feat(iree): fuse elementwise generic ops into iree_linalg_ext.scan ops
#39
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
|
Warning This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
This stack of pull requests is managed by Graphite. Learn more about stacking. |
AGindinson
force-pushed
the
artem/cpu-stack/fuse-scan
branch
from
92dccd2 to
7dd47ee
Compare
AGindinson
force-pushed
the
artem/cpu-stack/scan-op-semantics
branch
from
81c38ad to
03fbc4f
Compare
Enhance `iree_linalg_ext.scan` support in the pipeline by introducing a basic fusion pattern for generic ops that are determined as unary elementwise operations. Provided that the full shapes match between the GenericOp producer and the ScanOp consumer, we can simply repeat the unary operation directly during the input data indexation of a ScanOp reduction routine. This is essentially serving as a workaround for LLVMCPU pipeline, where stack-bound allocations tend to exceed the size limit in the presence of a ScanOp that just receives the pipeline's default tiling configuration. Further enhancements should include tiling config fine-tuning for LinalgExt operations, concious restrictions of work-group level tiling depending on the predicted size of stack-bound allocas within a dispatch. However, the fusion itself is beneficial, and we should look to support more producer cases, expanding the logic onto other pre-defined LinalgExt ops.
AGindinson
force-pushed
the
artem/cpu-stack/fuse-scan
branch
from
7dd47ee to
fb955a8
Compare
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Outdated suggestions cannot be applied.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.
Enhance
iree_linalg_ext.scansupport in the pipeline by introducing a basic fusion pattern for generic ops that are determined as unary elementwise operations. Provided that the full shapes match between the GenericOp producer and the ScanOp consumer, we can simply repeat the unary operation directly during the input data indexation of a ScanOp reduction routine.This is essentially serving as a workaround for LLVMCPU pipeline, where stack-bound allocations tend to exceed the size limit in the presence of a ScanOp that just receives the pipeline's default tiling configuration.
Further enhancements should include tiling config fine-tuning for LinalgExt operations, conscious restrictions of work-group level tiling depending on the predicted size of stack-bound allocas within a dispatch. However, the fusion itself is beneficial, and we should look to support more producer cases, expanding the logic onto other pre-defined LinalgExt ops.