This is a quick hack to run a POVray demo on Nordugrid.
You need (of course) to be able to run jobs on Nordugrid, and you need Python (2.* works for sure, earlier versions might work) with the Python Imaging Library installed.
You need the following files:
monolith.pov
The POVray model file.
front.png
An image of a cluster node, included by the model.
runpov.sh
A wrapper script to start POVray on a computation node.
mkjobs.py
A script to submit the rendering jobs. No error checking is
performed.
fetch.py
A script to fetch the output files when all jobs are finished. An
ngget command for each job is started concurrently. No error
checking is performed.
merge.py
A script to paste the partial images rendered by each job into a
composite image, called monolith.png. Run this in the same
directory as fetch.py.
The latter is actually the most complicated script of the three, since the partial images generated by POVray are not well-formed. If POVray is instructed, for example, to render 50 horizontal lines of a 1200 pixel tall scene in PNG format, the output file is a PNG file with a header claiming the image is 1200 pixels tall, but only containing image data for 50 lines. This causes most image handling tools to refuse to open the image, so the script needs to rewrite the binary header chunk of each file before they can be pasted together using the Imaging library. Sigh.
fetch.py and merge.py can be run repeatedly while jobs are still running. Any partial images that have not yet been rendered will turn up as black stripes in the composite image.
For your convenience, a tar archive containing the above files is available.
