My first goal was to improve my Python knowledge and implement The Game of Life on my desktop computer. I started with this so that I could implement all the features below to aid in understanding the game.

- Play
- Pause
- Reset
- Single Step
- Stagnant animation detection
- Auto Restart upon stagnant animation
- Image saving (in current directory)
- Speed Adjustment
- Grid Size Adjustment

x86_64 with Python3 and WX GUI

After understanding the basics of implementing The Game of Life, I moved on to implementing it in C on a STM32 using libopencm3. The implementation was easier than I expected as I just had to write a 12864 driver, implement Marsaglia’s Xorshift for randomization, and finally use C arrays/pointers to calculate cell states. This step was a lot of fun because the hardware is easily to procure and libopencm3 makes using all the STM32 peripherals a breeze.

STM32F103 with 12864 Display



Finally moving onto the most difficult implementation for me… using the iCE40UP5K FPGA to drive a Hub75 LED Panel while playing The Game of Life. This was my first project using an FPGA and Yosys, so I spent a good time in Icarus Verilog simulating how everything worked. Because simulation doesn’t have any LUT limitations, I initially implemented The Game of Life using wires and assign statements so the next state of the game could be determined in one clock cycle. This worked, but ate through the iCE40UP5K LUTs really fast so I could only display a 16x16 game. I then realized that I needed to use RAM in some form to store the current and future state of the game. This unveiled a whole new world of FPGAs to me and I went down the path of learning implicit and explicit BRAM/SPRAM. I really needed to use BRAM, but for some reason I could not get Yosys to infer BRAM correctly and the explicit BRAM definition also did not seem to be working. For this reason, I used explicit SPRAM and a lot of it. See the GitHub repository for details on the implementation, but in short it takes about 2000 clock cycles to determine the next state which seems like a lot, but at 24MHz you can play the game faster than the eye can see.

FPGA with Hub75 LED Matrix