Skip to content

Latest commit

 

History

History
217 lines (145 loc) · 12.2 KB

DEVELOP.md

File metadata and controls

217 lines (145 loc) · 12.2 KB
Raw

Developer Resources

This document aims to guide users with recommended and advanced workflows to build and use Holoscan SDK. This is generally not the simplest way to use the SDK, so make sure to review the project README before getting started.

⚠️ Disclaimer: we only recommend building the SDK from source if you are a developer of the SDK, or need to build the SDK with debug symbols or other options not used as part of the published packages.

Table of Contents

Building the SDK from source

Prerequisites

  • Prerequisites for each supported platform are documented in the user guide.
  • To build and run the SDK in a containerized environment (recommended) you'll need:

(Recommended) Using the run script

Call ./run build within the repository to build the build container and the CMake project.

  • If you encounter errors during the CMake build, you can execute ./run clear_cache to remove cache/build/install folders

  • Execute ./run build --help for more information

  • Execute ./run build --dryrun to see the commands that will be executed

  • That command can be broken-up in more granular commands also:

    ./run check_system_deps  # ensure the system is properly configured for building
./run build_image        # create the build Docker container
./run build              # run the CMake configuration, build, and install steps

Call the ./run launch command to start and enter the build container.

  • You can run from the install or build tree by passing the working directory as an argument (ex: ./run launch install)
  • Execute ./run launch --help for more information
  • Execute ./run launch --dryrun to see the commands that will be executed
  • Execute ./run launch --run-cmd "..." to execute a bash command directly in the container

Run the examples inside the container by running their respective commands listed within each directory README file.

Cross-compilation

While the Dockerfile to build the SDK does not currently support true cross-compilation, you can compile the Holoscan SDK for the developer kits (arm64) from a x86_64 host using an emulation environment.

  1. Install qemu
  2. Clear your build cache: ./run clear_cache
  3. Rebuild for linux/arm64 using --arch|-a or HOLOSCAN_BUILD_ARCH:
    • ./run build --arch arm64
    • HOLOSCAN_BUILD_ARCH=arm64 ./run build

You can then copy the install folder generated by CMake to a developer kit with a configured environment or within a container to use for running and developing applications.

(Advanced) Docker + CMake

The run script mentioned above is helpful to understand how Docker and CMake are configured and run, as commands will be printed when running it or using --dryrun. We recommend looking at those commands if you want to use Docker and CMake manually, and reading the comments inside the script for details about each parameter (specifically the build() and launch() methods).

(Advanced) Local environment + CMake

⚠️ Disclaimer: this method of building the SDK is not actively tested or maintained. Instructions below might go out of date.

Software Requirements

To build the Holoscan SDK on a local environment, the following versions of dev dependencies are needed (or tested). The last column refers to the stage (FROM) in the Dockerfile where respective commands can be found to build/install these dependencies.

Dependency Min version Needed by Dockerfile stage
CUDA 12.2 Core SDK base
gRPC 1.54.2 Core SDK grpc-builder
UCX 1.15.0 Core SDK ucx-builder
GXF 4.0 Core SDK gxf-downloader
MOFED 23.07 ConnectX mofed-installer
TensorRT 8.6.1 Inference operator base
ONNX Runtime 1.15.1 Inference operator onnxruntime-downloader
LibTorch 2.1.0 Inference operator
(torch plugin)		 torch-downloader-[x86_64|arm64]
TorchVision 0.16.0 Inference operator
(torch plugin)		 torchvision-downloader-[x86_64|arm64]
Vulkan SDK 1.3.216 Holoviz operator vulkansdk-builder
Vulkan loader and
validation layers		 1.3.204 Holoviz operator dev
spirv-tools 2022.1 Holoviz operator dev
V4L2 1.22.1 V4L2 operator dev
CMake 3.24.0 Build process build-tools
Patchelf N/A Build process build-tools

Note: refer to the Dockerfile for other dependencies which are not needed to build, but might be needed for:

  • runtime (openblas/mkl for torch, egl for headless rendering, cloudpickle for distributed python apps, cupy for some examples...)
  • testing (valgrind, pytest, xvfb...)
  • utilities (v4l-utils, ...)

For CMake to find these dependencies, install them in default system paths, or pass CMAKE_PREFIX_PATH, CMAKE_LIBRARY_PATH, and/or CMAKE_INCLUDE_PATH during configuration.

Build example

# Configure
cmake -S $source_dir -B $build_dir \
  -G Ninja \
  -D CMAKE_BUILD_TYPE=Release \
  -D CUDAToolkit_ROOT:PATH="/usr/local/cuda"

# Build
cmake --build $build_dir -j

# Install
cmake --install $build_dir --prefix $install_dir

The commands to run the examples are then the same as in the dockerized environment, and can be found in the respective source directory READMEs.

Runtime Container

There are multiple containers associated with Holoscan:

  • The build container generated by the top-level Dockerfile is designed to pull dependencies to build and test the SDK itself. The image does not contain the SDK itself, as it is mounted with during docker run to run the cmake build or run tests.
  • The development container available at NGC | Holoscan Container which includes all the development tools and libraries needed to build Holoscan applications.
    • This image is ~13 GB when uncompressed. However, once a Holoscan application is created, it does not need all those same development tools just to run an application.
  • To address this, a runtime container can now be generated with the runtime_docker/Dockerfile which contains only the runtime dependencies of the Holoscan SDK.
    • This Dockerfile is based on the CUDA-base image, which begins with Ubuntu:22.04 and installs the CUDA runtime and Compat package.
    • This image is ~8.7 GB on x86_64, and can be further reduced based on use cases (see below).

⚠️ Disclaimer: Currently iGPU is not supported by the runtime container

Generate the runtime container

The run script contains the command build_run_image to build the runtime Holoscan SDK image:

./run build_run_image

Once this image is built, it can be run exactly as the Holoscan development container on NGC is. Simply follow the 'Running the container' instructions beginning at step #3 at NGC | Holoscan Container, but replace ${NGC_CONTAINER_IMAGE_PATH} by holoscan-sdk-run-<arch>[-<gpu>] in step #4 (name outputted at the end of the above command).

Further Reducing Runtime Size

If you have a specific application you wish to deploy, you can further reduce this runtime image size in two ways:

  1. Targeting different stages of the runtime Dockerfile.

    1. add --cpp to the command above to not pull in python dependencies.
    2. add --cpp-no-mkl to the command above to not pull in MKL (x86_64-only libtorch dependency) in addition to the above.

  2. Modifying the Dockerfile

The runtime Dockerfile is thoroughly documented to indicate which dependency is used by which component of the Holoscan SDK. If you do not use some of these components (ex: Torch inference backend, ONNX Runtime inference backend, TensorRT inference backend, Python/Cupy, format_converter operator, etc...), comment out the appropriate line in the Dockerfile and run the build command above.

Utilities

Some utilities are available in the scripts folder, others closer to the built process are listed below:

Testing

Existing tests are using GTest for C++ and pytest for Python, and can be found under tests and python/tests respectively. The Holoscan SDK uses CTest as a framework to build and execute these tests.

Run the tests using the following command:

./run test

Note: Run run test --help to see additional options.

Linting

Run the following command to run various linting tools on the repository:

./run lint # optional: specify directories

Note: Run run lint --help to see the list of tools that are used. If a lint command fails due to a missing module or executable on your system, you can install it using python3 -m pip install <tool>.

Building the User Guide

The source of the user guide hosted at https://docs.nvidia.com/holoscan/sdk-user-guide is located in docs. It can be built with the following commands:

  • PDF: ./run build_pdf
  • HTML: ./run build_html (auto-reload: ./run live_html)

Run ./run help for more commands related to the user guide documentation.

VSCode

Visual Studio Code can be utilized to develop the Holoscan SDK. The .devcontainer folder holds the configuration for setting up a development container with all necessary tools and libraries installed.

The ./run script contains vscode and vscode_remote commands for launching Visual Studio Code in a container or from a remote machine, respectively.

  • To launch Visual Studio Code in a dev container, use ./run vscode.
  • To attach to an existing dev container from a remote machine, use ./run vscode_remote. For more information, refer to the instructions from ./run vscode_remote -h.

Once Visual Studio Code is launched, the development container will be built and the recommended extensions will be installed automatically, along with CMake being configured.

Configuring CMake in the Development Container

For manual configuration of CMake, open the command palette (Ctrl + Shift + P) and run the CMake: Configure command.

Building the Source Code in the Development Container

The source code in the development container can be built by either pressing Ctrl + Shift + B or executing Tasks: Run Build Task from the command palette (Ctrl + Shift + P).

Debugging the Source Code in the Development Container

To debug the source code in the development container, open the Run and Debug view (Ctrl + Shift + D), select a debug configuration from the dropdown list, and press F5 to initiate debugging.