[Intrinsics] Fix `nvvm_prmt` intrinsic attributes

[slydiman]

Summary

nvvm_prmt doesn't seem to be commutative. nvvm also sets IntrSpeculatable for it.
Here is the doc https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#data-movement-and-conversion-instructions-prmt

[tra]

Good catch. LGTM.

llvm/include/llvm/IR/IntrinsicsNVVM.td

class SHFL_INFO<bit sync, string mode, string type, bit return_pred> {
   list<LLVMType> ArgsTy = !if(sync,
     [llvm_i32_ty, OpType, llvm_i32_ty, llvm_i32_ty],
     [OpType, llvm_i32_ty, llvm_i32_ty]);
 }
 
 let TargetPrefix = "nvvm" in {
   def int_nvvm_prmt : GCCBuiltin<"__nvvm_prmt">,
       DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
-        [IntrNoMem, Commutative]>;
+        [IntrNoMem, IntrSpeculatable]>;
 
 //
 // Min Max
 //
 
   foreach operation = ["min", "max"] in {
     def int_nvvm_f # operation # _d :
       GCCBuiltin<!strconcat("__nvvm_f", operation, "_d")>,