Atropos single-threaded?

[effigies]

Watching htop, I don't think Atropos is ever spawning extra processes the way antsRegistration does, and its CPU usage never tops 100%.

Does someone else have any evidence to justify claiming these cores?

niworkflows/niworkflows/anat/ants.py

Lines 530 to 545 in 9b0f646

	atropos = pe.Node(
	Atropos(
	dimension=3,
	initialization="KMeans",
	number_of_tissue_classes=in_segmentation_model[0],
	n_iterations=3,
	convergence_threshold=0.0,
	mrf_radius=[1, 1, 1],
	mrf_smoothing_factor=0.1,
	likelihood_model="Gaussian",
	use_random_seed=use_random_seed,
	),
	name="01_atropos",
	n_procs=omp_nthreads,
	mem_gb=mem_gb,
	)

[oesteban]

I haven't found any issues in their repo mentioning multithreaded execution. In principle, ITK should make the multithreading almost automatically if the appropriate ITK environment variable is set. Nipype seems to at least have a notion of all the pertinent flags (https://github.com/nipy/nipype/blob/master/nipype/interfaces/ants/base.py#L13-L26), but I'm not sure whether the particular one that works with atropos is correctly set.

[effigies]

Sure, multithreading should be automatic, but I think it's a single-threaded job.