Data warehouses are very large databases and the crucial part of business intelligence. The performance of a data warehouse is an important aspect and its optimization is a difficult task. The emergence of the graphics processing unit (GPU) based computation in recent years has brought a potential in a range of scientific applications. The usage of GPUs in databases technologies, more precisely in the physical design phase, is a potential domain for optimization tasks. The Bitmap Join Indexes selection problem (BJISP) is important problem in physical design of data warehouse. The present work deal with a GPU-based parallel binary particle swarm optimization (GBPSO) method for solving the BJISP. Experiments have been done to show the efficiency of our contribution versus the best serial approach for solving the BJISP. Furthermore, scalability experiments were implemented to observe the behavior of the proposed method against the best comparable approach which is serial in nature. In all experiments, the GBPSO is found to be considerably more effective than the best competitor algorithm.