[Users] simfactory: memory request?

[Scott Hawley]

Thanks guys.  I was just concerned that I might need to explicitly request that, e.g. on a shared-memory machine like Nautilus.



On Aug 21, 2011, at 11:15 AM, Erik Schnetter wrote:

Ian

This functionality exists; the parameter Carpet::enable_no_storage provides it.

You could use the simulation factory to add this parameter to a
parameter file; simfactory calls this "appending" when you provide a
parameter file.

-erik

On Sat, Aug 20, 2011 at 2:00 PM, Ian Hinder <ian.hinder at aei.mpg.de<mailto:ian.hinder at aei.mpg.de>> wrote:

On 14 Aug 2011, at 05:41, Erik Schnetter wrote:

Scott

No, simfactory does not know that. Given Cactus's dynamic schedule,
and in particular given AMR, it is basically impossible to predict
this.

Simfactory does, however, know internally how much memory each system provides.

In my runs, the grid structure could be predicted based on the parameter file, as I don't regrid initially based on any computed data.  The memory usage is essentially constant within about 20% for the duration of the run.  I spend a lot of time trying different numbers of cores until the run fits, and it seem that this is something which could be estimated.  It would be very useful to be able to run Cactus on my workstation with a production parameter file, but have it not actually allocate the memory.  It could then tell me the total amount of memory that it would have allocated, and simfactory could use this to tell me how many cores to run on.  One could even imagine a future where this was done automatically, given a desired memory utilisation rate.  Is this really impossible with Carpet?  It that because of conceptual problems, or technical?  I think this would be a wonderful ability to have, and would save many people a large amount of time.




-erik

On Tue, Aug 9, 2011 at 5:02 PM, Scott Hawley <scott.hawley at belmont.edu<mailto:scott.hawley at belmont.edu>> wrote:

Does simfactory automatically know how much memory I'm going to request from the host machine?

I've looked in the Tutorial (perl) and the Advanced Tutorial, and haven't found such an option to set.

Thanks.


--
Scott H. Hawley, Ph.D.                  Asst. Prof. of Physics
Chemistry & Physics Dept                Office: Hitch 100D
Belmont University                      Tel:  +1-615-460-6206
Nashville, TN 37212 USA                 Fax: +1-615-460-5458
PGP Key at http://sks-keyservers.net



_______________________________________________
Users mailing list
Users at einsteintoolkit.org<mailto:Users at einsteintoolkit.org>
http://lists.einsteintoolkit.org/mailman/listinfo/users




--
Erik Schnetter <schnetter at cct.lsu.edu<mailto:schnetter at cct.lsu.edu>>   http://www.cct.lsu.edu/~eschnett/
_______________________________________________
Users mailing list
Users at einsteintoolkit.org<mailto:Users at einsteintoolkit.org>
http://lists.einsteintoolkit.org/mailman/listinfo/users

--
Ian Hinder
ian.hinder at aei.mpg.de<mailto:ian.hinder at aei.mpg.de>





--
Erik Schnetter <schnetter at cct.lsu.edu<mailto:schnetter at cct.lsu.edu>>   http://www.cct.lsu.edu/~eschnett/


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.einsteintoolkit.org/pipermail/users/attachments/20110823/a47b1da2/attachment.html