Johannes Pekkilä


Johannes at JSC
Johannes Pekkilä is a Master student at the Aalto University School of Science in Finland where he studies Computer Science. His academic interests are parallel computing, calculations on graphic processing units and computational physics (he even did his Bachelor’s on astrophysical fluid simulation). This year he will be spending his summer on particles simulations in the Supercomputer Center in Juelich.
Simulating the movement of electrically charged particles is a very computationally-demanding task. The naive solution for simulating those particles has quadratic complexity. That means that if the simulation of 10 particles takes an hour then the simulation of 20 particles takes 4 hours! In practice, one normally uses a more sophisticated method for the simulation, such as the fast multipole method (FMM), which lowers the time complexity to a linear one. Compared to a sequential solution, the performance of this method can be boosted further by using parallel programming techniques. One such a novel technique is writing programs which use computational abilities of graphics cards. And that is what Johannes’ project is about.
Johannes will combine his interests in physics and GPU computation and, hopefully, he will create an amazingly fast particle simulator. He is very excited to be at the Juelich Supercomputing Center, where he will be able to use (at least few of the) 150 state of the art nVidia K80 graphics cards, which offer insane speedups in problems which can be solved in a parallel fashion.
Besides GPUs and HPC, he also likes cycling, gaming and flight simulators. He also has a small telescope, which he likes to point at the sky at night. He even has seen some cool stuff from Jupiter and the Galilean moons to the M81 galaxy.
[…] personal computation machine boots up and a Git login prompt appears. Let’s see, username: johannes.pekkilae… I peer into the CUDA FMM repositry and the source code is jumbled with loop unrolls, […]