Add AMAX, AVG, NORM1, NORM2, MUL, MUL_NO_ZEROS reduction modes#325
Merged
Conversation
rsuderman
force-pushed
the
reduction_rest
branch
from
05be4e2 to
2a5541c
Compare
sjain-stanford
requested changes
Apr 16, 2026
Member
sjain-stanford
left a comment
sjain-stanford
left a comment
There was a problem hiding this comment.
Might need rebase / CI seems unclean.
Enable the remaining cuDNN reduction modes in ReductionAttr and add the corresponding MLIR schemas to the asm emitter: - NORM1 lowers to abs + sum.dim_IntList. - AMAX lowers to abs + amax. - AVG lowers to mean.dim (float dtypes only — torch.aten.mean.dim is not defined on integer tensors, so the sample skips int32 for AVG). - NORM2 lowers to mul + sum.dim_IntList + sqrt. - MUL lowers directly to torch.prims.prod. - MUL_NO_ZEROS uses aten.ne.Scalar to build an i1 mask, then aten.where.ScalarOther to substitute 1 for zero entries before feeding the result to torch.prims.prod, so zero inputs are excluded from the product. Extend samples/reduction/reduction_ops.cpp to exercise every new mode. Input data is built by a per-mode generateReductionInputData helper so MUL/MUL_NO_ZEROS get a non-trivial pattern (mostly 1s with a 2 and a 3, plus injected zeros for MUL_NO_ZEROS) that stays in range for fp16/int32, and the expected value is computed by the existing reference reduction loop rather than hardcoded. Add lit tests for each new mode under tests/lit/ and register them in tests/CMakeLists.txt. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Signed-off-by: Rob Suderman <rob.suderman@gmail.com>
Signed-off-by: Rob Suderman <rob.suderman@gmail.com>
sjain-stanford
approved these changes
May 7, 2026
Member
sjain-stanford
left a comment
sjain-stanford
left a comment
There was a problem hiding this comment.
LGTM modulo two more comments.
sjain-stanford
added a commit
that referenced
this pull request
May 8, 2026
I usually grep for `TODO` to quickly scan open items so adding this here. Nit cleanup from #325. Signed-off-by: Sambhav Jain <sambhav@alumni.stanford.edu>
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
2 participants
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.
Enable the remaining cuDNN reduction modes in ReductionAttr and add the corresponding MLIR schemas to the asm emitter:
Extend samples/reduction/reduction_ops.cpp to exercise every new mode. Input data is built by a per-mode generateReductionInputData helper so MUL/MUL_NO_ZEROS get a non-trivial pattern (mostly 1s with a 2 and a 3, plus injected zeros for MUL_NO_ZEROS) that stays in range for fp16/int32, and the expected value is computed by the existing reference reduction loop rather than hardcoded.
Add lit tests for each new mode under tests/lit/ and register them in tests/CMakeLists.txt.