Hashmaps are widely utilized data structures in many applications to perform a probe on key-value pairs. However, their performance tends to degrade with the increase in the dataset size, which leads to expensive off-chip memory accesses to perform bucket traversals associated with hash collision. In this work, we propose HashMem, a processing-in-memory (PIM) architecture designed to perform bucket traversals along the row buffers at the subarray level. Due to the inherent parallelism achieved with many concurrent subarray accesses and the massive bandwidth available within DRAM, the execution time related to bucket traversals is significantly reduced. We have evaluated two versions of HashMem, performance-optimized and area-optimized, which have a speedup of 49.1x/17.1x and 9.2x/3.2x over standard C++ map and hyper-optimized hopscotch map implementations, respectively.
@article{arxiv.2306.17721,
title = {HashMem: PIM-based Hashmap Accelerator},
author = {Akhil Shekar and Morteza Baradaran and Sabiha Tajdari and Kevin Skadron},
journal= {arXiv preprint arXiv:2306.17721},
year = {2023}
}
Comments
This paper was published in Fifth International Workshop on Domain-Specific System Architecture (DOSSA-5)