Max_step vs hipace.max_time

[lboult]

Hi,

I would like to use the hipace.max_time definition to define how long my HiPACE simulations run for rather than max_step. However, I find that if I don't include max_step in the prefile that the simulation crashes with the message "Please use more or equal time steps than number of ranks".

I assume this is because HiPACE assumes the default value of max_step (0) and if you're using more than 1 gpu it doesn't like it. But anyway, I guess it should be the case that if hipace.max_time is provided then this should override the default value of max_step, if I'm understanding correctly?

Cheers,
Lewis

P.S would it also make sense to have max_step be hipace.max_step to be more consistent?

[MaxThevenet]

Thanks @lboult for this issue. Could you share in input file, or just the relevant lines? I think the behavior you describe is what should be implemented already, so it would be good to see what dt, max_time and max_step you are using.

P.S would it also make sense to have max_step be hipace.max_step to be more consistent?

Could be done indeed, although we try to avoid such changes that require users to fix their old input scripts.

[AlexanderSinn]

Currently, it is set up in a way where the simulation ends when either max_step or max_time is reached. The default max_time is infinite, while the default max_step is zero. So when you specify max_time you also need to provide a very large max_step like 1000000 to be sure that there are sufficient steps to reach the time.

[AlexanderSinn]

A possible issue with having a max_step of infinity by default is that with an adaptive time step the beam might deplete before max_time which can result in a very large number of steps before the simulation ends. 

[lboult]

Hey guys,

Thanks for the responses. I see what Alex means now and perhaps this is the way it needs to say for the reasons you mention. Perhaps a note in the documentation about max_time might be useful but that's all :)

Cheers,
Lewis