|
|
|
@ -71,8 +71,7 @@ static inline void sad8_1_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h) |
|
|
|
|
"add %3, %%"REG_a" \n\t" |
|
|
|
|
" js 1b \n\t" |
|
|
|
|
: "+a" (len) |
|
|
|
|
: "r" (blk1 - len), "r" (blk2 - len), "r" ((x86_reg)stride) |
|
|
|
|
); |
|
|
|
|
: "r" (blk1 - len), "r" (blk2 - len), "r" ((x86_reg) stride)); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
static inline void sad8_1_mmxext(uint8_t *blk1, uint8_t *blk2, |
|
|
|
@ -92,8 +91,7 @@ static inline void sad8_1_mmxext(uint8_t *blk1, uint8_t *blk2, |
|
|
|
|
"sub $2, %0 \n\t" |
|
|
|
|
" jg 1b \n\t" |
|
|
|
|
: "+r" (h), "+r" (blk1), "+r" (blk2) |
|
|
|
|
: "r" ((x86_reg)stride) |
|
|
|
|
); |
|
|
|
|
: "r" ((x86_reg) stride)); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
static int sad16_sse2(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h) |
|
|
|
@ -117,8 +115,7 @@ static int sad16_sse2(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h) |
|
|
|
|
"paddw %%xmm0, %%xmm2 \n\t" |
|
|
|
|
"movd %%xmm2, %3 \n\t" |
|
|
|
|
: "+r" (h), "+r" (blk1), "+r" (blk2), "=r" (ret) |
|
|
|
|
: "r" ((x86_reg)stride) |
|
|
|
|
); |
|
|
|
|
: "r" ((x86_reg) stride)); |
|
|
|
|
return ret; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
@ -141,8 +138,7 @@ static inline void sad8_x2a_mmxext(uint8_t *blk1, uint8_t *blk2, |
|
|
|
|
"sub $2, %0 \n\t" |
|
|
|
|
" jg 1b \n\t" |
|
|
|
|
: "+r" (h), "+r" (blk1), "+r" (blk2) |
|
|
|
|
: "r" ((x86_reg)stride) |
|
|
|
|
); |
|
|
|
|
: "r" ((x86_reg) stride)); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
static inline void sad8_y2a_mmxext(uint8_t *blk1, uint8_t *blk2, |
|
|
|
@ -167,8 +163,7 @@ static inline void sad8_y2a_mmxext(uint8_t *blk1, uint8_t *blk2, |
|
|
|
|
"sub $2, %0 \n\t" |
|
|
|
|
" jg 1b \n\t" |
|
|
|
|
: "+r" (h), "+r" (blk1), "+r" (blk2) |
|
|
|
|
: "r" ((x86_reg)stride) |
|
|
|
|
); |
|
|
|
|
: "r" ((x86_reg) stride)); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
static inline void sad8_4_mmxext(uint8_t *blk1, uint8_t *blk2, |
|
|
|
@ -198,11 +193,11 @@ static inline void sad8_4_mmxext(uint8_t *blk1, uint8_t *blk2, |
|
|
|
|
"sub $2, %0 \n\t" |
|
|
|
|
" jg 1b \n\t" |
|
|
|
|
: "+r" (h), "+r" (blk1), "+r" (blk2) |
|
|
|
|
: "r" ((x86_reg)stride) |
|
|
|
|
); |
|
|
|
|
: "r" ((x86_reg) stride)); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
static inline void sad8_2_mmx(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, int stride, int h) |
|
|
|
|
static inline void sad8_2_mmx(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, |
|
|
|
|
int stride, int h) |
|
|
|
|
{ |
|
|
|
|
x86_reg len = -(x86_reg)stride * h; |
|
|
|
|
__asm__ volatile ( |
|
|
|
@ -236,8 +231,8 @@ static inline void sad8_2_mmx(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, int |
|
|
|
|
"add %4, %%"REG_a" \n\t" |
|
|
|
|
" js 1b \n\t" |
|
|
|
|
: "+a" (len) |
|
|
|
|
: "r" (blk1a - len), "r" (blk1b -len), "r" (blk2 - len), "r" ((x86_reg)stride) |
|
|
|
|
); |
|
|
|
|
: "r" (blk1a - len), "r" (blk1b - len), "r" (blk2 - len), |
|
|
|
|
"r" ((x86_reg) stride)); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h) |
|
|
|
@ -289,8 +284,8 @@ static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h) |
|
|
|
|
"add %4, %%"REG_a" \n\t" |
|
|
|
|
" js 1b \n\t" |
|
|
|
|
: "+a" (len) |
|
|
|
|
: "r" (blk1 - len), "r" (blk1 -len + stride), "r" (blk2 - len), "r" ((x86_reg)stride) |
|
|
|
|
); |
|
|
|
|
: "r" (blk1 - len), "r" (blk1 - len + stride), "r" (blk2 - len), |
|
|
|
|
"r" ((x86_reg) stride)); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
static inline int sum_mmx(void) |
|
|
|
@ -304,8 +299,7 @@ static inline int sum_mmx(void) |
|
|
|
|
"psrlq $16, %%mm6 \n\t" |
|
|
|
|
"paddw %%mm0, %%mm6 \n\t" |
|
|
|
|
"movd %%mm6, %0 \n\t" |
|
|
|
|
: "=r" (ret) |
|
|
|
|
); |
|
|
|
|
: "=r" (ret)); |
|
|
|
|
return ret & 0xFFFF; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
@ -314,8 +308,7 @@ static inline int sum_mmxext(void) |
|
|
|
|
int ret; |
|
|
|
|
__asm__ volatile ( |
|
|
|
|
"movd %%mm6, %0 \n\t" |
|
|
|
|
: "=r" (ret) |
|
|
|
|
); |
|
|
|
|
: "=r" (ret)); |
|
|
|
|
return ret; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
@ -323,55 +316,63 @@ static inline void sad8_x2a_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h) |
|
|
|
|
{ |
|
|
|
|
sad8_2_mmx(blk1, blk1 + 1, blk2, stride, h); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
static inline void sad8_y2a_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h) |
|
|
|
|
{ |
|
|
|
|
sad8_2_mmx(blk1, blk1 + stride, blk2, stride, h); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
#define PIX_SAD(suf) \ |
|
|
|
|
static int sad8_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
|
|
|
|
|
static int sad8_ ## suf(void *v, uint8_t *blk2, \
|
|
|
|
|
uint8_t *blk1, int stride, int h) \
|
|
|
|
|
{ \
|
|
|
|
|
av_assert2(h == 8); \
|
|
|
|
|
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
|
|
|
|
|
"pxor %%mm6, %%mm6 \n\t":);\
|
|
|
|
|
__asm__ volatile ( \
|
|
|
|
|
"pxor %%mm7, %%mm7 \n\t" \
|
|
|
|
|
"pxor %%mm6, %%mm6 \n\t" \
|
|
|
|
|
:); \
|
|
|
|
|
\
|
|
|
|
|
sad8_1_ ## suf(blk1, blk2, stride, 8); \
|
|
|
|
|
\
|
|
|
|
|
return sum_ ## suf(); \
|
|
|
|
|
} \
|
|
|
|
|
static int sad8_x2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
|
|
|
|
|
\
|
|
|
|
|
static int sad8_x2_ ## suf(void *v, uint8_t *blk2, \
|
|
|
|
|
uint8_t *blk1, int stride, int h) \
|
|
|
|
|
{ \
|
|
|
|
|
av_assert2(h == 8); \
|
|
|
|
|
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
|
|
|
|
|
__asm__ volatile ( \
|
|
|
|
|
"pxor %%mm7, %%mm7 \n\t" \
|
|
|
|
|
"pxor %%mm6, %%mm6 \n\t" \
|
|
|
|
|
"movq %0, %%mm5 \n\t" \
|
|
|
|
|
:: "m"(round_tab[1]) \
|
|
|
|
|
);\
|
|
|
|
|
:: "m" (round_tab[1])); \
|
|
|
|
|
\
|
|
|
|
|
sad8_x2a_ ## suf(blk1, blk2, stride, 8); \
|
|
|
|
|
\
|
|
|
|
|
return sum_ ## suf(); \
|
|
|
|
|
} \
|
|
|
|
|
\
|
|
|
|
|
static int sad8_y2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
|
|
|
|
|
static int sad8_y2_ ## suf(void *v, uint8_t *blk2, \
|
|
|
|
|
uint8_t *blk1, int stride, int h) \
|
|
|
|
|
{ \
|
|
|
|
|
av_assert2(h == 8); \
|
|
|
|
|
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
|
|
|
|
|
__asm__ volatile ( \
|
|
|
|
|
"pxor %%mm7, %%mm7 \n\t" \
|
|
|
|
|
"pxor %%mm6, %%mm6 \n\t" \
|
|
|
|
|
"movq %0, %%mm5 \n\t" \
|
|
|
|
|
:: "m"(round_tab[1]) \
|
|
|
|
|
);\
|
|
|
|
|
:: "m" (round_tab[1])); \
|
|
|
|
|
\
|
|
|
|
|
sad8_y2a_ ## suf(blk1, blk2, stride, 8); \
|
|
|
|
|
\
|
|
|
|
|
return sum_ ## suf(); \
|
|
|
|
|
} \
|
|
|
|
|
\
|
|
|
|
|
static int sad8_xy2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
|
|
|
|
|
static int sad8_xy2_ ## suf(void *v, uint8_t *blk2, \
|
|
|
|
|
uint8_t *blk1, int stride, int h) \
|
|
|
|
|
{ \
|
|
|
|
|
av_assert2(h == 8); \
|
|
|
|
|
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
|
|
|
|
|
__asm__ volatile ( \
|
|
|
|
|
"pxor %%mm7, %%mm7 \n\t" \
|
|
|
|
|
"pxor %%mm6, %%mm6 \n\t" \
|
|
|
|
|
::); \
|
|
|
|
|
\
|
|
|
|
@ -380,45 +381,55 @@ static int sad8_xy2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, i |
|
|
|
|
return sum_ ## suf(); \
|
|
|
|
|
} \
|
|
|
|
|
\
|
|
|
|
|
static int sad16_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
|
|
|
|
|
static int sad16_ ## suf(void *v, uint8_t *blk2, \
|
|
|
|
|
uint8_t *blk1, int stride, int h) \
|
|
|
|
|
{ \
|
|
|
|
|
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
|
|
|
|
|
"pxor %%mm6, %%mm6 \n\t":);\
|
|
|
|
|
__asm__ volatile ( \
|
|
|
|
|
"pxor %%mm7, %%mm7 \n\t" \
|
|
|
|
|
"pxor %%mm6, %%mm6 \n\t" \
|
|
|
|
|
:); \
|
|
|
|
|
\
|
|
|
|
|
sad8_1_ ## suf(blk1, blk2, stride, h); \
|
|
|
|
|
sad8_1_ ## suf(blk1 + 8, blk2 + 8, stride, h); \
|
|
|
|
|
\
|
|
|
|
|
return sum_ ## suf(); \
|
|
|
|
|
} \
|
|
|
|
|
static int sad16_x2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
|
|
|
|
|
\
|
|
|
|
|
static int sad16_x2_ ## suf(void *v, uint8_t *blk2, \
|
|
|
|
|
uint8_t *blk1, int stride, int h) \
|
|
|
|
|
{ \
|
|
|
|
|
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
|
|
|
|
|
__asm__ volatile ( \
|
|
|
|
|
"pxor %%mm7, %%mm7 \n\t" \
|
|
|
|
|
"pxor %%mm6, %%mm6 \n\t" \
|
|
|
|
|
"movq %0, %%mm5 \n\t" \
|
|
|
|
|
:: "m"(round_tab[1]) \
|
|
|
|
|
);\
|
|
|
|
|
:: "m" (round_tab[1])); \
|
|
|
|
|
\
|
|
|
|
|
sad8_x2a_ ## suf(blk1, blk2, stride, h); \
|
|
|
|
|
sad8_x2a_ ## suf(blk1 + 8, blk2 + 8, stride, h); \
|
|
|
|
|
\
|
|
|
|
|
return sum_ ## suf(); \
|
|
|
|
|
} \
|
|
|
|
|
static int sad16_y2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
|
|
|
|
|
\
|
|
|
|
|
static int sad16_y2_ ## suf(void *v, uint8_t *blk2, \
|
|
|
|
|
uint8_t *blk1, int stride, int h) \
|
|
|
|
|
{ \
|
|
|
|
|
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
|
|
|
|
|
__asm__ volatile ( \
|
|
|
|
|
"pxor %%mm7, %%mm7 \n\t" \
|
|
|
|
|
"pxor %%mm6, %%mm6 \n\t" \
|
|
|
|
|
"movq %0, %%mm5 \n\t" \
|
|
|
|
|
:: "m"(round_tab[1]) \
|
|
|
|
|
);\
|
|
|
|
|
:: "m" (round_tab[1])); \
|
|
|
|
|
\
|
|
|
|
|
sad8_y2a_ ## suf(blk1, blk2, stride, h); \
|
|
|
|
|
sad8_y2a_ ## suf(blk1 + 8, blk2 + 8, stride, h); \
|
|
|
|
|
\
|
|
|
|
|
return sum_ ## suf(); \
|
|
|
|
|
} \
|
|
|
|
|
static int sad16_xy2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
|
|
|
|
|
\
|
|
|
|
|
static int sad16_xy2_ ## suf(void *v, uint8_t *blk2, \
|
|
|
|
|
uint8_t *blk1, int stride, int h) \
|
|
|
|
|
{ \
|
|
|
|
|
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
|
|
|
|
|
__asm__ volatile ( \
|
|
|
|
|
"pxor %%mm7, %%mm7 \n\t" \
|
|
|
|
|
"pxor %%mm6, %%mm6 \n\t" \
|
|
|
|
|
::); \
|
|
|
|
|
\
|
|
|
|
|