Data Intensive Design for Multi-core Era☆

Abstract Modern multi-core design will continue Moore's law and facilitate platforms for both wired and wireless communications. It is also predicted that multi-core, as a popular computing platform, will need to have the capability to process “big data” with both high speed and high quality. However, slow dynamic interactions and data exchange are the two key challenges in the multi-core design. This paper discusses two new strategies that can potentially solve these two issues. First, we discuss a new “Communication-time Cost Minimization” (CCM) approach to reduce the communications between different cores. Then we discuss how to manage tasks and their associated data to achieve “Request Cancellation”to common structures such as memory or stack.We have demonstrated an average reduction of 50% in the communication- time cost, an average of 23.1% of throughput improvement, and a linear increase of request executions for the shared memory stack.

[1]  Nir Shavit,et al.  Elimination Trees and the Construction of Pools and Stacks , 1997, Theory of Computing Systems.

[2]  Roman L. Lysecky,et al.  Workload capacity considering NBTI degradation in multi-core systems , 2010, 2010 15th Asia and South Pacific Design Automation Conference (ASP-DAC).

[3]  Maged M. Michael,et al.  Nonblocking Algorithms and Preemption-Safe Locking on Multiprogrammed Shared Memory Multiprocessors , 1998, J. Parallel Distributed Comput..

[4]  G. Amdhal,et al.  Validity of the single processor approach to achieving large scale computing capabilities , 1967, AFIPS '67 (Spring).

[5]  David P. Rodgers,et al.  Improvements in multiprocessor system design , 1985, ISCA '85.

[6]  Anthony S. Wojcik,et al.  Afips Conference Proceedings , 1985 .

[7]  E.A. Lee,et al.  Synchronous data flow , 1987, Proceedings of the IEEE.

[8]  Nir Shavit,et al.  A scalable lock-free stack algorithm , 2004, SPAA '04.

[9]  Yu Cao,et al.  A self-evolving design methodology for power efficient multi-core systems , 2010, 2010 IEEE/ACM International Conference on Computer-Aided Design (ICCAD).