Port the low latency allgather kernels to AMD#162
Draft
erieaton-amd wants to merge 7 commits intoByteDance-Seed:mainfrom
Draft
Port the low latency allgather kernels to AMD#162erieaton-amd wants to merge 7 commits intoByteDance-Seed:mainfrom
erieaton-amd wants to merge 7 commits intoByteDance-Seed:mainfrom
Conversation
Signed-off-by: Eric Eaton <erieaton@amd.com>
Signed-off-by: Eric Eaton <erieaton@amd.com>
This file was not included in the CI and wasn't updated with some recent changes. Signed-off-by: Eric Eaton <erieaton@amd.com>
Signed-off-by: Eric Eaton <erieaton@amd.com>
There was a change in ROCm 7+ that makes it harder to match up the torch and amdsmi devices. This change to use KFD makes operations like sleep_async work again. Signed-off-by: Eric Eaton <erieaton@amd.com>
Signed-off-by: Eric Eaton <erieaton@amd.com>
Signed-off-by: Eric Eaton <erieaton@amd.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This builds on two other PRs which should be reviewed and merged first.
The low_latency_allgather.py kernels are copied and adapted for AMD devices. I needed to add the signal_op function, but rocshmem doesn't have it. It's a fairly simple function so I added the code to the wrapper file.
This work ran into an issue that there seems to be a race condition between benchmark iterations that stops the program from terminating. I've addressed this by changing the signal wait condition to GE instead of EQ. This allows the kernels to pass the correctness test and finish the benchmark, but slightly changes how they are expected to be used.