Language And Compilation Optimization for Next-gen High Performance Computing Research

A Joint Laboratory of Shanghai Enflame Technology and Shanghai Jiao Tong University

We are committed to building better programming languages to simplify the development of high-performance kernels. Our mission is to bridge the gap between performance and productivity, empowering developers to write efficient computing kernels without compromising on ease of use.

LancerLab brings together talented researchers and engineers from:

• Shanghai Jiao Tong University (SJTU) - Leading academic research in programming languages and compiler optimization
• Shanghai Enflame Technology - Industry expertise in AI accelerators and high-performance computing

Our team combines cutting-edge academic research with real-world industrial applications to push the boundaries of HPC programming.

Choreo is an extremely intuitive HPC kernel programming language designed to revolutionize how developers write high-performance computing code.

Choreo provides the most natural way to express tiled computations, making it extremely easy to write performant computing kernel code. Say goodbye to complex loop optimizations and memory management headaches.

The strongest support for dynamic shapes and logic in computing kernels. Choreo treats symbolic shapes as fundamental language constructs, enabling unprecedented flexibility in handling dynamic workloads.

Catch almost any bugs - including those pesky runtime bugs - at compile time. Experience the next level of debugging with comprehensive static analysis that saves hours of debugging time.

One codebase runs seamlessly across multiple platforms: GPUs, CPUs, and various private Domain-Specific Accelerators (DSAs). True portability without performance compromise.

We're constantly innovating! Stay tuned for more exciting tools and languages that will transform HPC development.

Interested in collaborating or learning more about our work? Feel free to explore our repositories and reach out!