Tilus: A New DSL for Powerful GPU Programming

Tilus is a powerful domain-specific language (DSL) for GPU programming. It offers thread-block-level granularity with tensors as the primary data type, explicit control over shared memory and register tensors, and low-precision types with arbitrary bit-widths (1 to 8 bits). Featuring automatic tuning, caching, and a Pythonic interface, Tilus simplifies GPU kernel development. Built upon research detailed in "Tilus: A Virtual Machine for Arbitrary Low-Precision GPGPU Computation in LLM Serving," it leverages insights from projects like Hidet, TVM, Triton, and Hexcute.