[Mlir-commits] [mlir] [MLIR][NVVM] Update prefetch.tensormap Op (PR #153134)

[Srinivasa Ravi]

================
@@ -2464,15 +2466,30 @@ def NVVM_PrefetchOp : NVVM_Op<"prefetch"> {
   }];
 }
 
-def NVVM_PrefetchTensorMapOp : NVVM_Op<"prefetch.tensormap",
-                    [DeclareOpInterfaceMethods<BasicPtxBuilderOpInterface>]>,
-  Arguments<(ins LLVM_AnyPointer:$tmaDescriptor, PtxPredicate:$predicate)> {
-  let assemblyFormat = "$tmaDescriptor (`,` `predicate` `=` $predicate^)? attr-dict `:` type(operands)";
+def NVVM_PrefetchTensorMapOp : NVVM_Op<"prefetch.tensormap", 
+    [DeclareOpInterfaceMethods<BasicPtxBuilderOpInterface>, NVVMRequiresSM<90>]> {
----------------
Wolfram70 wrote:

Actually I missed this earlier, but it looks like the NVGPU `tma.prefetch.descriptor` Op lowers to `prefetch.tensormap` (with support for predicates as well) so this lowering path might need to be changed in order to combine `prefetch` and `prefetch.tensormap`. So, should we retain them as separate Ops for now?

https://github.com/llvm/llvm-project/pull/153134