Affiliation:
1. University of Texas Pan American, Edinburg, TX, USA
Abstract
This paper proposes a buffered dual-access-mode cache to reduce power consumption for highly-associative caches in modern embedded systems. The proposed scheme consists of a MRU (most recently used) buffer table and a single cache structure to implement two accessing modes, phased mode and way-prediction mode. The proposed scheme shows better access time and lower power consumption than two popular low-power caches, phased cache and way-prediction cache. The authors used Cacti and SimpleScalar simulators to evaluate the proposed cache scheme by using SPEC benchmark programs. The experimental results show that the proposed cache scheme improves the EDP (energy delay product) up to 40% for instruction cache and up to 42% for data cache compared to way-prediction cache, which performs better than phased cache.
Reference21 articles.
1. Alipour, M., Moshari, K., & Bagheri, M. (2011). Performance per power optimum cache architecture for embedded applications, a design space exploration. In 2011 IEEE 2nd International Conference on Networked Embedded Systems for Enterprise Applications (NESEA) (pp. 1–6).
2. ARM. (2011). Cortex-A9 techinical manual. Retrieved from http://www.arm.com
3. Chung, E. Y., Kim, C. H., & Chung, S. W. (2008). An accurate and energy-efficient way determination technique for instruction caches by early tab matching. In Proceedings of the IEEE International Workshop on Electronic Design, Test and Applications (pp. 190–195).
4. Community, K. (2011). Kubuntu 10.10. Canonical Ltd., Retrieved from http://www.kubuntu.org/
5. Microprocessor Design Issues: Thoughts on the Road Ahead