Scientific Theme of the PASI
The focus of this Advanced Studies Institute is on scientific discovery by means of high-performance computing, HPC, using the latest many-core computer hardware, in particular graphics processors, or GPUs.
The use of GPUs for scientific computing has exploded in the last couple of years, as researchers realize that they can improve performance of their applications by 10 or 100 times. On the other hand, the technology has been developed with large amounts of R&D funding sustained by the market for video gaming hardware, and thus it is commodity technology available at much lower prices than traditional HPC hardware. Add to this a considerable improvement in power efficiency, compared with traditional HPC, and we have an enormous opportunity to satisfy the demand of scientists across the world for HPC capability.
An unexpected benefit is that this technology presents a special opportunity for scientists at institutions without large installed computer infrastructure. In particular, in medium-income developing countries, where there is an ample supply of highly trained scientists and engineers, but a lack of large-investment resources for research, there is a huge opportunity for increasing the scientific capacity by advanced training of researchers on the topics of this PASI.
This PASI provided bleeding-edge training to graduate students and postdoctoral scholars in the use of the modern hardware for HPC. It also dedicated a substantial amount of time to detailed discussions of fundamental algorithms for massively parallel computational applications. Another area of focus was familiarizing participants with the most recent developments in open source libraries and programming tools—examples of these include: PyCuda, OpenCurrent— and well-established parallel libraries, such as PETSc.
Areas of Application
All of the above were focused around a handful of particularly relevant application areas. These are scientific areas where high-impact advances are expected from scientific computing, accelerated by the new and emerging hardware architectures, including:
- Computational Physics: quantum chromodynamics; high-performance lattice field theory.
- Computational Fluid Dynamics: incompressible flows; flow in porous media; variable-viscosity flows.
- Computational Earth & Environmental Science: magma dynamics; carbon sequestration.
- Computational Biomolecular Science: simulation of large biomolecules.
The institute included talks, tutorials and hands-on sessions in a computer lab. The times allocated are as follows:
- Lectures: 12 sets of lectures, consisting of up to 5 one-hour slots each, for a total of 40h of instruction. There were also 2 keynote presentations.
- Tutorials: several tutorial sessions of 1.5–3h duration each.
- Laboratory: 3 or 4 open lab sessions for a total of 12h of hands-on instruction
- Discussion: several roundtable discussions.