MPI-Game-of-Life

A simple and rough MPI implementation of Conway's Game of Life.

Prerequisites

• GNU software (GCC 9, Make)
• MPI (tested with Open MPI 4.1.0)
• Python 3.x with numpy and Pillow installed

Developed and tested on a single machine with Ubuntu 20.04 LTS.

Compilation

A Makefile has been provided to build and compile the target executable. Simply run make in the top level of the project directory.

The target executable lifeMPI is created in the same location.

Usage

On a single machine

mpiexec [--use-hwthread-cpus] -np <#> ./lifeMPI <args> [opts]

• --use-hwthread-cpus enables use of logical processors (e.g. hyper-threaded cores). This is not likely to improve performance, and may negatively impact performance.

where

args :  nrows (uint) --> number of rows in the grid
        ncols (uint) --> number of columns in the grid
        ngens (uint) --> number of generations to simulate
        EITHER filename (string) OR prob_alive (float) --> initial state of the grid

opts :  --periodic --> enable periodic boundaries [default: disabled]
        --dump=<uint> --> num generations between output file dumps [default: 101]

Examples:

mpiexec -np 4 ./lifeMPI 256 256 500 0.10 --periodic

mpiexec -np 9 ./lifeMPI 247 331 250 grid.bin --dump=63 --periodic

NOTES

• The options --periodic and --dump=## must be specified after the 4 input arguments (but the order does not matter).
• The input file must be a raw binary file containing 0's and 1's represented by single bytes.
• The grid dimensions given (nrows, ncols) must match exactly the input binary file.
• A binary file can be easily generated from a CSV (or any other delimted text file) using a Python commnd, e.g.:

python -c "import numpy; numpy.loadtxt('grid.csv', dtype=numpy.uint8, delimiter=',').tofile('grid.bin')"

Visualisation

The program will dump binary files of the grid state for every generation to outfiles/ (directory created during the build process, and can be automatically cleaned with make clean).

An animated gif of the grid state for every generation can be created by running the provided script:

python scripts/animate_life.py

Testing

A verified python implemenation of Conway's Game of Life is provided (scripts/automata.py) for testing accuracy of the MPI implementation.

A simple test can be performed by running make test (or python scripts/test_life.py) which compares the final states of both implementations.