Building a compiler + HSL framework to turn @__tinygrad__ kernels into VHDL, and synthesize the perfect FPGA for a given compute graph.
Tinygrad UOps -> KernelIR (my custom IR) -> Amaranth hardware modules
When people say “Triton is a compiler”, they mean something very specific.
Triton is not just another way to write Python that happens to use CUDA underneath, it is a real compiler stack for GPU kernels! 🧵
Image credit: @PyTorch
Source: https://t.co/kORJDZy6vD