UK Manycore Developer Conference

Professor Mike Giles
December 16, 2013 - 08:30 to 17:15
Andrew Wiles Building, Mathematical Institute

Mathematical Institute
University of Oxford
Andrew Wiles Building
Radcliffe Observatory Quarter
Woodstock Road
Oxford
OX2 6GG

  • Conference
  • Registration required
  • Lunch provided

 

The conference covers high-performance computing on accelerators such as GPUs, FPGAs and the Xeon Phi, as well as embedded and mobile parallel systems. Application areas include computational science and engineering, finance, computer vision and beyond. The goal of the event is to strengthen the UK's expertise in many-core computing, in both industry and academia.

Time     Speaker / title
   
08:30 Registration
09:30 Mike Giles, University of Oxford
  Welcome
09:45 Harald Servat, Barcelona Supercomputing Centre
  Keynote address : Performance analysis & task-based heterogeneous programming
10:45 Garf Bowen, Ridgeway Kite Software
  A massively parallel architecture for reservoir simulation
11:15 Coffee
11:45 Ahmed Al-Jarro, UCL
  Scaling Computational Electromagnetics on EMERALD GPUs Using OpenACC
12:15 Eike Mueller, University of Bath
  Matrix-free multi-GPU Implementation of elliptic solvers for strongly anisotropic PDEs
   
12:45 Lunch
   
14:00 Alexey Kravets, ARM
  DSL-based approach to BLAS implementation for GPUs
14:30 Jorg Lotze, Xcelerit
  Designing for performance, portability, and scalability for CPUs, GPUs, and Xeon Phi - lessons from the trenches
15:00 Llion Marc Evans, Culham Centre for Fusion Energy
  Towards an Exascale many-core finite element program for the predictive modelling of nuclear reactor components
15:30 Coffee
16:00 Alastair Donaldson, Imperial College
  GPUVerify: Static Data Race Analysis for OpenCL and CUDA
16:30 Istvan Reguly, University of Oxford
  Acceleration of a Full-scale Industrial CFD Application with OP2
17:00 Mike Giles
  Closing remarks
17:15 Finish
   
19:00 Dinner (Branca restaurant directions)

The conference venue is the Andrew Wiles Building (the new home of the Mathematical Institute)

Travel and accommodation information is available here.

In addition to the conference on Dec 16th, the following morning there will be CUDA "masterclass" sessions to address the interests of CUDA developers.

Begin End Title
9:00 9:20 Nsight IDE
9:20 9:40 Use of new Kepler features
9:40 10:20 Profiling
10:20 10:50 Coffee
10:50 11:30 Tips for better performance
11:30 12:00 NVIDIA libraries
12:00 12:30 Open session addressing questions

They will be held in the same building as the conference.

The conference is again supported by the EPSRC-funded CCP on Algorithms and Software for Emerging Architecture: ASEArch CCP.

Early-bird registration is £20 for students and £40 for others -- this is open until Nov 22nd.

Late registration is £30 for students and £60 for others -- this is open until Dec 9th.