Performance-Portable Distributed KNN Using LSH and SYCL*

 

In the age of artificial intelligence, algorithms must efficiently cope with vast datasets. We propose a performance-portable implementation of locality-sensitive hashing (LSH), which is an approximate k-nearest neighbor (KNN) algorithm to speed up the classification on heterogeneous hardware.

Our new library provides a hardware independent, yet efficient and distributed implementation of the LSH algorithm using SYCL* and message passing interface (MPI).

The results show that our library can scale on multiple GPUs, achieving a speedup of up to 7.6x on eight GPUs. It supports different SYCL implementations—ComputeCpp, hipSYCL, DPC++—to target different hardware.

Speaker

Marcel Breyer is a PhD student at the University of Stuttgart, Germany. His main field of research is on performance portability on heterogeneous hardware, which includes new applications of SYCL. He has contributed performance-portable k-nearest neighbors implementations for vast datasets.

 

Product and Performance Information

1

Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.