[GPU][MTL] Resolve long token performance regression in MTL 125H plat… #28155
+2
−2
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
…form PR27831 enable MLP fusion in cldnn, it can improve performance, but it is not enabled in MTL 125H due to EU number is 112. So there should be no performance improvement, but PR26940, which integrate dynamic quantization, causes MTL 125H performance drop about 10% for 6K input token size. If we enable MLP fusion in MTL 125H, the performance regression will disappear.
|
Hi @riverlijunjie according to @isanghao , PR26940 should not affect MTL because the PR was for onednn case. Could you clarify how PR26940 affected perf? (And apart from the question, I think we can apply this change though) |
PR26940 updates dynamic_quantize for cldnn, it should take effect on MTL, am i right? |
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Outdated suggestions cannot be applied.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.
Details:
PR27831 enable MLP fusion in cldnn, it can improve performance, but it is not enabled in MTL 125H due to EU number is 112. So there should be no performance improvement, but PR26940, which integrate dynamic quantization, causes MTL 125H first token performance drop about 8% for 6K input token size. If we enable MLP fusion in MTL 125H, the performance regression will disappear.
test result:
Tickets: