[llvm] AMDGPU][True16][CodeGen] FP_Round f64 to f16 in true16 (PR #128911)

[Brox Chen via llvm-commits]

broxigarchen wrote:

> > I don't understand this change. You shouldn't have to make any of this code conditional on true16. Are you just avoiding the FP_TO_FP16? We can and should do that independently of true16
> 
> @arsenm so when should we use FP_TO_FP16? Is it only for targets where f16 is not legal? That would include AMDGPU subtargets where has16BitInsts() is false.
> 
> @broxigarchen I think we need to sort this out first, and then revisit the current patch.

Hi Matt and Jay. Thanks for the explaination! Tried to update this patch from the comments and please let me know if this still does not sound right

https://github.com/llvm/llvm-project/pull/128911