Skip to content

Latest commit

 

History

History
executable file
·
53 lines (35 loc) · 1.46 KB

README.md

File metadata and controls

executable file
·
53 lines (35 loc) · 1.46 KB
Raw

How to run on AUTH's HPC

  1. First edit the ASSETS variable inside the script.sh to point a directory that contains all the graph files (auto.mtx com-Youtube.mtx delaunay_n22.mtx great-britain_osm.mtx s12.mtx).

  2. Then run sbatch ./script.sh

  3. Finally you will get a slurm-jobid.out file that contains 3 runs from each graphs (the time metric is microseconds).

Commands make the project

Prerequisites

Commands to start the container

sudo docker run --rm --gpus all -it -e CUDBG_USE_LEGACY_DEBUGGER=1 --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -v $PWD:/cuda_project -w /cuda_project nvidia/cuda:12.0.1-devel-ubuntu20.04
  • -e means export CUDBG_USE_LEGACY_DEBUGGER=1
  • --cap-add=SYS_PTRACE --security-opt seccomp=unconfined is for gdb
  • -v $PWD:/cuda_project -w /cuda_project is for mounting current directory to /cuda_project in container
  • nvidia/cuda:12.0.1-devel-ubuntu20.04 is the image name (Works in Pop!_OS 22.04)
  • add -it before nvidia/cuda:12.0.1-devel-ubuntu20.04 to run the container in interactive mode

Commands to compile

Compile

make {ARGS}

with ARGS as:

  • BUILD_TYPE=debug or BUILD_TYPE=release
  • BUILD_ENV=container or BUILD_ENV=host

Run

./bin/fglt

or

./bin_debug/fglt