[Paraview] PV parallel mode loading

David E DeMarle dave.demarle at kitware.com
Thu Mar 25 15:40:12 EDT 2010


>> If so, and you can load the dataset on your personal computer in one
>> process, try -no 1. Extra processes add overhead. If the reader isn't
>
> Ok I do know that, I am not a programming newbie! lol
> But the loading time is about to 32 x the loading time on a single process,
> and this is not a normal behaviour.

Since you have 32 processes all reading the same file off of the same
filesystem it is likely to be at least 32 x slower.

> the overhead you are talking about should just be due to the fact there is a
> need to establish proper connections between data subsets so as to be able
> to properly track a streamline integration for instance, but this should not
> lead to such an overhead in loading the data I think.

There are _many_ potential sources of overhead in parallel. In your
case I am guessing the bottlenecks are IO to the filesystem, then RAM
contention, then way down at the bottom of the list, inteprocess
communication and finally computational load.

>
>> a parallel capable one, for example, you are temporarily loading
>> 32*133 million points.
>
> This is a .pvd file, might there be a problem?
>

If the pvd file references a single data file yes. You still have 32
processes hitting the same disk and RAM simultanously. If the pvd file
references 32 different "striped" files, the file IO will be lower
(each process hits a different file and takes up only 1/32 of the
whole dataset size) but still substantial.

Try fewer processes and let us know what happens. It would help to
know how many processors are on your machine and how much RAM.

>> David E DeMarle
>> Kitware, Inc.
>> R&D Engineer
>> 28 Corporate Drive
>> Clifton Park, NY 12065-8662
>> Phone: 518-371-3971 x109
>>
>>
>>
>> On Thu, Mar 25, 2010 at 2:01 PM, Fred Fred <stan1313 at hotmail.fr> wrote:
>> > I have a 133 millions points and 134 millions cells regular grid
>> > dataset,
>> > which I can load in a few seconds on my personnal computer.
>> > Now I launch PV in parallel with -no 32 on my single computer and ... I
>> > have
>> > to wait about 4 minutes before I can clic on the Apply button, and once
>> > I
>> > have done it I have to wait additional 3 minutes to have the dataset
>> > loaded!
>> > Any explaination?
>> >
>> >
>> > ________________________________
>> > Télécharger en toute sécurité sur Internet ? La solution avec Internet
>> > Explorer 8
>> > _______________________________________________
>> > Powered by www.kitware.com
>> >
>> > Visit other Kitware open-source projects at
>> > http://www.kitware.com/opensource/opensource.html
>> >
>> > Please keep messages on-topic and check the ParaView Wiki at:
>> > http://paraview.org/Wiki/ParaView
>> >
>> > Follow this link to subscribe/unsubscribe:
>> > http://www.paraview.org/mailman/listinfo/paraview
>> >
>> >
>
> ________________________________
> Acheter en ligne en toute sécurité ? Internet Explorer 8 vous protège
> gratuitement !


More information about the ParaView mailing list