[PATCH] D16213: [MC] Provide an MCTargetOptions to implementors of MCAsmBackendCtorTy, NFC

[Daniel Sanders via llvm-commits]

dsanders added a comment.

In https://reviews.llvm.org/D16213#487480, @joelkevinjones wrote:

> I'm trying to reconcile how this patch should interact with the set of proposed patches rooted at: https://reviews.llvm.org/D21465
>  The have similar goals, that is to support ABIs in a uniform fashion.
>
> Daniel, could you comment?
>
> Joel


I think your patch is the preferable approach but I can't reach a working state with just the approach that you're using. As a result, our patches end up working together to reach a working solution which is close to the ideal end result but has some short-term compromises in order to have something that works. The problem I have is that there are some areas affected by the N32/N64 ABI's that are impossible to move to the MCTargetOptions approach (see the llvm-dev thread for details). I've asked questions about those areas to see if Eric has a way to resolve them but those questions haven't been answered so I don't know if he sees something I don't. These problems mean that I see having unambiguous ABI information in the triple (like X86 and ARM do today) as the only possible way to get N32/N64 working properly. ABI information in the triple is sufficient by itself to make N32/N64 work but following the discussion on llvm-dev I've added extra patches on the end so that all targets end up using both the triple and MCTargetOptions (like ARM does today).

To summarize, I think both our patches should go ahead and I'll adapt my series to fit the changes you make.


https://reviews.llvm.org/D16213