[Mlir-commits] [mlir] [mlir][tensor] Implement folding logic for size 0 tensor and memref ops (PR #90814)

[Spenser Bauman]

sabauma wrote:

In our case this was occurring after some lowering patterns which generate a lot of `extract_slice` and `insert_slice` patterns. We do some folding during the lowering, but the fact that the slices are size-0 is not always obvious until after canonicalizing.

https://github.com/llvm/llvm-project/pull/90814