Provides a portable, very-high-performance implementation of LINGO similarities on SIMD CPUs and GPUs.
SIML ("Single-Instruction, Multiple-LINGO") is a library containing implementations of a fast SIMD algorithm for calculating the LINGO chemical similarity metric (described in an upcoming publication). This method, currently implemented for x86 CPUs (non-vectorized) and NVIDIA GPUs, is several times faster than existing LINGO implementations for the CPU, and two orders of magnitude faster when run on a GPU.
Mar 26, 2017
SIML includes source code for the multiset/sparse-vector algorithm to calculate LINGO similarities, with implementations in both C++/CUDA and Python/CUDA.See all Downloads