dnovillo wrote:

> @dnovillo Can you take ownership of the corresponding issue, and provide a 
> justification for the option? I forget why we needed it in DXC. Was it a 
> Vulkan validation issue to avoid warnings for mismatches between stages? If 
> we carry an option forward, I would like to make sure the reason for it is 
> still valid. I had opened the issue as a placeholder with the intent of 
> determining if it was still useful. I should have made that more explicit.

Well, I only ran into it because of 
[BigWheels](https://github.com/google/bigwheels). I couldn't get it to compile 
anything without the flag. I don't know whether this is a DXC feature that we 
definitely want to migrate. If it's not desirable to migrate this, I can just 
as easily adjust BigWheels's build setup. I haven't really done a survey on 
what other software packages may be using it. Anything that wants to do 
reflection?

> I believe the real issue is output variable that are never referenced. If a 
> vertex shader has an output that is never written to, then it might be 
> removed. If that variable at the same location is statically read in the 
> fragment shader, it will not be removed. Then the validation layers will 
> issue an error because of an interface mismatch.

I have a vague recollection of this being the case, yes.


https://github.com/llvm/llvm-project/pull/196404
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to