Get the repo:
  git clone https://github.com/cyclops-community/ctf.git

Installation on cori:
  copy "matmul-benchmark.cxx" under "examples" directory
  copy the updated Makefile to base ctf directory to replace the original Makefile
  module swap PrgEnv-intel PrgEnv-gnu  
  ./configure --install-dir=/global/homes/r/roguzsel/cb2/code/lib/ctf/install --build-dir=/global/homes/r/roguzsel/cb2/code/lib/ctf/build CXX=CC
  mkdir build install && cd build
  make
  make install
  make examples
    builds the benchmark code
    fails for some files (they are excluded in the updated makefile)

Run the multiplication code by (32 nodes and 4 task per node)
  srun -N 32 --ntasks-per-node 4 ./bin/matmul-benchmark <row-dim-size> <col-dim-size> <inner-dim-size> <A-path> <B-path> <C-path> <niters>
  The path to output file (<C-path>) is not used (commented out)
  <niters> is the number of times to repeat the multiplication

To convert the matrices in matrix market file format to the format required by
ctf use the provided python file:
  python ConvertMtxToCtf.py <matrix-market-file>
