|
Notes |
|
|
(0021489)
|
|
Koen van Mierlo
|
|
2010-07-27 04:44
|
|
|
I compiled paraview 3.8 with OSmesa 6.5.3 to check if this bug depends on the version of OSmesa. I get exactly the same output with OSmesa 6.5.3 so the version of OSmesa doesn't make a difference. |
|
|
|
(0021490)
|
|
Koen van Mierlo
|
|
2010-07-27 06:54
|
|
|
I have added output and error files from the 1 node (8 cores) and 2 nodes (16 cores) jobs. They seem to suggest that on the multiple node job the pvbatch process is killed as soon as 1 core has done the rendering and that it doesn't wait until all cores are done. |
|
|
|
(0022138)
|
|
Koen van Mierlo
|
|
2010-09-08 13:48
|
|
|
I thought that the bug could be caused by openmpi so I compiled a new version of openmpi (1.4.2). Unfortunately the bug is still there when I compile paraview 3.8.0 with osmesa 7.8.2 and openmpi 1.4.2 |
|
|
|
(0022252)
|
|
Alan Scott
|
|
2010-09-15 22:17
|
|
We are having problems with newer versions of Mesa. Please ask the users list (which is read by Kitware) for feedback on which version of Mesa to use. If this fixes your problem, please then mark this bug "resolved" with a comment stating what you did to fix it.
Alan |
|
|
|
(0022318)
|
|
Koen van Mierlo
|
|
2010-09-22 11:55
|
|
The paraview mailing list suggest to use either mesa 7.5.x or a development mesa from git. I don't know how to use git so instead I compiled osmesa 7.5.2 and the newest release candidate osmesa 7.8.3-rc1. Unfortunately the output is the same and the render error persists. I use the following command to configure / compile the osmesa library:
./configure --with-driver=osmesa --without-x --disable-glu --disable-egl --disable-gallium --prefix=/home/kvm/mesa752 --exec_prefix=/home/kvm/mesa752 |
|
|
|
(0022365)
|
|
thorenz
|
|
2010-09-27 10:05
|
|
Hello Koen,
I tried to get Paraview with Mesa to work on our cluster. After several attempts I got it work, here it works fine now with 96 processes in parallel.
Due to several dependencies, I can not use Mesa-7.5.x (or later), but have to use Mesa-7.2. I struggled a lot to get it to work, anyhow. The main clues for me were:
- NEVER go back in the ccmake process for Paraview. Never never never change an option to one value first and to another later. In that case, completely remove the contents from the build directory and start from scratch. Otherwise you can have strange side effects.
- Use different compilations for Mesa on the frontend and the nodes of the cluster (i.e. compile two versions of Mesa: With DRI (or whatever) for the frontend, without DRI, without X and with GLU for the nodes)
- Use two different compilations of Paraview for frontend and nodes: The first one with X, the second without X, without PARAVIEW_BUIL_QT_GUI, with OSMESA, without OPENGL_INCLUDE_DIR, without OPENGL_gl_LIBRARY, with OPENGL_glu_LIBRARY=qmesa-libglu
- Set LD_LIBRARY_PATH, so that the correct version of Mesa is loaded for frontend and nodes.
Hope it helps,
Carsten |
|
|
|
(0023752)
|
|
Koen van Mierlo
|
|
2010-12-07 09:20
|
|
|
Unfortunately this bug still exists with paraview 3.8.1 and OSmesa 7.9. Everything works fine when I submit a 1 node (8 core) pvbatch job to the cluster but the rendering goes wrong when I submit a 2 or more node job. |
|
|
|
(0037860)
|
|
Kitware Robot
|
|
2016-08-12 09:58
|
|
Resolving issue as `moved`.
This issue tracker is no longer used. Further discussion of this issue may take place in the current ParaView Issues page linked in the banner at the top of this page. |
|
4.png (398,606) 2010-07-26 15:34
CMakeCache.txt (119,549) 2010-07-26 15:35
Q38.py (1,948) 2010-07-26 15:36