Skip to content

Latest commit

 

History

History
57 lines (47 loc) · 1.33 KB

running_on_nusc_cluster.md

File metadata and controls

57 lines (47 loc) · 1.33 KB
  1. Download cmake 3.20 (already done):

  2. Download and unpack source files: https://github.com/friofry/motif_finder_cuda/archive/refs/tags/v1.0.zip

  3. Run the a6500g10q node:

qsub -I -l walltime=00:30:00 -l select=1:ngpus=4:ncpus=12:mem=40gb -q a6500g10q@vm-pbs2
  1. Unpack the sources:
unzip motif_finder_cuda-1.0.zip -d .
  1. Navigate to the sourcedir:
cd motif_finder_cuda-1.0
  1. Create build dir and navigate to it:
mkdir build
cd build
  1. Configure:

Easy way: ~/cmake-3.20.3-linux-x86_64/bin/cmake -DCMAKE_BUILD_TYPE=Release ..

If it doesn't work:

  • Add to CUDA_COMPILER to path:
export PATH="/opt/shared/nvidia/cuda-10.2.89/bin:$PATH";
  • Configure the project:
export CC=gcc-8; export CXX=g++-8; ~/cmake-3.20.3-linux-x86_64/bin/cmake -DCMAKE_CUDA_HOST_COMPILER=/usr/bin/g++-8 -DCUDA_TOOLKIT_ROOT_DIR=/opt/shared/nvidia/cuda-10.2.89/  -DCMAKE_BUILD_TYPE=Release ..
  1. Build:
~/cmake-3.20.3-linux-x86_64/bin/cmake --build . -j
  1. Run:
  • Copy test data:
cp ../motif_finder_gpu/init.ini ../motif_finder_gpu/test_12.fst motif_finder_gpu/
  • Run:
cd motif_finder_gpu
./motif_finder_gpu