High performance data analytics problems often use filters to approximately store or count a set of items while trading off accuracy for space-efficiency. Filters can also address the limited memory available on accelerators, such as GPUs. We present GQF, a feature-rich, GPU-optimized counting quotient filter, which provides both individual and bulk insertions, with different locking strategies to maximize performance and ensure correctness for each usage scenario. GQF has all the benefits of a standard counting quotient filter: good cache locality, high speed, and support for counting, deletions, and associating small values with items

Prashant Pandey
Prashant Pandey
Assistant Professor

My research interests lie at the intersection of Systems and Algorithms. I design and build theoretically well-founded data structures for big data problems in computational biology, streaming, and file systems.