This document discusses using OpenCL to accelerate database operations by offloading computation to GPUs. It presents PostgreSQL as an example database that can be extended with OpenCL using PgOpenCL. This allows SQL queries to launch OpenCL kernels on GPUs and retrieve results. Challenges include data transfer between CPU and GPU, runtime partitioning for large datasets, and concurrency without preemptive multitasking. Example use cases include spatial selections in GIS, DNA/protein matching, and accelerating database internal operations like joins and sorting.