[Dune] Suppressing output from .dgf parser

Eike Mueller e.mueller at bath.ac.uk
Tue Oct 9 21:23:41 CEST 2012


Hi Andreas and Andreas,

that's right, I'm using AluGrid<2,3>, and I'm getting the

ALUGrid.................: version 1.50 (parallel)

line in the config.log of my project. I thought the way we set it up it's ok to use the 2d ALUGrid now? To answer Andreas' first question, I want to run one simulation on 1000+ cores, so each of them should work on part of the grid. I set GRIDTYPE = PARALLELGRID_ALUGRID_CUBE in the Makefile.am. I did some runs on my local machine with

mpirun - n 24 geometricmultigrid

and there I get the same problem (i.e. the message is printed out by each process). However, the output looks ok, i.e. I get 24 .vtu files, and each of them contains a part of the grid, so this seems to work. I call loadbalance() after setting up the grid, though. Haven't checked what happens if I do not do this.

On HECToR I run out of memory on 96 cores, but this could also be for other reasons, and the grid is not that big (6*4*4 cells on 96 processes), however, if each process reads in the entire grid, it is not surprising. 

Thanks,

Eike


On 9 Oct 2012, at 20:11, Dedner, Andreas wrote:

> Hi.
> The problem is a different one:
> From the output it seems that Eike is using AluGrid<2,3>. i.e., 
> the 2d alugrid version and that does not run in parallel. 
> So you are simply getting 1000+ processes running the same problem and all 
> having rank=0.
> Best
> Andreas
> ________________________________________
> From: dune-bounces+a.s.dedner=warwick.ac.uk at dune-project.org [dune-bounces+a.s.dedner=warwick.ac.uk at dune-project.org] on behalf of Andreas Lauser [andreas.lauser at iws.uni-stuttgart.de]
> Sent: 09 October 2012 18:40
> To: dune at dune-project.org
> Subject: Re: [Dune] Suppressing output from .dgf parser
> 
> Hi Eike,
> 
> Do you want to run 1000 sequential simulations, or do you want to run one
> simulation on 1000 cores? If the latter, please check that the summary says
> that ALUGrid is compiled for parallel usage, i.e. that the corrosponding
> summary line reads:
> 
> ALUGrid.................: version 1.50 (parallel)
> 
> (In my simulations, the message is only printed once even for parallel runs,
> so you probably did not compile the ALU library for parallel usage...)
> 
> cheers
>  Andreas
> 
> Am Dienstag, 9. Oktober 2012, 17:41:37 schrieb Eike Mueller:
>> Dear DUNE list,
>> 
>> in my code I'm reading in a .dgf file as follows:
>> 
>>   typedef Dune::GridSelector::GridType HostGrid;
>>   Dune::GridPtr<HostGrid>
>> gridPtr(grid_param.gridFilename(),MPI_COMM_WORLD); HostGrid& hostgrid =
>> *gridPtr;
>> 
>> If I do this in parallel, each process prints out the following message:
>> 
>> Created serial ALUGrid<2,3,cube,nonconforming> from macro grid file
>> './cubedsphere_2x2.dgf.
>> 
>> (and also complains that I didn't specify the tolerance parameter).
>> 
>> Can I somehow switch off these messages or make only the master process
>> print out this information?
>> 
>> When running with 1000+ processes these messages make my output files pretty
>> hard to read...
>> 
>> Thanks a lot,
>> 
>> Eike
> --
> Andreas Lauser
> Department of Hydromechanics and Modelling of Hydrosystems
> University of Stuttgart
> Pfaffenwaldring 61
> D-70569 Stuttgart
> Phone: (+49) 711 685-64719
> Fax: (+49) 711 685-60430
> www.hydrosys.uni-stuttgart.de
> 
> _______________________________________________
> Dune mailing list
> Dune at dune-project.org
> http://lists.dune-project.org/mailman/listinfo/dune
> 
> 
> 
> _______________________________________________
> Dune mailing list
> Dune at dune-project.org
> http://lists.dune-project.org/mailman/listinfo/dune





More information about the Dune mailing list