Parallel tasks only use 1 core per node #72
Description
Is your feature request related to a problem? Please describe.
Parallel tasks (i.e. the parallel profile type, the parallel homogenous and the parallel total profile type) simulate some number of flops for each node but only use one core per node. This is frustrating when trying to accurately simulate energy, or when trying to combine the parallel task type with a usage profile type where it is required to have more than one core per host.
Describe the solution you'd like
I am thinking to add an optional "utilization" field to the parallel task types. This will be used the same way the utilization field is used in the usage-trace profile to estimate how many cores to use during the computation.
Describe alternatives you've considered
I've tried simulating the same behavior using the usage trace profile, but in addition to being quite tedious (you have to create a trace for each rank of the computation), it leads to high overheads and crashes during the simulation.