[GlobalOpt] Support index types when hoisting constexprs #22920
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Outdated suggestions cannot be applied.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.
Avoids crashes for operations like
tensor.dim/tensor.from_elementswhen hoisting constant expressions.The only
HoistableTypeInterface's that are feasible to instantiate are the existing ones, for general tensors and for scalar indices.tensor<index>types need explicit handling, as default utils for bitwidth assessment do not (and should not) assume the exact size of the index type.The approach taken is to patch up the common
RankedTensorTypepaths inHoistableTypeInterfaceandConstExpranalyses. The conservative assumption of 64-bit indices is wrapped around the common helpers likeIREE::Util::getRoundedPhysicalStorageSizeandIREE::Util::getTypeBitWidth- and therefore, isolated from the rest of the codebase.Instead of the 64-bit assumption, an alternative approach would be to access module-level data layout information from within
ConstExprAnalysis. Achieving this forHoistableTypeInterfaceseems less clear.