https://hgpu.org/?p=24835
An Investigation of Atomic Synchronization for Sort-Based Group-By Aggregation on GPUs