Originally named QODA (Quantum Optimized Device Architecture), CUDA (Compute Unified Device Architecture) Quantum was a proprietary and closed-source parallel computer platform and application programming interface.
Launched by Nvidia as QODA in July 2022, the programming language was renamed CUDA Quantum on March 21, 2023, and is currently available as an open-source project through Apache License 2.0. For high-performance simulation, the language makes use of Nvidia cuQuantum SDK, which holds a separate license.
CUDA Quantum links GPUs and quantum processors in future hybrid systems and is designed to accelerate workflows such as quantum simulation, quantum machine learning, quantum chemistry, and other applications, such as artificial intelligence, health, finance, chemistry, logistics, and other high-performance computing needs.
Quantum computers are programmed in equivalence with assembly code, which has a very high learning curve for those who aren't quantum engineers already incorporating quantum computing into their workflows. With CUDA Quantum, Nvidia hopes to remove that barrier by enabling the programming of hybrid quantum systems in a model familiar to scientific computing developers, and interoperable with other computing applications.
CUDA Quantum offers kernel-based programming and can be used with C, C++, Python, or Fortran. CUDA-powered GPUs (graphics processing units) also support programming frameworks such as OpenMP, OpenACC, OpenCL, and HIP (heterogeneous-computing interface for portability) by compiling such code to CUDA.
The CUDA platform is accessible to software developers through CUDA-accelerated libraries, compiler directives, and extensions to CUDA C/C++ and CUDA Fortran. Third-party wrappers are also available for Common Lisp, Fortran, Haskell, Java, Julia, Lua, Mathematica, MATLAB, Perl, Python, R, and Ruby.
CUDA Quantum is designed to be flexible and scalable, easily integrated with modern GPU-accelerated applications, high performing, productive, and open platform, in that it will connect to any type of QPU (quantum processing unit) backend.
Recommended Resources
Hosted on the Python Package Index (PyPI), the CUDA Quantum Python API is a Python library that provides bindings for the CUDA Quantum toolkit, enabling high-performance simulation of quantum processors using Nvidia's quantum SDK. The API is available on PyPI and can be installed via pip. However, it is not a standalone library; it requires the installation of the CUDA Quantum toolkit and the cuQuantum SDK. Documentation, statistics, and installation instructions are given.
https://pypi.org/project/cuda-quantum/
The Python Package Index (PyPI) discusses the NVIDIA cuQuantum SDK, a high-performance library for quantum information science and other high-performance applications. It consists of cuStateVec for state vector computations, and cuTensorNet for tensor network computations, as well as Python APIs via cuQuantum Python. Installation instructions are provided, along with a project description, release history, project links, statistics, licensure data, and maintainers.
https://pypi.org/project/cuquantum/
Getting Started with Accelerated Computing in CUDA C/C++
The Nvidia Deep Learning Institute offers an eight-hour online course designed to teach students who meet the stated prerequisites to accelerate and optimize existing C/C++ CPU-only applications using the essential CUDA techniques and Nsight Systems and to understand an iterative style of CUDA development. The prerequisites are stated, along with suggested resources, tools, libraries, and frameworks that are used, the price, and related available training.
https://courses.nvidia.com/courses/course-v1:DLI+S-AC-04+V1/
With a unified and open programming model, CUDA Quantum is an open-source platform for integrated quantum-classical programming. An overview of its features, prerequisites, and running CUDA Quantum is set forth, with links to the language's documentation and the next steps involved in running a CUDA Quantum application. CUDA Quantum is an open-source project, licensed under Apache License 2.0, making use of the Nvidia cuQuantum SDK for high-performance simulation under a separate license.
https://catalog.ngc.nvidia.com/orgs/nvidia/containers/cuda-quantum
CUDA Quantum is a programming model and toolchain for using quantum acceleration in heterogeneous computing architectures available in C++ and Python. Full and searchable documentation for creating applications with CUDA Quantum is provided here, beginning with a section on getting started, and learning the basics, continuing on to advanced topics, examples, tutorials, simulator and hardware backends, specifications, API references, and other versions.
https://nvidia.github.io/cuda-quantum/
CUDA is a parallel computing platform and programming model developed by Nvidia for general computing on graphical processing units, allowing developers to greatly speed up computing applications by harnessing the power of GPUs. The CUDA Toolkit includes GPU-accelerated libraries, a compiler, development tools, and the CUDA runtime, which may be downloaded from the site. Other resources on the Nvidia Developer site include videos, code samples, hands-on labs, tutorials, and webinars.
https://developer.nvidia.com/cuda-zone
Hosted on GitHub, files are available for C++ and Python support for the CUDA Quantum programming model for heterogeneous quantum-classical workflows. An introduction to the repository is provided, with CUDA Quantum documentation and installation instructions, the latest packages, and for building CUDA Quantum from source. Licensure data and information for those interested in developing quantum applications with CUDA Quantum are included. Contributors are noted.
https://github.com/NVIDIA/cuda-quantum
