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

[llvmlistbot at llvm.org]

================
@@ -211,31 +213,61 @@ func.func @softmax(%arg0: tensor<2x16x32xf32>, %dst: tensor<2x16x32xf32>) -> ten
 // CHECK-SAME:           %[[ARG0:[a-zA-Z0-9_]+]]: tensor<2x16x32xf32>, %[[DST:[a-zA-Z0-9_]+]]: tensor<2x16x32xf32>) -> tensor<2x16x32xf32> {
 // CHECK-DAG:        %[[D1:.+]] = tensor.empty() : tensor<2x16xf32>
 // CHECK-DAG:        %[[CST:.+]] = arith.constant -3.40282347E+38 : f32
-// CHECK:        %[[D2:.+]] = linalg.fill ins(%[[CST]] : f32) outs(%[[D1]] : tensor<2x16xf32>) -> tensor<2x16xf32>
+// CHECK:        %[[D2:.+]] = linalg.generic {indexing_maps = [#[[$MAP2]], #[[$MAP3]]],
----------------
MaheshRavishankar wrote:

I am not sure why this becomes a generic op. Can we preserve the original behavior of this being a fill op?

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