TileLang is a programming language and compiler infrastructure built on TVM for expressing and optimizing high-performance deep learning operators. It provides high-level abstractions for tensor programs, automatic tuning, debugging tools, and specialized operators such as GEMM, sparse matrix multiplication, and FlashMLA. Developers use it to generate efficient code for modern GPUs with minimal manual optimization effort.
In the Developer Tools space, TileLang takes a focused approach. It focuses on writing and optimizing high-performance GPU kernels and deep learning operators without low-level CUDA or manual tuning. It is built as an open-source project for machine learning compiler developers and performance engineers. TileLang is open source under the Open Source license. TileLang is available on the web and the command line, and it can be self-hosted.
Behind TileLang is TileLang Project, and it first shipped in 2024. The project is developed in the open on GitHub with 6.9k stars and 322 commits in the last 90 days. Among its 6 catalogued features are auto-Tuning, Performance Analyzer, and Layout Visualization.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Same category — not a similarity match