If you've joined our #rustseq GitHub organization, please check out the issues on the "planning-and-discussions" repo for a link to join our Zulip channel! If you're interested in joining the #rustseq organization but aren't a part of it yet, please reach out!

@nomad421 @pierre_marijon @rustseq The result will mostly be 1 simple streaming impl, and 1 fast SIMD buffered impl. I like @zommiommy's suggestion of putting the rest directly in benches/. That way it's not in the way of users, but easily accessible for testing (eg on different hardware)

@nomad421 @pierre_marijon it would be nice to have a good storyline for this for @rustseq crates, even if just for me

@Rutseq That's good, strong Rustseq .. but we aren't mutuals

@curious_coding @pierre_marijon @rustseq Is there a clear "winner" implementation? If so, I feel that's the only one that belongs in the lib. If there are different winners along different axes, fearure flags feel right. For historical but sub-optimal implementations, maybe a feature branch?

Excited to see this take off 🚀 !!!

@curious_coding @nomad421 @rustseq @zommiommy I only kept the best version in lib/crates and the others in the benches folder. This is viable as long as there aren't too many or they aren't too long. But you could create a module to store alternatives and use an inline function and features to let the user choose version.