GPU Game of Life

A CUDA-enabled version of Conway’s Game of Life written in C++.

Features

• Pause and un-pause the simulation
• Click on cells to toggle their state
• Pan the grid
• Zoom in and out
• Clear the existing grid
• Randomize the existing grid
• Display current UPS/FPS
• Pre-built test cases
• Stretch Goal: Control the speed of the simulation
• Stretch Goal: Deploy “templates” of commonly-known entities in the Game of Life, e.g. gliders

Demonstration

Compilation

The makefile is in src/GPU-Game-of-Life/

make CPU=1 to compile for the CPU; make CPU=0 or just make to compile for the GPU

The project depends on SFML 2.5.1. Edit SFML_root in the makefile to point to wherever you place the compiled files.

You may also need to update cclibraries, gpp, and cudaroot if compiling on a system other than Linux.

Running the Program:

./life blockWidth blockHeight gridWidth gridHeight [numPasses] [p] [t<1|2>]

Parameters:

• blockWidth, blockHeight, gridWidth, and gridHeight must be positive integers.
• Non-square grids will be stretched to fit the window if using the GUI.
• blockWidth and blockHeight will be ignored if running on the CPU.
• If numPasses (a positive integer) is given, the GUI will be disabled. Instead, a starting grid will be generated and the program will execute exactly numPasses updates (or kernel calls). Note: Timings are averaged by ignoring the first iteration since it is abnormally slow on the GPU.
• If p is given and numPasses is given, the program will print the grid after each iteration.
• If t# is given, the program will no longer start with a random grid. Instead, it will contain a pre-defined test setup. This can be used to verify CPU / GPU correctness. The grid may be expanded automatically beyond the given arguments if it is too small.
  • t1: The grid will contain a light-weight space ship that should move horizontally across the grid.
  • t2: The grid will contain a Gosper glider gun.