Faster alpha blending #123
Merged
Conversation
This file contains hidden or 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
…th better tolerances than libwebp blending!
…tests with better tolerances than libwebp blending!" This reverts commit 68ea852.
|
Ah, test fail in big endian. Well, that's gonna be fun to debug. Thanks, C. |
|
I've fixed the endianness bug (good thing we have big-endian CI!), so this is now ready for review. |
|
Senpai has noticed my PR! ❤️ |
|
I just realized that this PR updated some of the reference images without editing CREDITS.md which documents how they were generated. Specifically, they were previously produced by |
|
Oh, yes, that is correct. I did update the animated image sample because it differed very slightly. |
This was referenced Jun 28, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Switch to a direct translation of libwebp alpha blending, plus a tweak suggested by @awxkee in #119 to improve the precision of this approximation.
It's not an exact match of the f64 method, but it is more accurate than libwebp itself, and without losing performance compared to the libwebp method (at least on my machine). The inaccuracies in both approximations show up only on very low alpha values, which makes them imperceptible.
This PR includes a test verifying this approximation against the floating-point reference.
libwebp is BSD licensed, which is not a viral license. This is also not the first instance of adapting code from libwebp based on the comments in the repo.
There's a case to be made for refactoring away the C-isms and also experimenting with
u16instead ofu32for intermediate values so that the outer loop could blend two pixels at once, but that would require much more work, and this is already a dramatic performance improvement on the sample from #119. Combined with #122 this reduces the end-to-end decoding time by 27%.