Sequoia Processing on All Cores
In January it was announced that the Sequoia supercomputer at Lawrence Livermore National Laboratory (LLNL) successfully ran a simulation on one million of its computing cores. Now the second fastest supercomputer in the world has unleashed all 1,572,864 of its cores on a single model concerning the 'fast ignition' method of nuclear fusion.
Utilizing multiple cores in any machine can be quite challenging as what are normally minor inefficiencies and bugs can explode in size to the point of crashing the system. After more than a decade of work though, researchers at LLNL, the University of California, Los Angeles and Portugal's Instituto Superior Tecnico had successfully crafted the particle-in-cell simulation to be able to run on as many cores as can be thrown at it. As the number of cores increase the researchers found code's efficiency to be 75%, but as the simulation grew in size it was 97% efficient. That means that while a 4000 core computer cluster would take a year to run the same simulation that Sequoia could complete in a day, a problem 400 times larger would finish in the same amount of time.
The simulation modeled the interactions between tens of billion to trillions of particles within a plasma similar to what is created for fast ignition nuclear fusion. This method of fusion is similar to others which use lasers to heat and compress their fuel, but adds an ultrafast and powerful laser to actually ignite the fuel.