[Mlir-commits] [mlir] [MLIR][Arith] Add canonicalization rules for int-to-float of integer extension (PR #185386)

[Jakub Kuderski]

kuhar wrote:

> Happy to do that, but are there any instructions?

It's mostly a straightforward arith to llvm ir translation. I've tried the first one https://alive2.llvm.org/ce/z/W3ue2f (by hand)

> Also: it seems like those canonicalization patterns undo the results of arith-emulate-wide-int. Is that a problem? Should I just remove [`--canonicalize` from the test file](https://github.com/llvm/llvm-project/blob/main/mlir/test/Dialect/Arith/emulate-wide-int-canonicalization.mlir)?

Seems like we won't need this test anymore?

https://github.com/llvm/llvm-project/pull/185386