Commit a61824af authored by michael's avatar michael

rewrite h261 loop filter

no malloc(64) memcpy free stuff
no filter1 A->B then filter2 A->B (yes not B->A)
no incorrect rouding after the 1d filter


git-svn-id: file:///var/local/repositories/ffmpeg/trunk@3177 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b
parent 0960ee86
...@@ -2360,32 +2360,29 @@ static void h263_h_loop_filter_c(uint8_t *src, int stride, int qscale){ ...@@ -2360,32 +2360,29 @@ static void h263_h_loop_filter_c(uint8_t *src, int stride, int qscale){
} }
} }
static void h261_v_loop_filter_c(uint8_t *dest,uint8_t *src, int stride){ static void h261_loop_filter_c(uint8_t *src, int stride){
int i,j,xy,yz; int x,y,xy,yz;
int res; int temp[64];
for(i=0; i<8; i++){
for(j=1; j<7; j++){ for(x=0; x<8; x++){
xy = j * stride + i; temp[x ] = 4*src[x ];
yz = j * 8 + i; temp[x + 7*8] = 4*src[x + 7*stride];
res = (int)src[yz-1*8] + ((int)(src[yz+0*8]) * 2) + (int)src[yz+1*8]; }
res +=2; for(y=1; y<7; y++){
res >>=2; for(x=0; x<8; x++){
dest[xy] = (uint8_t)res; xy = y * stride + x;
yz = y * 8 + x;
temp[yz] = src[xy - stride] + 2*src[xy] + src[xy + stride];
} }
} }
}
static void h261_h_loop_filter_c(uint8_t *dest,uint8_t *src, int stride){ for(y=0; y<8; y++){
int i,j,xy,yz; src[ y*stride] = (temp[ y*8] + 2)>>2;
int res; src[7+y*stride] = (temp[7+y*8] + 2)>>2;
for(i=1; i<7; i++){ for(x=1; x<7; x++){
for(j=0; j<8; j++){ xy = y * stride + x;
xy = j * stride + i; yz = y * 8 + x;
yz = j * 8 + i; src[xy] = (temp[yz-1] + 2*temp[yz] + temp[yz+1] + 8)>>4;
res = (int)src[yz-1] + ((int)(src[yz]) *2) + (int)src[yz+1];
res+=2;
res>>=2;
dest[xy] = (uint8_t)res;
} }
} }
} }
...@@ -3325,8 +3322,7 @@ void dsputil_init(DSPContext* c, AVCodecContext *avctx) ...@@ -3325,8 +3322,7 @@ void dsputil_init(DSPContext* c, AVCodecContext *avctx)
c->h263_h_loop_filter= h263_h_loop_filter_c; c->h263_h_loop_filter= h263_h_loop_filter_c;
c->h263_v_loop_filter= h263_v_loop_filter_c; c->h263_v_loop_filter= h263_v_loop_filter_c;
c->h261_h_loop_filter= h261_h_loop_filter_c; c->h261_loop_filter= h261_loop_filter_c;
c->h261_v_loop_filter= h261_v_loop_filter_c;
c->try_8x8basis= try_8x8basis_c; c->try_8x8basis= try_8x8basis_c;
c->add_8x8basis= add_8x8basis_c; c->add_8x8basis= add_8x8basis_c;
......
...@@ -261,8 +261,7 @@ typedef struct DSPContext { ...@@ -261,8 +261,7 @@ typedef struct DSPContext {
void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale); void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale); void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
void (*h261_v_loop_filter)(uint8_t *dest,uint8_t *src, int stride); void (*h261_loop_filter)(uint8_t *src, int stride);
void (*h261_h_loop_filter)(uint8_t *dest,uint8_t *src, int stride);
/* (I)DCT */ /* (I)DCT */
void (*fdct)(DCTELEM *block/* align 16*/); void (*fdct)(DCTELEM *block/* align 16*/);
......
...@@ -64,44 +64,13 @@ void ff_h261_loop_filter(H261Context * h){ ...@@ -64,44 +64,13 @@ void ff_h261_loop_filter(H261Context * h){
uint8_t *dest_y = s->dest[0]; uint8_t *dest_y = s->dest[0];
uint8_t *dest_cb= s->dest[1]; uint8_t *dest_cb= s->dest[1];
uint8_t *dest_cr= s->dest[2]; uint8_t *dest_cr= s->dest[2];
uint8_t *src;
CHECKED_ALLOCZ((src),sizeof(uint8_t) * 64 ); s->dsp.h261_loop_filter(dest_y , linesize);
s->dsp.h261_loop_filter(dest_y + 8, linesize);
for(i=0; i<8;i++) s->dsp.h261_loop_filter(dest_y + 8 * linesize , linesize);
memcpy(src+i*8,dest_y+i*linesize,sizeof(uint8_t) * 8 ); s->dsp.h261_loop_filter(dest_y + 8 * linesize + 8, linesize);
s->dsp.h261_v_loop_filter(dest_y, src, linesize); s->dsp.h261_loop_filter(dest_cb, uvlinesize);
s->dsp.h261_h_loop_filter(dest_y, src, linesize); s->dsp.h261_loop_filter(dest_cr, uvlinesize);
for(i=0; i<8;i++)
memcpy(src+i*8,dest_y+i*linesize + 8,sizeof(uint8_t) * 8 );
s->dsp.h261_v_loop_filter(dest_y + 8, src, linesize);
s->dsp.h261_h_loop_filter(dest_y + 8, src, linesize);
for(i=0; i<8;i++)
memcpy(src+i*8,dest_y+(i+8)*linesize,sizeof(uint8_t) * 8 );
s->dsp.h261_v_loop_filter(dest_y + 8 * linesize, src, linesize);
s->dsp.h261_h_loop_filter(dest_y + 8 * linesize, src, linesize);
for(i=0; i<8;i++)
memcpy(src+i*8,dest_y+(i+8)*linesize + 8,sizeof(uint8_t) * 8 );
s->dsp.h261_v_loop_filter(dest_y + 8 * linesize + 8, src, linesize);
s->dsp.h261_h_loop_filter(dest_y + 8 * linesize + 8, src, linesize);
for(i=0; i<8;i++)
memcpy(src+i*8,dest_cb+i*uvlinesize,sizeof(uint8_t) * 8 );
s->dsp.h261_v_loop_filter(dest_cb, src, uvlinesize);
s->dsp.h261_h_loop_filter(dest_cb, src, uvlinesize);
for(i=0; i<8;i++)
memcpy(src+i*8,dest_cr+i*uvlinesize,sizeof(uint8_t) * 8 );
s->dsp.h261_v_loop_filter(dest_cr, src, uvlinesize);
s->dsp.h261_h_loop_filter(dest_cr, src, uvlinesize);
fail:
av_free(src);
return;
} }
static int h261_decode_block(H261Context *h, DCTELEM *block, static int h261_decode_block(H261Context *h, DCTELEM *block,
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment