Idea: Allow multiple input files to be specified.
Each input file can have its own regions etc., and then the runtime/mapper can just run each at appropriate scale. Idea would be something like launching X copies of the main task (at current) except with different input files, and these spread throughout the machine
This is something for far in the future (obviously) but represents a common use case in some particle methods, and with an appropriate mapper could enable better machine usage than individual jobs.
Idea: Allow multiple input files to be specified. Each input file can have its own regions etc., and then the runtime/mapper can just run each at appropriate scale. Idea would be something like launching X copies of the main task (at current) except with different input files, and these spread throughout the machine
This is something for far in the future (obviously) but represents a common use case in some particle methods, and with an appropriate mapper could enable better machine usage than individual jobs.