
+ 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"
"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");
+ MOVNTQ" %%mm0, (%0) \n\t"
+ ::"r"(d),"r"(s),"m"(blue_15mask):"memory");
d += 4;
s += 12;
}
|
original bits
*/
-static inline void RENAME(rgb15tobgr24)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb15tobgr24)(const uint8_t *src, uint8_t *dst, int src_size)
{
const uint16_t *end;
const uint16_t *mm_end;
mm_end = end - 7;
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"
+ 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"
"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"
+ "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"
"por %%mm5, %%mm3 \n\t"
:"=m"(*d)
- :"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r), "m"(mmx_null)
+ :"r"(s),"m"(mask15b),"m"(mask15g),"m"(mask15r), "m"(mmx_null)
:"memory");
/* borrowed 32 to 24 */
__asm__ volatile(
STORE_BGR24_MMX
- :"=m"(*d)
- :"m"(*s)
+ :: "r"(d), "m"(*s)
:"memory");
d += 24;
s += 8;
}
}
-static inline void RENAME(rgb16tobgr24)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb16tobgr24)(const uint8_t *src, uint8_t *dst, int src_size)
{
const uint16_t *end;
const uint16_t *mm_end;
mm_end = end - 7;
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"
+ 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"
"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"
+ "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"
"por %%mm4, %%mm3 \n\t"
"por %%mm5, %%mm3 \n\t"
:"=m"(*d)
- :"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r),"m"(mmx_null)
+ :"r"(s),"m"(mask16b),"m"(mask16g),"m"(mask16r),"m"(mmx_null)
:"memory");
/* borrowed 32 to 24 */
__asm__ volatile(
STORE_BGR24_MMX
- :"=m"(*d)
- :"m"(*s)
+ :: "r"(d), "m"(*s)
:"memory");
d += 24;
s += 8;
"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" \
+ 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)
+static inline void RENAME(rgb15to32)(const uint8_t *src, uint8_t *dst, int src_size)
{
const uint16_t *end;
const uint16_t *mm_end;
mm_end = end - 3;
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"
+ 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"
"psrlq $2, %%mm1 \n\t"
"psrlq $7, %%mm2 \n\t"
PACK_RGB32
- :"=m"(*d)
- :"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r)
+ ::"r"(d),"r"(s),"m"(mask15b),"m"(mask15g),"m"(mask15r)
:"memory");
d += 16;
s += 4;
}
}
-static inline void RENAME(rgb16to32)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(rgb16to32)(const uint8_t *src, uint8_t *dst, int src_size)
{
const uint16_t *end;
const uint16_t *mm_end;
mm_end = end - 3;
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"
+ 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"
"psrlq $3, %%mm1 \n\t"
"psrlq $8, %%mm2 \n\t"
PACK_RGB32
- :"=m"(*d)
- :"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r)
+ ::"r"(d),"r"(s),"m"(mask16b),"m"(mask16g),"m"(mask16r)
:"memory");
d += 16;
s += 4;
}
}
-static inline void RENAME(shuffle_bytes_2103)(const uint8_t *src, uint8_t *dst, long src_size)
+static inline void RENAME(shuffle_bytes_2103)(const uint8_t *src, uint8_t *dst, int src_size)
{
x86_reg idx = 15 - src_size;
const uint8_t *s = src-idx;
PREFETCH" 32(%1, %0) \n\t"
"movq (%1, %0), %%mm0 \n\t"
"movq 8(%1, %0), %%mm1 \n\t"
-# if COMPILE_TEMPLATE_MMX2
+# if COMPILE_TEMPLATE_MMXEXT
"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)
+static inline void RENAME(rgb24tobgr24)(const uint8_t *src, uint8_t *dst, int src_size)
{
unsigned i;
x86_reg mmx_size= 23 - src_size;
}
static inline void RENAME(yuvPlanartoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
- long width, long height,
- long lumStride, long chromStride, long dstStride, long vertLumPerChroma)
+ int width, int height,
+ int lumStride, int chromStride, int dstStride, int vertLumPerChroma)
{
- long y;
+ int y;
const x86_reg chromWidth= width>>1;
for (y=0; y<height; y++) {
//FIXME handle 2 lines at once (fewer prefetches, reuse some chroma, but very likely memory-limited anyway)
* (If this is a problem for anyone then tell me, and I will fix it.)
*/
static inline void RENAME(yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
- long width, long height,
- long lumStride, long chromStride, long dstStride)
+ int width, int height,
+ int lumStride, int chromStride, int dstStride)
{
//FIXME interpolate chroma
RENAME(yuvPlanartoyuy2)(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride, 2);
}
static inline void RENAME(yuvPlanartouyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
- long width, long height,
- long lumStride, long chromStride, long dstStride, long vertLumPerChroma)
+ int width, int height,
+ int lumStride, int chromStride, int dstStride, int vertLumPerChroma)
{
- long y;
+ int y;
const x86_reg chromWidth= width>>1;
for (y=0; y<height; y++) {
//FIXME handle 2 lines at once (fewer prefetches, reuse some chroma, but very likely memory-limited anyway)
* (If this is a problem for anyone then tell me, and I will fix it.)
*/
static inline void RENAME(yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
- long width, long height,
- long lumStride, long chromStride, long dstStride)
+ int width, int height,
+ int lumStride, int chromStride, int dstStride)
{
//FIXME interpolate chroma
RENAME(yuvPlanartouyvy)(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride, 2);
* Width should be a multiple of 16.
*/
static inline void RENAME(yuv422ptouyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
- long width, long height,
- long lumStride, long chromStride, long dstStride)
+ int width, int height,
+ int lumStride, int chromStride, int dstStride)
{
RENAME(yuvPlanartouyvy)(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride, 1);
}
* Width should be a multiple of 16.
*/
static inline void RENAME(yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
- long width, long height,
- long lumStride, long chromStride, long dstStride)
+ int width, int height,
+ int lumStride, int chromStride, int dstStride)
{
RENAME(yuvPlanartoyuy2)(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride, 1);
}
* (If this is a problem for anyone then tell me, and I will fix it.)
*/
static inline void RENAME(yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
- long width, long height,
- long lumStride, long chromStride, long srcStride)
+ int width, int height,
+ int lumStride, int chromStride, int srcStride)
{
- long y;
+ int y;
const x86_reg chromWidth= width>>1;
for (y=0; y<height; y+=2) {
__asm__ volatile(
}
#endif /* !COMPILE_TEMPLATE_AMD3DNOW */
-#if COMPILE_TEMPLATE_MMX2 || COMPILE_TEMPLATE_AMD3DNOW
-static inline void RENAME(planar2x)(const uint8_t *src, uint8_t *dst, long srcWidth, long srcHeight, long srcStride, long dstStride)
+#if COMPILE_TEMPLATE_MMXEXT || COMPILE_TEMPLATE_AMD3DNOW
+static inline void RENAME(planar2x)(const uint8_t *src, uint8_t *dst, int srcWidth, int srcHeight, int srcStride, int dstStride)
{
- long x,y;
+ int x,y;
dst[0]= src[0];
SFENCE" \n\t"
:::"memory");
}
-#endif /* COMPILE_TEMPLATE_MMX2 || COMPILE_TEMPLATE_AMD3DNOW */
+#endif /* COMPILE_TEMPLATE_MMXEXT || COMPILE_TEMPLATE_AMD3DNOW */
#if !COMPILE_TEMPLATE_AMD3DNOW
/**
* FIXME: Write HQ version.
*/
static inline void RENAME(uyvytoyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
- long width, long height,
- long lumStride, long chromStride, long srcStride)
+ int width, int height,
+ int lumStride, int chromStride, int srcStride)
{
- long y;
+ int y;
const x86_reg chromWidth= width>>1;
for (y=0; y<height; y+=2) {
__asm__ volatile(
* FIXME: Write HQ version.
*/
static inline void RENAME(rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
- long width, long height,
- long lumStride, long chromStride, long srcStride)
+ int width, int height,
+ int lumStride, int chromStride, int srcStride)
{
- long y;
+ int y;
const x86_reg chromWidth= width>>1;
for (y=0; y<height-2; y+=2) {
- long i;
+ int i;
for (i=0; i<2; i++) {
__asm__ volatile(
"mov %2, %%"REG_a" \n\t"
"1: \n\t"
PREFETCH" 64(%0, %%"REG_d") \n\t"
PREFETCH" 64(%1, %%"REG_d") \n\t"
-#if COMPILE_TEMPLATE_MMX2 || COMPILE_TEMPLATE_AMD3DNOW
+#if COMPILE_TEMPLATE_MMXEXT || COMPILE_TEMPLATE_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 COMPILE_TEMPLATE_MMX2 || COMPILE_TEMPLATE_AMD3DNOW
+#if COMPILE_TEMPLATE_MMXEXT || COMPILE_TEMPLATE_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"
#if !COMPILE_TEMPLATE_AMD3DNOW
static void RENAME(interleaveBytes)(const uint8_t *src1, const uint8_t *src2, uint8_t *dest,
- long width, long height, long src1Stride,
- long src2Stride, long dstStride)
+ int width, int height, int src1Stride,
+ int src2Stride, int dstStride)
{
- long h;
+ int h;
for (h=0; h < height; h++) {
- long w;
+ int w;
#if COMPILE_TEMPLATE_SSE2
__asm__(
#if !COMPILE_TEMPLATE_AMD3DNOW
static inline void RENAME(vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
uint8_t *dst1, uint8_t *dst2,
- long width, long height,
- long srcStride1, long srcStride2,
- long dstStride1, long dstStride2)
+ int width, int height,
+ int srcStride1, int srcStride2,
+ int dstStride1, int dstStride2)
{
- x86_reg y;
- long x,w,h;
+ x86_reg x, y;
+ int w,h;
w=width/2; h=height/2;
__asm__ volatile(
PREFETCH" %0 \n\t"
x=0;
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"
+ PREFETCH" 32(%1,%2) \n\t"
+ "movq (%1,%2), %%mm0 \n\t"
+ "movq 8(%1,%2), %%mm2 \n\t"
+ "movq 16(%1,%2), %%mm4 \n\t"
+ "movq 24(%1,%2), %%mm6 \n\t"
"movq %%mm0, %%mm1 \n\t"
"movq %%mm2, %%mm3 \n\t"
"movq %%mm4, %%mm5 \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])
+ MOVNTQ" %%mm0, (%0,%2,2) \n\t"
+ MOVNTQ" %%mm1, 8(%0,%2,2) \n\t"
+ MOVNTQ" %%mm2, 16(%0,%2,2) \n\t"
+ MOVNTQ" %%mm3, 24(%0,%2,2) \n\t"
+ MOVNTQ" %%mm4, 32(%0,%2,2) \n\t"
+ MOVNTQ" %%mm5, 40(%0,%2,2) \n\t"
+ MOVNTQ" %%mm6, 48(%0,%2,2) \n\t"
+ MOVNTQ" %%mm7, 56(%0,%2,2)"
+ :: "r"(d), "r"(s1), "r"(x)
:"memory");
}
for (;x<w;x++) d[2*x]=d[2*x+1]=s1[x];
x=0;
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"
+ PREFETCH" 32(%1,%2) \n\t"
+ "movq (%1,%2), %%mm0 \n\t"
+ "movq 8(%1,%2), %%mm2 \n\t"
+ "movq 16(%1,%2), %%mm4 \n\t"
+ "movq 24(%1,%2), %%mm6 \n\t"
"movq %%mm0, %%mm1 \n\t"
"movq %%mm2, %%mm3 \n\t"
"movq %%mm4, %%mm5 \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])
+ MOVNTQ" %%mm0, (%0,%2,2) \n\t"
+ MOVNTQ" %%mm1, 8(%0,%2,2) \n\t"
+ MOVNTQ" %%mm2, 16(%0,%2,2) \n\t"
+ MOVNTQ" %%mm3, 24(%0,%2,2) \n\t"
+ MOVNTQ" %%mm4, 32(%0,%2,2) \n\t"
+ MOVNTQ" %%mm5, 40(%0,%2,2) \n\t"
+ MOVNTQ" %%mm6, 48(%0,%2,2) \n\t"
+ MOVNTQ" %%mm7, 56(%0,%2,2)"
+ :: "r"(d), "r"(s2), "r"(x)
:"memory");
}
for (;x<w;x++) d[2*x]=d[2*x+1]=s2[x];
static inline void RENAME(yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
uint8_t *dst,
- long width, long height,
- long srcStride1, long srcStride2,
- long srcStride3, long dstStride)
+ int width, int height,
+ int srcStride1, int srcStride2,
+ int srcStride3, int dstStride)
{
x86_reg x;
- long y,w,h;
+ int y,w,h;
w=width/2; h=height;
for (y=0;y<h;y++) {
const uint8_t* yp=src1+srcStride1*y;
:"memory");
}
for (; x<w; x++) {
- const long x2 = x<<2;
+ const int x2 = x<<2;
d[8*x+0] = yp[x2];
d[8*x+1] = up[x];
d[8*x+2] = yp[x2+1];
}
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)
+ int width, int height,
+ int lumStride, int chromStride, int srcStride)
{
- long y;
- const long chromWidth= -((-width)>>1);
+ int y;
+ const int chromWidth= -((-width)>>1);
for (y=0; y<height; y++) {
RENAME(extract_even)(src, ydst, width);
#if !COMPILE_TEMPLATE_AMD3DNOW
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)
+ int width, int height,
+ int lumStride, int chromStride, int srcStride)
{
- long y;
- const long chromWidth= -((-width)>>1);
+ int y;
+ const int chromWidth= -((-width)>>1);
for (y=0; y<height; y++) {
RENAME(extract_even)(src, ydst, width);
#endif /* !COMPILE_TEMPLATE_AMD3DNOW */
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)
+ int width, int height,
+ int lumStride, int chromStride, int srcStride)
{
- long y;
- const long chromWidth= -((-width)>>1);
+ int y;
+ const int chromWidth= -((-width)>>1);
for (y=0; y<height; y++) {
RENAME(extract_even)(src+1, ydst, width);
#if !COMPILE_TEMPLATE_AMD3DNOW
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)
+ int width, int height,
+ int lumStride, int chromStride, int srcStride)
{
- long y;
- const long chromWidth= -((-width)>>1);
+ int y;
+ const int chromWidth= -((-width)>>1);
for (y=0; y<height; y++) {
RENAME(extract_even)(src+1, ydst, width);
yuyvtoyuv422 = RENAME(yuyvtoyuv422);
#endif /* !COMPILE_TEMPLATE_SSE2 */
-#if COMPILE_TEMPLATE_MMX2 || COMPILE_TEMPLATE_AMD3DNOW
+#if COMPILE_TEMPLATE_MMXEXT || COMPILE_TEMPLATE_AMD3DNOW
planar2x = RENAME(planar2x);
-#endif /* COMPILE_TEMPLATE_MMX2 || COMPILE_TEMPLATE_AMD3DNOW */
+#endif /* COMPILE_TEMPLATE_MMXEXT || COMPILE_TEMPLATE_AMD3DNOW */
rgb24toyv12 = RENAME(rgb24toyv12);
yuyvtoyuv420 = RENAME(yuyvtoyuv420);
projects / ffmpeg / blobdiff