Move some branches outside looped code. Should improve the generated asm (and
thus performance) slightly. Patch by Sebastian Vater <cdgs.basty googlemail com>. Originally committed as revision 22975 to svn://svn.ffmpeg.org/ffmpeg/trunk
This commit is contained in:
parent
473147bed0
commit
fe51b5ce50
@ -140,16 +140,23 @@ static int decode_frame_ilbm(AVCodecContext *avctx,
|
||||
return -1;
|
||||
}
|
||||
|
||||
if (avctx->pix_fmt == PIX_FMT_PAL8) {
|
||||
for(y = 0; y < avctx->height; y++ ) {
|
||||
uint8_t *row = &s->frame.data[0][ y*s->frame.linesize[0] ];
|
||||
memset(row, 0, avctx->pix_fmt == PIX_FMT_PAL8 ? avctx->width : (avctx->width * 4));
|
||||
memset(row, 0, avctx->width);
|
||||
for (plane = 0; plane < avctx->bits_per_coded_sample && buf < buf_end; plane++) {
|
||||
if (avctx->pix_fmt == PIX_FMT_PAL8) {
|
||||
decodeplane8(row, buf, FFMIN(s->planesize, buf_end - buf), avctx->bits_per_coded_sample, plane);
|
||||
} else { // PIX_FMT_BGR32
|
||||
decodeplane32((uint32_t *) row, buf, FFMIN(s->planesize, buf_end - buf), avctx->bits_per_coded_sample, plane);
|
||||
buf += s->planesize;
|
||||
}
|
||||
}
|
||||
} else { // PIX_FMT_BGR32
|
||||
for(y = 0; y < avctx->height; y++ ) {
|
||||
uint8_t *row = &s->frame.data[0][y*s->frame.linesize[0]];
|
||||
memset(row, 0, avctx->width << 2);
|
||||
for (plane = 0; plane < avctx->bits_per_coded_sample && buf < buf_end; plane++) {
|
||||
decodeplane32((uint32_t *) row, buf, FFMIN(s->planesize, buf_end - buf), avctx->bits_per_coded_sample, plane);
|
||||
buf += s->planesize;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
@ -173,10 +180,11 @@ static int decode_frame_byterun1(AVCodecContext *avctx,
|
||||
return -1;
|
||||
}
|
||||
|
||||
if (avctx->codec_tag == MKTAG('I','L','B','M')) { //interleaved
|
||||
if (avctx->pix_fmt == PIX_FMT_PAL8) {
|
||||
for(y = 0; y < avctx->height ; y++ ) {
|
||||
uint8_t *row = &s->frame.data[0][ y*s->frame.linesize[0] ];
|
||||
if (avctx->codec_tag == MKTAG('I','L','B','M')) { //interleaved
|
||||
memset(row, 0, avctx->pix_fmt == PIX_FMT_PAL8 ? avctx->width : (avctx->width * 4));
|
||||
memset(row, 0, avctx->width);
|
||||
for (plane = 0; plane < avctx->bits_per_coded_sample; plane++) {
|
||||
for(x = 0; x < s->planesize && buf < buf_end; ) {
|
||||
int8_t value = *buf++;
|
||||
@ -193,13 +201,36 @@ static int decode_frame_byterun1(AVCodecContext *avctx,
|
||||
}
|
||||
x += length;
|
||||
}
|
||||
if (avctx->pix_fmt == PIX_FMT_PAL8) {
|
||||
decodeplane8(row, s->planebuf, s->planesize, avctx->bits_per_coded_sample, plane);
|
||||
}
|
||||
}
|
||||
} else { //PIX_FMT_BGR32
|
||||
for(y = 0; y < avctx->height ; y++ ) {
|
||||
uint8_t *row = &s->frame.data[0][y*s->frame.linesize[0]];
|
||||
memset(row, 0, avctx->width << 2);
|
||||
for (plane = 0; plane < avctx->bits_per_coded_sample; plane++) {
|
||||
for(x = 0; x < s->planesize && buf < buf_end; ) {
|
||||
int8_t value = *buf++;
|
||||
unsigned length;
|
||||
if (value >= 0) {
|
||||
length = value + 1;
|
||||
memcpy(s->planebuf + x, buf, FFMIN3(length, s->planesize - x, buf_end - buf));
|
||||
buf += length;
|
||||
} else if (value > -128) {
|
||||
length = -value + 1;
|
||||
memset(s->planebuf + x, *buf++, FFMIN(length, s->planesize - x));
|
||||
} else { // noop
|
||||
continue;
|
||||
}
|
||||
x += length;
|
||||
}
|
||||
decodeplane32((uint32_t *) row, s->planebuf, s->planesize, avctx->bits_per_coded_sample, plane);
|
||||
}
|
||||
}
|
||||
}
|
||||
} else {
|
||||
for(y = 0; y < avctx->height ; y++ ) {
|
||||
uint8_t *row = &s->frame.data[0][y*s->frame.linesize[0]];
|
||||
for(x = 0; x < avctx->width && buf < buf_end; ) {
|
||||
int8_t value = *buf++;
|
||||
unsigned length;
|
||||
|
Loading…
x
Reference in New Issue
Block a user