Skip to content

Latest commit

 

History

History
209 lines (165 loc) · 8.86 KB

BUILD.md

File metadata and controls

209 lines (165 loc) · 8.86 KB

The stitchEm project is built with CMake. The user interface is developed with Qt. The stitching is done on the GPU and requires either CUDA or OpenCL.

Several 3rd party libraries are required to build the project. On Linux and macOS these can be installed with a package manager. Windows currently requires a manual installation and building of the required libraries.

Using ccache to speed up recompilation is recommended, but not required, on Linux and macOS. Using ninja as the build system for quick builds is recommended, but not recquired. A list of IDEs including Qt Creator, Xcode and Visual Studio are supported through their CMake support or CMake generators.

Building with macOS:

To install the dependencies on macOS, use either the MacPorts or Homebrew package manager.

Using MacPorts

# Install build tools
sudo port install CMake bison doxygen yasm ninja ccache

# Install library dependencies
sudo port install opencv glew gsed jpeg libpng \
          tiff faac faad2 ceres-solver glfw glm OpenEXR \
          ffmpeg +gpl2 +librtmp +nonfree

# Configure build
cmake -DCREATE_BOX_PACKAGE=OFF \
      -DGPU_BACKEND_CUDA=ON -DGPU_BACKEND_OPENCL=ON \
      -DQt5_DIR=~/Qt/5.9.6/clang_64/lib/cmake/Qt5 \
      -DMACPORTS=ON \
      -G Ninja \
      stitchEm

# Build!
ninja

Using Homebrew

See the run instructions in the MacOpenCL builder in build.yml for a full OpenCL-build with Homebrew dependencies.

Creating a binary OpenCL release on macOS

git clone https://github.com/stitchEm/stitchEm.git

# checkout master branch, create a tag in the naming scheme Studio-v2.4.1-macOS

# Using Qt 5.12.10 in this example:
echo 5.12.10 > stitchEm/qt.version

# path of the build directory is hardcoded in script that creates the package later
mkdir cmake_build
cd cmake_build

# BUILD_MACOSX_BUNDLE=ON is required to get an .app bundle
# DISABLE_OPENCL_SPIR=ON is important, otherwise there will be kernel compilation errors
# from SPIR incompatibilities

cmake -DBUILD_MACOSX_BUNDLE=ON \
      -DGPU_BACKEND_CUDA=OFF \
      -DGPU_BACKEND_OPENCL=ON \
      -DDISABLE_OPENCL_SPIR=ON \
      -DQt5_DIR=~/Qt/5.12.10/clang_64/lib/cmake/Qt5 \
      -DMACPORTS=ON \
      -G Ninja \
      -DCMAKE_BUILD_TYPE=Release \
      ../stitchEm

ninja

cd ../stitchEm/apps/dmg
env QT_INSTALL=/Users/{$USER}/Qt ./make_studio_dmg_opencl.sh

# on succes, a VideoStitch-Studio.dmg should have been created in the current directory

# -> extract app to another folder, and zip for upload to a new GitHub release
# (GitHub doesn't allow .dmg uploads)

Building with Linux:

# Installing the dependencies on Ubuntu:
sudo apt install
  bison \
  clang-tools \
  cmake \
  doxygen \
  flex \
  libceres-dev \
  libeigen3-dev \
  libfaac-dev \
  libfaad-dev \
  libglew-dev \
  libglfw3-dev \
  libglm-dev \
  libmp3lame-dev \
  libopencv-dev \
  libopenexr-dev \
  libpng-dev \
  libportaudio-ocaml-dev \
  librtmp-dev \
  libturbojpeg0-dev \
  libx264-dev \
  ninja-build \
  ocl-icd-opencl-dev \
  opencl-headers \
  portaudio19-dev \
  qt5-default \
  qtmultimedia5-dev \
  qttools5-dev \
  swig \
  wget \
  xxd \


# Install CUDA: https://developer.nvidia.com/cuda-downloads

# Configure build
cmake -DGPU_BACKEND_CUDA=ON -DGPU_BACKEND_OPENCL=OFF -DRTMP_NVENC=OFF \
      -G Ninja \
      stitchEm

ninja

Building on windows

You need visual studio 2017, QT >= 5.9 and CUDA 10 Install vcpkg Then installs all of this:

./vcpkg install --triplet x64 ceres eigen3 ffmpeg[avresample,core,gpl,x264,opencl] gflags glfw3 glog libjpeg-turbo liblzma libpng librtmp libwebp mp3lame opencl opencv3 openexr opengl openssl openvr portaudio protobuf tiff x264 zlib glm
./vcpkg install glew:x64-windows-static

clone the repo and create a directory next to it, and configure with cmake:

git clone https://github.com/stitchEm/stitchEm.git
mkdir build
cd build
cmake -G "Visual Studio 15 2017 Win64" \
    -DCMAKE_TOOLCHAIN_FILE=PATH_TO_YOUR_vcpkg_REPOSITORY\scripts\buildsystems\vcpkg.cmake \
    -DQt5_DIR=PATH_TO_QT_5\msvc2017\lib\cmake\Qt5 \
    -DGPU_BACKEND_CUDA=ON \
    -DGPU_BACKEND_OPENCL=OFF \
    -DCUDA_LOCAL_ARCH_ONLY=ON \
    -DVCPKG_ROOT="PATH_TO_YOUR_vcpkg_REPOSITORY" \
    -DVCPKG_TARGET_TRIPLET="x64-windows" \
    -DMAGEWELL_PATH="PATH_TO_MAGEWELL_SDK" \
    -DXIMEA_PATH="PATH_TO_XIMEA_API" \
    -DDECKLINK_PATH="PATH_TO_DECK_LINK_OVR" \
    -DINTEL_MEDIA_SDK_PATH="PATH_TO_INTEL_SDK" \
    ..\stitchEm\

Then open the generated project with visual studio and build it.

CMake flags

Global options

Option Default Usage
BUILD_STATIC_LIB OFF
ON for Windows		 Static library for unit tests
BUILD_IO_PLUGINS ON I/O plugins
BUILD_APPS ON VideoStitch Studio, Vahana VR
GENERATE_BINDINGS OFF libvideostitch Python bindings
CMAKE_CXX_INCLUDE_WHAT_YOU_USE undefined Linux/macOS. Specifies the path to the "include what you use" tool.

libvideostitch options

Option Default Usage
GPU_BACKEND_CUDA ON Build CUDA backend
GPU_BACKEND_OPENCL OFF Build OpenCL backend
GPU_BACKEND_DEFAULT CUDA If both backend builds are enabled, set the default one (for tests and static lib)
AUTOCROP_CMD ON

OpenCL backend

Option Default Usage
DISABLE_OPENCL_SPIR OFF Turn off SPIR compilation, put OpenCL source code in lib binary

CUDA backend

Option Default Usage
FASTMATH ON CUDA compilation flag
LINEINFO OFF CUDA compilation flag
CUDA_LOCAL_ARCH_ONLY OFF Recommended to turn ON for local dev builds to reduce build time
CUDA_TARGET_ARCH undefined Build for specific GPU arch that is different from local GPU. A list of target capabilies can be set (e.g. "52,50").

UI (APPS)

Option Default Usage
Qt5_DIR undefined Qt5 install path

macOS

Option Default Usage
MACPORTS ON ON if you want to use MacPorts, OFF to use Homebrew