Add CUDAGuard to ensure correct device #5113
Closed
+99
−16
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
Contributor
cthi
changed the title
Summary: X-link: facebookresearch/FBGEMM#2119 If the input tensors use a device that differs from the current device, it would cause the wrong device to be used for things such as workspace allocation (when using `cutlass::device_memory::allocation`) and kernel to run on the wrong stream. Either would break the kernel. As a fix we add the `CUDAGuard` to ensure correct device is used. - `cutlass::device_memory::allocation` is a wrapper around [`cudaMalloc`](https://github.com/NVIDIA/cutlass/blob/2252254ce2c3f11ef5cfff9721ebbe7bd62cf8cb/tools/util/include/cutlass/util/device_memory.h#L56), but this would bypass PyTorch CCA. We replace all usages with torch tensor allocation instead which would be less error prone and allow proper memory reuse. Differential Revision: D86768064
cthi
force-pushed
the
export-D86768064
branch
from
c611606 to
722f8b6
Compare
Contributor
|
This pull request has been merged in 62bdc5f. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
If the input tensors use a device that differs from the current device, it would cause the wrong device to be used for things such as workspace allocation (when using
cutlass::device_memory::allocation) and kernel to run on the wrong stream. Either would break the kernel. As a fix we add theCUDAGuardto ensure correct device is used.cutlass::device_memory::allocationis a wrapper aroundcudaMalloc, but this would bypass PyTorch CCA. We replace all usages with torch tensor allocation instead which would be less error prone and allow proper memory reuse.Differential Revision: D86768064