[Mlir-commits] [mlir] [mlir][ArmSME] Add optional padding and mask operands to tile_load (PR #69195)

[Andrzej Warzyński]

================
@@ -90,10 +106,38 @@ func.func @arm_sme_move_vector_to_tile_slice_f32__bad_vector_type(%vector : vect
   return %0 : vector<[4]x[4]xf32>
 }
 
+//===----------------------------------------------------------------------===//
+// arm_sme.move_tile_slice_to_vector
+//===----------------------------------------------------------------------===//
+
 // -----
 
 func.func @arm_sme_move_tile_slice_to_vector__bad_result_type(%tile : vector<[4]x[4]xf32>, %tile_slice_index : index) -> vector<[2]xf64> {
   // expected-error at +1 {{op failed to verify that type of 'result' matches type of 'tile' slice}}
   %0 = arm_sme.move_tile_slice_to_vector %tile[%tile_slice_index] : vector<[2]xf64> from vector<[4]x[4]xf32>
   return %0 : vector<[2]xf64>
 }
+
+//===----------------------------------------------------------------------===//
+// arm_sme.tile_load
+//===----------------------------------------------------------------------===//
+
+// -----
+
+func.func @arm_sme_tile_load__bad_padding_type(%src : memref<?x?xf64>, %pad : f32, %mask : vector<[2]x[2]xi1>) {
+  %c0 = arith.constant 0 : index
+  // expected-note at -2 {{prior use here}}
----------------
banach-space wrote:

[nit] We can probably skip this line

https://github.com/llvm/llvm-project/pull/69195