Sparse efficient implementation of Conway's Game of Life.
First clone this repo by running:
git clone https://github.com/davidmallasen/GameOfLife_sparse.git
cd GameOfLife_sparse
Install Boost C++ Libraries. In Ubuntu 20.04 you can simply sudo apt install libboost-dev.
Install GoogleTest. In Ubuntu 20.04 you can simply sudo apt install libgtest-dev.
Install GPerftools and Google pprof. In Uuntu 20.04 you can simply sudo apt install google-perftools libgoogle-perftools-dev.
To build the application and run an example execute:
mkdir build && cd build && cmake .. && make
./bin/main.out
The default initial conditions can be adjusted by modifying the initial_population variable in src/main.cpp.
When running main.out you will also generate the VTK representation of each step. You can visualize this data for example using Paraview.
To create an animation from the .png files that can be generated using Paraview run convert -delay 10 -loop 0 *.png animation.gif.
To run the tests execute ./bin/tests.out in the build directory. The four unit tests check the following:
- SimpleDie: Checks that cells with less than 3 neighbors should die.
- SimpleLive: Checks that positions with 3 neighbors should create a new cell.
- SimpleExample: Checks a simple example that covers all the rules over one step.
- IterativeGlider: Executes a small glider over 3 time steps.
To activate the VTK visualization of the performance test declare #define VTK 1 in src/perf_test.cpp (not recommended when actually analyzing the performance).
To measure the execution time of the performance test, declare #define TIMING 1 in src/perf_test.cpp.
Then, run:
./bin/perf.out
To create a profile and visualize the hotspots run the following in the build directory (remember to set the defines back to 0 to avoid their overhead):
env CPUPROFILE=line20000.prof CPUPROFILE_FREQUENCY=1000 ./bin/perf.out
google-pprof --text ./bin/perf.out line20000.prof
If you want to sort by cumulative time instead of the flat value, add the -cum option to the google-pprof command.
To check that there are no memory leaks change the number of steps in src/perf_test.cpp to something smaller like 200, recompile and run HEAPCHECK=normal ./bin/perf.out.
First you will need to install the Linux perf tools (sudo apt install linux-tools-$(uname -r) linux-tools-generic). You will also have to either run the commands with superuser mode (sudo) or enable counter values collection without root permission: sudo sh -c 'echo 1 >/proc/sys/kernel/perf_event_paranoid'.
To obtain stats on the instructions and the caches run:
perf stat -e cycles,instructions,cache-references,cache-misses,L1-dcache-loads,L1-dcache-load-misses,L1-dcache-stores,L1-icache-load-misses,LLC-loads,LLC-load-misses,LLC-stores -B ./bin/perf.out
The list of all pre-defined events to be used with -e can be seen with perf list.
Note: Linux perf tools may not give all values if used in a virtual machine.