NAIS Postdoctoral Researchers

Chris Fensch Informatics Forum 1.02 University of Edinburgh c.fensch@ed.ac.uk

My research interests revolve around the design of many-core architecture and its implications on programmability. In particular, I am interested in: * Patterns in parallel applications and exploiting these patterns in both software (programmability) and hardware (coherence mechanisms). * Memory consistency and coherence model for many-core architectures. * Transactional Memory and its usage in many-core architetures for consistency. * Interaction of hardware and software in many-core architectures. [1]Nick Barrow-Williams, Christian Fensch and Simon Moore, Proximity Coherence for Chip Multiprocessors, nternational Conference on Parallel Architectures and Compilation Techniques (PACT '10), pp. 123-134 (2010) [2]Nick Barrow-Williams, Christian Fensch and Simon Moore, A Communication Characterisation of Splash-2 and Parsec, IEEE International Symposium on Workload Characterization (IISWC '09), pp. 86-97 (2009)

Emad Noorizadeh University of Edinburgh Mayfield Road Edinburgh, EH9 3JZ e.noorizadeh@ed.ac.uk

Emad’s research is in the areas of molecular dynamics, mathematical modelling and stochastic processes. In particular developing sampling methods that enhance results for simulation of biomolecules or materials with minimal random perturbation to their dynamics [1,2]. Methods for computation of rare events in molecular systems and understanding transition for reactive processes. [1]B. Leimkuhler, E. Noorizadeh and O. Penrose: Comparing the efficiencies of stochastic isothermal molecular dynamics methods. Journal of Statistical Physics, 143, 5 (2011) [2]B.Leimkuhler, E. Noorizadeh and F. Theil: A gentle stochastic thermostat for molecular dynamics. Journal of Statistical Physics, 135, 2 (2009)

Richard Rankin Department of Mathematics and Statistics University of Strathclyde Glasgow richard.a.rankin@strath.ac.uk

I work on obtaining error estimators for finite element approximations of partial differential equations. Error estimators can be used to tell where in the finite element mesh the highest sources of error are so that more degrees of freedom can be placed in these areas. This allows far greater accuracy to be attained for a lot less computational work. That these estimators also provide guaranteed bounds on the error or a quantity of interest is important if one is to know whether a computation has given them sufficient accuracy or not, in which case the computation must be continued on a more refined mesh. Moreover, it is advantageous that the computation of these estimators can be parallelised. I use hybrid MPI/OpenMP to parallelise the computation of the estimators. [1]M. Ainsworth and R. Rankin, Fully computable error bounds for conforming and nonconforming finite element approximations in planar elasticity, Internat. J. Numer. Methods Engrg., vol. 82, no. 9, pp. 1114-1157 (2010) [2]M. Ainsworth and R. Rankin, Realistic computable error bounds for three dimensional finite element analyses in linear elasticity, Comput. Methods Appl. Mech. Engrg., vol. 200, no. 21-22, pp. 1909-1926 (2011) [3]M. Ainsworth, A. Allendes, G. R. Barrenechea and R. Rankin, Computable error bounds for nonconforming Fortin-Soulie finite element approximation of the Stokes problem, IMA J. Numer. Anal., (in press)

Wang Zheng IF1.02 Informatics Forum 10 Crichton Street Edinburgh zheng.wang@ed.ac.uk

I am interested in is using smart compiler analysis and predictive modelling techniques to map parallelism onto multi-cores. I am also interested in developing an automatic approach that can automatically adapts to the evolution of computer architectures. Recently we have developed innovative approaches to this problem using machine learning where it outperforms hand-craft approaches[1][2][3]. [1]Z. Wang and M. O'Boyle, Partitioning Streaming Parallelism for Multi-cores: A Machine Learning Based Approach, In 19th Intl. Conference on Parallel Architectures and Compilation Techniques (PACT), September, 2010, 307-318. [2]G. Tournavitis, Z. Wang, B. Franke, and M. O'Boyle, Towards a Holistic Approach to Auto-Parallelization - Integrating Profile-Driven Parallelism Detection and Machine-Learning Based Mapping, In ACM SIGPLAN 2009 Conference on Programming Language Design and Implementation (PLDI), June 2009, 177-187. [3] Z. Wang and M. O'Boyle, Mapping Parallelism to Multi-cores: A Machine Learning Based Approach, In 14th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming (PPoPP), February 2009, 75-84.