This week the CUDA Developer Zone features an interview with Monica Syal, an Aerospace Engineer at Advanced Rotorcraft Technology (ART). Syal is working on the development of a real-time rotorcraft brownout simulation facility for flight simulator applications. Brownout dust clouds develop because of rotor downwash flow, which impinges upon the ground and uplifts dust particles. The underlying physics is basically a dual-phase fluids problem, one fluid phase being the air and the other being the dust.
The simulation of the individual particle motions in the dust clouds is equivalent to an N-body problem, where the number of bodies (or particles) is very large, in this case of the order of 1014. We are using several techniques to expedite such simulations, some of these being smart algorithms (e.g., fast multipole methods), particle clustering algorithms, and high-performance parallel computing techniques. An obvious way to achieve the needed computational accelerations by using parallel computing is to conduct the simulations using as many computing resources as possible. The number of cores used in a CPU is relatively few and the CPUs are optimized for serial processing. On the other hand, a GPU consists of hundreds of cores, which can be used to parallelize the computationally intensive parts of the simulations. Therefore, we decided to use high-end Tesla GPUs to conduct these simulations. This has provided us with about two orders of magnitude speedup in the computational time compared to the serial execution of the code.
This research will help enhance flight safety and reduce the large number of brownout related accidents that occur in both military and civil rotorcraft flight operations. Read the Full Story.
