RepoMirrors
/
ffmpeg
mirror of https://git.ffmpeg.org/ffmpeg.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

lavfi/dctdnoiz: move DC normalization out of loops. 

Make code slightly faster, simpler, clearer.

The filter is still slow as hell, and that change won't cause any
visible performance improvement (it still takes more than one minute to
process a single 1080p frame on a Core 2 here).
This commit is contained in:
Clément Bœsch 2013-05-26 22:32:24 +02:00
parent bd89b2b22a
commit b439ece51c
1 changed files with 13 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
libavfilter/vf_dctdnoiz.c
View File

@ -82,9 +82,10 @@ static float *dct_block(DCTdnoizContext *ctx, const float *src, int src_linesize
av_dct_calc(ctx->dct, line); av_dct_calc(ctx->dct, line);
column = ctx->tmp_block + y; column = ctx->tmp_block + y;
for (x = 0; x < BSIZE; x++) { column[0] = line[0] * (1. / sqrt(BSIZE));
*line *= x == 0 ? 1. / sqrt(BSIZE) : sqrt(2. / BSIZE); column += BSIZE;
*column = *line++; for (x = 1; x < BSIZE; x++) {
*column = line[x] * sqrt(2. / BSIZE);
column += BSIZE; column += BSIZE;
} }
} }
@ -92,8 +93,9 @@ static float *dct_block(DCTdnoizContext *ctx, const float *src, int src_linesize
column = ctx->tmp_block; column = ctx->tmp_block;
for (x = 0; x < BSIZE; x++) { for (x = 0; x < BSIZE; x++) {
av_dct_calc(ctx->dct, column); av_dct_calc(ctx->dct, column);
for (y = 0; y < BSIZE; y++) column[0] *= 1. / sqrt(BSIZE);
column[y] *= y == 0 ? 1. / sqrt(BSIZE) : sqrt(2. / BSIZE); for (y = 1; y < BSIZE; y++)
column[y] *= sqrt(2. / BSIZE);
column += BSIZE; column += BSIZE;
} }
@ -111,18 +113,18 @@ static void idct_block(DCTdnoizContext *ctx, float *dst, int dst_linesize)
float *tmp = ctx->tmp_block; float *tmp = ctx->tmp_block;
for (y = 0; y < BSIZE; y++) { for (y = 0; y < BSIZE; y++) {
for (x = 0; x < BSIZE; x++) block[0] *= sqrt(BSIZE);
block[x] *= x == 0 ? sqrt(BSIZE) : 1./sqrt(2. / BSIZE); for (x = 1; x < BSIZE; x++)
block[x] *= 1./sqrt(2. / BSIZE);
av_dct_calc(ctx->idct, block); av_dct_calc(ctx->idct, block);
block += BSIZE; block += BSIZE;
} }
block = ctx->block; block = ctx->block;
for (y = 0; y < BSIZE; y++) { for (y = 0; y < BSIZE; y++) {
for (x = 0; x < BSIZE; x++) { tmp[0] = block[y] * sqrt(BSIZE);
tmp[x] = block[x*BSIZE + y]; for (x = 1; x < BSIZE; x++)
tmp[x] *= x == 0 ? sqrt(BSIZE) : 1./sqrt(2. / BSIZE); tmp[x] = block[x*BSIZE + y] * (1./sqrt(2. / BSIZE));
}
av_dct_calc(ctx->idct, tmp); av_dct_calc(ctx->idct, tmp);
for (x = 0; x < BSIZE; x++) for (x = 0; x < BSIZE; x++)
dst[x*dst_linesize + y] += tmp[x]; dst[x*dst_linesize + y] += tmp[x];