[llvm] [LLVM] SDAG and GISel support for folding fabs into fp_round source modifiers (PR #204861)

[Jay Foad via llvm-commits]

jayfoad wrote:

I'm not sure this belongs in generic code. Is it beneficial for any targets other than AMDGPU?

For fneg, this kind of thing is implemented in the AMDGPU backend with fnegFoldsIntoMI/fnegFoldsIntoOp. Maybe fabs should follow the same pattern?

https://github.com/llvm/llvm-project/pull/204861