[CCLBackend] Using parallel memcpy for inference_all_reduce #4404
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
delock
requested review from
RezaYazdaniAminabadi,
awan-10,
jeffra,
cmikeh2 and
arashb
as code owners
tjruwase
approved these changes
Oct 3, 2023
mauryaavinash95
pushed a commit
to mauryaavinash95/DeepSpeed
that referenced
this pull request
Oct 9, 2023
…dai#4404) * use parallel version of memcpy * include max buf size to 16MB per rank * support any input buffer size * fix format error
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Outdated suggestions cannot be applied.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.
This PR introduce a parallel version of memcpy and use it in inference_all_reduce. This allows memcpy fully utilize host memory bandwidth and improve performance. For some typical message size, this PR improve their performance as follows: (Tested on 2 socket 4th Gen Xeon Scalable system)
32KB: 26us -->23us
128KB: 59us --> 29us
512KB: 210us --> 40us
Besides, we improved handling of large message size. Now a fixed size shared memory can support any message size. This would help first token of a long input sequence.