Support lowering VAArgOp

[ChuanqiXu9]

Close #862

I made this by introducing VaArgOp to mlir and use that op in clangir. I feel this is the most natural way to do this. I don't think we should handle different architecture in clangir. It is the job of LLVM or MLIR. I understand that we hope we don't touch any other workdirs except CIR. But it looks pretty better to do the proper thing in the proper places.

I'll try to upstream the patch to support MLIR later.

[ChuanqiXu9]

CC @ghehg @sitio-couto since I can't add reviewers.

[bcardosolopes]

since I can't add reviewers.

Just fixed that.

[bcardosolopes]

I'll try to upstream the patch to support MLIR later.

Thanks for fixing this, we indeed need that op in the LLVM dialect to make progress. We are very afraid of the technical debt though, specially the ones outside CIR stuff. The LLVM part should be done upstream first and cherry-picked here. It should also be in a different PR from the actual lib/CIR changes (so our rebaser can make into an empty commit once it kicks in against upstream, cc @lanza).

EDIT: perhaps this won't be needed at all.

[ghehg]

Thanks for working on this. This is necessary along right direction. I agree with @bcardosolopes, it might be better if we have LLVM dialect part of change landed in upstream first, then other parts of this PR can just rebase on it, and go. This way we don't have to sync changes of upstream PR and that of this PR. Let me know what you think.

[sitio-couto]

Hey @ChuanqiXu9!

I really like this approach—it's clean and a great contribution to the LLVM Dialect.

I don't think we should handle different architecture in clangir.

AFAIK, we need to handle targets in CIR to some extent, since LLVM IR doesn’t cover all the details. Still, this is a good abstraction.

One concern I have is that, for some reason, the OG Clang doesn’t use the LLVM va_arg instruction when lowering C/C++.
Maybe it's just technical debt. Maybe it's something more complex, like optimizations (choosing between reg or mem like the example above).
By emitting a different IR than the original CodeGen, we might be abstracting something we shouldn’t, like an optimization or argument-passing correctness.

Just something to keep in mind, otherwise LGTM!

[bcardosolopes]

One concern I have is that, for some reason, the OG Clang doesn’t use the LLVM va_arg instruction when lowering C/C++.

Yea, same feedback in the issue. The PR should work along the lines of Aarch64 here.