[Mlir-commits] [mlir] [MLIR][Vector] Add support for inner-parallel masked multi-reductions (PR #126722)

[Manupa Karunaratne]

================
@@ -41,3 +41,23 @@ func.func @vector_multi_reduction_parallel_middle(%arg0: vector<3x4x5xf32>, %acc
 //        ALL-SAME: %[[INPUT:.+]]: vector<3x4x5xf32>, %[[ACC:.+]]: vector<4xf32>
 // INNER-REDUCTION: vector.transpose %[[INPUT]], [1, 0, 2] : vector<3x4x5xf32> to vector<4x3x5xf32>
 //  INNER-PARALLEL: vector.transpose %[[INPUT]], [0, 2, 1] : vector<3x4x5xf32> to vector<3x5x4xf32>
+
+// -----
+
+func.func @vector_masked_multi_reduction(%arg0: vector<2x4xf32>, %acc: vector<2xf32>, %mask: vector<2x4xi1>) -> vector<2xf32> {
----------------
manupak wrote:

done

https://github.com/llvm/llvm-project/pull/126722