Repository navigation
perf(renderer): a nebula of patches stays under the opacity ceiling; Cassiopeia A's zoom goes from a 291 ms frame to 51 ms on an iPad - #1359
Merged
Conversation
…Cassiopeia A's zoom out repainted every patch in one 291 ms frame
alowpoly
enabled auto-merge
October 5, 2026 21:28
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
A nebula drawn as patches (Cassiopeia A, the Homunculus) sat at opacity 1 since the Eta Carinae change, so its bank crossed 1 on every zoom out and Safari repainted every patch in one frame. Patches go back under the 0.999 ceiling that sheets already keep, on the stack and on the bank's root.
Proves the change: iPad, the zoom program: Cassiopeia A's longest frame 291 ms as served, 51 ms under the ceiling; the Homunculus's 133 ms zoom-out frame is gone.
The reason patches were at 1 was a wedge of the Homunculus going black for a step of a drag at its nearest view, seen in headless Chrome. I looked for it and did not find it: at the nearest view, on the nebula's page and around its star, the picture under the ceiling matched the picture at 1 at every step (580 steps in Chrome on a GPU, 90 in the software-rendered shell, 11 on an iPad).
Clean: the image-layer and catalogue-bank tests (12 pass), lint on the two files.
Limits: the iPad numbers come from the ceiling applied to the live page, not from this build. The Homunculus still drops 94 to 106 ms zooming in. The wedge's original camera places are not known, so this is "not reproduced", not "cannot happen".