Skip to content

Conversation

@MamziB
Copy link
Contributor

@MamziB MamziB commented Jun 30, 2022

OSC UCX: Fix the flush issue in dynamic win accumulate

Signed-off-by: Mamzi Bayatpour [email protected]
Co-authored-by: Tomislav Janjusic [email protected]

Signed-off-by: Mamzi Bayatpour  <[email protected]>
Co-authored-by: Tomislav Janjusic <[email protected]>
Copy link
Contributor

@devreal devreal left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The changes here are important. However, blocking for operations to complete inside a nonblocking operation is not desirable. We should change this to starting the GET operations, returning right away, and then performing the accumulate and PUT inside the completion callback. That way we don't waste time waiting for transfers.

@MamziB
Copy link
Contributor Author

MamziB commented Jun 30, 2022

@devreal Thanks for the note. I agree we should not do blocking Get inside Accumulate. Once we are done with stabilizing the code, we are planning to re-design the Accumulate. With the last fix, we have around a 90% pass rate for RMA mpich tests. So we are almost there.

@janjust
Copy link
Contributor

janjust commented Jul 1, 2022

bot:retest

@janjust janjust merged commit 698be2c into open-mpi:main Jul 1, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants