Adaptive Tag Comparison for Hybrid Cache

Seong Jae Eom,Eui-young Chung

2023 20TH INTERNATIONAL SOC DESIGN CONFERENCE, ISOCC(2023)

Cited 0|Views13
No score
Abstract
This paper presents a proposal for an STT-RAM (Spin Transfer Torque Magnetic Random Access Memory) and SRAM (Static Random Access Memory) hybrid cache architecture and adaptive tag comparison for this architecture, considering the drawbacks of high-power consumption and the high cost of cache size expansion in conventional SRAM-based cache architectures for performance enhancement. Although the hybrid cache structure allows for an increase in memory size relative to the area, it has the drawback of increased tag comparison. The proposed adaptive tag comparison method involves dynamically changing the flag associated with the memory state and the value of threshold, which determines the memory state based on access count, in real-time to make decisions about tag comparison. By adopting the hybrid cache architecture, an average power reduction of 51.01% is achieved. And applying adaptive tag comparison, further power consumption is reduced by an average of 34.12% without degradation of performance, through an average reduction of 53.14% in tag comparison overhead compared to the hybrid cache scheme.
More
Translated text
Key words
Hybird Cache,GPGPU,Adaptive Tag Comparison
AI Read Science
Must-Reading Tree
Example
Generate MRT to find the research sequence of this paper
Chat Paper
Summary is being generated by the instructions you defined