run getdp backstage

[Christophe Geuzaine]

Lin Ji wrote:
> 
> Dear Christophe,
>     I want to get back to a discussion we had earlier this month about
> post-processing data format and the -split flag. Sometimes, I do need to store
> the result at so many time steps e.g. in the case of high frequency source. If
> the -split flag only makes the result stored in seperate files for each time
> step I want to record, there would be too many files (e.g. 2000 in one of my
> example). 

But what is the problem of having 2000 files? You may simply put them in
a separate directory by using the '-name' option, e.g. getdp test -msh
test.msh -name out/test -solve... I often play with several thousands of
files, and, at least under Unix, I never had any problem.

> In fact, the data file .res is so big (2.1G) that my current
> computer (512 M) can not postprocessing it. So, maybe you can update the
> program so that the user can ask how many files they want to split. Better
> yet, the result is still stored in one big file. But, the program can read the
> result piece by piece and postprocessing one piece at a time.

If you split the files, you can choose wich time step you want to
analyse, and do exactly what you want (I think). For example, let's say
you have your 2000 result files stored in out/test-*.res. If you want to
post-process the time steps 1000 to 1100 by steps of 5, you simply do

getdp test -res out/test-01[01]?[05].res -pos ...

Tell me if you need more info on this: but I think it's really simple
this way to treat extremely big results. The same applies for the
'-restart' flag : using separate files, you can simply give getdp the
few preceding solutions (2 in your case) to read before restarting the
computation, i.e.

getdp test -res out/test-01999.res out/test-02000.res -restart


Christophe


-- 
Christophe Geuzaine

Tel: 32 (0) 4 366 37 10    http://geuz.org
Fax: 32 (0) 4 366 29 10    mailto:Christophe.Geuzaine at ulg.ac.be