distributed-ranges icon indicating copy to clipboard operation
distributed-ranges copied to clipboard

Multi-node tests on Borealis

Open lslusarczyk opened this issue 2 years ago • 1 comments

lslusarczyk avatar Nov 07 '23 09:11 lslusarczyk

Suspended, passed to @lslusarczyk The status:

  • qsub + mpirun working, running the multinode benchmarks (single node temporary disabled in the branch)
  • plotter generating only part of figures
  • not proper control (--nodes param not properly interpreted, --different_devices switches to multinode instead) Current state of code:
  1. https://github.com/mateuszpn/distributed-ranges/tree/multinode
  2. https://github.com/intel-sandbox/libraries.runtimes.hpc.dds.dr-ci/tree/dev/multi-node

For quick start:

  • clone the 2) repo into ci directory
  • cd ci
  • run the script:
rm -rf dr
git clone https://github.com/mateuszpn/distributed-ranges.git dr && cd dr && git switch multinode && cd ..
/usr/bin/python3.10 -m venv venv
source venv/bin/activate
pip install dr/src-python/drbench
CXX=icpx CC=icx MKL_DIR=${MKLROOT}/lib/cmake/mkl TBB_DIR=${TBBROOT}/lib/cmake/tbb cmake -B dr/build -DENABLE_SYCL=on dr
scripts/borealis-reserve-nodes.sh 8
scripts/borealis-benchmark-multi.sh

mateuszpn avatar Nov 28 '23 11:11 mateuszpn