AaronBallman wrote:

Oof, this is tough (at least for me), we seem to have reasonably strong support 
for `noipa` implying `noinline` but we also have someone pointing out that 
they've already needed these to be separated and that the workaround with 
`always_inline` may not be sufficient in all circumstances.

In terms of compatibility with GCC, they document:

>  This attribute implies noinline, noclone and no_icf attributes

We don't expose `noclone` or `no_icf` as Clang attributes, but does LLVM have 
the concepts to map to in IR? e.g., is the argument about compatibility with 
GCC somewhat moot because we're not going to be compatible anyway?

https://github.com/llvm/llvm-project/pull/207502
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to