Skip to content
forked from tile-ai/tilelang

Extended TileLang as a unified DSL to enable high-performance kernel development for Near-Memory Computing, Distributed Memory AI Accelerators, and Networked Accelerators.

License

Notifications You must be signed in to change notification settings

SUNMMIO/Tilelang

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Tile Language for Distributed Memory

PyPI version Ask DeepWiki Discord

The Goal

The goal of this open source project is to extend Tilelang (https://tilelang.com/) as a unified DSL (domain-specific language) to enable high-performance kernel development for Near-Memory Computing, Distributed Memory AI Accelerators, and Networked Accelerators.

Near-memory computing and distributed memory systems have become key approaches to address the huge computing demand of AI, while networked accelerators further promote the decoupling and coordination of computational resources. To support efficient programming for such emerging heterogeneous computing architectures, we need a unified domain-specific language (DSL) aimed at enabling high-performance kernel development. The design goal is to abstract away underlying hardware differences among different Near-Memory Accelerators or Networked Accelerators, and provide a unified interface, allowing developers to focus on algorithm optimization rather than hardware adaptation. The language likely incorporates key techniques such as tensor tiling, dataflow scheduling, memory layouting, and communication-aware compilation, supporting automatic code generation and optimization to achieve efficient kernel execution across various advanced AI acceleration architectures. Through this unified language framework, we aim to significantly reduce the complexity of cross-platform AI operator development, improving both development efficiency and system performance.

Latest News

  • 10/30/2025 📦:

About

Extended TileLang as a unified DSL to enable high-performance kernel development for Near-Memory Computing, Distributed Memory AI Accelerators, and Networked Accelerators.

Topics

Resources

License

Code of conduct

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 57.2%
  • Python 41.8%
  • Cuda 0.4%
  • Cython 0.2%
  • Shell 0.2%
  • CMake 0.2%