Work-Efficient Query Evaluation with PRAMs
This paper addresses two major bottlenecks—excessive work and scattered output memory—in parallel evaluation of relational queries on the CRCW PRAM model. We propose the first constant-time, weakly work-efficient (O(T^{1+ε})) parallel evaluation framework. Methodologically, we integrate the Goldberg–Zwick approximate prefix-sum algorithm with compactification techniques, leveraging practical assumptions such as sorting or numerical constraints to achieve operator-level parallelization for acyclic queries, semijoin algebra, and worst-case-optimal joins. Our contributions are threefold: (i) the first constant-time, weakly work-efficient parallel evaluation for these three fundamental query classes; (ii) work complexity asymptotically approaching that of the optimal sequential algorithms; and (iii) substantial improvements over naive parallel approaches, coupled with highly compact output memory layout.