NAG Numerical Routines for GPUs (Beta)

NAG has been at the forefront of numerical algorithm provision for nearly four decades and during that time has delivered algorithms in many different guises addressing numerous application areas. One area where NAG is currently collaborating with Oxford University Professor Mike Giles, is the provision of routines optimized for Graphics Processing Units (GPUs).

GPUs have traditionally been used for 3D gaming acceleration on personal computers but are rapidly evolving for use in general purpose scientific computations. This type of computing has been termed ‘personal supercomputing’ (PSC).

Part of the collaborative research undertaken by Professor Giles and NAG experts has been the development of a set of fully functional parallelized numerical routines which, when used on a GPU architecture, result in significant performance gains compared to traditional CPU systems.

How can I gain access to NAG Numerical Routines for GPUs prototype?

The NAG Numerical Routines for GPUs prototype are available as a beta release. Currently, the prototype includes components for Monte Carlo simulation, comprising a pseudorandom number generator (L'Ecuyer, MRG32k3a), a quasi-random sequence generator (Sobol with extension to 50,000 dimensions) and a Brownian bridge constructor. The random number generators provide three output distributions (uniform, exponential and Normal) implemented in both single and double precisions.

Eligibility?

It is available free to academics subject to them signing a collaborative agreement. NAG is also keen to engage with commercial clients for assessment and further development of numerical routines for GPUs

To obtain a copy of the software please complete the online registration of interest form here or contact us directly for more information.

Ongoing work

NAG is continuing to develop further routines for GPUs including univariate and multi-variate distributions, copulas amongst others.

Acknowledgements

We would like to thank the Technology Strategy Board (TSB) and the Smith Institute for their support in sponsoring this project and EPSRC for supporting Professor Giles’ academic research