In this video from the EuroMPI’19, Mathieu Luisier presents: Massively Parallel Computer Aided Design of Nano-Transistors: When Physics Lets You Down.
Moore’s scaling law is slowly coming to an end because of the difficulties to further decrease the dimensions of transistors, the active components of all integrated circuits, and because scaling does not provide as much performance benefit as before. In particular, managing heat dissipation at the nanometer scale has become one of the most critical challenges that the semiconductor industry is currently facing. In this context, physics-based computer aided design (CAD) tools have proved their usefulness at supporting experimental activities and at mitigating serious conceptual flaws. Nowadays, atomistic quantum mechanical simulators are needed for that purpose. They rely on the iterative solution of complex non-linear systems of equations, taking advantage of parallel programming models and modern HPC infrastructures. These are the keys to investigate always larger transistor structures with a continuously improving accuracy. In this talk, the capabilities of a state-of-the-art quantum mechanical device simulator will be briefly reviewed, insisting on the developed multi-level parallelization scheme and highlighting the fact that physics can lead to wrong assumptions regarding the best-suited distribution of the workload. This problem will be addressed through a data-centric transformation of the code.
Mathieu Luisier is Associate Professor of Computational Nanoelectronics at ETH Zurich, Switzerland. He received a degree in electrical engineering and a PhD in 2003 and 2007, respectively, both from ETH Zurich. After a one year post-doc at the same institution, he joined in 2008 the Network for Computational Nanotechnology at Purdue University, USA, as a research assistant professor. In 2011 he returned to ETH Zurich to become assistant professor. His current research interests focus on the modeling of nanoscale devices and on the development of advanced simulation tools. He won an honorable mention at the ACM Gordon Bell Prize for High Performance Computing in 2011. In 2013, he received a Starting Grant from the European Research Council (ERC).
