Skip to content

experiment(v0.0.3): skip empty per-draw GPU sync-request mutex lock - #24

Draft
Framerr wants to merge 1 commit into
nx2from
experiment/v003-sync-request-fastpath
Draft

Framerr wants to merge 1 commit into
nx2from
experiment/v003-sync-request-fastpath

Conversation

@Framerr

@Framerr Framerr commented Nov 8, 2026

Copy link
Copy Markdown
Owner

Independent v0.0.3-based performance test. Isolated port of the GPU::TickWork portion of notsapinho/Eglantine a1008ebd, omitting its unrelated RtDump static-init change. An atomic count avoids locking the synchronization request mutex when no operation is queued. Request accounting is updated under the same queue mutex as operations, and queued operations still use the original locking and fence-completion path.

Author measured around 0.7% GPU split back-thread cost for the empty lock; no Frameropia FPS increase is proven. No DMA, texture, shader or photo change. Test performance, startup and first three Pokopia photos over several restarts, plus normal game stability. Keep distinct from #23 and v0.0.3. Do not merge until validated.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant