[Mlir-commits] [mlir] [mlir][vector] Refactor linearize.mlir (PR #86648)

Balaji V. Iyer. llvmlistbot at llvm.org
Tue Mar 26 08:38:46 PDT 2024


================
@@ -95,10 +92,11 @@ func.func @test_index_no_linearize(%arg0: vector<2x2xindex>, %arg1: vector<2x2xi
 
 // vectorizable operation (arith.mulf) with tensor result types.
 
-func.func @nonvec_result(%arg0: tensor<2x2xf32>, %arg1: tensor<2x2xf32>) -> (tensor<2x2xf32>, tensor<2x2xf32>) {
-    // CHECK: %[[MULF:.*]] = arith.mulf %arg0, %arg1 : tensor<2x2xf32>
-    // CHECK128: %[[MULF:.*]] = arith.mulf %arg0, %arg1 : tensor<2x2xf32>
-    // CHECK0: %[[MULF:.*]] = arith.mulf %arg0, %arg1 : tensor<2x2xf32>
+// ALL-LABEL: test_tensor_no_linearize
+func.func @test_tensor_no_linearize(%arg0: tensor<2x2xf32>, %arg1: tensor<2x2xf32>) -> (tensor<2x2xf32>, tensor<2x2xf32>) {
+    // DEFAULT: %[[MULF:.*]] = arith.mulf %arg0, %arg1 : tensor<2x2xf32>
----------------
bviyer wrote:

Can't you replace all these 3 with ALL?

https://github.com/llvm/llvm-project/pull/86648


More information about the Mlir-commits mailing list