Comparing and Surveying Parallel Hashing Algorithms

On January 30, 2018, Brent Lessley, a fifth-year Ph.D. student at the Department of Computer and Information Science presented his area exam by surveying and comparing parallel hashing algorithms that are specially designed to run in the Graphical Processing Unit (GPU). 

The design patterns Lessley used in his area exam research are particularly suitable for GPU architectures. Hashing themes within his work led him to explore all the related research on GPU hashing.”Hashing is a surprisingly difficult task to perform efficiently when the hash table is accessed by thousands of threads in parallel,” said Lessley, “moreover, normal thread synchronization techniques and ways of accessing memory need to be fine-tuned for best hashing performance on the GPU.” Lessley looked into all those performance criteria to discover which worked best. 

To conduct this research Lessley categorized over 100 studies and after thorough reading, he differentiated each study within a category. “Completing a comprehensive survey of an area, no matter how narrow in scope, requires lots of paper reading and time,” said Lessley, “There is just so much research out there!”

After graduation Lessley hopes to work in the high-performance computing or data analysis fields. “It would be cool to continue designing parallel algorithm solutions for widely-used problems,” said Lessley.