pkubaj wrote:

> What's the story with the runtimes? Since I left feedback here I came across 
> the following comment of yours in the FreeBSD review 
> (https://reviews.freebsd.org/D57388#1314874):
> 
> > compiler-rt part can't be upstreamed, because this is a hack -
> > compiler-rt does not support kf type ldbl functions that powerpc64le
> > needs. Upstream usually just uses libgcc (from GCC) for that.
> 
> We can't just go changing ABI if it's going to break runtimes, and any 
> requisite support needs to go upstream first.

The story is that compiler_rt lacks proper support. It's not FreeBSD-specific, 
the issue is also present on Linux. On Linux it's also done the hacky way - 
libgcc does sed-renaming TF symbols to KF and clang links with -lgcc. The 
proper way would be to write independent code in compiler_rt, but that is 
another project on its own (probably much bigger than this). As much as I'd 
like it to happen, I understand that this may not make much sense since TF 
functions work the same as KF, so simple renaming works fine and doesn't 
introduce new bugs.

https://github.com/llvm/llvm-project/pull/201298
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to