discovered_02_2015

WWW.HZDR.DE discovered 02.15 RESEARCH could be significantly detrimental to GPU computing power." So for optimal performance, the graphics processors must work as independently of each other as possible. The fact that thousands of cards work together processing the models poses a challenge: How could the slow data transfer between individual graphics cards be compensated for? The simple, but effective solution is: New calculations are initiated while the transfer to other cards is taking place. This ensures that waiting times are put to good use and all graphics cards are operating simultaneously. The flexibility of the code makes it suitable for use in the new Helmholtz extreme lab in Hamburg Heiko Burau is the inventor of PIConGPU: After winning the student competition "Jugend forscht", in 2009 he received the opportunity to research his topic "Graphic Card Programming" at the HZDR. After just six weeks, the then 17-year-old had already programmed the first version of the code for a single GPU in collaboration with other young researchers in the group. Bit by bit other students also began to participate and added on to PIConGPU until it ultimately became one of the most powerful codes for laser plasma physics that there is today. Whether it's the cosmic jets of astrophysics, the study of laser- driven fusion, or laser particle acceleration - now just about any calculation involving high-energy plasma can be carried out with this code. Data obtained this way can, for example, help to make accelerator systems for proton therapies used against cancer more compact and cost-effective. Thanks to its high level of flexibility, in the future the code will also include atomic physics models important for studying processes in plasmas at the new Helmholtz International TITAN: Currently the second fastest computer in the world, super computer "Titan" resides at the Oak Ridge National Laboratory in the US state of Tennessee. Photo: ORNL | U.S. Dept. of Energy Beamline for Extreme Fields (HIBEF) at the European XFEL in Hamburg. As part of this process considerably more parameters will be taken into account, right down to the quantum physics level. This does make the code run more slowly, but at the same time allows for simulations that wouldn't have been feasible before. Bussmann's dream is to use this with the analysis of experimental data: "That way one and the same code could pave the way from the model to experimental testing and onward to an improved model." The group leader is proud that his trust in the capabilities of his junior researchers has paid off: "It isn't unusual for me to be skeptically confronted at conferences with: 'With students? That would never work.' But it has worked - not despite, but rather because of the students." This fostering of outstanding junior scientists continues to this day: In September 2015, Daniel Grassinger – another winner of the "Jugend forscht" contest – worked as a new intern in the group. He developed a new process that can be used to read the data from different codes in the same open format, OpenPMD. As a result of his work the predictive capability of the HZDR code can now be compared with other plasma codes for the first time, thus working to improve them. External researchers are also very eager to see such an

Seitenübersicht