[tuner] add padding_conv attribute along IGEMM supprot for conv #2701
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Outdated suggestions cannot be applied.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.
Context
paddingattribute: A general padding configuration used for most operations (matmul, IGEMM, and other linalg operations). Specified in IGEMM/GEMM dimension space (after convolution is transformed to IGEMM). Used byGPUPadOperandsPassto pad operations in their transformed GEMM form.padding_convattribute: A specialized padding configuration for convolutions before IGEMM transformation. Specified in the original convolution dimension space (before transformation). Used byGPUPadConvsPassto pad convolution operations in their original form by mapping IGEMM padding sizes back to convolution dimensions.Key difference: padding works in GEMM space (M, N, K dimensions), while padding_conv works in the original convolution space (batch, spatial, channel dimensions).
Reference PR from IREE side: iree-org/iree#21470
What this PR does
This PR adds support for the
padding_convattribute for convolution operations when using the IGEMM transformation path. This enables padding convolutions in their original dimension space before IGEMM transformation.Issue: #2687