Bulk-bitwise processing-in-memory (PIM), an emerging computational paradigm utilizing memory arrays as computational units, has been shown to benefit database applications. This paper demonstrates how GROUP-BY and JOIN, database operations not supported by previous works, can be performed efficiently in bulk-bitwise PIM used for relational database analytical processing. We develop a gem5 simulator and show that our hardware modifications, on the Star Schema Benchmark and compared to previous works, improve, on average, execution time by $1.83\times$, energy by $4.31\times$, and the system's lifetime by $3.21\times$. We also achieved a speedup of $4.65\times$ over MonetDB, a modern state-of-the-art in-memory database.

翻译：批量位处理内存计算（bulk-bitwise processing-in-memory, PIM）是一种利用内存阵列作为计算单元的新兴计算范式，已被证明可应用于数据库领域。本文展示了如何在用于关系数据库分析处理的批量位PIM中高效执行GROUP-BY和JOIN操作——这些数据库操作此前的研究尚未支持。我们开发了一个gem5模拟器，并表明：在星型模式基准测试中，与先前工作相比，我们的硬件修改平均将执行时间提升了$1.83\times$，能耗降低了$4.31\times$，系统寿命延长了$3.21\times$。此外，与现代最先进的内存数据库MonetDB相比，我们实现了$4.65\times$的加速比。