[Storage_ipc] Option II: Provides IPC extensions for 3rd devices. #126373
+104
−25
Conversation
This file contains 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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/126373
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit b7ead53 with merge base b24ad7e ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
legionGIT
reviewed
May 20, 2024
mengpenghui
force-pushed
the
ipc_without_device
branch
from
33bdf3c
to
b7ead53
Compare
drisspg
added
the
triaged
albanD
reviewed
May 24, 2024
There was a problem hiding this comment.
Thanks for the PR, that sounds good in principle but I think the implementation needs a bit of work for two things:
- the aten level hooks should not take as input and return PyObject*. This is the c++ layer and should not deal with arg-parsing in there. I think in this case, you want to refactor the arg-parsing from the cuda code to be shared. And then call into a c++ only hook.
- The hook should be generic, I think the AcceleratorHookInterface is the right place to do that. This way it can be shared by CUDA/PrivateUse1/others
- The python layer should be generic using acceleratorhooks so that you don't need if/else for each device.
| @@ -294,7 +294,7 @@ struct C10_API StorageImpl : public c10::intrusive_ptr_target { | |||
| bool resizable_; | |||
| // Identifies that Storage was received from another process and doesn't have | |||
| // local to process cuda memory allocation | |||
| bool received_cuda_; | |||
| bool received_device_; | |||
There was a problem hiding this comment.
Comment on the line above needs updating as well
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
I am working on IPC-related interfaces to support third-party devices.
This is another option for IPC support for third party devices.
In this way, third-party devices and CUDA will use the same interface, and the device distinction is hidden in the backend. Users also need to manually set the device type (although these are internal interfaces).
another plan: #125122
Do you think which is feasible? Looking forward to your suggestions
Fixes #124902
cc @albanD