UCT/ROCM: bring ROCm fixes over to v1.13 branch #8235
Merged
Conversation
This file contains 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
also minor reorganization of the uct_rocm_base_agents structure to reduce the number of cache misses. (cherry picked from commit 7baac93)
fix the approach used to identify ROCm memory type. ROCm memory type is as of right now of type HSA_EXT_POINTER_TYPE_HSA with the owner agent being a GPU. (cherry picked from commit 4a876b5)
use the correct remote_agent handle if information is available. Using the local_agent for both src and destination is just the backup plan in case the remote_agent info is not available, e.g. in case the user limited the visibility of devices for a/some/all processes. (cherry picked from commit 79ea940)
yosefe
approved these changes
May 18, 2022
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Outdated suggestions cannot be applied.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.
What
This pr brings three fixes from the master repository over to the v1.13 branch.
The three Prs are:
UCT/ROCM: increase max. number of hsa agents
(cherry picked from commit 7baac93)
UCT/ROCM: fix memory type detection
(cherry picked from commit 4a876b5)
UCT/ROCM/IPC: use remote_agent if available
(cherry picked from commit 79ea940)
Why ?
All three commits are required for correct execution of rocm/ucx jobs.
How ?