GPUMemSort: A High Performance Graphics Co-processors Sorting Algorithm for Large Scale In-Memory Data

In this paper, we present a GPU-based sorting algorithm,GPUMemSort, which achieves high performance insorting large-scale in-memory data by take advantage ofGPU processors. It consists of two algorithms: an in-corealgorithm, which is responsible for sorting data in GPUglobal memory efficiently, and an out-of-core algorithm,which is responsible for dividing large-scale data intomultiple chunks that fit GPU global memory.GPUMemSort is implemented based on NVIDIA’s CUDAframework and some critical and detailed optimizationmethods are also presented. The tests of differentalgorithms have been run on multiple data sets. Theexperimental results show that our in-core sorting canoutperform other comparison-based algorithms andGPUMemSort is highly effective in sorting large-scale inmemorydata