Requirements

OS requirement :

1. Ubuntu == 20.04
2. Nvidia Driver Version == 495.29.05
3. CUDA versuib == 11.5

Python Env requirement :

1. python == 3.9
2. pyqubo == 1.2.0
3. numpy == 1.23.3
4. tqdm == 4.64.1

Build Project

1. Make sure your host env as same as OS requirement
2. Download Dockerfile
3. Build Docker Image

   sudo docker build -t gpuda . --no-cache
   

4. Docker Run gpuda

   sudo docker run -it -p 8080:22 --gpus all gpuda bash
   

Run MaxCut sample on Docker

reference paper http://people.brunel.ac.uk/~mastjjb/jeb/bqp.pdf

1. go to gpu_kernel sub directory

   cd ./gpu_kernel
   

2. Run maxcut sample to solve Maxcut test data located at ./data/g05

   python maxcut_sample.py
   

3. Check output file "g05hist10000.txt" 10000 denotes it runs 10000 iterations and it will execute 100 runs for each test file then generate report in one line summary. The summary line include 'test_file_name 0.xxxx [0.yy ...]' , xxxx is average rnu time(in seconeds), yy is percentage which is i% closed to solution(i is arrary index, 0 means DA find exact answer or event better as solution and 1 means there is 1% delta)