Table 2. Parallel execution of the data analysis.

Processing step(s)	Work distributed to MPI ranks	Host and architecture	Nodes employed	CPU cores/node	Total MPI ranks	Total wall time (h)
(a)–(c) Spotfinding, indexing, refinement and integration with DIALS	Conventional data reduction on independent images	Linux server, AMD Opteron 6300	1	64	64	5.5
(d) Energy-dependent geometrical profile modeling for each spot, and profile-based refinement of crystal orientation	Each image processed independently	cori.nersc.gov, Knights Landing	400	68	6800	30.0
(e) Refinement of background and scale parameters	Each image processed independently	cori.nersc.gov, Knights Landing	32	68	1088	8.9
(f) Refinement of the scattering factors	Sum the independent contributions from each spot for each iteration