[clang] [llvm] [Clang][AMDGPU] Add __builtin_amdgcn_cvt_off_f32_i4 (PR #133741)

[Jay Foad via llvm-commits]

================
@@ -1578,6 +1578,11 @@ class MovDPP8Pattern<Predicate Pred, Instruction Inst, ValueType vt> : GCNPat <
   let OtherPredicates = [Pred];
 }
 
+def : GCNPat <
----------------
jayfoad wrote:

Can you do this in the definition of V_CVT_OFF_F32_I4, instead of using a separate Pat?

https://github.com/llvm/llvm-project/pull/133741