Jan Wassenberg (jan.wassenberg.delete@this.gmail.com) on May 22, 2022 10:56 pm wrote:
> Jörn Engel (joern.delete@this.purestorage.com) on May 22, 2022 1:25 pm wrote:
> > Heh! Same here. At some point I implemented expand. But the code is fairly
> > tricky and I noticed that I didn't need it after all before finishing it.
> Oh no :) If you are able to publish it under something other than AGPL, I'd appreciate
> having a look, no sense reinventing tricky wheels completely from scratch.
>
> > Interesting. Clearly you see a lot more hardware variety than I do.
> :) Like it or not, we are moving into a heterogeneous future.
>

Shipping RISC-V hardware with vectors? Or experimental boards?

> > There actually isn't very much to it, people could recreate it from scratch in a couple of days.
> You might be underestimating the value here. I have a list of about two dozen wrappers
> and their value is proportional to the number of compilers/targets/ops supported, plus
> all the compiler bugs they shield users from. There is a huge difference between something
> whipped up in a couple of days, vs something maintained for years.
>
> > But if you know someone that wants a C vector library and doesn't mind a "here you go, have a nice
> > life" attitude or might even step in as the future maintainer, send them my way.
> Sure, I'll ask internally.