CUDASW++

CUDASW++ (compute unified device architecture) 3.0 is a software tool that improves the speed and performance of the Smith-Waterman algorithm for protein sequence database search. It achieves this by coupling CPU and GPU SIMD instructions and carrying out concurrent CPU and GPU computations. The algorithm employs SSE-based vector execution units as accelerators for CPU computation and uses CUDA PTX SIMD video instructions for GPU computation to gain more data parallelism beyond the SIMT execution model. The workload is automatically distributed over CPUs and GPUs based on their respective compute capabilities. CUDASW++ 3.0 has demonstrated significant speedups over its predecessor, CUDASW++ 2.0, as well as over other top-performing tools such as SWIPE and BLAST+.