https://github.com/bogner approved this pull request.

LGTM with some misgivings. The issues I have with this are pre-existing so I 
think figuring out what to do about them is better handled in follow ups.

I think that this PR points out some rather serious problems with how the HLSL 
toolchain handling works in general. We're lumping in various target specific 
logic into general "HLSL" handling but then keying off specific targets to work 
out what we're doing with them. There's a very poor separation of concerns 
between the language and the target, as evidenced by how many target specific 
checks there are in the HLSL logic and how much of the HLSL logic is stuff that 
could only ever be relevant for the DirectX backend. I really think we need to 
step back and reconsider how a lot of this is done and how we can do these 
abstractions in a less adhoc way.

https://github.com/llvm/llvm-project/pull/188150
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to