[ROCm] Fix for breakage in XLA Conv Op functionality #46219
Merged
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Outdated suggestions cannot be applied.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.
The following commit breaks Conv Op functionality (in the XLA backend) for ROCm platform.
8684c6b
The cause seems to be that the
scratch_size
field in the newGpuConvDescriptor
is not getting correctly populated in the new MLIR path. It is being used correctly in the convolution runner code.declaration:
8684c6b#diff-6453912dbc4ee715a56da9d7b218b52795dea2aa631a482101fc6d58c573d9ccR122-R135
use (get access) in conv runner:
8684c6b#diff-a01181d08b28a9c7432f22439622f16725126184283a73822c70b2151098a8adR277
set access in non-MLIR(?) based path:
8684c6b#diff-a01181d08b28a9c7432f22439622f16725126184283a73822c70b2151098a8adR450
This commit merely adds the missing "set" in the MLIR based path
thanks to @ekuznetsov139 for identifying the fix
/cc @chsigg @cheshire @nvining-work