mtrofin wrote:

> alwaysinline can't always meet the needs -- it causes a particular function 
> to be inlined into all callers, which is not selective based on calling 
> context. What this does is make inlining happen under a particular call tree, 
> and fully flatten a call tree.
> 
> +@mtrofin hope you can see the merit of the extension -- we have real use 
> case for it in our internal codebase where we don't want any calls for some 
> paths and this is a simple, natural extension for existing flatten attribute.

Sorry, I only saw this now. I second @efriedma-quic 's feedback that a RFC 
seems appropriate. An aspect that I hope the RFC could cover is whether this 
could be initially implemented LLVM-side only (e.g. via some manifest file 
saying which functions should be treated this way) - effectively a gradual 
introduction.

https://github.com/llvm/llvm-project/pull/165777
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to