Skip to content
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

Clarify where GPU related states are stored for WebGPU interop (MLContext, MLCommandEncoder and MLGraph) #342

Closed
huningxin opened this issue Feb 15, 2023 · 2 comments

Comments

@huningxin
Copy link
Contributor

Split from #341, where @wacky6 mentioned

From the current spec, it's unclear where GPU related states are stored (are they on MLCommandEncoder, or MLGraph itself).

I'd agree the spec should improve in this area.

AFAIK, for GPU context, MLContext should maintain a reference to GPU device, MLGraphBuilder created from GPU MLContext should also refer to GPU device to create and compile GPU operators, MLGraph should maintain the compiled GPU operators and persistent/constant GPU resources, MLCommandEncoder should create the command list from GPU device and record operators initialization and compute commands.

@wchao1115 should know more details than me, please feel free to chime in.

@bbernhar
Copy link

bbernhar commented May 30, 2024

@huningxin is the group still pursuing this approach? If not, can we close in favor of #688?

@huningxin
Copy link
Contributor Author

Yes, this one is obsoleted by #688. Closing it.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

4 participants