Towards Parallel Execution of Scientific Applications - Softcover

Kristensen, Mads Ruben Burgdorff

Softcover

ISBN 10: 384430195X ISBN 13: 9783844301953

Publisher: LAP LAMBERT Academic Publishing, 2012

Synopsis

Rapid prototyping of numerically expressed problems is essential for a broad range of research areas. Finding the solution for computational scientific and engineering problems often requires experimenting with various algorithms and different parameters using the feedback from several iterations. Therefore, being able to quickly prototype the solution is critical for a timely and successful scientific discovery. In this book, I have explored the possibility of seamlessly executing sequential scientific applications in parallel. The idea is to introduce implicit data parallelism in order to provide a high-productivity and high-performance framework. I introduce two new projects, DistNumPy and Bohrium, that strive to provide a high-performance back-end for Numerical Python (NumPy) without reducing the high-productivity of Python/NumPy. I present several performance studies that demonstrate good scalable performance on a variety of architectures: from a small Ethernet Linux cluster with 32 CPU-cores to the Cray XE-6 supercomputer Hopper with 1536 CPU-cores.

About this title

About the Author

Mads R. B. Kristensen is a Postdoc at the Niels Bohr Institute, University of Copenhagen, Denmark. His primary research area is High Performance Computing where he is focusing on parallel runtime systems that seamlessly utilize parallel architectures.

Edition details

• Publisher: LAP LAMBERT Academic Publishing
• Publication date: 2012
• ISBN 10: 384430195X
• ISBN 13: 9783844301953
• Binding: Paperback
• Language: English
• Number of pages: 204