From: Michael Niedermayer <michael@niedermayer.cc>
To: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org>
Subject: [FFmpeg-devel] [PATCH 11/11] avcodec/ffv1enc: Heuristic to select fixed in float multipliers
Date: Mon, 24 Mar 2025 18:39:45 +0100
Message-ID: <20250324173945.2574461-11-michael@niedermayer.cc> (raw)
In-Reply-To: <20250324173945.2574461-1-michael@niedermayer.cc>
Sponsored-by: Sovereign Tech Fund
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
---
libavcodec/ffv1enc.c | 30 ++++++++++++++++++++++++------
1 file changed, 24 insertions(+), 6 deletions(-)
diff --git a/libavcodec/ffv1enc.c b/libavcodec/ffv1enc.c
index c3b3b926adf..004ea73e06b 100644
--- a/libavcodec/ffv1enc.c
+++ b/libavcodec/ffv1enc.c
@@ -1408,17 +1408,30 @@ static void encode_float32_remap(FFV1Context *f, FFV1SliceContext *sc,
s.i = 0;
s.p = p;
- s.mul_count = 1;
-
+ for(int v = 0; v< 512; v++) {
+ if (v >= 0x378/8 && v <= 23 + 0x378/8) {
+ s.mul[v] = -(0x800080 >> (v - 0x378/8));
+ } else
+ s.mul[v] = -1;
+ }
for (int i= 0; i<s.pixel_num; i++) {
int64_t val = unit[p][i].val;
if (val != last_val) {
av_assert2(last_val < val);
for(int si= 0; si < FF_ARRAY_ELEMS(score_tab); si++) {
int64_t delta = val - last_val;
- int mul = last_val < 0 ? 1 : (1<<si);
- int64_t cost = FFMAX((delta + mul/2) / mul, 1);
- score_tab[si] += log2(cost) + fabs(delta - cost*mul);
+ int mul;
+ int64_t cost;
+
+ if (last_val < 0) {
+ mul = 1;
+ } else if (si + 1 == FF_ARRAY_ELEMS(score_tab)) {
+ mul = -s.mul[ (last_val + 1) >> (32-9) ];
+ } else
+ mul = 1<<si;
+
+ cost = FFMAX((delta + mul/2) / mul, 1);
+ score_tab[si] += log2(cost) + log2(fabs(delta - cost*mul)+1);
}
last_val = val;
}
@@ -1428,7 +1441,12 @@ static void encode_float32_remap(FFV1Context *f, FFV1SliceContext *sc,
if (score_tab[si] < score_tab[ best_index ])
best_index = si;
}
- s.mul[0] = -1 << best_index;
+ 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, unit, &s, 1, 1);
--
2.48.1
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
next prev parent reply other threads:[~2025-03-24 17:41 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-03-24 17:39 [FFmpeg-devel] [PATCH 01/11] avformat/nut: Add GBR(A)PF16/32LE/BE Michael Niedermayer
2025-03-24 17:39 ` [FFmpeg-devel] [PATCH 02/11] avcodec/raw: " Michael Niedermayer
2025-03-24 17:39 ` [FFmpeg-devel] [PATCH 03/11] avcodec/ffv1enc: remap mode 2 is for floats only Michael Niedermayer
2025-03-25 9:38 ` Michael Niedermayer
2025-03-24 17:39 ` [FFmpeg-devel] [PATCH 04/11] avcodec/ffv1enc: remap table size is 1<<bits_per_raw_sample not fixed 65536 Michael Niedermayer
2025-03-24 17:39 ` [FFmpeg-devel] [PATCH 05/11] avcodec/ffv1dec: compute end instead of hardcoding it and test for fltmap correctly Michael Niedermayer
2025-03-24 17:39 ` [FFmpeg-devel] [PATCH 06/11] avcodec/ffv1dec: replace literal 65535 by what it is Michael Niedermayer
2025-03-24 17:39 ` [FFmpeg-devel] [PATCH 07/11] avcodec/ffv1enc: eliminate if() Michael Niedermayer
2025-03-24 17:39 ` [FFmpeg-devel] [PATCH 08/11] avcodec/ffv1enc: introduce step to simplify code Michael Niedermayer
2025-03-24 17:39 ` [FFmpeg-devel] [PATCH 09/11] avcodec/ffv1enc: Simplify last_val update Michael Niedermayer
2025-03-24 17:39 ` [FFmpeg-devel] [PATCH 10/11] avcodec/ffv1enc: Factorize s.last_val update out Michael Niedermayer
2025-03-24 17:39 ` Michael Niedermayer [this message]
2025-03-28 12:24 ` [FFmpeg-devel] [PATCH 01/11] avformat/nut: Add GBR(A)PF16/32LE/BE Michael Niedermayer
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20250324173945.2574461-11-michael@niedermayer.cc \
--to=michael@niedermayer.cc \
--cc=ffmpeg-devel@ffmpeg.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git