Hi Alexei,

Is there a setting to control the number of parallel weight optimization threads in phenix.refine?

nproc parameter controls number of CPUs to be used in weight optimization.

Also, is there a way to force phenix.refine to perform global real-space refinement (rsrg)?

No. There is some built-in logic, phenix.refine does it when these conditions are met and it can un-do the result if it is not an improvement. Is there particular need to do this in your context? Pavel