Test "ElementwiseToAffine" support for ops with both tensor and scalar operands

[AlexanderViand-Intel]

The ElementwiseToLinalg pass in MLIR only supports operations where all operands are tensors and this requirement has been left in place for the ElementwiseToAffine pass. However, the reason that this limitation is necessary for the linalg version doesn't exist in the affine version, which should already support scalar operands as-is (well, after replacing the all_of in isElementwiseMappableOpOnRankedTensors with any_of) . However, polynomial operations have a SameOperandsAndResultType constraint, so to test whether this feature actually works as expected, we'd need to either allow mixed tensor-scalar operations and/or simply test this pass with a different ElementwiseMappable op.

[github-actions[bot]]

This issue has 1 outstanding TODOs:

• lib/Transforms/ElementwiseToAffine/ElementwiseToAffine.cpp:30: Test ElementwiseToAffine with any_of constraints

This comment was autogenerated by todo-backlinks