While all my colleagues are completely absorbed in the last fine tuning of their .ppt presentations (good luck guys!) for IAHR 2007 Conference in Venice, I spend my days (and nights) scanning the web in search of interesting stuff for you, the readers. Here’s what I found.
Did you know that your video card Graphics Processing Unit (GPU) has maybe more transistors, i.e. more flops, than your powerful CPU? [It depends, of course, but in average...] Well, now you know.
What *I* know, is the thought you just had: using the GPU for your super-heavy, three-day-long simulation.
Unfortunately, it is not so easy, due to a basic difference between CPUs and GPUs:
The CPU is capable of quickly processing all sorts of tasks whereas the GPU is capable of processing very quickly a certain type of task. For the latter, this has to be in the form of a problem composed of independent elements, because of the massive parallelization of GPUs. [Damien Triolet, www.behardware.com]
It means that it will be more difficult to program and compile for a GPU, but the benefits in performance are maybe worth an effort. A new market for GPU hardware vendors (namely, ATI/AMD and nVIDIA) is growing, and we, the simulation-addicted, will have to assess the new solutions that they will be selling. For a deeper look into the technical stuff take a look here and here.
According to this book table of contents (see chapter 47) they have already explored the case of “Flow Simulation with Complex Boundaries”, among other very intriguing and actual problems.
For the curious (and probably unsatisfied) reader, the last link: http://www.gpgpu.org/
Hydraulic modeling is out
The “dernier cri” is the global, social, economic, politic and military modeling; take a look here:
http://www.theregister.co.uk/2007/06/23/sentient_worlds/