I am out this week, so I am not following this thread closely, but it's not clear from the description of your hardware that you are ever going to get an improvement from parallel rendering.
You said you have 8 cores. Are these 8 cores in the same machine? If so, I don't think running in parallel will help with your rendering. Probably all 8 instances of MPI are rendering with the same GPU, which means all the rendering instructions from the parallel job fight with each other to get queued up on the same GPU, which will make the overall rendering slower, not faster. Somewhere in the thread I thought I read that you have 2 GPUs. Unless you have set up something special, one of them is probably being ignored. Even if you go through the pain of managing to set up a parallel ParaView to use both GPUs, I doubt you will get much if any speed up. The parallel rendering, which was designed for cluster computing, has an overhead; buffers are read back, pixels are transferred, and blending operations are performed. You are unlikely to gain much of that back with only 2 GPUs. -Ken Sent from my iPad so blame autocorrect. On Dec 27, 2016, at 12:36 PM, Chuck Atkins <[email protected]<mailto:[email protected]>> wrote: In my case, the dataset consists only of random numbers and it's in a single partition. I will try to generate a dataset in parallel in a few days and render it as you told me. 5M point's is actually a pretty small dataset that should be easy to handle from a rendering standpoint, i.e. you should be fine handling it on a single node. Most of your time is likely spent in I/O and data conversion rather than rendering. How are you getting your data into ParaView, i.e. what file format are you using? There is likely a better way that my address most of the performance issues you are having and remove the need for parallel rendering entirely. Just for some context, I generated 5M random points in a 100x100x100 cube with a single field data assigned to each point. CSV File: * File size: 177M * Read file: 29 sec * Convert to points: 0.1 sec VTK XML (binary encoding of data arrays): * File Size: 130M * Read File: 1.4 sec. In both cases, once the file is read, the first frame render time with the basic on-board Intel graphics card on my laptop is only 0.15 sec with the subsequent frame render times being 0.04 sec. You can see from this that the bulk of the time is spent in file I/O and data conversion, especialy with using something like CSV. This was with ParaView 5.2. With 4.4, the I/O times are the same but the first-frame render time is much slower: 0.6 sec. Still much closer to a usable frame rate. than 100 sec. - Chuck _______________________________________________ Powered by www.kitware.com<http://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 Search the list archives at: http://markmail.org/search/?q=ParaView Follow this link to subscribe/unsubscribe: http://public.kitware.com/mailman/listinfo/paraview
_______________________________________________ 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 Search the list archives at: http://markmail.org/search/?q=ParaView Follow this link to subscribe/unsubscribe: http://public.kitware.com/mailman/listinfo/paraview
