[Mlir-commits] [mlir] [MLIR][Linalg] Ternary Op & Linalg select (PR #91461)

[llvmlistbot at llvm.org]

================
@@ -1924,3 +1924,25 @@ func.func @fill_tensor(%f: f32, %v: vector<2x4xf32>) -> (tensor<f32>, tensor<vec
   %1 = linalg.fill ins(%v : vector<2x4xf32>) outs(%e1 : tensor<vector<2x4xf32>>) -> tensor<vector<2x4xf32>>
   return %0, %1: tensor<f32>, tensor<vector<2x4xf32>>
 }
+
+// -----
+
+// CHECK-LABEL: func @select_dynamic
+func.func @select_dynamic(%arg0: memref<?x?x?xi1>, %arg1: memref<?x?x?xf32>, %arg2: memref<?x?x?xf32>, %arg3: memref<?x?x?xf32>) {
+  // CHECK: linalg.select
+  // CHECK-SAME: ins(%{{.+}}, %{{.+}}, %{{.+}} : memref<?x?x?xi1>, memref<?x?x?xf32>, memref<?x?x?xf32>)
+  // CHECK-SAME: outs(%{{.+}} : memref<?x?x?xf32>)
+  linalg.select ins(%arg0, %arg1, %arg2 : memref<?x?x?xi1>, memref<?x?x?xf32>, memref<?x?x?xf32>) outs(%arg3: memref<?x?x?xf32>)
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @select_static
+func.func @select_static(%arg0: memref<4x8x16xi1>, %arg1: memref<4x8x16xf32>, %arg2: memref<4x8x16xf32>, %arg3: memref<4x8x16xf32>) {
----------------
ciyongch wrote:

Better to cover `tensor` besides `memref`?

https://github.com/llvm/llvm-project/pull/91461