Expose maximum_frame_latency #4899
Conversation
|
I just want to emphasize what a huge difference changing going from 3 to 2 is on Mac, at least in windowed mode on a 60 Hz monitor. It takes the application from feeling sluggish to feeling responsive (using When the CPU is fast (e.g. producing a frame in 1ms), Interestingly though, when the CPU is too slow (e.g. taking 20ms to prepare a frame), then So: I'm planning on making |
On most platforms this is merely a change in narrative, on DX12 it actually has a practical effect already since we use it to directly set frame latency
|
code looks mostly now like we want it to be I think! For sure way more expressive of the stated goal and in close alignment with Need to update changelog and test whether |
|
Got around to test with Vulkan on windows of when GPU/CPU ending up to working in sync. On DX12 I'm hitting some odd bug making frame time expand to 1 full second, so I don't have test results from that so far. Not the result I was hoping for. Need to repeat the same on Mac to better understand stuff. |
Sounds like you might be hitting |
|
Same behavior on Metal. The thing I don't get yet is how one |
Setting `desired_maximum_frame_latency` to a low value should theoretically lead to lower latency in winit apps using `egui-wgpu` (e.g. in `eframe` with `wgpu` backend). * Replaces #3714 * See also gfx-rs/wgpu#4899 ---- It seems like `desired_maximum_frame_latency` has no effect on my Mac. I lowered my monitor refresh-rate to 30Hz to test, and can see no difference between `desired_maximum_frame_latency` of `0` or `3`. Before when experimenting with changing the global `DESIRED_NUM_FRAMES` in `wgpu` I saw a huge difference, so I wonder what has changed. I verified that `set_maximum_drawable_count` is being called with either `1` or `2`, but I perceive no difference between the two.
Connections
Immediateas initially proposed hereDescription
Previously, the swapchain buffer count was hardcoded to be always 3 no matter what is supported. This PR exposes the desired swap chain count as a minimal extension to the wgc presentation setup.
Among other things this allows one now to switch between "Game Mode" and "Classic Game Mode" as classified here
https://www.intel.com/content/www/us/en/developer/articles/code-sample/sample-application-for-direct3d-12-flip-model-swap-chains.html
Prime motivation for this is to reduce perceived lag in egui applications where a buffer count of 2 makes usually more sense (assuming a stable and low workload).
Testing
Experimented with buffer count of 2 in egui example on Mac & Windows. Gives a lot less latency.
Checklist
cargo fmt.cargo clippy. If applicable, add:--target wasm32-unknown-unknown--target wasm32-unknown-emscriptencargo xtask testto run tests.CHANGELOG.md. See simple instructions inside file.