An Algorithm to Enhance Cache Efficiency in Multi-core Processors

The Cache efficiency is considered to be one of the major challenges in multi-core processors. Hence, using cache space in such processors should be meticulously managed by each of the cores. This paper addresses the issue of cache re-access and proposes an algorithm which divides the last level of cache into local and global share for the cores. The rationale behind the proposed algorithm is to activate or deactivate the ways of cache for any intended core. Consequently, the collision between cores is reduced and each of the cores can use the cache space dynamically. To simulate the proposed algorithm, the researchers used three groups of applications and the obtained results were examined and evaluated in two stages. The first phase is involved with the number of active ways in cache for each core. It should be highlighted that the proposed algorithm, in the merged state, was able to enhance the active ways up to ۱۹%. In the second phase, cache miss rate was taken into consideration and it was observed that about ۷% improvement was achieved in this stage.

Keywords:

Asymmetric multicore-processors , en , Processor performance , Cache performance , Shared and dedicated structure , Partitioning of the last level of cache