Repository navigation
Only render views that need to be rendered (with external texture & platform view support) - #179874
Conversation
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
19da943 to
71bab07
Compare
|
@9AZX, it seems that |
71bab07 to
5a3ac8c
Compare
eaab6aa to
0059b33
Compare
My bad! I didn't know that web_ui has his own |
|
@9AZX FYI there are failures |
08205c4 to
cfbdf5b
Compare
bdb2a9f to
faadca6
Compare
dddcf84 to
8b00990
Compare
d4624c5 to
e7f9828
Compare
|
autosubmit label was removed for flutter/flutter/179874, because - The status or check suite Google testing has failed. Please fix the issues identified (or deflake) before re-applying this label. |
|
A reason for requesting a revert of flutter/flutter/179874 could not be found or the reason was not properly formatted. Begin a comment with 'Reason for revert:' to tell the bot why this issue is being reverted. |
|
Reason for revert: #191251 |
|
welp. that didn't take long. |
|
Successfully created revert PR: #191252 |
Is there a sad emoji? The closest I found in the list was "confused" which didn't sound right. Anyway... 😢 |
I am wondering a bit if something like this should have been caught in presubmit checks... |
Benchmarks only run after submit because we record them in a dashboard by their true final commit. During pre-submit they don't have a commit key to report against. We could potentially run them in "don't submit/dry run" mode maybe in pre-submit, though. I'm not sure how the web benchmarks are configured or whether they have such a mode. We also briefly looked at some other benchmark dashboards at least one of which had a nice feature that you could run your benchmarks in pre-submit and get a report on any potential regression. It would take some effort to upgrade all of our benchmarks to a new system, though. :( |
|
We can't run benchmarks on every PR since that requires running on a phone. If we had an analogue that was just running benchmarks on a machine, I'd entertain a test in presubmit. The only other option is using |
Fixes #164320
Based on @knopp work in #164322
This PR adds a dirty-view compositing optimization that skips compositing for views that don't have any dirty render objects, so it should improves performance in multi-view scenarios by avoiding unnessacery work.
When a texture has new content, the framework now gets notified via a new onTextureFrameAvailable callback from the engine. This marks the texture's render object as needing paint, so its view is included in compositing even when nothing else changed.
On Android, texture callbacks now use
markTextureFrameAvailableinstead ofscheduleEngineFrame. This is actually more efficient since it reuses the existing layer tree rather than regenerating it. Other platforms already used the correct path.It also forces all views to composite when returning from background, ensuring the display is properly refreshed after the app was paused.
On MacOS:
Screen.Recording.2025-12-15.at.11.54.17.mov
Pre-launch Checklist
///).