R. Govindarajan received
his B.Sc. degree in Mathematics from Madras University in
1981 and B.E. (Electronics and Communication) and Ph.D. (Computer
Science) degrees from the Indian Institute of Science, Bangalore
in 1984 and 1989 respectively. He has held postdoctoral research
positions and visiting faculty positions at Universities in
USA and Canada. Since 1995, he has been with the Supercomputer
Education and Research Centre and the Department of Computer
Science and Automation, Indian Institute of Science, Bangalore.
Currently he is a professor and the chairman of the Supercomputer
Education and Research Centre. His research interests are
in the areas of High Performance Computing, Compilation Techniques,
and Computer Architecture. He has more than 120 publications
in international journals and refereed conferences in these
areas. He is a Fellow of the Indian National Academy of Engineering,
Senior member of IEEE, and a Member of ACM and IEEE Computer
Society.
|
Title:
The Role of Accelerators in HPC Systems: Challenges and
Opportunities
Abstract:
Rapid advancements in multi-core processor architectures along
with low-cost, low-latency, high-bandwidth interconnects have
made clusters of multi-core machines a common resource in
today's world. Accelerators such as Graphics Processing Units,
Many Integrated Cores, Specialized accelerators built using
FPGAs, etc. are playing an increasing role in rapidly increasing
the performance of HPC systems. While the theoretical performance
and sustained performance reported on specific benchmark programs
are promising, achieving a similar performance on real-world
applications present many challenges. In this talk we will
describe the different types of parallelism that can be exploited
in heterogeneous accelerator-based multi-core architectures.
We will discuss the how the accelerator architectures are
evolving taking specific examples. We present various work
(including our own) on automatically compiling programs written
in different high-level languages for heterogeneous GPU-based
architectures. We conclude by discussing the main challenges
ahead and future research directions in accelerator-based
architecture.
http://www.serc.iisc.ernet.in/~govind/index.html
|