TY - GEN
T1 - Communication-aware parallelization strategies for high performance applications
AU - Ashraf, Imran
AU - Bertels, Koen
AU - Khammassi, Nader
AU - Le Lann, Jean Christophe
N1 - Publisher Copyright:
© 2015 IEEE.
PY - 2015/10/27
Y1 - 2015/10/27
N2 - With the advent of multicore processor architectures and the existence of a huge legacy code base, the need for efficient and scalable parallel zing compilers is growing. Where multi-core processors were seen as the way forward to address the known challenges such as the memory, power and ILP wall, efficient parallelization to make use of the multiple cores, is still an open issue. In this paper, we present two complementary tools, MCROF and XPU which provide an alternative development path to parallelise applications and that address the challenges of identifying potential parallelism and exploiting it in a different way. The MCROF tool provides a detailed profile of the data flowing inside an application and the XPU programming paradigm provides an intuitive and simple interface to express parallelism as well as the necessary runtime support. We demonstrate through two different use cases that better performance up to 4× can be achieved than available commercial compilers.
AB - With the advent of multicore processor architectures and the existence of a huge legacy code base, the need for efficient and scalable parallel zing compilers is growing. Where multi-core processors were seen as the way forward to address the known challenges such as the memory, power and ILP wall, efficient parallelization to make use of the multiple cores, is still an open issue. In this paper, we present two complementary tools, MCROF and XPU which provide an alternative development path to parallelise applications and that address the challenges of identifying potential parallelism and exploiting it in a different way. The MCROF tool provides a detailed profile of the data flowing inside an application and the XPU programming paradigm provides an intuitive and simple interface to express parallelism as well as the necessary runtime support. We demonstrate through two different use cases that better performance up to 4× can be achieved than available commercial compilers.
KW - Data-communication profiling
KW - Multicore
KW - Parallel Programming
KW - Program Parallelization
UR - https://www.scopus.com/pages/publications/84956992348
U2 - 10.1109/ISVLSI.2015.89
DO - 10.1109/ISVLSI.2015.89
M3 - Conference contribution
AN - SCOPUS:84956992348
T3 - Proceedings of IEEE Computer Society Annual Symposium on VLSI, ISVLSI
SP - 539
EP - 544
BT - Proceedings - IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2015
PB - IEEE Computer Society
T2 - IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2015
Y2 - 8 July 2015 through 10 July 2015
ER -