[Carpet] Job ian2.23638.0 killed

Ryoji Takahashi ryoji at cct.lsu.edu
Tue May 30 18:14:14 CEST 2006


i activated CarpetLib::print_memstats_every=10?
but now, code was terminated during dumping of hdf5 ...

for this i am rquestion 16GBx8, concerning in mike ... i should be fit.


Erik Schnetter wrote:

> Ryoji,
>
> the emailed error messages indicate that you used more than 1.6 GB  
> per processor.  The job output says that this particular jobs was  
> killed after about 311 iterations.  There is nothing obvious that was  
> happening, or that should have been happening according to the  
> parameter file, at this time.
>
> Several things may be going on.  It can be that Carpet's memory usage  
> is not evenly distributed, e.g. due to I/O on the root processor, so  
> that there is still memory available while the root process uses more  
> than 1.6 GB.  It could also be that your simulation just uses too  
> much memory, and after some time e.g. the C libraries I/O buffers  
> grow to an extent to exceed this limit.
>
> We need to know more about the memory consumption of your job.  Can  
> you use "top" or a similar tool to monitor the job while it is  
> running?  Can you use the parameter CarpetLib::print_memstats_every=10?
>
> Does this happen with every job of yours?
>
> -erik
>
>------------------------------------------------------------------------
>
>_______________________________________________
>developers mailing list
>developers at lists.carpetcode.org
>http://lists.carpetcode.org/listinfo/developers
>  
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: /archives/developers/attachments/20060530/ac5d9612/attachment.htm 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: d3.0-256-0.05.out.0.tgz
Type: application/x-gzip
Size: 23655 bytes
Desc: not available
Url : /archives/developers/attachments/20060530/ac5d9612/attachment.bin 


More information about the developers mailing list