[Mlir-commits] [mlir] [mlir][vector] Support more mask types in foldTransferFullMask() (PR #96761)

[Cullen Rhodes]

================
@@ -868,14 +868,16 @@ func.func @canonicalize_broadcast_shapecast_to_shapecast(%arg0: vector<3x4xf32>)
 // -----
 
 // CHECK-LABEL: fold_vector_transfer_masks
-func.func @fold_vector_transfer_masks(%A: memref<?x?xf32>) -> (vector<4x8xf32>) {
+func.func @fold_vector_transfer_masks(%A: memref<?x?xf32>) -> (vector<4x8xf32>, vector<4x[4]xf32>) {
   // CHECK: %[[C0:.+]] = arith.constant 0 : index
   %c0 = arith.constant 0 : index
   // CHECK: %[[F0:.+]] = arith.constant 0.000000e+00 : f32
   %f0 = arith.constant 0.0 : f32
 
   %mask = vector.constant_mask [8, 4] : vector<8x4xi1>
 
+  %mask_splat = arith.constant dense<true> : vector<4x[4]xi1>
----------------
c-rhodes wrote:

nit: splat could be false, all true would be more accurate
```suggestion
  %all_true_mask = arith.constant dense<true> : vector<4x[4]xi1>
```

https://github.com/llvm/llvm-project/pull/96761