Tex Riddell

Please disregard the prior request for a repro. We have a repo and understand why this happens. The root cause is that our intermediate high-level intrinsic functions have a collision...

I've been really frustrated with what I think might be this same bug for a very long time and finally figured out what reliably repros this for me. Plus, I...

The issue was detected when the NodeOutputArray subscript operator returned a reference to a NodeOutput object, which was not correct. For returning the record struct itself (not the object), like...

To be clear, this no longer impacts the same work graph case, it *might* impact some other pre-existing scenario, it isn't a regression, therefore shouldn't be considered an SM 6.8...

@anupamachandra > @llvm-beanz I thought the issue was that the subscript operator for NodeOutputArray was defined as `T& operator versus T operator. That is not the case, it does return...

> @tex3d Thank you for the details. I went back to the original change that added the subscript operator for NodeOutputArray types and it returns a value and not a...

Dang, it looks like type annotations are still not added for these subscript operators when they return values as well. I was able to crash the current code due to...

@anupamachandra I put up a draft PR that adds missing type annotations by iterating methods on node objects, as well as making sure that any `Type` template parameter for something...

This should only be used when compiling to a pre-release shader model. Selecting the internal validator won't sign the shader, and it will require developer mode with experimental shader models...

Unfortunately, this one is not sound language-wise, and tricky to support. It's likely going to be considered by-design. The language semantics are such that you could not use GetAttributeAtVertex anywhere...