falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
Browse Source

x86: motion_est: K&R formatting cosmetics

tags/n2.3
Diego Biurrun 11 years ago
parent
a36947c167
commit
f8bbebecfd
1 changed files with 161 additions and 150 deletions
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. +161
    -150
      libavcodec/x86/motion_est.c

+ 161
- 150
libavcodec/x86/motion_est.c View File

@@ -31,18 +31,18 @@

#if HAVE_INLINE_ASM

DECLARE_ASM_CONST(8, uint64_t, round_tab)[3]={
0x0000000000000000ULL,
0x0001000100010001ULL,
0x0002000200020002ULL,
DECLARE_ASM_CONST(8, uint64_t, round_tab)[3] = {
0x0000000000000000ULL,
0x0001000100010001ULL,
0x0002000200020002ULL,
};

DECLARE_ASM_CONST(8, uint64_t, bone)= 0x0101010101010101LL;
DECLARE_ASM_CONST(8, uint64_t, bone) = 0x0101010101010101LL;

static inline void sad8_1_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
{
x86_reg len= -(stride*h);
__asm__ volatile(
x86_reg len = -(stride * h);
__asm__ volatile (
".p2align 4 \n\t"
"1: \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
@@ -71,14 +71,13 @@ 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,
int stride, int h)
{
__asm__ volatile(
__asm__ volatile (
".p2align 4 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
@@ -92,14 +91,13 @@ 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)
{
int ret;
__asm__ volatile(
__asm__ volatile (
"pxor %%xmm2, %%xmm2 \n\t"
".p2align 4 \n\t"
"1: \n\t"
@@ -116,16 +114,15 @@ static int sad16_sse2(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)
"movhlps %%xmm2, %%xmm0 \n\t"
"paddw %%xmm0, %%xmm2 \n\t"
"movd %%xmm2, %3 \n\t"
: "+r" (h), "+r" (blk1), "+r" (blk2), "=r"(ret)
: "r" ((x86_reg)stride)
);
: "+r" (h), "+r" (blk1), "+r" (blk2), "=r" (ret)
: "r" ((x86_reg) stride));
return ret;
}

static inline void sad8_x2a_mmxext(uint8_t *blk1, uint8_t *blk2,
int stride, int h)
{
__asm__ volatile(
__asm__ volatile (
".p2align 4 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
@@ -141,14 +138,13 @@ 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,
int stride, int h)
{
__asm__ volatile(
__asm__ volatile (
"movq (%1), %%mm0 \n\t"
"add %3, %1 \n\t"
".p2align 4 \n\t"
@@ -167,14 +163,13 @@ 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,
int stride, int h)
{
__asm__ volatile(
__asm__ volatile (
"movq "MANGLE(bone)", %%mm5 \n\t"
"movq (%1), %%mm0 \n\t"
"pavgb 1(%1), %%mm0 \n\t"
@@ -198,14 +193,14 @@ 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= -(stride*h);
__asm__ volatile(
x86_reg len = -(stride * h);
__asm__ volatile (
".p2align 4 \n\t"
"1: \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
@@ -236,15 +231,15 @@ 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)
{
x86_reg len= -(stride*h);
__asm__ volatile(
"movq (%1, %%"REG_a"), %%mm0 \n\t"
x86_reg len = -(stride * h);
__asm__ volatile (
"movq (%1, %%"REG_a"), %%mm0 \n\t"
"movq 1(%1, %%"REG_a"), %%mm2 \n\t"
"movq %%mm0, %%mm1 \n\t"
"movq %%mm2, %%mm3 \n\t"
@@ -256,7 +251,7 @@ static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
"paddw %%mm3, %%mm1 \n\t"
".p2align 4 \n\t"
"1: \n\t"
"movq (%2, %%"REG_a"), %%mm2 \n\t"
"movq (%2, %%"REG_a"), %%mm2 \n\t"
"movq 1(%2, %%"REG_a"), %%mm4 \n\t"
"movq %%mm2, %%mm3 \n\t"
"movq %%mm4, %%mm5 \n\t"
@@ -289,14 +284,14 @@ 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)
{
int ret;
__asm__ volatile(
__asm__ volatile (
"movq %%mm6, %%mm0 \n\t"
"psrlq $32, %%mm6 \n\t"
"paddw %%mm0, %%mm6 \n\t"
@@ -304,129 +299,145 @@ static inline int sum_mmx(void)
"psrlq $16, %%mm6 \n\t"
"paddw %%mm0, %%mm6 \n\t"
"movd %%mm6, %0 \n\t"
: "=r" (ret)
);
return ret&0xFFFF;
: "=r" (ret));
return ret & 0xFFFF;
}

static inline int sum_mmxext(void)
{
int ret;
__asm__ volatile(
__asm__ volatile (
"movd %%mm6, %0 \n\t"
: "=r" (ret)
);
: "=r" (ret));
return ret;
}

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);
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);
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)\
{\
assert(h==8);\
__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)\
{\
assert(h==8);\
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
"pxor %%mm6, %%mm6 \n\t"\
"movq %0, %%mm5 \n\t"\
:: "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)\
{\
assert(h==8);\
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
"pxor %%mm6, %%mm6 \n\t"\
"movq %0, %%mm5 \n\t"\
:: "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)\
{\
assert(h==8);\
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
"pxor %%mm6, %%mm6 \n\t"\
::);\
\
sad8_4_ ## suf(blk1, blk2, stride, 8);\
\
return sum_ ## suf();\
}\
\
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":);\
\
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)\
{\
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
"pxor %%mm6, %%mm6 \n\t"\
"movq %0, %%mm5 \n\t"\
:: "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)\
{\
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
"pxor %%mm6, %%mm6 \n\t"\
"movq %0, %%mm5 \n\t"\
:: "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)\
{\
__asm__ volatile("pxor %%mm7, %%mm7 \n\t"\
"pxor %%mm6, %%mm6 \n\t"\
::);\
\
sad8_4_ ## suf(blk1 , blk2 , stride, h);\
sad8_4_ ## suf(blk1+8, blk2+8, stride, h);\
\
return sum_ ## suf();\
}\
#define PIX_SAD(suf) \
static int sad8_ ## suf(void *v, uint8_t *blk2, \
uint8_t *blk1, int stride, int h) \
{ \
assert(h == 8); \
__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) \
{ \
assert(h == 8); \
__asm__ volatile ( \
"pxor %%mm7, %%mm7 \n\t" \
"pxor %%mm6, %%mm6 \n\t" \
"movq %0, %%mm5 \n\t" \
:: "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) \
{ \
assert(h == 8); \
__asm__ volatile ( \
"pxor %%mm7, %%mm7 \n\t" \
"pxor %%mm6, %%mm6 \n\t" \
"movq %0, %%mm5 \n\t" \
:: "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) \
{ \
assert(h == 8); \
__asm__ volatile ( \
"pxor %%mm7, %%mm7 \n\t" \
"pxor %%mm6, %%mm6 \n\t" \
::); \
\
sad8_4_ ## suf(blk1, blk2, stride, 8); \
\
return sum_ ## suf(); \
} \
\
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" \
:); \
\
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) \
{ \
__asm__ volatile ( \
"pxor %%mm7, %%mm7 \n\t" \
"pxor %%mm6, %%mm6 \n\t" \
"movq %0, %%mm5 \n\t" \
:: "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) \
{ \
__asm__ volatile ( \
"pxor %%mm7, %%mm7 \n\t" \
"pxor %%mm6, %%mm6 \n\t" \
"movq %0, %%mm5 \n\t" \
:: "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) \
{ \
__asm__ volatile ( \
"pxor %%mm7, %%mm7 \n\t" \
"pxor %%mm6, %%mm6 \n\t" \
::); \
\
sad8_4_ ## suf(blk1, blk2, stride, h); \
sad8_4_ ## suf(blk1 + 8, blk2 + 8, stride, h); \
\
return sum_ ## suf(); \
} \

PIX_SAD(mmx)
PIX_SAD(mmxext)
@@ -448,17 +459,17 @@ av_cold void ff_dsputil_init_pix_mmx(DSPContext *c, AVCodecContext *avctx)
c->pix_abs[1][2] = sad8_y2_mmx;
c->pix_abs[1][3] = sad8_xy2_mmx;

c->sad[0]= sad16_mmx;
c->sad[1]= sad8_mmx;
c->sad[0] = sad16_mmx;
c->sad[1] = sad8_mmx;
}
if (INLINE_MMXEXT(cpu_flags)) {
c->pix_abs[0][0] = sad16_mmxext;
c->pix_abs[1][0] = sad8_mmxext;

c->sad[0] = sad16_mmxext;
c->sad[1] = sad8_mmxext;
c->sad[0] = sad16_mmxext;
c->sad[1] = sad8_mmxext;

if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
c->pix_abs[0][1] = sad16_x2_mmxext;
c->pix_abs[0][2] = sad16_y2_mmxext;
c->pix_abs[0][3] = sad16_xy2_mmxext;
@@ -468,7 +479,7 @@ av_cold void ff_dsputil_init_pix_mmx(DSPContext *c, AVCodecContext *avctx)
}
}
if (INLINE_SSE2(cpu_flags) && !(cpu_flags & AV_CPU_FLAG_3DNOW)) {
c->sad[0]= sad16_sse2;
c->sad[0] = sad16_sse2;
}
#endif /* HAVE_INLINE_ASM */
}

Write Preview
Loading…
Cancel
Save