[Mlir-commits] [mlir] [MLIR][Vector] Generalize broadcast lowering for single-element vector to nD (PR #206501)

[Andrzej Warzyński]

================
@@ -95,6 +95,42 @@ func.func @broadcast_stretch(%arg0: vector<1xf32>) -> vector<4xf32> {
   return %0 : vector<4xf32>
 }
 
+// CHECK-LABEL: func @broadcast_single_elem_vec1d_to_vec2d
+// CHECK-SAME: %[[A:.*0]]: vector<1xf32>
+// CHECK:      %[[T0:.*]] = vector.extract %[[A]][0] : f32 from vector<1xf32>
+// CHECK:      %[[T1:.*]] = vector.broadcast %[[T0]] : f32 to vector<16x1xf32>
+// CHECK:      return %[[T1]] : vector<16x1xf32>
+
+func.func @broadcast_single_elem_vec1d_to_vec2d(%arg0: vector<1xf32>) -> vector<16x1xf32> {
----------------
banach-space wrote:

In order to follow the effective naming convention in this file, please use this:
```suggestion
func.func @broadcast_vec2d_from_vec1d_single_element(%arg0: vector<1xf32>) -> vector<16x1xf32> {
```

Similar suggestion for the other tests.

https://github.com/llvm/llvm-project/pull/206501