Maximising the Multi-Core Advantage -€“ Easier Parallel Programming With GPUs

Presented by

Timothy Lanfear, NVIDIA, Simon McIntosh Smith

About this talk

: In the past five years, NVIDIA has driven the use of graphics processing units (GPUs) for general-purpose computing tasks from nowhere to a mainstream technology. Today, three of five most powerful supercomputers in the world are powered by GPUs and we see many examples where new science and technology is enabled by faster, better, more accurate simulation applications running on GPU accelerated computers. How can a working computational engineer or scientist take advantage of this new computing paradigm? The simplest way is to use an application that has been ported to GPUs; more and more commercial applications in the fields of structural mechanics, fluid dynamics, computational chemistry, bioinformatics, etc. are being ported to GPUs. For those using high-productivity scripting languages like MATLAB and Mathematica, the latest releases of these packages are accelerated by GPUs. Libraries offer a route to exploit the benefits of a GPU with minimal effort. NVIDIA and our partners provide a broad spectrum of libraries for dense and sparse linear algebra, Fourier transforms, random number generation, image processing, video encoding and decoding etc. Simply replacing calls to libraries running on a CPU with calls to libraries written for the GPU will give an immediate boost in performance.

Related topics:

More from this channel

Upcoming talks (0)
On-demand talks (33)
Subscribers (3740)
The new KTN will focus on Knowledge Transfer as a stimulus to economic growth. It will seek to deliver improved UK industrial performance by facilitating the development and take up of information and communications technologies and their adoption as key enablers in other industries.