Dynamic Filter Cache for Low Power Instruction Memory Hierarchy

Authors:
Kugan Vivekanandarajah;Thambipillai Srikanthan;Saurav Bhattacharyya
Affiliations:
Nanyang Technological University, Singapore;Nanyang Technological University, Singapore;Nanyang Technological University, Singapore
Venue:
DSD '04 Proceedings of the Digital System Design, EUROMICRO Systems
Year:
2004

Citing 0
Cited 4

Low power design of precomputation-based content-addressable memory

IEEE Transactions on Very Large Scale Integration (VLSI) Systems
A tagless cache design for power saving in embedded systems

The Journal of Supercomputing
Survey of Low-Energy Techniques for Instruction Memory Organisations in Embedded Systems

Journal of Signal Processing Systems
Design Space Exploration of Distributed Loop Buffer Architectures with Incompatible Loop-Nest Organisations in Embedded Systems

Journal of Signal Processing Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

Filter cache(FC) is effective in achieving energy saving at the expense of some performance degradation.The energy savings, here, comes from repeated execution of tiny loops from energy efficient FC.The absence of cacheable loops leads to performance degradation in such FC structures. Therefore, we propose a simple dynamic FC scheme, which detects the opportunity for use of the FC and enables (or disables) it dynamically.Thus providing (slightly reduced) energy savings at minimal performance degradation. A combination of the predictive filter cache with the above schemes reduces the performance and energy penalty.For the benchamrks simulated with 256 Byte FC, the average performance degradation is 1.13% with the proposed scheme compared to 2.47% with just the predictive filter cache.With the same configuration, the resulting energy reduction is 42.77%.Finally, the proposed dynamic filter cache scheme is inherently simple and hence it lends well for VLSI efficient implementation.