This optional keyword allows users to perform the different tasks in a mg-para parallel run asynchronously. Specifically, a processor masquerades as process rank in a parallel focusing run and provides output (data files and energies/forces) appropriate to that processor's local partition. The user must then assemble the results after all processes complete. First, this option is useful for scheduling on-demand resources: this makes it easy for users to backfill into the available processes in a queue. Second, this option is useful for running on limited resources: this enables users without access to large parallel machines to still perform the same calculations. The syntax is
where rank is the integer ID of the particular processor to masquerade as. Processor IDs range from 0 to N-1, where N is the total number of processors in the run (see pdime). Processor IDs are related to their position in the overall grid by p = nx ny k + nx j + i where nx is the number of processors in the x-direction, ny is the number of processors in the y-direction, nz is the number of processors in the z-direction, i is the index of the processor in the x-direction, j is the index of the processor in the y-direction, k is the index of the processor in the z-direction, and p is the overall rank of the processor.