pub fn process_affinity_cpus() -> Vec<usize>Expand description
The logical CPUs this process may actually run on, ascending.
This is the affinity mask, not the machine: a pool sized to the
machine inside a taskset- or cpuset-confined container
oversubscribes the slice it was given by exactly the factor it was
confined by, and every worker then time-slices against every other.
Linux reads sched_getaffinity; everywhere else (macOS included,
which has no equivalent) this is 0..available_parallelism(), i.e.
today’s answer.