Scientific computing


Scientific Computing & Many-Core Acceleration

High Performance Computing (HPC) is a key part of Scientific Computing.  With the growth of multicore CPUs & many-core GPUs (graphics processors) HPC is concerned with algorithms & implementation techniques for shared-memory and distributed-memory parallel computing.  We focus on the use of GPUs, and work across the University to exploit the benefits offered by them, both on single systems or the Emerald GPU system.

Energy-efficiency is also a major concern in HPC, as the three-year electricity cost can now exceed the purchase price of a system. We are looking at software techniques to assess energy efficiency, and new algorithmic approaches for improving it.

In 2012, OeRC led the procurement of Emerald, the largest GPU system in the UK.  as part of a collaboration between Oxford, Bristol, Southampton and UCL.

NVIDIA has recognised Oxford University, and in particular OeRC, as a CUDA Center of Excellence in GPU computing.

If you would like to find out more about Scientific Computing or collaborate with us in this area please get in touch.

Current Projects

Focusing on enabling real-time processing of time-domain radio astronomy data by exploiting powerful many-core processors.
The OPS project is developing an open-source framework for the execution of multi-block structured mesh applications on clusters of GPUs or multi-core CPUs.
The OP2 project is developing an open-source framework for the execution of unstructured grid applications on clusters of GPUs or multi-core CPUs.
Supporting interaction and collaboration between UK numerical analysts, computer scientists and developers and users of software and HPC.
Investigating the latest many-core technologies for delivering energy and cost efficient radio astronomy HPC.
Modern systems are increasingly energy efficient but carbon emissions, constantly increasing energy prices and recent changes in legislations drive a desire to minimise the overall energy consumption of computing.
Oxford University has been named by NVIDIA as a CUDA Centre of Excellence, recognising the excellent research on GPU computing within many departments in the university.
The ARTEMIS group in Oxford design, develop and exploit real-time processors for the discovery of pulsars and Fast Radio Bursts, using many-core technology and intelligent algorithms.

Past Projects

The goals were to help each other with these exciting new technologies, build bridges between application scientists and computing experts, maintain/develop links with international experts, and prepare joint proposals for future research funding.
Existing electricity distribution management systems (DMS) have been designed using operational and algorithmic procedures that are highly centralised.
The Oxford e-Research Centre has a 96 core Dell cluster installed with MS Cluster Compute Server version 1. The cluster is arranged with both thick (quad dual core) nodes and thin (dual dual core) nodes.
Flamingo is a general-purpose auto-tuning framework for software optimisation developed by Ben Spencer. Flamingo is a general-purpose auto-tuning framework for software optimisation.
This pilot will deliver a public-private hybrid cloud architecture and two demonstrators providing higher level services for the support of research.
Exploring which fields share common numerical algorithmic bases to determine areas in which the UK should invest in algorithm development
The purpose of CRISP is to create synergies and develop common solutions for an initial group of eleven European Strategy Forum on Research Infrastructure preparatory phase projects in the field of Physics, Astronomy, and Analytical Facilities.
Algorithms and Software for Emerging Architectures (ASEArch) is a new EPSRC-funded Collaborative Computational Project (CCP), led by the Oxford e-Research Centre, in collaboration with Bristol University and STFC staff at both Daresbury and RAL.