Restart save frequency

Questions and remarks about Code_Saturne usage
Forum rules
Please read the forum usage recommendations before posting.
Post Reply
knewlands
Posts: 42
Joined: Mon Mar 26, 2012 10:53 am

Restart save frequency

Post by knewlands » Fri Oct 12, 2012 6:48 pm

Hello,

I would like to ask a relatively basic question about the frequency with which restart files can be saved. I read on the user manual that ntsuit can be set to a desired saving period, but I can't seem to find where I can define ntsuit.

I would like to save the restart files with the same frequency as the output post-processing files - is this possible through the use of ntsuit>0?

Thank you,

Kristin

Yvan Fournier
Posts: 2871
Joined: Mon Feb 20, 2012 3:25 pm

Re: Restart save frequency

Post by Yvan Fournier » Fri Oct 12, 2012 8:00 pm

Hello,

Yes, ntsuit can be defined in usipes or usipsu (or through the GUI).

If ntsuit is > 0, it defines a checkoint frequency (0 defines of default based on ntmabs/4, -1 checkpointing at the end only, and -2 deactivation).

Cheers,

Yvan

knewlands
Posts: 42
Joined: Mon Mar 26, 2012 10:53 am

Re: Restart save frequency

Post by knewlands » Mon Oct 15, 2012 4:00 pm

Hello Yvan,

Thank you for your reply, I have been able to set ntsuit > 0 in the GUI. However, it would seem that this only defines the period at which checkpoints are saved. Is this such that in the event of a crash there would still be a restart that prevents the loss of the entire calculation?

My interest in setting ntsuit > 0 originated from the need to run a calculation and then be able to return to a certain stage of interest and restart from there, as opposed to repeating the simulation from the start or only being able to restart from ntpabs. In other words, let's assume I initially have a calculation with ntmabs=20000 and post-processing every 'n' time steps = 5000. Having looked at the results when this first calculation is over I notice some interesting flow phenomena between 10000 and 15000 time steps and I would therefore like to be able to restart from the 10000th time step and increase the amount of output files saved only for a short interval of time. Is this possible without having to also re-run the first 10000 time steps before the interesting phenomena started? Or would I have to split the calculation into smaller intervals of ntmabs to achieve this?

I apologise if this is a basic question, but it would be very useful for me to understand the best procedure as, so far, I've only managed to restart a calculation from the end of a previous one.

Thank you,

Kristin

Yvan Fournier
Posts: 2871
Joined: Mon Feb 20, 2012 3:25 pm

Re: Restart save frequency

Post by Yvan Fournier » Mon Oct 15, 2012 7:40 pm

Hello,

Currently, new checkpoints overwrite previous checkpoints, to limit generated output size. We could maybe add an option for generating a series of checkpoints instead.

Also, in version 3.0, there will be a possibilty using a "control_file" (sucessor to the current "ficstp") to force a checkpoint at a specific time step, or change the frequency, without stopping the calculation. This would still overwrite existing files, but if you notice something interesting, nothing prevents you from moving checkpoint to checkpoint.<id> manually (as long as you have sufficient time between checkpoints).

Regards,

Yvan

MartianDuan
Posts: 74
Joined: Fri Aug 02, 2013 4:26 pm

Re: Restart save frequency

Post by MartianDuan » Mon Dec 08, 2014 4:53 pm

Hi Yvan,

I had check the manual, but it did not offer some exemple. Can the lines in 'control_file' be simply like:

max_time_step <14000>
checkpoint_time_step <400>

The file is attached. the extention '.log' is to make the file can be uploaded.
Thanks very much.

Yu
Yvan Fournier wrote:Hello,

Currently, new checkpoints overwrite previous checkpoints, to limit generated output size. We could maybe add an option for generating a series of checkpoints instead.

Also, in version 3.0, there will be a possibilty using a "control_file" (sucessor to the current "ficstp") to force a checkpoint at a specific time step, or change the frequency, without stopping the calculation. This would still overwrite existing files, but if you notice something interesting, nothing prevents you from moving checkpoint to checkpoint.<id> manually (as long as you have sufficient time between checkpoints).

Regards,

Yvan
Attachments
control_file.log
(49 Bytes) Downloaded 57 times

Yvan Fournier
Posts: 2871
Joined: Mon Feb 20, 2012 3:25 pm

Re: Restart save frequency

Post by Yvan Fournier » Tue Dec 16, 2014 11:23 am

Hello,

Sorry, I missed the "<" and ">" characters in your syntax, which need to be removed.

In many computing documentations (such as Unix/Linux man pages), we use "<", ">" to enclose mandatory arguments, and "[", "]" for optional arguments. We use this same convention.

Your syntax should be:
max_time_step 14000
checkpoint_time_step 400
Regards,

Yvan

Post Reply