#systemverilog

Tiny-gpu is a minimal GPU implementation in Verilog designed to teach you how GPUs work from the ground up. Built with fewer than 15 fully documented files, it includes working matrix addition and multiplication kernels with complete simulation support. The project simplifies GPU architecture by removing graphics-specific complexity and focusing on core principles like parallelization, memory management, and hardware acceleration that apply to modern GPUs and AI accelerators. This benefits you by providing an accessible learning resource where proprietary GPU details are normally hidden—you can study the actual hardware design, understand how parallel processing works, and see real kernel execution traces. The 11-instruction instruction set architecture enables practical proof-of-concept demonstrations without overwhelming complexity.

https://github.com/adam-maj/tiny-gpu