[Mlir-commits] [mlir] [mlir][arith] Add `arith.fptofp` op (PR #188041)

[Matthias Springer]

matthias-springer wrote:

> Like when making a rewriter pattern you would catch the entire FpToFp op and do multiple if and validation checks to match it.

Yes, when lowering to LLVM, you have to check if it's going to turn into an `fext` or `ftrunc`. But it's a simple check.

On the other hand, you may be able to get rid of these checks when coming from the frontend. E.g., PyTorch does not distinguish between FP extensions/truncations: `x_f32.to(torch.float16)`.


https://github.com/llvm/llvm-project/pull/188041