HPC on a Multi-core Laptop
Hui Lan Jiang · 2009
The acceleration of computer science is pushing multi-core technology into rapid development. Currently, processors with two or four cores have been commonly used during people’s daily life. And it is also believed that processors with even more cores will be ubiquitous in the near future. So how to efficiently utilize those powerful processors has become a major challenge. Although a lot of high-end HPC technologies such as OpenMP and MPI have been developed to implement parallel computing, which is one of the best methods to make use of multi-core processors, they are mostly designed to be used on the national supercomputers rather than on the desktop-level hardware. Therefore, this study will focus on how useful these high-end HPC technologies are on the desktop-level machine and try to find out the most efficient parallel method by analyzing the performance of a number of benchmarks with different parallel techniques and different operating systems. The comparison of performance of such benchmarks run in the desktop-level machine and in high performance computers will also be included. In this project, the desktop-level machine to be used is the Dell M6400 laptop with a Intel® CoreTM 2 Extreme Quad Core QX9300 processor inside, and the operating systems to be used are Windows Vista and Scientific Linux. For the high performance computer, we will use ness parallel computer which has two back-end X4600 SMP nodes and is based on AMD Opteron processors running Linux. In ness, each node contains eight dual-core processors.