[Paraview-developers] [EXTERNAL] Paraview is almost wonderful (not dreadful as I implied previously)

[Scott, W Alan]

John,
One thing you may want to try is the new, nifty, slick view/Memory Inspector.  From my experience running large data in parallel, often the issue is that one node just runs out of memory.  This screen shows the highest memory used on any of the nodes.  From my viewpoint, this command is much more useful and accurate than the View/ Statistics Inspector.

Alan

-----Original Message-----
From: paraview-developers-bounces at paraview.org [mailto:paraview-developers-bounces at paraview.org] On Behalf Of Biddiscombe, John A.
Sent: Monday, February 06, 2012 9:49 AM
To: Sebastien Jourdain
Cc: paraview-developers at paraview.org
Subject: [EXTERNAL] [Paraview-developers] Paraview is almost wonderful (not dreadful as I implied previously)

Sebastien (Also Utkarsh who emailed me off list)

I would like to point out that after I patched the realloc problem, (and disabled the same call in trilinos, which between you and me was a real pain in the bum to find) - which strictly speaking is a libc/kernel problem on our linux installations not really a paraview bug - everything has been much much better. I was able to load a large dataset and generate an animation.

in fact another  job just completed on the cluster and I am now viewing one timestep from a half a billion particle dataset !!! hoorah. At last. 

So despite my constant wining (and trust me, paraview was crashing constantly on our cluster), it is now wonderful and I will recommend it to all my friends (both of them).

Paraview still crashes randomly from time to time, but the frequency is very much reduced now, once I have the state set up, I avoid clicking anything and am able to generate some animations. Users generally report to me that paraview crashes all the time, my own experience is that it crashes all the time, and the fact that I have been seriously considering using visit instead, tells me something...

My apologies for the slightly over-the-top title of my earlier post. 

JB
PS. Paraview just crashed on me again and I lost all my particles. segfault. <sigh>. Back to totalview again. 


-----Original Message-----
From: Sebastien Jourdain [mailto:sebastien.jourdain at kitware.com]
Sent: 06 February 2012 16:50
To: Biddiscombe, John A.
Cc: paraview-developers at paraview.org
Subject: Re: [Paraview-developers] paraview completely unusable (as usual)

Hi John,

thanks for reporting the issue, we'll look into that during the day.

Seb

On Mon, Feb 6, 2012 at 6:08 AM, Biddiscombe, John A. <biddisco at cscs.ch> wrote:
> The reason is because of the commit below. It seems like it fails to work after N calls on most machines. I fixed this a long time ago on my local branches, but forgot to reapply the fix when I switched to the standard paraview release branch (and master last week). After tracking that down, I found that trilinos also uses the dodgy realloc call and this was bringing down my plugins with segfaults at random and hard to reproduce times.
>
> Can I suggest a CMake option to disable this realloc call.
>
> ta
>
> JB
>
> commit c869c3d5875f503e757b64f2fd1ec349aee859bf
> Author: Sebastien Jourdain <sebastien.jourdain at kitware.com>
> Date:   Mon Jun 28 15:02:21 2010 -0400
>
>    ENH: Change the behaviour of DataArray.Resize(newSize)
>
>    Now, if the newSize is bigger than the current size, the call of
> Resize(newSize) will produce an array of that exact newsize and not 
> big
>
>
> --
> John Biddiscombe,                            email:biddisco @ cscs.ch 
> http://www.cscs.ch/ CSCS, Swiss National Supercomputing Centre  | Tel:
> +41 (91) 610.82.07 Via Cantonale, 6928 Manno, Switzerland      | Fax:
> +41 (91) 610.82.82
>
>
> _______________________________________________
> Paraview-developers mailing list
> Paraview-developers at paraview.org
> http://public.kitware.com/mailman/listinfo/paraview-developers
_______________________________________________
Paraview-developers mailing list
Paraview-developers at paraview.org
http://public.kitware.com/mailman/listinfo/paraview-developers