Rtcontrib-BSDF and DAYSIM-pmap: where are we heading?

Hi Lars,

Thanks for starting the conversation... It may work best if the originators/maintainers of the various software explain what they see as the strengths and weaknesses of what they've developed, and those who have used it can chime in with their experiences.

Rtcontrib is essentially an interface to rtrace that runs it in such a way that individual ray contributions can be tallied. Because there are so many ways one might want to do this, the command-line interface is quite complicated, as is interpreting the output. There is no interface directly to weather files or data, and a lot of other tools and scripts are needed to make it truly useful. That said, here is a quick list of strengths and weaknesses, with mention of related tools for rtcontrib:




o Records contributions or coefficients between source rays and scene surfaces.

o Separates the notion of ray-tracing from the notion of generating images or values.

o Can compute daylight coefficients or lighting contribution components.

o May also be used to track flux through optical systems.

o Output is binary or ascii data files or component/coefficient pictures.

o Use with genklemsamp, genskyvec, and dctimestep for annual simulations with BSDFs (illuminance values, animations, etc.)

o Use with pcomb for combining component images.


o Difficult to use in a stand-alone fashion.

o Uses Monte Carlo path tracing, which does not make use of irradiance caching.

o Results can be noisy if parameters are set improperly.

o Won't explicitly do forward ray-tracing, but with some effort it is possible to trace flux and create a candlepower distribution. (See 2006 Radiance workshop presentation below.)

o Less useful for single time-point calculations.

2005 Workshop talk: http://www.radiance-online.org/radiance-workshop4/cd/website/PDF/Ward_rtcontrib.pdf
2006 Workshop talk: http://www.radiance-online.org/radiance-workshop5/2006_Radiance_Workshop/Presentations/Ward_rtcontrib.mov
Axel's tutorial: http://luminance.londonmet.ac.uk/learnix/docs/rtcontrib_lesson.html

Comments welcome as always.


From: "Lars O. Grobe" <grobe@gmx.net>
Date: May 16, 2010 8:27:25 PM PDT


I may be wrong, but I see two different primary applications. As far as I understood, rtcontrib is optimized for operation with weather data. Especially patch-based sky models seam to be a perfect application. The photonmap is currently not all that efficient in these applications. On the other hand, it allows to include a forward-raytracing step without modifications in the scene (well, maybe adding photon-ports to optimize). Maybe it is because of I never did it, but modelling a facade with non-planar light shelves sounds scary to me to attempt using rtcontrib, while almost trivial using the photon-map. On the other hand, I admire the potential of rtcontrib when used with sky patches or a bunch of light sources. Maybe it is because there is no such compact documentation available for using rtcontrib with redirecting structures as the photon-map manual. So - from my perspective - it would be perfect to have both. Pmap to easily include a very generic forward raytracing step. Rtcontrib as an optimization for those who know what they are doing.

Cheers, Lars.