Rob has an international reputation for his work in High Performance Computing (HPC), machine learning, complex dynamical systems and high energy physics. He has been a lifelong participant in HPC starting his research career in 1984 as the youngest staff scientist in the Theoretical Division at Los Alamos National Laboratory. Following that, Rob focused on commercial applications until deciding to return to research at the National Energy Research Supercomputer Center (NERSC) at Lawrence Berkeley National Laboratory. In 2005 he joined Pacific Northwest National Laboratory (PNNL). He has been invited speaker both nationally and internationally with over 80 publications in peer-reviewed journals and other media: in magazines such as Innovation: America's Journal of Technology Commercialization through his regular HPC column in Scientific Computing, CUDA tutorials on the Doctor Dobb's Journal website, and OpenCL tutorials on The Code Project.
As a recognized leader in GPU computing, Rob has recently completed a book "CUDA Application Design and Development" in partnership with NVIDIA, the dominant manufacturer of GPGPUs and creator of the CUDA software development platform. In relation to the present and future prospects for GPU computing, Rob comments:
"The tremendous draw of GPU computing is the application speed. In my Scientific Computing article, "Redefining what is possible", I examined the top 100 reported application speedups over conventional processors. These are real applications that solve problems in finance, medicine, Internet technology, biology, chemistry, and a wide variety of other fields. Just as computer spreadsheets and the Internet changed how people work with computers, so is GPU technology changing how business and research is performed. Very simply, a 10-times speedup makes the computer interactive, a 100-times speedup completely changes the rules as computational problems that would have taken a year or more can finish in a few days.
There are two remarkable facts about GPU hardware technology:
- Anyone from Dublin to Dubai can purchase these "supercomputers", over ¼ billion CUDA-enabled GPUs have been sold worldwide.
- The newest generation of tablets and smart phones contain an ultra-low power GPU that is not a supercomputer but rather is part of a SoC (System on a Chip) that is faster than most current laptops.
There are two outstanding facts about GPU software technology:
- NVIDIA's CUDA software platform was first introduced in February 2007. Now it is taught at 442 universities and academic institutions world-wide.
- CUDA has now matured to be a viable language for all application development; not just GPU development.
GPU technology is changing the world in many ways. Very simply, now is the time for small, quality organizations like ICHEC to shine. High-performance supercomputing is no longer limited to big money, big research organizations. Instead, GPU technology is bringing "Supercomputing to the masses", which is the topic of my freely available online CUDA tutorial series."
Robert M. Farber
Research Consultant ICHEC