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

[Adam Siemieniuk]

================
@@ -250,6 +282,8 @@ module attributes {transform.with_named_sequence} {
 
     %2 = transform.structured.match ops{["linalg.softmax"]} in %arg1 : (!transform.any_op) -> !transform.any_op
     %3 = transform.structured.decompose_interface %2 : (!transform.any_op) -> !transform.any_op
+    %4 = transform.structured.generalize %3: (!transform.any_op) -> !transform.any_op
----------------
adam-smnk wrote:

> generic lowering is still valid

It used to be decomposed to generic but now the decomposition is always to named ops and you have to rely on separate pass to have that representation. So, I still think it's unrelated to this test.


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