Hi,
is there any plan on allowing a restart of a given simulation using different resources (i.e. number of partitions)?
Say, for example, that I’m running with cuda with X GPUs and want to switch to openmp with Y CPUs
Regards
Hi,
is there any plan on allowing a restart of a given simulation using different resources (i.e. number of partitions)?
Say, for example, that I’m running with cuda with X GPUs and want to switch to openmp with Y CPUs
Regards
I’m not sure what you mean?
As noted in the documentation you can use pyfr partition
to repartition a mesh and associated solution files. You may then restart using these repartitioned mesh/solution files.
Regards, Freddie.
When using rk45, for example, do I have to repartition not only the *.pyfrm file but also the latest 4 *.pyfrs files?
No, RK methods don’t use the time history in the interpolation, that’s there advantage. The RK integration only requires a single solution and then takes a series of substeps, these get stored and used to build the next sub step.
Got it!
As always, very grateful for the help!