This is a naive implementation of the Near-Far Pile Single Source Shortest Path Algorithm by Davidson et al.
- CUDA 7.5
- Thrust 1.8.1
- CUSP 0.5.1
- Install Docker and NVIDIA for Docker
- Build image
docker build . --pull --rm f "Dockerfile" -t naive-near-far-gpu:latest
- Image will automatically download libraries and compile the software
- Make sure CUDA 7.5 is installed
- Run the build script
sh build.sh
- The build script downloads CUSP
- Command:
./main `[m|d]` <file>
- Example
./main m graph.mtx
- First argument: file type.
m
for Matrix Market *.mtx files,d
for Dimacs - Second argument: path to file
- Create a local folder to put your graphs
- Set this local folder as a volume in Docker
- Run command
docker run -v <local-path-to-graphs>:/<docker-path> --name near-far --gpus all near-far-pile:latest ./main `[m|d]` <docker-path>/<file>
- Copy "distance.mtx" file from Docker to local folder:
docker cp near-far:distance.mtx .
This project is licensed under GPLv3