Variable-way set associative cache design for embedded system applications
Ramy E. Aly, B.R. Nallamilli, Magdy Bayoumi
Abstract
Ramy E. Aly, B.R. Nallamilli, Magdy Bayoumi
Abstract
Variable-way set associative cache is proposed as a new technique to maximize the cache performance especially for embedded applications or to reduce the power consumption with the same performance. Static profiling is used to determine the sets' behavior of the set-associative cache. Variable-way set-associative can be used in high-performance or low-power operation modes. Each set in the proposed design basically has different associativity to maximize the total performance for the cache size or reduce the power consumption. The proposed architecture is simulated on SimpleScalar simulator and tested on several Spec2000 Benchmarks. The results show on average 2% reduction in the miss rate at the high-performance mode and up to 43% reduction of the power consumption at low-power mode.
OpenAlex reports 13 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
Variable-way set associative cache is proposed as a new technique to maximize the cache performance especially for embedded applications or to reduce the power consumption with the same performance. Static profiling is used to determine the sets' behavior of the set-associative cache. Variable-way set-associative can be used in high-performance or low-power operation modes. Each set in the proposed design basically has different associativity to maximize the total performance for the cache size or reduce the power consumption. The proposed architecture is simulated on SimpleScalar simulator and tested on several Spec2000 Benchmarks. The results show on average 2% reduction in the miss rate at the high-performance mode and up to 43% reduction of the power consumption at low-power mode.
Key concepts: Cache, Computer science, Associative property, Cache algorithms, Parallel computing, Cache invalidation, CPU cache, Cache pollution