mirror of https://git.ffmpeg.org/ffmpeg.git
sizeof(s->block) isnt 64*6*2 anymore bugfix
mpeg12 decoding optimization Originally committed as revision 364 to svn://svn.ffmpeg.org/ffmpeg/trunk
This commit is contained in:
parent
cf8039b2cf
commit
649c00c96d
|
@ -30,6 +30,7 @@ void (*get_pixels)(DCTELEM *block, const UINT8 *pixels, int line_size);
|
|||
void (*put_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size);
|
||||
void (*add_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size);
|
||||
void (*gmc1)(UINT8 *dst, UINT8 *src, int srcStride, int h, int x16, int y16, int rounder);
|
||||
void (*clear_blocks)(DCTELEM *blocks);
|
||||
|
||||
op_pixels_abs_func pix_abs16x16;
|
||||
op_pixels_abs_func pix_abs16x16_x2;
|
||||
|
@ -866,6 +867,11 @@ void block_permute(INT16 *block)
|
|||
}
|
||||
#endif
|
||||
|
||||
void clear_blocks_c(DCTELEM *blocks)
|
||||
{
|
||||
memset(blocks, 0, sizeof(DCTELEM)*6*64);
|
||||
}
|
||||
|
||||
void dsputil_init(void)
|
||||
{
|
||||
int i, j;
|
||||
|
@ -890,6 +896,7 @@ void dsputil_init(void)
|
|||
put_pixels_clamped = put_pixels_clamped_c;
|
||||
add_pixels_clamped = add_pixels_clamped_c;
|
||||
gmc1= gmc1_c;
|
||||
clear_blocks= clear_blocks_c;
|
||||
|
||||
pix_abs16x16 = pix_abs16x16_c;
|
||||
pix_abs16x16_x2 = pix_abs16x16_x2_c;
|
||||
|
|
|
@ -40,11 +40,13 @@ extern void (*get_pixels)(DCTELEM *block, const UINT8 *pixels, int line_size);
|
|||
extern void (*put_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size);
|
||||
extern void (*add_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size);
|
||||
extern void (*gmc1)(UINT8 *dst, UINT8 *src, int srcStride, int h, int x16, int y16, int rounder);
|
||||
extern void (*clear_blocks)(DCTELEM *blocks);
|
||||
|
||||
|
||||
void get_pixels_c(DCTELEM *block, const UINT8 *pixels, int line_size);
|
||||
void put_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size);
|
||||
void add_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size);
|
||||
void clear_blocks_c(DCTELEM *blocks);
|
||||
|
||||
/* add and put pixel (decoding) */
|
||||
typedef void (*op_pixels_func)(UINT8 *block, const UINT8 *pixels, int line_size, int h);
|
||||
|
|
|
@ -156,6 +156,7 @@ static int h263_decode_frame(AVCodecContext *avctx,
|
|||
if (s->mb_y && !s->h263_pred) {
|
||||
s->first_gob_line = h263_decode_gob_header(s);
|
||||
}
|
||||
|
||||
s->block_index[0]= s->block_wrap[0]*(s->mb_y*2 + 1) - 1;
|
||||
s->block_index[1]= s->block_wrap[0]*(s->mb_y*2 + 1);
|
||||
s->block_index[2]= s->block_wrap[0]*(s->mb_y*2 + 2) - 1;
|
||||
|
@ -183,28 +184,8 @@ static int h263_decode_frame(AVCodecContext *avctx,
|
|||
s->y_dc_scale = 8;
|
||||
s->c_dc_scale = 8;
|
||||
}
|
||||
|
||||
#ifdef HAVE_MMX
|
||||
if (mm_flags & MM_MMX) {
|
||||
asm volatile(
|
||||
"pxor %%mm7, %%mm7 \n\t"
|
||||
"movl $-128*6, %%eax \n\t"
|
||||
"1: \n\t"
|
||||
"movq %%mm7, (%0, %%eax) \n\t"
|
||||
"movq %%mm7, 8(%0, %%eax) \n\t"
|
||||
"movq %%mm7, 16(%0, %%eax) \n\t"
|
||||
"movq %%mm7, 24(%0, %%eax) \n\t"
|
||||
"addl $32, %%eax \n\t"
|
||||
" js 1b \n\t"
|
||||
: : "r" (((int)s->block)+128*6)
|
||||
: "%eax"
|
||||
);
|
||||
}else{
|
||||
memset(s->block, 0, sizeof(s->block));
|
||||
}
|
||||
#else
|
||||
memset(s->block, 0, sizeof(s->block));
|
||||
#endif
|
||||
clear_blocks(s->block[0]);
|
||||
|
||||
s->mv_dir = MV_DIR_FORWARD;
|
||||
s->mv_type = MV_TYPE_16X16;
|
||||
if (s->h263_msmpeg4) {
|
||||
|
|
|
@ -1025,6 +1025,23 @@ static void sub_pixels_xy2_mmx( DCTELEM *block, const UINT8 *pixels, int line
|
|||
} while(--h);
|
||||
}
|
||||
|
||||
static void clear_blocks_mmx(DCTELEM *blocks)
|
||||
{
|
||||
asm volatile(
|
||||
"pxor %%mm7, %%mm7 \n\t"
|
||||
"movl $-128*6, %%eax \n\t"
|
||||
"1: \n\t"
|
||||
"movq %%mm7, (%0, %%eax) \n\t"
|
||||
"movq %%mm7, 8(%0, %%eax) \n\t"
|
||||
"movq %%mm7, 16(%0, %%eax) \n\t"
|
||||
"movq %%mm7, 24(%0, %%eax) \n\t"
|
||||
"addl $32, %%eax \n\t"
|
||||
" js 1b \n\t"
|
||||
: : "r" (((int)blocks)+128*6)
|
||||
: "%eax"
|
||||
);
|
||||
}
|
||||
|
||||
static void just_return() { return; }
|
||||
|
||||
void dsputil_init_mmx(void)
|
||||
|
@ -1049,7 +1066,8 @@ void dsputil_init_mmx(void)
|
|||
get_pixels = get_pixels_mmx;
|
||||
put_pixels_clamped = put_pixels_clamped_mmx;
|
||||
add_pixels_clamped = add_pixels_clamped_mmx;
|
||||
|
||||
clear_blocks= clear_blocks_mmx;
|
||||
|
||||
pix_abs16x16 = pix_abs16x16_mmx;
|
||||
pix_abs16x16_x2 = pix_abs16x16_x2_mmx;
|
||||
pix_abs16x16_y2 = pix_abs16x16_y2_mmx;
|
||||
|
|
|
@ -1402,7 +1402,7 @@ static int mpeg_decode_slice(AVCodecContext *avctx,
|
|||
}
|
||||
|
||||
for(;;) {
|
||||
memset(s->block, 0, sizeof(s->block));
|
||||
clear_blocks(s->block[0]);
|
||||
ret = mpeg_decode_mb(s, s->block);
|
||||
dprintf("ret=%d\n", ret);
|
||||
if (ret < 0)
|
||||
|
|
Loading…
Reference in New Issue