-
Notifications
You must be signed in to change notification settings - Fork 1k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Let Player figure out the backend part of IDs #792
Labels
Comments
I tries to see where this would be going. A blanket |
bors bot
added a commit
that referenced
this issue
Jul 17, 2020
803: Player-based GPU test framework r=cwfitzgerald a=kvark **Connections** Closes #786 **Description** This change adds a GPU-based testing by re-using the Player from tracing infrastructure - #289. It converts the player into a lib + binary, and adds an integration test into the crate that implements RON-specified testing. Current implementation has a few requirements/gotchas that are listed in `test.rs`: * in all the IDs, the backend is `Empty` * all expected buffers have `MAP_READ` usage on them * last action is `Submit` I believe it's workable, and we can improve it down the road (e.g. with #792). **Testing** MUHAHAHA `cargo test` nails it Co-authored-by: Dzmitry Malyshau <kvarkus@gmail.com>
kvark
added a commit
to kvark/wgpu
that referenced
this issue
Jun 3, 2021
792: Use implicit layout for hello-compute r=kvark a=kvark Co-authored-by: Dzmitry Malyshau <kvarkus@gmail.com>
No longer relevant; we removed the backend in IDs in #6263. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Labels
Is your feature request related to a problem? Please describe.
When API traces are serialized, we are saving each ID as a triple of (index, epoch, backend), e.g.:
This is verbose and redundant, also incorrect from the player point of view. The player picks the backend independently and just wants to treat all the IDs as belonging to this backend. So we don't want to have the backend specified in a trace, other than maybe once at the top for informational reasons only.
Describe the solution you'd like
@lachlansneff suggested to have a custom serializer/deserializer implementation for traces. It would behave just like the standard one from
serde
, but omit the backend on serialization and recover it on deserialization, based on a global setting.Ideally, it would look something like:
Describe alternatives you've considered
A text-based search&replace would also work, but it's a bit awkward.
Besides, we don't want to pollute traces with irrelevant data.
Additional context
We could live without this for the matter of replaying traces manually, but it now it's needed for #786
The text was updated successfully, but these errors were encountered: