Update of cost model for z17 [No change to default policy]#3396
Open
AlexandreEichenberger wants to merge 1 commit intoonnx:mainfrom
Open
Update of cost model for z17 [No change to default policy]#3396AlexandreEichenberger wants to merge 1 commit intoonnx:mainfrom
AlexandreEichenberger wants to merge 1 commit intoonnx:mainfrom
Conversation
Signed-off-by: Alexandre Eichenberger <alexe@us.ibm.com>
AlexandreEichenberger
changed the title
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Revised version from #3383 which updates the cost model for some missing z17 operations, tuned the heuristics to be more inline with LLM-ish models, and fix a compiler warning.
Part of the changes are also enforcing the placement decision when lowering ONNX to ZHigh... as currently some "CPU" placed operations where still moved to NNPA.
Still believe we should change the default placement heuristics, can delay/discuss this in the original PR