Im learning zig for kernel bypass packet work. Aside from some noise, and one bug filed on cross compliation (prob simple to fix) bug, it works like I expect. Export of zig code to clib works fine. Comptime is great. Compile time superb. Like TB I will import no other modules ... and avoid all async.
But this link has got me concerned. When I get more ibverbs work done ill see how good/bad it is in rust.
In this kind of extremely low latency high throughput work there is hardly any (mostly none) MT data structures. Memory is pre-allocated for one pinned thread. Atomics are not much used either. So rust isn't going to help on design much. But I think rust/c interoperability is worse than zig.
In fairness to zig many of these issues were closed when zig 0.16.0 was released, which is what im using.
But this link has got me concerned. When I get more ibverbs work done ill see how good/bad it is in rust.
In this kind of extremely low latency high throughput work there is hardly any (mostly none) MT data structures. Memory is pre-allocated for one pinned thread. Atomics are not much used either. So rust isn't going to help on design much. But I think rust/c interoperability is worse than zig.
In fairness to zig many of these issues were closed when zig 0.16.0 was released, which is what im using.
We'll see