[Mlir-commits] [mlir] [mlir][linalg] raise generic to named ops. (PR #110421)

[Andrzej Warzyński]

================
@@ -0,0 +1,11 @@
+// RUN: mlir-opt %s -linalg-generalize-named-ops | mlir-opt --linalg-specialize-generic-ops | FileCheck %s
+
+// CHECK-LABEL: linalg_transpose
+// CHECK-SAME:  %[[A:.+]]: tensor<16x64xf32>, %[[Out:.+]]: tensor<64x16xf32>
+// CHECK-NOT:   linalg.generic
+// CHECK:  %transposed = linalg.transpose ins(%[[A]] : tensor<16x64xf32>) outs(%[[Out]] : tensor<64x16xf32>) permutation = [1, 0]
+//
+func.func @linalg_transpose(%A: tensor<16x64xf32>, %Out: tensor<64x16xf32>) -> tensor<64x16xf32> {
----------------
banach-space wrote:

[nit] Skip `linalg` in func name (repeating info already available)

https://github.com/llvm/llvm-project/pull/110421