[Mlir-commits] [mlir] [mlir][ArmSME] Lower transfer_write + transpose to vertical store (PR #71181)

[Andrzej Warzyński]

================
@@ -398,6 +429,17 @@ func.func @transfer_write_2d__fixed(%vector : vector<16x16xi8>, %dest : memref<?
   return
 }
 
+// -----
+
+// CHECK-LABEL: @transfer_write_2d__out_of_bounds
+// CHECK: vector.transfer_write
+// CHECK-NOT: arm_sme.tile_store
+func.func @transfer_write_2d__out_of_bounds(%vector : vector<[4]x[4]xf32>, %dest : memref<?x?xf32>) {
+  %c0 = arith.constant 0 : index
+  vector.transfer_write %vector, %dest[%c0, %c0] : vector<[4]x[4]xf32>, memref<?x?xf32>
----------------
banach-space wrote:

[nit] Adding `in_bounds = [false, false]` would make this example a bit clearer.

https://github.com/llvm/llvm-project/pull/71181