[Gate-users] Query - multi-thread with with shared memory

[Maikol Salas Ramirez]

Dear Gate users,

I sent this email (below) some weeks ago, I would like to try again to see
if someone has experience.
I use HTCondor for parallel simulation, but in the normal HTCondor
configuration the simulation geometry is charged in each CPU, now this is
not a problem when the geometry demands less than 1GB of memory, but for
more complex geometries (4-10GB memory) mean that I need for a 8CPU
computer between 32-80GB RAM memory, it is difficult to have this memory
(only google cloud or amazon with high memory computers).

I explored the "partitionable and dynamic slots" configuration in HTCondor
but I am facing some problem.

I would like if someone has an idea how to deal with something like that.

I know that the best way is to simplify the geometry, but it is my last
resource in this case.

Best regards

2018-04-20 12:39 GMT+02:00 Maikol Salas Ramirez <mmsalas at gmail.com>:

> Dear Gate users,
>
> I am dealing with complex geometry, and it demands a big memory to run. I
> would like to ask, if it is possible to configure HTcondor or any
> other batch queuing system to do multi-thread with with shared memory.
> It mean share the geometry located in the memory.
>
> I found the GEANT4 has some feature in this direccion:
> https://twiki.cern.ch/twiki/bin/view/Geant4/Geant4MTAdvandedTopicsForAppli
> cationDevelopers
>
> But I am not sure if Gate get this.
>
> I appreciate any help.
>
> Best regards
> Maikol
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opengatecollaboration.org/pipermail/gate-users/attachments/20180503/e9eb5fdc/attachment.html>