[Mlir-commits] [mlir] [mlir]Add a check to ensure bailing out when reducing to a scalar (PR #129694)

[Andrzej Warzyński]

================
@@ -222,6 +222,13 @@ func.func @vector_multi_reduction(%v : vector<4x6xf32>, %acc: vector<4xf32>) ->
 //       CHECK:   %[[V2:.*]] = vector.insert_strided_slice %[[R5]], %[[V1]] {offsets = [2], strides = [1]} : vector<2xf32> into vector<4xf32>
 //       CHECK:   return %[[V2]] : vector<4xf32>
 
+func.func @negative_vector_multi_reduction(%v: vector<4x2xf32>, %acc: f32) -> f32 {
----------------
banach-space wrote:

Please add a comment explaining what makes this a negative test (IIUC, the result is already 100% unrolled).

https://github.com/llvm/llvm-project/pull/129694