Skip to content

Conversation

@davidwendt
Copy link
Contributor

@davidwendt davidwendt commented Jan 15, 2026

Description

Experimental script for checking thrust calls are execution policy nosync compliant.
If a thrust call returns a value, then nosync is not required.

Checklist

  • I am familiar with the Contributing Guidelines.
  • New or existing tests cover these changes.
  • The documentation is up to date with these changes.

@davidwendt davidwendt self-assigned this Jan 15, 2026
@davidwendt davidwendt added 2 - In Progress Currently a work in progress libcudf Affects libcudf (C++/CUDA) code. improvement Improvement / enhancement to an existing function non-breaking Non-breaking change labels Jan 15, 2026
@copy-pr-bot
Copy link

copy-pr-bot bot commented Jan 15, 2026

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

@davidwendt
Copy link
Contributor Author

/ok to test

@davidwendt
Copy link
Contributor Author

/ok to test

@davidwendt
Copy link
Contributor Author

/ok to test

@davidwendt
Copy link
Contributor Author

/ok to test

@davidwendt
Copy link
Contributor Author

/ok to test

@davidwendt davidwendt changed the title [EXPER] Use smarter thrust execution-policy checker [EXPER] Use improved thrust execution-policy checker Jan 16, 2026
@davidwendt
Copy link
Contributor Author

/ok to test

@davidwendt
Copy link
Contributor Author

/ok to test

@davidwendt
Copy link
Contributor Author

/ok to test

@bdice
Copy link
Contributor

bdice commented Jan 20, 2026

CCCL already guarantees that if a value is returned, a sync is required even with a nosync policy, regardless of whether that returned value is used or discarded. See docs linked here (also copied below for simplicity).

//! Execution policy type, optionally prescribing with CUDA stream to use for execution. Omits any optional stream
//! synchronization. The CUDA stream is still synchronized when required for algorithmic correctness.

That means we can simplify our usage to "always nosync." We never have to think about the differences and the correctness is self-enforcing.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

2 - In Progress Currently a work in progress improvement Improvement / enhancement to an existing function libcudf Affects libcudf (C++/CUDA) code. non-breaking Non-breaking change

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants