scottk/UltraGrid
1
0
Fork 0
mirror of https://github.com/outbackdingo/UltraGrid.git synced 2026-03-22 06:40:26 +00:00
Code Issues Packages Projects Releases Wiki Activity

SSE optimized vc_deinterlace_ex

Browse Source 
now faster than vc_deinterlace

+ fixed possible error that has there been perhaps always when dst_pitch
  > src_linesize - after inner cycle, dst was incremented by
  (dst_pitch+src_linesize), not 2xdst_pitch
This commit is contained in:
Martin Pulec
2023-01-11 12:15:26 +01:00
parent 35eda1f7fa
commit ecb20902b7
1 changed files with 20 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
src/video_codec.c
View File

@@ -694,19 +694,31 @@ static void vc_deinterlace_unaligned(unsigned char *src, long src_linesize, int
*/
bool vc_deinterlace_ex(codec_t codec, unsigned char *src, size_t src_linesize, unsigned char *dst, size_t dst_pitch, size_t lines)
{
(void) codec;
if (codec != RGB && codec != RGBA && codec != UYVY) {
return false;
}
for (size_t y = 0; y < lines; y += 2) {
for (size_t x = 0; x < src_linesize; ++x) {
int val = (*src + src[src_linesize] + 1) >> 1;
*dst = dst[dst_pitch] = val;
src++;
dst++;
unsigned char *s = src + y * src_linesize;
unsigned char *d = dst + y * dst_pitch;
size_t x = 0;
#ifdef __SSSE3__
for ( ; x < src_linesize / 16; ++x) {
__m128i i1 = _mm_lddqu_si128((__m128i const*)(const void *) s);
__m128i i2 = _mm_lddqu_si128((__m128i const*)(const void *) (s + src_linesize));
__m128i res = _mm_avg_epu8(i1, i2);
_mm_storeu_si128((__m128i *)(void *) d, res);
_mm_storeu_si128((__m128i *)(void *) (d + dst_pitch), res);
s += 16;
d += 16;
}
#endif
x *= 16;
for ( ; x < src_linesize; ++x) {
int val = (*s + s[src_linesize] + 1) >> 1;
*d = d[dst_pitch] = val;
s++;
d++;
}
src += src_linesize;
dst += dst_pitch;
}
return true;
}