<br>OK. Now it makes sense.<br><br><div><span class="gmail_quote"><br></span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div style="direction: ltr;">
<div style="direction: ltr;"><span class="q"><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div style="direction: ltr;">1. You said if you open the same file and isosurface on 1 processor, it is faster, right?
</div></blockquote></span></div><div style="direction: ltr;"><div><br>reading the image (a legacy vtk file saved from a tiff in ImageJ) when running in MPI is very slow, about 5 times slower. again, i still have no idea how to set up paraview to read a tiff directly from the disk... i am first converting to vtk using a plugin for ImageJ.
</div></div></div></blockquote><div><br><br>It is slower because the legacy reader is not capable of reading a piece of the data. So what happens is that each process is reading the whole file and throwing away that part that is not needed. So you are hitting the same disk with 4 simultaneous reads of the same file. Try this. After opening the file, save it as a pvti file. Load this file. Is it faster? Even this might be slower than reading with 1 processor since you are reading from the same disk and causing contention.
<br></div><br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div style="direction: ltr;"><div style="direction: ltr;"><div> rendering the isosurface is not faster in single cpu mode, but is also not noticeably faster in mpi mode. seems like the slowest part is the PainterPolyDataMapper, which only usea a single CPU even if running in MPI.
</div></div></div></blockquote><div><br><br>This is also expected. You are rendering locally, meaning you are rendering everything on the first node. This should be as slow (or as fast) as running on 1 node. If you want to make use of parallel rendering, you have to enable compositing. Of course, you have to running paraview on a real cluster with graphics cards for this to work.
<br></div><br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div style="direction: ltr;"><div style="direction: ltr;"><span class="q"><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div style="direction: ltr;">2. Please verify that you are rendering locally by checking off the Composite flag in 3D View Properties.
</div></blockquote></span></div><div style="direction: ltr;"><div><br>yes, this is true... i am rendering locally. perhaps also of interest, the Advanced Render Parameters have some funny behavior: if "Use Immediate Mode Rendering" is on, and "Use Triangle Strips" is off, when I click to disable Immediate Mode, Use Triangle is automatically activated. Perhaps that is intentional, but thought if not, it might be of interest.
</div></div></div></blockquote><div><br><br>This is the intended behavior. Rendering using display lists without triangle strips is potentially dangerous because the graphics card might run out of memory. <br></div><br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div style="direction: ltr;"><div style="direction: ltr;"><span class="q"><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div style="direction: ltr;">
3. When you load the image and you only see the wireframe outline, does it render slow?
</div></blockquote></span></div><div style="direction: ltr;"><div><br>it actually does not show a wireframe outline. i do not see anything in the axes until i render an isosurface. i see the box that defines the limits of the dataset, but i see nothing else until i actually render the isosurface.
</div></div></div></blockquote><div><br><br>That's what I meant, the box. <br></div><br><br></div><br>