Add pattern to convert generic conv ops to IGEMM #19798
Merged
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
nirvedhmeshram
force-pushed
the
generic_conv_igemm_final
branch
2 times, most recently
from
22ed0c0 to
c94c8b6
Compare
Max191
reviewed
Jan 24, 2025
compiler/src/iree/compiler/Dialect/LinalgExt/Transforms/test/conv2d_to_im2col.mlir
Show resolved
Hide resolved
nirvedhmeshram
force-pushed
the
generic_conv_igemm_final
branch
3 times, most recently
from
f70dc62 to
afcd4e5
Compare
nirvedhmeshram
requested review from
hanhanW,
MaheshRavishankar and
antiagainst
as code owners
nirvedhmeshram
force-pushed
the
generic_conv_igemm_final
branch
from
afcd4e5 to
5f8ffad
Compare
Max191
approved these changes
Jan 29, 2025
compiler/src/iree/compiler/Codegen/Dialect/GPU/TargetUtils/ConfigUtils.cpp
Show resolved
Hide resolved
compiler/src/iree/compiler/Codegen/Dialect/GPU/TargetUtils/ConfigUtils.cpp
Show resolved
Hide resolved
compiler/src/iree/compiler/Dialect/LinalgExt/Transforms/ConvertConv2DToIm2ColOp.cpp
Outdated
Show resolved
Hide resolved
compiler/src/iree/compiler/Dialect/LinalgExt/Transforms/ConvertConv2DToIm2ColOp.cpp
Outdated
Show resolved
Hide resolved
compiler/src/iree/compiler/Dialect/LinalgExt/Transforms/ConvertConv2DToIm2ColOp.cpp
Show resolved
Hide resolved
compiler/src/iree/compiler/Dialect/LinalgExt/Transforms/test/conv2d_to_im2col.mlir
Show resolved
Hide resolved
Max191
approved these changes
Jan 29, 2025
Signed-off-by: Nirvedh Meshram <[email protected]>
Signed-off-by: Nirvedh Meshram <[email protected]>
nirvedhmeshram
force-pushed
the
generic_conv_igemm_final
branch
from
80af6ec to
b86e844
Compare
ita9naiwa
pushed a commit
to ita9naiwa/iree
that referenced
this pull request
Feb 4, 2025
This PR removes the named op patterns to convert convs to IGEMM and replaces them with a generic pattern that works for all supported convs. A new utility function that populates the shared details required for setting lowering config and doing the IGEMM computation is added. The PR is currently using a default true flag `iree-gpu-use-tile-and-fuse-generic-convolution` . The idea is that since a lot more convolutions will go down the IGEMM path with this PR if any of them run into issues we can turn the flag off by default rather then needing to revert the whole PR. If after some time we find that there are no issues then we can drop the flag and have this happening always. --------- Signed-off-by: Nirvedh Meshram <[email protected]> Signed-off-by: Hyunsung Lee <[email protected]>
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 removes the named op patterns to convert convs to IGEMM and replaces them with a generic pattern that works for all supported convs. A new utility function that populates the shared details required for setting lowering config and doing the IGEMM computation is added.
The PR is currently using a default true flag
iree-gpu-use-tile-and-fuse-generic-convolution. The idea is that since a lot more convolutions will go down the IGEMM path with this PR if any of them run into issues we can turn the flag off by default rather then needing to revert the whole PR. If after some time we find that there are no issues then we can drop the flag and have this happening always.