-
Notifications
You must be signed in to change notification settings - Fork 397
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
Marking a CUDA custom call as command buffer-compatible has no effect #14889
Comments
copybara-service bot
pushed a commit
that referenced
this issue
Jul 17, 2024
…tion and lookup + Use xla:util error constructors instead of absl::XyzError to automatically capture error stack trace Fix for #14889 PiperOrigin-RevId: 653278962
copybara-service bot
pushed a commit
to tensorflow/tensorflow
that referenced
this issue
Jul 17, 2024
…tion and lookup + Use xla:util error constructors instead of absl::XyzError to automatically capture error stack trace Fix for openxla/xla#14889 PiperOrigin-RevId: 653278962
copybara-service bot
pushed a commit
that referenced
this issue
Jul 17, 2024
…tion and lookup + Use xla:util error constructors instead of absl::XyzError to automatically capture error stack trace Fix for #14889 PiperOrigin-RevId: 653319684
copybara-service bot
pushed a commit
to tensorflow/tensorflow
that referenced
this issue
Jul 17, 2024
…tion and lookup + Use xla:util error constructors instead of absl::XyzError to automatically capture error stack trace Fix for openxla/xla#14889 PiperOrigin-RevId: 653319684
Hi @ezhulenev, |
Closing, per @phu0ngng 's report that this is fixed. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
This seems to happen because this piece of logic only looks at registrations for generic platform
gpu
:xla/xla/service/gpu/command_buffer_scheduling.cc
Lines 162 to 165 in 5e9ce97
CUDA
are not taken into account.@ezhulenev has suggested offline that the fix might be to do platform name canonicalization more thoroughly.
A quick way to repro is to modify the JAX
cuda_custom_call
test as follows:then run the following script (you'll need Nsight Systems CLI installed), which will show whether each kernel was launched as part of a CUDA graph or not:
The text was updated successfully, but these errors were encountered: