[clang] [llvm] [HLSL] Implement `WaveReadLaneAt` intrinsic (PR #111010)

[Chris B via cfe-commits]

================
@@ -87,6 +87,7 @@ class CGHLSLRuntime {
   GENERATE_HLSL_INTRINSIC_FUNCTION(SDot, sdot)
   GENERATE_HLSL_INTRINSIC_FUNCTION(UDot, udot)
   GENERATE_HLSL_INTRINSIC_FUNCTION(WaveIsFirstLane, wave_is_first_lane)
+  GENERATE_HLSL_INTRINSIC_FUNCTION(WaveReadLaneAt, wave_readlaneat)
----------------
llvm-beanz wrote:

It is probably worth naming this `wave_readlane` to align a bit with `amdgcn.readlane`. I'm not sure what the equivalent NV PTX intrinsic is... probably some crazy bundle of shuffles which confuses me.

https://github.com/llvm/llvm-project/pull/111010