Hiroyuki Tomiyama, Ritsumeikan University, Japan


OpenCL Compiler and Runtime Library for Embedded Manycore SoCs


The OpenCL language is now widely used for parallel computing on GPUs, Cell B./E., and multicore personal computers, but it is rarely used for real-time embedded systems. One reason for it is that most of current OpenCL compilers do not support task-parallel execution, which is essential for embedded systems, although the language itself supports it. Another reason is that the runtime overhead of performance is significant. This talk describes our OpenCL compiler and runtime library developed for real-time embedded systems. Unlike existing OpenCL compilers, our OpenCL framework explicitly supports not only task-parallel execution in an application but also parallel execution of multiple applications. Another important feature is its low runtime overhead, which is enabled by design-time context creation and mapping.


Hiroyuki Tomiyama received his Ph.D. degree in computer science from Kyushu University in 1999. From 1999 to 2001, he was a visiting postdoctoral researcher with the Center of Embedded Computer Systems, University of California, Irvine. Then, he worked for Institute of Systems & Information Technologies/KYUSHU as a researcher and Nagoya University as an associate professor. In 2010, he moved to Ritsumeikan University as a full professor. His research interests include system-level design methodology for embedded systems and MPSoC. He was General Co-Chair of MPSoC 2010, and is now Editor-in-Chief of IPSJ Transactions on System LSI Design Methodology.

* If you wish to modify any information or update your photo, please contact the web chairalain.fourmigue@polymtl.ca