std.pool

The pool of threads that parallel for runs on. A parallel for gives the same result with any number of threads, so the number changes only how fast a program runs.

CalxOutput

Tested

import std.pool.{set_workers}

fn main() {
    set_workers(4)
    var squares = [0; 8]
    parallel for k in 0usize..8 {
        squares[k] = i32.wrap(k * k)
    }
    println("{squares[7]}")
}

Item	Does
set_workers(n: usize)	Makes each parallel for run on n threads, counting the thread that reaches it. With 1, every parallel for runs on that thread alone. Panics if n is 0, or if called inside a parallel for
unsafe workers() -> usize	How many threads each parallel for runs on: the default below, or what set_workers set. It’s unsafe, as reading the clock is, since it differs from machine to machine: a program can size its own threads by it, such as threads that compress what a server sends, but what it prints with it isn’t the same everywhere

The default is the number of logical processors the program may run on: GetActiveProcessorCount on Windows, and the CPUs in the program’s affinity mask on Linux, so a container limited to some CPUs gets that many.

CalxOutput

Tested

import std.pool.{set_workers, workers}

fn main() {
    set_workers(3)
    println("{unsafe { workers() }}")
    println("{unsafe { workers() } > 0}")
}

Setting it doesn’t start or stop threads. The next parallel for that needs more starts them, and threads beyond the number just don’t take part.

std.pool is written in Calx: std/pool.calx. The pool itself is in the runtime, runtime/calx_rt.c.