Stream: t-compiler/wg-llvm

Topic: good-next-issues

dlrobertson (Mar 25 2019 at 17:06, on Zulip):

Anyone know of other good issues to work on that would relate to this WG?

dlrobertson (Mar 25 2019 at 17:06, on Zulip):

I have some interest in RISC-V, so I was considering looking into ways I could contribute to RISC-V support in LLVM. I also am interested in va_arg. At some point there was some discussion on improving va_arg in LLVM. This would be helpful to support Rust defined C-variadics on more architectures.

rkruppe (Mar 25 2019 at 18:16, on Zulip):

I believe that providing a portable unified va_list construct in LLVM IR will require shaving a herd of yaks, too many for one person to reasonably do (in such a huge project as LLVM). Also IIRC it's an open question whether "improving va_list" from llvm's perspective means that, or whether it would mean removing the vestiges that exist today and going all in on low-level manipulation of ABI specific structures.

dlrobertson (Mar 25 2019 at 18:48, on Zulip):

will require shaving a herd of yaks

hmmm yeah that may be a bit much :)

Jake Goulding (Mar 25 2019 at 23:19, on Zulip):

If you are interested in contributing to newer aspects of LLVM, you could help out with AVR world too (

dlrobertson (Mar 26 2019 at 01:50, on Zulip):

Nice, good suggestion. There are some upstream issues that are unassigned

Nikita Popov (Apr 14 2019 at 09:45, on Zulip):

A nice issue for trailing_zeros / cttz: The change itself should be rather simple, but there might be some disagreement on compile time tradeoff.

dlrobertson (Apr 18 2019 at 15:11, on Zulip):

For anyone else following along

dlrobertson (Apr 18 2019 at 15:12, on Zulip):

I'd like to add more tests for some of the other paths that are improved by this... I'll update the diff with them sometime this evening

Last update: Jul 03 2020 at 16:55UTC