[Mlir-commits] [mlir] [MLIR][Transforms] add eliminate-explicit-rounding pass (PR #93443)

[Jakub Kuderski]

kuhar wrote:

> In traditional compiler world(saying LLVM), fast-math is a combination of below flags:
> Clearly, the rounding elimination doesn't belong to any of them. Adding this optimization into fast-math will make fast-math definition different between LLVM and MLIR. This will create confusion.

The design of the arith dialect is not tied to llvm (see this [RFC](https://discourse.llvm.org/t/rfc-define-precise-arith-semantics/65507/33). The other lowering path, SPIR-V, defines a different set of fast math flags:
https://registry.khronos.org/SPIR-V/specs/unified1/SPIRV.html#_fp_fast_math_mode.

https://github.com/llvm/llvm-project/pull/93443