CPU load ?

[zigster64]

Just running the basic example, Im seeing 100% CPU on 1 core.

?? Not sure if I have configured something wrong, or if Im on a particularly bad release of zig-master with a broken std.net.

Anyway - ApplePie is idling in var connection = stream.accept() catch |err| switch (err) { in src/server.zig ... so that suggest that zig std.net is eating all the CPU.

Curious, will have a deeper play tomorrow.

[zigster64]

followup - tried with different build options (as this is actually a benchmarking exersize comparing output to Go / Crystal / Rust)

When built with -Drelease-fast=true ... the benchmarking results using siege are absurdly good compared to regular debug build (which is slightly better than Go / Crystal)

Wasnt expecting that. (fast build has about 1% of the error rate, and is reporting 10x the transactions processed - weird !)

[zigster64]

Cancel that - I tried the same with https://github.com/Vexu/routez and got the exact same results

... 100% CPU on 1 core when waiting for socket connections through zig std.net
... massive benchmark improvement when building with -Drelease-fast=true

So the "problem" - if there is one - is higher up the chain. Nothing to do with ApplePie.

Also - changing the io_mode to .blocking solves the CPU issue
So with .blocking io_mode, and -Drelease-fast=true
The benchmark results v Go very very good.

closing

[Luukdegram]

Hi @zigster64, Thank you for reporting this nonetheless.
accept() should be a blocking call (theoretically) but it isn't atm. Currently, there are quite some issues upstream with regards to the event loop and I believe there's an idea to improve it soon, but I'm not sure what the exact details of those are. Just wanted to let you know :)

[zigster64]

ziglang/zig#8224 is on the radar now :)