Conference Paper

Using fine grain multithreading for energy efficient computing

DOI: 10.1145/1229428.1229480 Conference: Proceedings of the 12th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPOPP 2007, San Jose, California, USA, March 14-17, 2007
Source: DBLP

ABSTRACT We investigate extremely fine-grain multithreading as a means for improving energy efficiency of single-task program execution. Our work is based on low-overhead threads executing an explicitly par- allel program in a register-sharing context. The thread-based paral- lelism takes the place of instruction-level parallelism, allowing us to use simple and more energy-efficient in-order pipelines while re- taining performance that is characteristic of classical out-of-order processors. Our evaluation shows that in energy terms, the paral- lelized code running over in-order pipelines can outperform both plain in-order and out-of-order processors. Categories and Subject Descriptors C.1.4 (Computer Systems

Download full-text


Available from: Assaf Schuster, Apr 17, 2014
26 Reads
  • [Show abstract] [Hide abstract]
    ABSTRACT: What do the architectures of a future exascale computing system and a future battery-operated embedded system have in common? At first glance, their requirements and challenges seem unrelated. However, discussions and collaboration on the projects revealed not only similar requirements, but many common power and packaging issues as well.
    Computing in Science and Engineering 01/2011; 12(6-12):20 - 29. DOI:10.1109/MCSE.2010.95 · 0.99 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Multi-threaded processor designs enable high performance of a single processor core by exploiting both the thread-level and instruction-level parallelism. The performance gain is, however, at the cost of increasing energy consumption, which is not desirable to embedded systems. This paper investigates the energy efficiency of varied multi-threaded processor designs (with the coarse-grained and fine-grained thread switching schemes and different thread numbers). Our experiments, based on a six-stage PISA processor, show that in terms of potential energy saving the coarse-grained design is better than the fine-grained design. Furthermore, for the coarse-grained design, the thread number for the optimal energy efficiency is closely related to the memory access delay. When the memory access latency is small, the low-thread processor appears more energy efficient, When the memory delay increases the high-thread processor becomes superior.
    Proceedings of the 2013 IEEE 16th International Conference on Computational Science and Engineering; 12/2013