[llvm] [CodeGen][AMDGPU] Support type-aware intrinsic feature checks (PR #205536)

[Keshav Vinayak Jha via llvm-commits]

keshavvinayak01 wrote:

I assume you're talking about #201470. The merged version provides the static `TargetFeatures` mapping, feature-expression checking, intrinsic-ID cache, and generic `SelectionDAG/GlobalISel` diagnostics. It does not include overload-sensitive support, which is precisely what this PR adds. It's also already rebased on those changes.

https://github.com/llvm/llvm-project/pull/205536