[flang-commits] [flang] [mlir] [RFC][mlir] Conditional support for fast-math attributes. (PR #125620)

[Ben Vanik via flang-commits]

benvanik wrote:

This feels very odd to me - like an implementation detail of LLVM leaking way higher up into the stack than it should. `arith.select` is a conditional move op and should never change the value of either operand. Ops like `arith.select` and block arguments are changed repeatedly throughout many pipelines and preserving (or even knowing how to set) fast math information is not feasible (or useful). If lowering into LLVM needs this information then it should perform analysis to determine the required information - it is much easier to do so at the time of lowering than it is to keep that information present and valid through layers as high up as tensor graph programming all the way down to LLVM that may be passing through half a dozen intermediate dialects.

https://github.com/llvm/llvm-project/pull/125620