[llvm] AMDGPU: Improve codegen for intrinsic llvm.fptrunc.round (PR #104486)

[Changpeng Fang via llvm-commits]

changpeng wrote:

> Title is misleading, this doesn't change codegen at all?
The patch changes the instruction definition, and adds the cases for two additional rounding modes. I was thinking, the whole process for intrinsic to generate assembly code is called code generation. But codegen may have a specific meaning. 

https://github.com/llvm/llvm-project/pull/104486