[llvm] [AMDGPU] Implement llvm.lrint intrinsic lowering (PR #98931)

[Matt Arsenault via llvm-commits]

================
@@ -786,13 +786,17 @@ void TargetLoweringBase::initActions() {
                      Expand);
 
   // These library functions default to expand.
-  setOperationAction({ISD::FCBRT,      ISD::FLOG,    ISD::FLOG2,  ISD::FLOG10,
-                      ISD::FEXP,       ISD::FEXP2,   ISD::FEXP10, ISD::FFLOOR,
-                      ISD::FNEARBYINT, ISD::FCEIL,   ISD::FRINT,  ISD::FTRUNC,
-                      ISD::LROUND,     ISD::LLROUND, ISD::LRINT,  ISD::LLRINT,
-                      ISD::FROUNDEVEN, ISD::FTAN,    ISD::FACOS,  ISD::FASIN,
-                      ISD::FATAN,      ISD::FCOSH,   ISD::FSINH,  ISD::FTANH},
-                     {MVT::f32, MVT::f64, MVT::f128}, Expand);
+  setOperationAction(
+      {ISD::FCBRT, ISD::FLOG,   ISD::FLOG2,  ISD::FLOG10,     ISD::FEXP,
+       ISD::FEXP2, ISD::FEXP10, ISD::FFLOOR, ISD::FNEARBYINT, ISD::FCEIL,
+       ISD::FRINT, ISD::FTRUNC, ISD::LROUND, ISD::LLROUND,    ISD::FROUNDEVEN,
+       ISD::FTAN,  ISD::FACOS,  ISD::FASIN,  ISD::FATAN,      ISD::FCOSH,
+       ISD::FSINH, ISD::FTANH},
+      {MVT::f32, MVT::f64, MVT::f128}, Expand);
+
+  // Unless the target expands, default LRINT to LibCall.
----------------
arsenm wrote:

This should be more a FIXME move all of those to libcall (although really this would make more sense to query RuntimeLibcalls for each of these to make the decision 

https://github.com/llvm/llvm-project/pull/98931