Skip to content

Latest commit

 

History

History
194 lines (144 loc) · 8.13 KB

Dev_Setup.md

File metadata and controls

194 lines (144 loc) · 8.13 KB

Setting up your Environment

This document contains a guide for how to set up your development environment if you want to contribute to the code in this repository.

Working inside the provided development container (recommended)

This repository defines a development container to facilitate working in a controlled environment that does not depend on, or interfere with, other software that is installed on your system. We recommend you work within it. To do so, please follow the instructions for enabling Docker on your respective operating system, and then proceed to set up your IDE following the recommendations for developing with VS Code and Docker.

Enabling Docker on Mac and Linux

Docker Desktop is a free one-click-install application for Mac, Linux, or Windows that contains the Docker Engine needed to use our development container, as well as a GUI to manage your containers, images and volumes. It is not necessary to log into an account to use docker if you don't want to create one.

On Linux, it is possible and sufficient to install only the engine. However, we recommend installing Docker Desktop regardless of your operating system for a consistent experience across all platforms.

Enabling Docker on Windows

If you are working with a Windows machine, please enable the Windows Subsystem for Linux by running the command wsl --install. The subsystem gives you an easy way to work with Docker containers, as well as to develop and test code locally for a variety of Linux distributions.

Please install Docker Desktop on Windows to manage your containers. It is not necessary to log into an account if you don't want to create one. Make sure to select "Use WSL 2 instead of Hyper-V" during installation.

Developing with VS Code and Docker

VS Code is a robust cross-platform IDE that integrates well with Docker through the use of an extension. Extensions let you add languages and others tools to support specialized development workflows. To work with our development container please install the Development Containers extension.

Create a local clone of your fork of the CUDA-Q repository, navigate to that folder, and open VS Code by executing the command code .. You should now get a pop up offering to open the folder in the container. Please select this option to work in a pre-configured environment that contains all necessary dependencies to build CUDA-Q:

  • On Mac/Linux: Select the option Open Folder in Container
  • On Windows: Select the option Clone Repository in Container Volume for the best performance. You can also select Open Folder in Container, which will ensure that changes you made inside the container environment are visible also outside the container volume. For more information, see also this guide.

After you have selected the above option to launch the container, the VS Code window will reload and you should see a green box in the lower left corner that states Development Container: cudaq-dev.

The container also defines which extensions will be loaded automatically when launching it. This does not impact your VS Code configuration outside the container environment.

You should now be all set to build CUDA-Q and run tests. Please open a terminal in VS Code and follow the build instructions to confirm that everything works as expected.

Language server support for LLVM and MLIR dialects in VS Code

The development container in this repository is configured with extensions to facilitate working with MLIR and LLVM files. The MLIR extension requires a language server to work properly. The cudaq-lsp-server tool in this repository extends the MLIR Language Server to add support for CUDA-Q specific dialects. It recognizes files with extensions .mlir and .qke.

The cudaq-lsp-server is built when running the CUDA-Q build script as described in Building CUDA-Q from Source. If you customize the installation location, you will need to either add that location to your path, or edit the value of mlir.server_path in your workspace settings for the extension to work properly.

Working in your own environment

If you have followed the instructions for working inside the provided development container (recommended) you are all set and can proceed to Getting Started with Developing CUDA-Q.

If you do not leverage the container definition provided in this repository, you will need to manually install all prerequisites for building the code on this repository, and configure your preferred IDE for working with Python, C++, MLIR, and LLVM components. To install all necessary build and runtime dependencies, please follow the instructions in our official guide for building and installing CUDA-Q from source. The guide details how to get a fully featured CUDA-Q installation that can be deployed on a system other than the one used for the build. For development purposes, components/certain prerequisites can be omitted as needed, and additional tools and flags for debugging purposes may be beneficial.

Working on macOS

macOS is supported for development builds on ARM64 (Apple silicon) only. Intel Macs are not supported. GPU support is not available on macOS. The development container is not available on macOS, so you will need to set up your environment manually.

Prerequisites

  1. Xcode Command Line Tools:

    xcode-select --install

    Verify with xcode-select -p and clang --version.

  2. Homebrew (package manager for installing dependencies):

    /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

  3. Core development dependencies:

    While the build scripts will install cmake/ninja if they are not detected, we recommend installing and managing these with brew:

    brew install cmake ninja

    If you do not use brew they will be installed at ~/.local/bin and we recommend adding it to your path:

    export PATH="$PATH:$HOME/.local/bin"

  4. OpenMP (for multi-threaded CPU simulation):

    If you build with ./scripts/build_cudaq.sh -p (which we recommend), OpenMP (libomp) is built automatically as part of LLVM. However, if you're building without the -p flag, you'll need OpenMP installed separately (follow any brew specific instructions):

    brew install libomp

  5. Python virtual environment:

    python3 -m venv ~/.venv/cudaq
source ~/.venv/cudaq/bin/activate
pip install --upgrade pip
pip install -r requirements-dev.txt

Environment Variables

Note: Most environment variables like LLVM_INSTALL_PREFIX, AWS_INSTALL_PREFIX, and SDKROOT are only needed during the initial CMake configure. Once set, their values are cached in CMakeCache.txt and persist across subsequent builds.

On macOS, the SDK path for system headers is auto-detected via xcrun --show-sdk-path and baked into nvq++ at build time, so you don't need to set SDKROOT manually.

If you run build_cudaq.sh or scripts/install_prerequisites.sh first, all necessary environment variables will be set automatically during the initial configure.

Once set up, proceed to Building CUDA-Q from Source.