I'm still very busy playing with ray tracing as there are just so many opportunities to optimize the performance. It also involves a lot of converting seemingly simple serial algorithms to parallel versions - sorts / searches etc. Couple that with a full time job and sometimes I just need a break. A break in my book is defined as "a different project" :p
Not that I'm stopping work on raytracing and rendering - far from it, in fact the alternatives I'm looking at include a lot of similar processes / algorithms. The three topics in the title are what I'm looking at being largely inspired by the great work done by the guys who attended the Tesla Personal Supercomputer launch last month.
The problem is that both Seismic Analysis and Tomography require very specialized data sets in order to do any meaningful work. Although it is probably possible to construct an artifical data set for either topic it will not give real world results. While at the Tesla lauch I spoke with Steve Purves from ffA. They do really impressive work with CUDA to analyse vast quantities of collected seismic data. It is really nice to see a commercial company adopt new technology so quickly. Understandably as the data costs a small fortune to obtain there is almost no set available for the hobbyist to play with. Steve did mention the possibility of an old sample set being available from a company in the US for about 600 (I haven't confirmed this yet or the type of license agreement they have in place)
Tomography was my next option but again data sets are hard to get hold of. Although I don't think the data is as expensive to obtain as seismic, it is medical images and protected by all sorts of privacy issues. Non-medical CT scans would also do nicely but as yet I have not found any available data sets (please drop me a note if you know of any).
So with the absence of sufficient data I have settled on playing with CFD whenever I need a break from ray tracing. My normal method of starting a new project is to sit with a pencil and paper and work out some mathematical principles and draw some pictures. I started off by looking at the lattice-Boltzman method which is pretty ideal for parallel processing. And although it clearly does produce good results as always I feel there is a better way. For now I'll stick to the lattice-Boltzman method and see how things progress on paper.
If anyone does know where to get sample data for seismic / tomography, please let me know.
No comments:
Post a Comment