The RapidMind Development Platform for Cell, GP-GPU, and Multicore CPUs

Massively multi-core processors such as GPUs and the Cell BE have the potential to deliver high performance computation to many applications. However, these processors require parallel programming on several levels, use some novel programming models, and native code written for one will not execute on the other. The RapidMind development platform enables portable access to the power of these processors. It provides a uniform, simple, safe, data-parallel programming model and takes care of most of the low-level details of mapping programs to each hardware target. It combines a dynamic compiler with a runtime streaming execution manager, and provides a single system image for computations running on any number of cores. The interface to this system is embedded inside ISO standard C++, where it can capture arbitrary computations specified at runtime. The use of a dynamic compiler means that high-level C++ abstractions can be used without sacrificing performance, while maintaining portability among current--and future--processors. This seminar will introduce and demonstrate the use of the RapidMind platform on GPUs (for both visualization and general-purpose computation), the Cell BE (where it runs on both IBM blades and the PS3) and on multicore CPU (x86 processor targets are under development). Comparative performance results will be presented for the GPU, the Cell BE, and multicore CPUs. On GPUs up to a 30x speedup over tuned CPU code has been achieved. On the Cell, we can match or exceed the performance of vendor tools in tuned applications, but with much less developer effort. On CPUs, we have doubled performance over native tools (i.e. 8x speedup on a 4-core relative to icc on one). About the speaker Michael McCool is an Associate Professor in the School of Computer Science at the University of Waterloo and co-founder and Chief Scientist of RapidMind Inc. Prof. McCool's current research efforts are targeted at enabling high-performance parallel applications by the development of advanced programming technologies. Research interests include interval analysis, Monte Carlo and quasi-Monte Carlo numerical methods, optimization, simulation, sampling, cellular automata, real-time computer graphics, vision, image processing, hardware design, and programming languages and development platforms. He has degrees in both Computer Engineering (B.A.Sc. with Math option, Waterloo, 1989, Sir Sandford Medal) and Computer Science (M.Sc. in 1991 and Ph.D. in 1995, Toronto). Organiser(s): Miguel Angel Marquina Computing Seminars / IT Department