[cfe-dev] Adding AVR target to Clang

Joerg Sonnenberger via cfe-dev cfe-dev at lists.llvm.org
Sat Nov 26 11:42:04 PST 2016


On Sat, Nov 26, 2016 at 02:46:56PM +0000, via cfe-dev wrote:
>   I submitted a patch yesterday (https://reviews.llvm.org/D27123) that
>   adds bare bones Clang support for the experimental AVR target in
>   LLVM.

Please always remember to put at least one mailing list on the
subscriber list, i.e. cfe-commits in this case. Otherwise the review has
very low visibility. It would help to include a reference to whatever
ABI document you are using as base. E.g. if you want to support C++, it
would be nice to include getEHDataRegisterNumber(). If you support
atomic ops, make sure to provide __GCC_HAVE_SYNC_COMPARE_AND_SWAP_*
builtins. Any reason to default to unsigned wchar_t? I know that a few
platforms have done that, but it is normally creating unnecessary
problems. Why do you default to DWARF2? Otherwise, just as jroelofs
said, lots of missing testing.

>   So, is it ok to submit Clang patches for an experimental LLVM target?

Yes. Clang patches are not fundamentally tied to LLVM targets, so it is
perfectly fine to have the frontend support for in-tree LLVM targets.

>   If yes, who would be the right person to review them? From
>   CODE_OWNERS, I found and added Richard Smith as the reviewer, is that
>   ok?

Yes, it is.

Joerg



More information about the cfe-dev mailing list