Hybrid KNN-join: Parallel nearest neighbor searches exploiting CPU and GPU architectural features

K Nearest Neighbor (KNN) joins are used in scientific domains for data analysis, and building blocks of several well-known algorithms. KNN-joins find the KNN all points a dataset. This paper focuses on hybrid CPU/GPU approach low-dimensional KNN-joins, where GPU may not yield substantial performance gains over parallel CPU We utilize work queue that prioritizes computing high density regions GPU, low CPU, thereby taking advantage each architecture’s relative strengths. Our approach, HybridKNN-Join, effectively augments state-of-the-art multi-core algorithm. propose optimizations (i) maximize query throughput by assigning large batches work; (ii) increase workload granularity to optimize utilization; and, (iii) limit load imbalance between architectures. compare HybridKNN-Join one two reference implementations. Compared implementations, we algorithm performs best larger workloads (dataset size K). The methods employed this show promise general division other