bound check elimination (BCE)?

[barko]

Is there a world in which koka could support the elimination of bound-checks? Golang seems to have some support for this. BCE would make koka appealing for the implementation of numerical algorithms.

[TimWhiting]

I don't think there is any theoretical or practical reason not to. It's just a matter of time. Other things have priority at the moment.

As a side note, my PhD research is about control flow analysis which is a more general technique which could enable us to remove many unnecessary checks. Bounds check elimination would be a subset of what the technique could handle. Of course there would be some engineering effort and buy-in required to use that particular approach in Koka.

There are of course less powerful analyses that can do bounds-check elimination in the subset of cases that golang and other languages handle. Until we have more numerical algorithms implemented in Koka using vectors, we won't be able to tell if we are targeting common use-cases or wasting engineering effort though. What idiomatic Koka code should look like is still something that is unclear at the moment.

If you are interested in this, I would first write some algorithms for Koka and benchmark them. Feel free to join the community in Discord [https://discord.gg/CEkHHnTqEv] to get some feedback and help. Also feel free to check out some of the community packages: https://github.com/koka-community, and if you want to contribute to the community packages or create a new one feel free to say so and I can give access.

[barko]

Thanks for you thoughtful response.

I see you are giving me homework ;-) I accept.

Good luck with the research.

[TimWhiting]

Thanks! Academics do tend to give homework :). Also, I can't do everything and my PhD too :).