mirror of https://git.ffmpeg.org/ffmpeg.git
VP8: idct_mb optimizations
Currently uses AV_RL32 instead of AV_RL32A, as the latter doesn't exist yet.
This commit is contained in:
parent
17cf7c68ed
commit
62457f9052
|
@ -1421,17 +1421,17 @@ static av_always_inline void idct_mb(VP8Context *s, uint8_t *dst[3], VP8Macroblo
|
|||
if (mb->mode != MODE_I4x4) {
|
||||
uint8_t *y_dst = dst[0];
|
||||
for (y = 0; y < 4; y++) {
|
||||
uint32_t nnz4 = AV_RN32A(s->non_zero_count_cache[y]);
|
||||
uint32_t nnz4 = AV_RL32(s->non_zero_count_cache[y]);
|
||||
if (nnz4) {
|
||||
if (nnz4&~0x01010101) {
|
||||
for (x = 0; x < 4; x++) {
|
||||
int nnz = s->non_zero_count_cache[y][x];
|
||||
if (nnz) {
|
||||
if (nnz == 1)
|
||||
if ((uint8_t)nnz4 == 1)
|
||||
s->vp8dsp.vp8_idct_dc_add(y_dst+4*x, s->block[y][x], s->linesize);
|
||||
else
|
||||
else if((uint8_t)nnz4 > 1)
|
||||
s->vp8dsp.vp8_idct_add(y_dst+4*x, s->block[y][x], s->linesize);
|
||||
}
|
||||
nnz4 >>= 8;
|
||||
if (!nnz4)
|
||||
break;
|
||||
}
|
||||
} else {
|
||||
s->vp8dsp.vp8_idct_dc_add4y(y_dst, s->block[y], s->linesize);
|
||||
|
@ -1442,19 +1442,19 @@ static av_always_inline void idct_mb(VP8Context *s, uint8_t *dst[3], VP8Macroblo
|
|||
}
|
||||
|
||||
for (ch = 0; ch < 2; ch++) {
|
||||
uint32_t nnz4 = AV_RN32A(s->non_zero_count_cache[4+ch]);
|
||||
uint32_t nnz4 = AV_RL32(s->non_zero_count_cache[4+ch]);
|
||||
if (nnz4) {
|
||||
uint8_t *ch_dst = dst[1+ch];
|
||||
if (nnz4&~0x01010101) {
|
||||
for (y = 0; y < 2; y++) {
|
||||
for (x = 0; x < 2; x++) {
|
||||
int nnz = s->non_zero_count_cache[4+ch][(y<<1)+x];
|
||||
if (nnz) {
|
||||
if (nnz == 1)
|
||||
if ((uint8_t)nnz4 == 1)
|
||||
s->vp8dsp.vp8_idct_dc_add(ch_dst+4*x, s->block[4+ch][(y<<1)+x], s->uvlinesize);
|
||||
else
|
||||
else if((uint8_t)nnz4 > 1)
|
||||
s->vp8dsp.vp8_idct_add(ch_dst+4*x, s->block[4+ch][(y<<1)+x], s->uvlinesize);
|
||||
}
|
||||
nnz4 >>= 8;
|
||||
if (!nnz4)
|
||||
break;
|
||||
}
|
||||
ch_dst += 4*s->uvlinesize;
|
||||
}
|
||||
|
|
Loading…
Reference in New Issue