[Codegen][llvmgpu] Compute gemmC size when C promotion is done in padding matmul #19307
+555
−38
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
Signed-off-by: Nirvedh <[email protected]>
Signed-off-by: Nirvedh <[email protected]>
- This is a followup of iree-org#19271 - This take gemmC promotion into consideration in computing shared memory consumption Signed-off-by: jerryyin <[email protected]>
jerryyin
requested review from
MaheshRavishankar,
qedawkins,
kuhar,
Groverkss and
antiagainst
as code owners
jerryyin
changed the title
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Outdated suggestions cannot be applied.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 depends on #19271. Please review the last commit only.
The existing implementation of #19271 doesn't take gemmC into consideration when computing shared memory size. Though in condition of #19271, gemmC always get promoted and we ended always allocating the C tensor in shared memory. Ignoring C tensor will severely underestimate the amount of shared memory used and eventually cause
deduceMMASchedule()to pick a tile size too large and go beyond the space limit for shared memory.6c9ffd2 addressed this by adding
calculateResultSharedMemoryUsedInBytes()and apply it to matmul tiling size derive process.