rtcontrib results (OSX)

Hi,

I'm getting strange results from rtcontrib on OSX. Firstly, the images combined with dctimestep are very speckly, and many of the patch images are very speckly as well, some with only a handful of dots. Do I need to direct rays from the sky patch directly to the windows and ignore all other directions? How can I do that? I'm unsure which parameters affect this most.

The other issue is that there are strange colours in different rays. The combined result has bands of colour whereas the model has almost no colour in it at all. The materials are all grey. What might cause this? Command sequence and image below.

vwrays -ff -vf views/EIn.vf -x 800 -y 800 | rtcontrib -i -ab 4 -ds 0.05 -dj .7 -ad 4096 -lw 2e-4 -n 2 -ffc $(vwrays -d -vf views/EIn.vf -x 800 -y 800) -e MF:2 -f reinhart.cal -b rbin -V+ -fo -o images/patchesEI/p%03d.hdr -m sky_glow -w octrees/top1.oct
  
Third issue: I've been running this successfully, but sometimes when I run it, the patch images are created, but then nothing seems to happen. There is one rtcontrib process and four rtrace processes visible in the activity monitor, but none of them are using any of the CPUs and the patch images are stuck on 4KB. Normally when I run it I get two rtrace processes using about 95% of the CPUs. (This is on a MacBook Pro with two cores). I cntrl-C out and then try again and get the same thing.

Thanks in advance.

John.

Hi John,

Running rtcontrib is a bit tricky. Thanks for sticking with it...

I notice a couple of things in your command line. First, you are using the -V+ option, when I think you might want -V-. The latter is the default (coefficients rather than partial values), and more appropriate in a DC calculation. You might get absolute errors otherwise if your sky isn't normalized. Second, you forgot to include the -bn option, which is very important. This might be behind your stalls as well as the unexpected colors. Using reinhart.cal, you are provided with the constant "Nrbins" for this setting, which is 578 for MF:2. Here is how I would modify your command:

vwrays -ff -vf views/EIn.vf -x 800 -y 800 | rtcontrib -i -ab 4 -ds 0.05 -dj .7 -ad 4096 -lw 2e-4 -n 2 -ffc $(vwrays -d -vf views/EIn.vf -x 800 -y 800) -e MF:2 -f reinhart.cal -b rbin -bn Nrbins -V- -fo -o images/patchesEI/p%03d.hdr -m sky_glow -w octrees/top1.oct

The stalls might also be caused by running up against your open file limit. You can try running "ulimit -n 2500" and seeing if that works before starting rtcontrib. The default limit on open files is 256 on most systems.

Good luck!
-Greg

···

From: John Ford <[email protected]>
Date: October 25, 2011 11:13:52 PM PDT

Hi,

I'm getting strange results from rtcontrib on OSX. Firstly, the images combined with dctimestep are very speckly, and many of the patch images are very speckly as well, some with only a handful of dots. Do I need to direct rays from the sky patch directly to the windows and ignore all other directions? How can I do that? I'm unsure which parameters affect this most.

The other issue is that there are strange colours in different rays. The combined result has bands of colour whereas the model has almost no colour in it at all. The materials are all grey. What might cause this? Command sequence and image below.

vwrays -ff -vf views/EIn.vf -x 800 -y 800 | rtcontrib -i -ab 4 -ds 0.05 -dj .7 -ad 4096 -lw 2e-4 -n 2 -ffc $(vwrays -d -vf views/EIn.vf -x 800 -y 800) -e MF:2 -f reinhart.cal -b rbin -V+ -fo -o images/patchesEI/p%03d.hdr -m sky_glow -w octrees/top1.oct
  
Third issue: I've been running this successfully, but sometimes when I run it, the patch images are created, but then nothing seems to happen. There is one rtcontrib process and four rtrace processes visible in the activity monitor, but none of them are using any of the CPUs and the patch images are stuck on 4KB. Normally when I run it I get two rtrace processes using about 95% of the CPUs. (This is on a MacBook Pro with two cores). I cntrl-C out and then try again and get the same thing.

Thanks in advance.

John.