PoreSpy settings and parallel execution

PoreSpy exposes its process-wide configuration as porespy.settings. The settings.ncores value controls the number of threads passed to the edt.edt distance-transform backend and is also used as the default worker count by selected PoreSpy routines. It can be changed at any time:

import porespy as ps

ps.settings.ncores = 4

The setting persists until it is explicitly changed. Changes apply to subsequent EDT calls even when a module obtained its EDT callable before the change. A direct call can still override it with parallel=....

By default, PoreSpy uses the number of logical CPUs available to the current process. CPU affinity is preferred when the operating system supports it; the logical CPU count is used as a fallback. On Linux, a cgroup CPU quota can further reduce the default. The result is always at least one. Assigning None to settings.ncores selects this default again, while assigning a positive integer provides an explicit value.

PoreSpy prefers pyedt when that optional backend is installed. Its edt function does not accept the parallel keyword, so ncores is not passed to it. The separate parallel network-extraction path that requires pyedt is configured with its threads option.

settings.ncores does not configure every parallel facility in PoreSpy. Chunk decomposition (divs and overlap), Dask scheduling, and Numba threading have their own controls. Their documentation identifies the appropriate settings for each algorithm.