xenia-project/FFmpeg
1
0
Fork 0
mirror of https://github.com/xenia-project/FFmpeg.git synced 2025-01-21 03:15:54 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

mmx2 optimization of huffyuv median encoding

Browse Source 
Originally committed as revision 2372 to svn://svn.ffmpeg.org/ffmpeg/trunk
This commit is contained in:
Michael Niedermayer 2003-10-13 17:27:30 +00:00
parent 8c812d7317
commit 8470540344
4 changed files with 70 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
libavcodec/dsputil.c
View File

@ -2526,6 +2526,24 @@ static void diff_bytes_c(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w){
dst[i+0] = src1[i+0]-src2[i+0];
}
static void sub_hfyu_median_prediction_c(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top){
int i;
uint8_t l, lt;
l= *left;
lt= *left_top;
for(i=0; i<w; i++){
const int pred= mid_pred(l, src1[i], (l + src1[i] - lt)&0xFF);
lt= src1[i];
l= src2[i];
dst[i]= l - pred;
}
*left= l;
*left_top= lt;
}
#define BUTTERFLY2(o1,o2,i1,i2) \
o1= (i1)+(i2);\
o2= (i1)-(i2);
@ -3007,6 +3025,7 @@ void dsputil_init(DSPContext* c, AVCodecContext *avctx)
c->add_bytes= add_bytes_c;
c->diff_bytes= diff_bytes_c;
c->sub_hfyu_median_prediction= sub_hfyu_median_prediction_c;
c->bswap_buf= bswap_buf;
#ifdef HAVE_MMX

5
libavcodec/dsputil.h
View File

@ -234,6 +234,11 @@ typedef struct DSPContext {
/* huffyuv specific */
void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
/**
* subtract huffyuv's variant of median prediction
* note, this might read from src1[-1], src2[-1]
*/
void (*sub_hfyu_median_prediction)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top);
void (*bswap_buf)(uint32_t *dst, uint32_t *src, int w);
/* (I)DCT */

33
libavcodec/huffyuv.c
View File

@ -153,25 +153,6 @@ static inline void add_median_prediction(uint8_t *dst, uint8_t *src1, uint8_t *d
*left_top= lt;
}
//FIXME optimize
static inline void sub_median_prediction(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top){
int i;
uint8_t l, lt;
l= *left;
lt= *left_top;
for(i=0; i<w; i++){
const int pred= mid_pred(l, src1[i], (l + src1[i] - lt)&0xFF);
lt= src1[i];
l= src2[i];
dst[i]= l - pred;
}
*left= l;
*left_top= lt;
}
static inline void add_left_prediction_bgr32(uint8_t *dst, uint8_t *src, int w, int *red, int *green, int *blue){
int i;
int r,g,b;
@ -999,9 +980,9 @@ static int encode_frame(AVCodecContext *avctx, unsigned char *buf, int buf_size,
lefttopy= p->data[0][3];
lefttopu= p->data[1][1];
lefttopv= p->data[2][1];
sub_median_prediction(s->temp[0], p->data[0]+4, p->data[0] + fake_ystride+4, width-4 , &lefty, &lefttopy);
sub_median_prediction(s->temp[1], p->data[1]+2, p->data[1] + fake_ustride+2, width2-2, &leftu, &lefttopu);
sub_median_prediction(s->temp[2], p->data[2]+2, p->data[2] + fake_vstride+2, width2-2, &leftv, &lefttopv);
s->dsp.sub_hfyu_median_prediction(s->temp[0], p->data[0]+4, p->data[0] + fake_ystride+4, width-4 , &lefty, &lefttopy);
s->dsp.sub_hfyu_median_prediction(s->temp[1], p->data[1]+2, p->data[1] + fake_ustride+2, width2-2, &leftu, &lefttopu);
s->dsp.sub_hfyu_median_prediction(s->temp[2], p->data[2]+2, p->data[2] + fake_vstride+2, width2-2, &leftv, &lefttopv);
encode_422_bitstream(s, width-4);
y++; cy++;
@ -1011,7 +992,7 @@ static int encode_frame(AVCodecContext *avctx, unsigned char *buf, int buf_size,
if(s->bitstream_bpp==12){
while(2*cy > y){
ydst= p->data[0] + p->linesize[0]*y;
sub_median_prediction(s->temp[0], ydst - fake_ystride, ydst, width , &lefty, &lefttopy);
s->dsp.sub_hfyu_median_prediction(s->temp[0], ydst - fake_ystride, ydst, width , &lefty, &lefttopy);
encode_gray_bitstream(s, width);
y++;
}
@ -1021,9 +1002,9 @@ static int encode_frame(AVCodecContext *avctx, unsigned char *buf, int buf_size,
udst= p->data[1] + p->linesize[1]*cy;
vdst= p->data[2] + p->linesize[2]*cy;
sub_median_prediction(s->temp[0], ydst - fake_ystride, ydst, width , &lefty, &lefttopy);
sub_median_prediction(s->temp[1], udst - fake_ustride, udst, width2, &leftu, &lefttopu);
sub_median_prediction(s->temp[2], vdst - fake_vstride, vdst, width2, &leftv, &lefttopv);
s->dsp.sub_hfyu_median_prediction(s->temp[0], ydst - fake_ystride, ydst, width , &lefty, &lefttopy);
s->dsp.sub_hfyu_median_prediction(s->temp[1], udst - fake_ustride, udst, width2, &leftu, &lefttopu);
s->dsp.sub_hfyu_median_prediction(s->temp[2], vdst - fake_vstride, vdst, width2, &leftv, &lefttopv);
encode_422_bitstream(s, width);
}

39
libavcodec/i386/dsputil_mmx.c
View File

@ -583,6 +583,43 @@ static void diff_bytes_mmx(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w){
for(; i<w; i++)
dst[i+0] = src1[i+0]-src2[i+0];
}
static void sub_hfyu_median_prediction_mmx2(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top){
int i=0;
uint8_t l, lt;
asm volatile(
"1: \n\t"
"movq -1(%1, %0), %%mm0 \n\t" // LT
"movq (%1, %0), %%mm1 \n\t" // T
"movq -1(%2, %0), %%mm2 \n\t" // L
"movq (%2, %0), %%mm3 \n\t" // X
"movq %%mm2, %%mm4 \n\t" // L
"psubb %%mm0, %%mm2 \n\t"
"paddb %%mm1, %%mm2 \n\t" // L + T - LT
"movq %%mm4, %%mm5 \n\t" // L
"pmaxub %%mm1, %%mm4 \n\t" // max(T, L)
"pminub %%mm5, %%mm1 \n\t" // min(T, L)
"pminub %%mm2, %%mm4 \n\t"
"pmaxub %%mm1, %%mm4 \n\t"
"psubb %%mm4, %%mm3 \n\t" // dst - pred
"movq %%mm3, (%3, %0) \n\t"
"addl $8, %0 \n\t"
"cmpl %4, %0 \n\t"
" jb 1b \n\t"
: "+r" (i)
: "r"(src1), "r"(src2), "r"(dst), "r"(w)
);
l= *left;
lt= *left_top;
dst[0]= src2[0] - mid_pred(l, src1[0], (l + src1[0] - lt)&0xFF);
*left_top= src1[w-1];
*left = src2[w-1];
}
#define LBUTTERFLY2(a1,b1,a2,b2)\
"paddw " #b1 ", " #a1 " \n\t"\
"paddw " #b2 ", " #a2 " \n\t"\
@ -1699,6 +1736,8 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
SET_QPEL_FUNC(qpel_pixels_tab[1][14], qpel8_mc23_mmx2)
SET_QPEL_FUNC(qpel_pixels_tab[1][15], qpel8_mc33_mmx2)
#endif
c->sub_hfyu_median_prediction= sub_hfyu_median_prediction_mmx2;
} else if (mm_flags & MM_3DNOW) {
c->put_pixels_tab[0][1] = put_pixels16_x2_3dnow;
c->put_pixels_tab[0][2] = put_pixels16_y2_3dnow;