[llvm] AMDGPU: Remove flat/global fmin/fmax intrinsics (PR #105642)

[Gheorghe-Teodor Bercea via llvm-commits]

doru1004 wrote:

> > The problem is that no one knows what options to use to make the atomicRMW closer to the old intrinsic. You do not document that anywhere.
> 
> [It's in AMDGPUUsage](https://llvm.org/docs/AMDGPUUsage.html#llvm-ir-metadata)

I agree but that doesn't say what combination of those options to use to be as close as possible to the previous intrinsic. So if we can document that then we will be all good!

https://github.com/llvm/llvm-project/pull/105642