I am very pleasantly surprised that this got resolved so quickly (that it
got resolved at all!). I feel very happy to collaborate with so many awesome
people.
-berk
On Sat, Feb 6, 2010 at 3:29 PM, Bart Janssens wrote:
> On Saturday 06 February 2010 05:05:50 pm burlen wrote:
> > > Are you using t
On Saturday 06 February 2010 05:05:50 pm burlen wrote:
> > Are you using the stock 7.5.1 source code of Mesa?
>
> Yes, I was but I recently updated to Mesa 7.7.
>
> > Could you also post the compiler flags and
> > configure options you use for mesa?
>
> There are two ways to build Mesa. The a
Are you using the stock 7.5.1 source code of Mesa?
Yes, I was but I recently updated to Mesa 7.7.
Could you also post the compiler flags and
configure options you use for mesa?
There are two ways to build Mesa. The autotools way and the old way. By
autotools way I mean "./configure && make
On Saturday 06 February 2010 04:09:41 am you wrote:
> maybe I shouldn't have opened my mouth, mixing libraries doesn't
> necessarily explain the more than 4 procs thing... unless when you run
> with 4 procs they are being scheduled to the same node, and running with
> more causes another node to
maybe I shouldn't have opened my mouth, mixing libraries doesn't
necessarily explain the more than 4 procs thing... unless when you run
with 4 procs they are being scheduled to the same node, and running with
more causes another node to come into the mix.
How are process being assigned to node
Also, make sure you PV configuration uses the opengl and osmesa from
your build. To make it easy I always run ccmake the first time with
these options on the command line:
-DVTK_OPENGL_HAS_OSMESA=ON
-DOPENGL_INCLUDE_DIR=/u/burlen/apps/Mesa-7.5.1-Intel/include
-DOPENGL_gl_LIBRARY=/u/burlen/app
Getting the
exact same OSMesa setup as someone who has this working on many processors
could help to confirm this.
A possibility is that when run on the compute nodes open gl libraries
other than the one you expect are used. That can cause a strange crash.
I have seen it before. If you're n
On Friday 05 February 2010 07:48:45 pm Jeff Baumes wrote:
> Just a thought: Have you tried an entirely clean rebuild? Someone got
> an error similar to this before and it went away after a rebuild.
>
Nope, sorry, keeps happening after a full rebuild (even threw away the CMake
cache). One thing I
Just a thought: Have you tried an entirely clean rebuild? Someone got
an error similar to this before and it went away after a rebuild.
Jeff
On Fri, Feb 5, 2010 at 10:32 AM, Bart Janssens wrote:
> On Thursday 04 February 2010 09:33:24 pm you wrote:
>> What about Valgrind? Any chance you can star
On Thursday 04 February 2010 09:33:24 pm you wrote:
> What about Valgrind? Any chance you can start one of the processes such
> that you can run it in Valgrind?
>
I ran things in valgrind, but this showed exactly the same crash. I have
recompiled mesa with TLS, so the PTHREADS block in the ASM c
In that case, I agree with:
> Somehow, I get the feeling this is not going to be easy... :)
What about Valgrind? Any chance you can start one of the processes such that
you can run it in Valgrind?
-berk
On Thu, Feb 4, 2010 at 2:31 PM, Bart Janssens wrote:
> On Thursday 04 February 2010 07:51:5
On Thursday 04 February 2010 07:51:58 pm Berk Geveci wrote:
> Just a hunch but is there any chance you can try another version of
> OpenMPI? Hopefully a newer version?
>
Sorry, already tried that :) First attempt was with 1.3.2, as supplied by
CentOS 5.4, and now all nodes are using 1.4.1.
Chee
Just a hunch but is there any chance you can try another version of OpenMPI?
Hopefully a newer version?
-berk
On Thu, Feb 4, 2010 at 11:16 AM, Bart Janssens wrote:
> On Wednesday 03 February 2010 09:58:53 pm Berk Geveci wrote:
> > Any chance you can attach with a debugger to one of the nodes tha
On Wednesday 03 February 2010 09:58:53 pm Berk Geveci wrote:
> Any chance you can attach with a debugger to one of the nodes that
> crash? It is hard to tell what is wrong from the errors you posted. We
I have recompiled with debugging symbols and attached a backtrace from the
first pvserver proc
Hi Bart,
Any chance you can attach with a debugger to one of the nodes that crash? It
is hard to tell what is wrong from the errors you posted. We have definitely
run ParaView with OpenMPI with 4 core and more. Also, what do you mean by
"simple VTK file"? Small file? Big file?
-berk
On Tue, Feb
Hi all,
I've been trying to run pvserver using 8 processes. It accepts the connection,
but when loading a simple VTK file it crashes. Exactly the same line with -np 4
does work. The problem appears on both paraview SVN and 3.6.2. The machine has
8 cores, below is the full command and output:
m
16 matches
Mail list logo