Design of a Scalable Hash Table on a GPU

We investigate the scalability of an existing lockless hash table when it is implemented on a GPU. This lockless hash table is an essential part of a model checker, a tool that can be used to prove the correctness of software that performs critical tasks. We show that our implementation of the hash table on a GPU is scalable and that a GPU can lookup twice as much per second as a CPU when running the same OpenCL implementation. We also propose a design for integrating our implementation with a model checker. We argue that this design will lead to better performance of the model checker.