[Mlir-commits] [mlir] [mlir][tosa][tosa-to-linalg] Fix rescale with double rounding failing validation (PR #184787)

[Sayan Saha]

================
@@ -114,11 +118,15 @@ void mlir::tosa::registerTosaToLinalgPipelines() {
       [](OpPassManager &pm) {
         TosaToLinalgOptions tosaToLinalgOptions;
         TosaToLinalgNamedOptions tosaToLinalgNamedOptions;
+        TosaAttachTargetOptions tosaAttachTargetOptions;
+        tosaAttachTargetOptions.profiles = {"pro_int", "pro_fp"};
+        // TODO: Populate all the extensions that this conversion supports
+        tosaAttachTargetOptions.extensions = {"doubleround"};
         TosaValidationOptions validationOptions;
         validationOptions.strictOpSpecAlignment = false;
         validationOptions.allowInvalidOpDatatypeCombinations = false;
----------------
sahas3 wrote:

In https://github.com/llvm/llvm-project/pull/184787/changes#diff-f4196f31a3236cd12178e88c8d22e512874a01bbdda5bfcf1fe9ab5bc635f37eL40 the default for `TosaValidationOptions` is to create the options with `{false, false}` which is same here. But for `TosaAttachTargetOptions`, the default is `std::nullopt` but the pipeline creates the options with a specific configuration. Can you elaborate why the pipeline differs from the default for this specific options?

https://github.com/llvm/llvm-project/pull/184787