[Bugfix][NCCL] Release NCCL thread_local resources in destructor #17078
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
Prior to this commit, allocations performed by `ncclCommInitRank` had no corresponding call to `ncclCommDestroy`. While `ncclCommDestroy` does occur in the `CCLThreadLocalContext::Clear` method, there are no calls into this method. On worker processes, the failure to call `ncclCommDestroy` typically had little effect. Any destruction would occur shortly before the process closes, and so resources would be reclaimed by the OS when the process terminates. However, worker0 of a Disco session is a separate thread, rather than a separate process. While this allows it to easily receive data from the controller thread, resources allocated by worker0 are not reclaimed by the OS until the entire process terminates. As a result, the `CCLThreadLocalContext` leaked GPU memory, as the `ncclCommInitRank` call at the start of each `tvm.runtime.disco.ProcessSession` was never de-allocated. The increase in GPU memory usage was about 1 gigabyte for each `ProcessSession`. This commit updates `CCLThreadLocalContext` to have a destructor that calls the `Clear` method. For worker0, this is called when the thread is joined to the main thread.
vinx13
approved these changes
Jun 12, 2024
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Outdated suggestions cannot be applied.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.
Prior to this commit, allocations performed by
ncclCommInitRankhad no corresponding call toncclCommDestroy. WhilencclCommDestroydoes occur in theCCLThreadLocalContext::Clearmethod, there are no calls into this method. On worker processes, the failure to callncclCommDestroytypically had little effect. Any destruction would occur shortly before the process closes, and so resources would be reclaimed by the OS when the process terminates.However, worker0 of a Disco session is a separate thread, rather than a separate process. While this allows it to easily receive data from the controller thread, resources allocated by worker0 are not reclaimed by the OS until the entire process terminates. As a result, the
CCLThreadLocalContextleaked GPU memory, as thencclCommInitRankcall at the start of eachtvm.runtime.disco.ProcessSessionwas never de-allocated. The increase in GPU memory usage was about 1 gigabyte for eachProcessSession.This commit updates
CCLThreadLocalContextto have a destructor that calls theClearmethod. For worker0, this is called when the thread is joined to the main thread.