sequence-alignment-gpu

Run on Barkla (no prerequisites required)

   1. Get code base:
          cd
          tar -xf ~sgrszafa/sequence-alignment-gpu.tgz
          cd sequence-alignment-gpu
   2. Load Cuda and a C compiler:
          module load libs/nvidia-cuda/10.1.168/bin && module load compilers/gcc/8.3.0
   3. Run:
          // Will run on Nvidia Quadro P4000 (intended for lighter use).
          ./test
          ./benchmark
          ./alignSequence

          // Will run on Nvidia V100 (or P100) using sbatch.
          // Fill out barkla_alignSequence.sh with your desired input sequences.
          sbatch barkla_runBenchmark.sh
          sbatch barkla_runTest.sh
          sbatch barkla_alignSequence.sh

Run locally

   1. Check prerequisites.
   2. Unzip code
   3. Compile with:
          make -j
   4. Run:
          // Main program
          Usage: ./alignSequence [-d|-p] [-c|-g] [--global|--local] [-s <file>] [--gap-penalty <int>] <file> <file>
                 -d, --dna             - align dna sequences (default)
                 -p, --protein         - align protein sequence
                 -c, --cpu             - use cpu device (default)
                 -g, --gpu             - use gpu device
                 --global              - use global alignment (default)
                 --local               - use local alignment
                 -s, --score-matrix    - next argument is a score matrix file
                 --gap-penalty         - next argument is a gap open penalty (default 5)

          // test suite
          ./test

          // performance tests
          ./benchmark

Prerequisites

   - Nvidia GPU with compute capability >3.0 (Kepler, and later)
   - CUDA (tested with 10.1 and 11.1)
   - A C++14 compliant host compiler (tested with clang and gcc)
   - Make