[Mlir-commits] [mlir] [mlir][tosa] Fix no op clamp canonicalization when nan_mode = ignore (PR #229887)

[Sayan Saha]

================
@@ -297,6 +297,16 @@ func.func @clamp_f32_is_noop(%arg0: tensor<4xf32>) -> tensor<4xf32> {
 
 // -----
 
+// CHECK-LABEL: @clamp_f32_nan_ignore_not_noop
+func.func @clamp_f32_nan_ignore_not_noop(%arg0: tensor<4xf32>) -> tensor<4xf32> {
----------------
sahas3 wrote:

Is there a test that integer clamp with `ignore` as `nan_mode` still folds since the new check is float specific?

https://github.com/llvm/llvm-project/pull/229887