Overview / Usage

SW# is a software released in 2013 for biological sequence alignment. It is able to compute pairwise alignments as well as database similarity searches, for both protein and DNA sequences. This software allows configuring the algorithm to be used for different alignments (SW, NW, HW, OV) as well as open/extension penalties, and also the substitution matrix (BLOSUM45, BLOSUM50, BLOSUM62, among others, for proteins; and match/mismatch values for DNA). As it combines CPU and GPU computation, it allows configuring the number of CPU threads and GPU devices to be used.

We have completely migrated this software to SYCL by using oneAPI dpct, applying modifications in the sections that the tool could not migrate. We have been able to run on both NVIDIA and Intel GPUs, as well as Intel CPUs.

Methodology / Approach

Our migration process consisted of 5 stages: (1) running the dpct tool to generate the first version of the code, (2) modifying the migrated code based on dpct warnings to obtain the first executable version, (3) fixing runtime errors to obtain the first functional version, (4) verifying the correctness of the results and (5) optimizing the resulting code

Technologies Used

Intel LLVM (https://github.com/intel/llvm)

oneAPI 2022.1

Repository

https://github.com/ManuelCostanzo/swsharp_oneapi

Other links

• Migrating CUDA to oneAPI: A Smith-Waterman Case Study