Tensor Computing Application Programming Interface (TCAPI)

TCAPI is an open-source, lightweight, application-oriented interface for portable, high-performance tensor-network computations. By separating tensor-network algorithms from the tensor-computing frameworks that execute tensor computations, it allows tensor-network applications written in C++ or Python to target diverse hardware platforms and software back ends through a concise set of well-defined tensor operations. To begin, see the Get Started section, then consult the specification for detailed API semantics.

Maintainers

Rong-Yang Sun (sun-rongyang@outlook.com)
Hidehiko Kohshiro
Tomonori Shirakawa

Available Implementations

C++

https://github.com/r-ccs-cms/tensor-computing-interface-backend-cytnx (backend: Cytnx)

Publication

If you use TCAPI in research software or publications, please cite the following paper:

Rong-Yang Sun, Tomonori Shirakawa, Hidehiko Kohshiro, D. N. Sheng, and Seiji Yunoki, “Tensor Computing Application Programming Interface: An Application-Oriented, Lightweight API for Portable High-Performance Tensor Network Computations,” arXiv:2512.23917 [quant-ph] (2025). https://doi.org/10.48550/arXiv.2512.23917

License

The TCAPI specification, examples, and the source code of this website are licensed under the Apache License 2.0.

Acknowledgments

The maintainers thank D. N. Sheng and Seiji Yunoki for their guidance and collaboration on the TCAPI paper, arXiv:2512.23917. They also acknowledge helpful discussions with Q. Zhao, C. Li, H. Xu, and S. Niu.