[Tuner] Add support for TileAndFuse and multi-dim contractions #745
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Outdated suggestions cannot be applied.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 adds support for multi-dimensional (multiple M/N/K/batch) contraction ops using the TileAndFuse pipeline. The PR adds a new constraint set for the TileAndFuse pipeline, which allows for multiple M, N, or K contraction dimensions. A new flag is also added called
--codegen-pipeline
, which indicates whether to usellvmgpu_vector_distribute
orllvmgpu_tile_and_fuse
. Multi dim contraction ops do not currently work with the LLVMGPUVectorDistribute pipeline, but this can be extended as a followup. Also, tuning for convolution on the TileAndFuse pipeline is not yet implemented, but it should be a simple extension now that the TileAndFuse contraction tuning is in place.