| Summary: | [Cocoa] rVFC() callbacks have up to 60ms extra latency | ||||||
|---|---|---|---|---|---|---|---|
| Product: | WebKit | Reporter: | Jer Noble <jer.noble> | ||||
| Component: | Media | Assignee: | Jer Noble <jer.noble> | ||||
| Status: | NEW --- | ||||||
| Severity: | Normal | CC: | dino, eric.carlson, jean-yves.avenard, mjs, peng.liu6, webkit-bug-importer, youennf | ||||
| Priority: | P2 | Keywords: | InRadar | ||||
| Version: | WebKit Nightly Build | ||||||
| Hardware: | Unspecified | ||||||
| OS: | Unspecified | ||||||
| Attachments: |
|
||||||
|
Description
Jer Noble
2022-01-25 08:59:01 PST
Correction: 60ms extra latency. Created attachment 449934 [details]
Patch
Unfortunately, Performance testing shows this patch to have a negligible effect on latency. It appears latency is more greatly determined by when in a display refresh cycle the frame becomes available. The good news is that even without this patch, frame display latencies are in the order of 9-16ms. Is it 60ms or 9-16ms? The former seems fatal to use of the feature, the latter seems maybe ok? (In reply to Maciej Stachowiak from comment #5) > Is it 60ms or 9-16ms? The former seems fatal to use of the feature, the > latter seems maybe ok? Potentially 60ms, because that's the timer which AVPlayerItemVideoOutput uses to purge samples. In practice, I've never seen it higher than 16ms on this particular machine. |