Improving main memory hash joins on Intel Xeon Phi processors

摘要: Modern processor technologies have driven new designs and implementations in main-memory hash joins. Recently, Intel Many Integrated Core (MIC) co-processors (commonly known as Xeon Phi) embrace emerging x86 single-chip many-core techniques. Compared with contemporary multi-core CPUs, Phi has quite different architectural features: wider SIMD instructions, many cores hardware contexts, well lower-frequency in-order cores. In this paper, we experimentally revisit the state-of-the-art join algorithms on co-processors. particular, study two camps of algorithms: hardware-conscious ones that advocate careful tailoring to underlying architectures hardware-oblivious omit such tailoring. For each camp, impact features software optimizations comparison results CPUs. Our experiments show major findings Phi, which are quantitatively from those First, behavior CPU, calls for optimization tuning Phi. Second, oblivious can outperform conscious a wide parameter window. These further shed light design implementation query processing new-generation technologies.