[llvm-dev] Efficient Green Thread Context-Switching

[Joshua Thomas Wise via llvm-dev]

I could live with that. You’re right that “newstack" is probably more use-case-dependant. A “preserve_none” calling convention would be more general than swapstack, which is perhaps more preferred for LLVM. I don’t think that would be a big deal for swapstack-seekers, because I think the stack pointer swapping could be achieved using @llvm.read_register() and @llvm.write_register().

I’m glad to know other people are in support of this functionality.

> On Mar 27, 2020, at 2:20 PM, Jon Chesterfield via llvm-dev <llvm-dev at lists.llvm.org> wrote:
> 
> I propose that the LLVM devs consider adding “swapstack” and “newstack” into the LLVM project officially.
> 
> Thanks for the reference. I was going to suggest the 'preserve_none' calling convention, applied to asm which does the stack pointer swapping, would be equivalent to swapstack. However that calling convention doesn't appear to exist - it would be one where the caller is required to save all live values to the stack before jumping.
> 
> 'preserve_none' may be a useful calling convention independent of green threads. Functions which will always use all the available registers benefit from not having to save/restore any of them, if the caller knew to preserve any live values. A green thread context switch is roughly a call that preserved nothing.
> 
> Newstack I'm not as sure about. Green thread libraries are likely to put a lot of effort into how stacks are sized, allocated, reused. That seems better as a runtime call.
> 
> Thanks
> 
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20200327/e2153b1b/attachment.html>