virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2025-08-10 06:10:52 +02:00
Code Issues Releases Activity

avcodec/ffv1enc: Better heuristic for selecting mul values.

Browse Source 
This increases the search space from the hardcoded 17 tables
to ~ 23^512 tables

The mul_count choice is chosen by bruteforce

All testcases tried, improve

Sponsored-by: Sovereign Tech Fund
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
This commit is contained in:
Michael Niedermayer
2025-03-28 01:52:15 +01:00
parent 948a78365b
commit 2f0500f22c
1 changed files with 49 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

85
libavcodec/ffv1enc.c
View File

@@ -1409,52 +1409,65 @@ static void encode_float32_remap(FFV1Context *f, FFV1SliceContext *sc,
av_assert0 (s.pixel_num <= 65536); av_assert0 (s.pixel_num <= 65536);
for (int p= 0; p < 1 + 2*f->chroma_planes + f->transparency; p++) { for (int p= 0; p < 1 + 2*f->chroma_planes + f->transparency; p++) {
float score_tab[16] = {0}; float score_sum[2] = {0};
int64_t last_val = -1; int mul_all[2][513];
int best_index = 0;
s.rc = sc->c; s.rc = sc->c;
s.i = 0; s.i = 0;
s.p = p; s.p = p;
for(int v = 0; v< 512; v++) { for (int mul_count= 1; mul_count<=512; mul_count+=511) {
if (v >= 0x378/8 && v <= 23 + 0x378/8) { float score_tab_all[513][23] = {0};
s.mul[v] = -(0x800080 >> (v - 0x378/8)); int64_t last_val = -1;
} else int *mul_tab = mul_all[mul_count>>9];
s.mul[v] = -1; int last_mul_index = -1;
} score_sum[mul_count>>9] += log2(mul_count);
for (int i= 0; i<s.pixel_num; i++) { for (int i= 0; i<s.pixel_num; i++) {
int64_t val = sc->unit[p][i].val; int64_t val = sc->unit[p][i].val;
if (val != last_val) { int mul_index = (val + 1LL)*mul_count >> 32;
av_assert2(last_val < val); if (val != last_val) {
for(int si= 0; si < FF_ARRAY_ELEMS(score_tab); si++) { float *score_tab = score_tab_all[(last_val + 1LL)*mul_count >> 32];
int64_t delta = val - last_val; av_assert2(last_val < val);
int mul; for(int si= 0; si < FF_ARRAY_ELEMS(*score_tab_all); si++) {
int64_t cost; int64_t delta = val - last_val;
int mul;
int64_t cost;
if (last_val < 0) { if (last_val < 0) {
mul = 1; mul = 1;
} else if (si + 1 == FF_ARRAY_ELEMS(score_tab)) { } else {
mul = -s.mul[ (last_val + 1) >> (32-9) ]; mul = (0x10001LL)<<si >> 16;
} else }
mul = 1<<si;
cost = FFMAX((delta + mul/2) / mul, 1); cost = FFMAX((delta + mul/2) / mul, 1);
score_tab[si] += log2(cost) + log2(fabs(delta - cost*mul)+1); score_tab[si] += log2(cost);
if (mul > 1)
score_tab[si] += log2(fabs(delta - cost*mul)+1) * (1 + (mul_count > 1));
if (mul_index != last_mul_index)
score_tab[si] += 0.5*log2(mul);
}
} }
last_val = val; last_val = val;
last_mul_index = mul_index;
} }
for(int i= 0; i<mul_count; i++) {
int best_index = 0;
float *score_tab = score_tab_all[i];
for(int si= 1; si < FF_ARRAY_ELEMS(*score_tab_all); si++) {
if (score_tab[si] < score_tab[ best_index ])
best_index = si;
}
mul_tab[i] = -((0x10001LL)<<best_index >> 16);
score_sum[mul_count>>9] += score_tab[ best_index ];
}
mul_tab[mul_count] = 1;
s.mul_count = mul_count;
memcpy(s.mul, mul_all[s.mul_count>>9], sizeof(*s.mul)*(s.mul_count+1));
score_sum[mul_count>>9] = encode_float32_remap_segment(sc, &s, 0, 0);
} }
for(int si= 1; si < FF_ARRAY_ELEMS(score_tab); si++) {
if (score_tab[si] < score_tab[ best_index ]) s.mul_count = score_sum[0] <= score_sum[1] ? 1 : 512;
best_index = si; memcpy(s.mul, mul_all[s.mul_count>>9], sizeof(*s.mul)*(s.mul_count+1));
}
if (best_index + 1 < FF_ARRAY_ELEMS(score_tab)) {
s.mul[0] = -1 << best_index;
s.mul_count = 1;
} else {
s.mul_count = 512;
}
s.mul[s.mul_count] = 1;
encode_float32_remap_segment(sc, &s, 1, 1); encode_float32_remap_segment(sc, &s, 1, 1);