high performance computing on graphics processing units: hgpu.org

This paper introduces the development of a new GPU-based database to accelerate data retrieval. The main goal is to explore new ways of handling complex data types and managing data and workloads in massively parallel databases. This paper presents three novel innovations to create an efficient virtual database engine that executes the majority of database operations directly on the GPU. The GPU database executes a subset of SQLite’s SELECT queries, which are typically the most computationally expensive operations in a transactional database. This database engine extends existing research by exploring methods of table caching on the GPU, handling irregular and complex data types, and executing multiple table joins and managing the resulting workload on the GPU. The GPU database discussed in this paper is implemented on a consumer grade GPU to demonstrate the high-performance computing benefits of relatively inexpensive hardware. Advances are compared both to existing CPU standards and to alternate implementations of the GPU database.