[Mlir-commits] [mlir] [MLIR][Linalg] Add aggregate ops decomposition pass and softmax decom… (PR #97582)

[Adam Siemieniuk]

================
@@ -0,0 +1,113 @@
+// RUN: mlir-opt %s -split-input-file -linalg-decompose-named-ops | FileCheck %s
+// RUN: mlir-opt %s -split-input-file -linalg-decompose-named-ops -linalg-generalize-named-ops | FileCheck %s --check-prefix=GENERALIZECHECK
+
+func.func @softmax(%arg0: tensor<2x16x32xf32>, %dst: tensor<2x16x32xf32>) -> tensor<2x16x32xf32> {
----------------
adam-smnk wrote:

Could you also add a separate case with dynamic shapes?

https://github.com/llvm/llvm-project/pull/97582