Repository navigation
Use pinned memory for split and ensure_started - #2414
Merged
Merged
Conversation
The shared completion storage of `split` and `ensure_started` was allocated with `cudaMallocManaged`. The continuation task reads it on the host as soon as the kernel that enqueues the task links it into the queue, so the read overlaps that kernel. On devices where `cudaDevAttrConcurrentManagedAccess` is 0, the host must not access managed memory while a kernel is running, and the access fails with SIGSEGV. Allocate the storage from the context's pinned memory resource instead, as the other users of continuation_task already do. Because that resource is a pool that does not synchronize with the device on deallocation, `split` now waits for its completion-copy event before the storage is released.
ericniebler
approved these changes
Nov 5, 2026
Collaborator
|
/ok to test 7e54987 |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The following tests crash in WSL:
This is because
cudaDevAttrConcurrentManagedAccessis0and how the shared completion storage is allocated causesSIGSEGVin the continuation when a kernel is still running. This PR fixes the issue by using pinned host memory instead.