New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
minor optimization for enc_heuristics #3536
Merged
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
jyrkialakuijala
added
CI:full
Label to attach to a PR to run the full CI workflow and not just the regular PR workflows
encoder
Related to the libjxl encoder
encoder quality
Quality tuning of lossy encoding
labels
Apr 26, 2024
szabadka
previously approved these changes
Apr 30, 2024
``` Before: Encoding kPixels Bytes BPP E MP/s D MP/s Max norm SSIMULACRA2 PSNR pnorm BPP*pnorm QABPP Bugs ---------------------------------------------------------------------------------------------------------------------------------------- jxl:d0.1 20315 17771024 6.9979205 1.072 10.153 0.35756577 95.39413774 55.88 0.11792287 0.825214859523 6.998 0 jxl:d0.5 20315 7252143 2.8557679 1.278 14.595 0.78009917 91.43862540 47.13 0.33393541 0.953642033483 2.944 0 jxl:d0.8 20315 5363198 2.1119342 1.180 15.134 1.11808133 88.71449750 44.49 0.47287594 0.998682857933 2.498 0 jxl:d1 20315 4603082 1.8126137 1.202 16.286 1.34310821 86.92441895 43.26 0.56190222 1.018511676188 2.539 0 jxl:d1.5 20315 3493018 1.3754898 1.200 15.527 1.86482998 82.75997012 41.13 0.76257997 1.048920975049 2.679 0 jxl:d2.0 20315 2839864 1.1182891 1.222 16.554 2.39538877 78.76238966 39.63 0.95207627 1.064696518711 2.823 0 jxl:d2.5 20315 2390345 0.9412763 1.234 15.872 2.85224490 74.97863810 38.44 1.12701990 1.060837154023 2.820 0 jxl:d3 20315 2083852 0.8205847 1.182 16.514 3.25656987 71.69784136 37.60 1.28318000 1.052957893138 2.793 0 jxl:d5 20315 1396548 0.5499363 1.151 11.355 4.58515346 60.33355388 35.30 1.81020579 0.995497945222 2.624 0 jxl:d10 20315 842090 0.3316004 1.182 12.256 7.63349125 39.09399046 32.54 2.92331122 0.969371201332 2.517 0 Aggregate: 20315 3362708 1.3241760 1.189 14.238 1.89851036 74.90381740 41.09 0.75239306 0.996300812242 2.959 0 After: Encoding kPixels Bytes BPP E MP/s D MP/s Max norm SSIMULACRA2 PSNR pnorm BPP*pnorm QABPP Bugs ---------------------------------------------------------------------------------------------------------------------------------------- jxl:d0.1 20315 17771024 6.9979205 1.080 10.130 0.35756577 95.39413774 55.88 0.11792287 0.825214859523 6.998 0 jxl:d0.5 20315 7252143 2.8557679 1.298 15.061 0.78009917 91.43862540 47.13 0.33393541 0.953642033483 2.944 0 jxl:d0.8 20315 5367111 2.1134751 1.174 15.263 1.11498186 88.72067125 44.51 0.47189587 0.997340150865 2.494 0 jxl:d1 20315 4606615 1.8140049 1.184 16.058 1.33978720 86.93643661 43.28 0.56051359 1.016774427528 2.535 0 jxl:d1.5 20315 3495106 1.3763120 1.174 15.498 1.86243967 82.75276197 41.14 0.76111748 1.047535130323 2.673 0 jxl:d2.0 20315 2841131 1.1187880 1.223 16.434 2.39270883 78.72248340 39.64 0.95068901 1.063619476080 2.823 0 jxl:d2.5 20315 2391176 0.9416036 1.205 16.258 2.85363662 74.90745005 38.44 1.12566896 1.059933899631 2.821 0 jxl:d3 20315 2084298 0.8207603 1.196 16.639 3.27099277 71.57564628 37.60 1.28291115 1.052962585207 2.801 0 jxl:d5 20315 1396215 0.5498052 1.188 11.067 4.60528606 60.08180640 35.30 1.81269081 0.996626849726 2.631 0 jxl:d10 20315 841857 0.3315087 1.153 11.835 7.65826483 38.55099789 32.51 2.92552904 0.969838209567 2.531 0 Aggregate: 20315 3363578 1.3245185 1.186 14.217 1.89943602 74.74511614 41.09 0.75185110 0.995840722234 2.960 0 ``` The results are somewhat smoother with ringing artefacts usually disappearing. SSIMULACRA2 doesn't like it much, probably because the variance is reduced as a consequence and compression tends to reduce variance by itself. For the eye, the new results look slightly better.
jyrkialakuijala
force-pushed
the
epfcontrol
branch
from
May 13, 2024 09:12
5831e28
to
4233235
Compare
mo271
approved these changes
May 13, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Labels
CI:full
Label to attach to a PR to run the full CI workflow and not just the regular PR workflows
encoder quality
Quality tuning of lossy encoding
encoder
Related to the libjxl encoder
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The results are somewhat smoother with ringing artefacts usually disappearing. SSIMULACRA2 doesn't like it much, probably because the variance is reduced as a consequence and compression tends to reduce variance by itself. For the eye, the new results look slightly better.
Description
Pull Request Checklist
./ci.sh lint
for automatic code formatting.Please review the full contributing guidelines for more details.