Bail out of reserve_space on under-transfer WRITE flows#733
Merged
bosilca merged 1 commit intoICLDisco:masterfrom Mar 13, 2025
Merged
Bail out of reserve_space on under-transfer WRITE flows#733bosilca merged 1 commit intoICLDisco:masterfrom
bosilca merged 1 commit intoICLDisco:masterfrom
Conversation
We must not call transfer_ownership with a WRITE flow under transfer. We cannot distinguish between stage-in and eviction so we just bail out and come back later. Signed-off-by: Joseph Schuchart <joseph.schuchart@stonybrook.edu>
Contributor
Author
|
Alternative: reset the transfer flag and set the host version to 0. The epilogue of the w2r task will ignore the host copy if it detects that the device version has changed (https://github.com/ICLDisco/parsec/blob/master/parsec/mca/device/transfer_gpu.c#L319) |
5 tasks
Contributor
|
Interestingly that fixes the assert for POTRF with the lowmem parameters (ctest), but not for GEMM lowmem_mpi. Everything else works, so overall it is improving the current state, but I had expected and hoped it would fix both. |
This comment was marked as resolved.
This comment was marked as resolved.
This comment was marked as resolved.
This comment was marked as resolved.
This comment was marked as outdated.
This comment was marked as outdated.
This was referenced Mar 12, 2025
abouteiller
approved these changes
Mar 13, 2025
Contributor
abouteiller
left a comment
abouteiller
left a comment
There was a problem hiding this comment.
Overall improvement over master, case that is still deadlocking #741 also deadlock in master
Pass all other ctests and large-size PO/GEMM
bosilca
approved these changes
Mar 13, 2025
10 tasks
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
3 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.
We must not call transfer_ownership with a WRITE flow under transfer. We cannot distinguish between stage-in and eviction so we just bail out and come back later.
The assert that otherwise triggers is here: https://github.com/ICLDisco/parsec/blob/master/parsec/data.c#L420