Building with CMake

Building with CMake


Assuming that you have cloned the source repository and selected an appropriate release branch, the following instructions will walk you through the CMake and configuration and build process.

CMake is a cross-platform, open-source build system. CMake controls the compilation process using platform independent configuration files in order to generate Makefiles and workspaces that are native to the platform of choice.

The process involves the following steps:

  1. Locate & build the requisite dependencies
  2. Configure & run CMake to generate Makefiles / MSVC solution / XCode project
  3. Run the build from make / MSVC / XCode

Step 1: Dependencies

CMake will adapt the build based on which dependencies have been successfully discovered and will disable certain features and code examples accordingly.

Please refer to the documentation of each of the dependency packages for specific build and installation instructions.



Step 2: Configuring CMake

One way to configure CMake is to use the CMake GUI. In many cases CMake can fall back on default standard paths in order to find the packages that OpenSubdiv depends on. For non-standard installations however, a complete set of override variables is available. The following sub-section lists some of these variables. For more specific details, please consult the source of the custom CMake modules in the OpenSubdiv/cmake/ folder.

Useful Build Options

The following configuration arguments can be passed to the CMake command line.


-DCMAKE_INSTALL_PREFIX=[base path to install OpenSubdiv (default: Current directory)]
-DCMAKE_LIBDIR_BASE=[library directory basename (default: lib)]

-DCMAKE_PREFIX_PATH=[semicolon-separated list of directories specifying installation prefixes to be searched by the find_package() command (default: empty list)]

-DOSD_CUDA_NVCC_FLAGS=[CUDA options, e.g. --gpu-architecture]

-DPTEX_LOCATION=[path to Ptex]
-DICC_LOCATION=[path to Intel's C++ Studio XE]

-DNO_LIB=1        // disable the opensubdiv libs build (caveat emptor)
-DNO_EXAMPLES=1   // disable examples build
-DNO_TUTORIALS=1  // disable tutorials build
-DNO_REGRESSION=1 // disable regression tests build
-DNO_PTEX=1       // disable PTex support
-DNO_DOC=1        // disable documentation build
-DNO_OMP=1        // disable OpenMP
-DNO_TBB=1        // disable TBB
-DNO_CUDA=1       // disable CUDA
-DNO_OPENCL=1     // disable OpenCL
-DNO_OPENGL=1     // disable OpenGL
-DNO_CLEW=1       // disable CLEW wrapper library

Environment Variables

The paths to Ptex, GLFW, other dependencies can also be specified through the following environment variables:


Automated Script

The GUI solution will probably become a burden for active developpers who tend to re-run the configuration step fairly often. A scripted solution can save a lot of time. Here is a typical workflow:

git clone <folder>
cd <folder>
mkdir build
cd build
source ../../cmake_setup

Where cmake_setup is a configuration script.

Here is an example CMake configuration script for a full typical windows-based build that can be run in GitShell :


# Replace the ".." with a full path to the root of the OpenSubdiv source tree if necessary
"c:/Program Files (x86)/CMake 2.8/bin/cmake.exe" \
    -G "Visual Studio 15 2017 Win64" \
    -D "GLFW_LOCATION:string=c:/Program Files/glfw-2.7.7.bin.WIN64" \
    -D "OPENCL_INCLUDE_DIRS:string=c:/ProgramData/NVIDIA Corporation/NVIDIA GPU Computing SDK 4.2/OpenCL/common/inc" \
    -D "_OPENCL_CPP_INCLUDE_DIRS:string=c:/ProgramData/NVIDIA Corporation/NVIDIA GPU Computing SDK 4.2/OpenCL/common/inc" \
    -D "OPENCL_LIBRARIES:string=c:/ProgramData/NVIDIA Corporation/NVIDIA GPU Computing SDK 4.2/OpenCL/common/lib/x64/OpenCL.lib" \
    -D "PTEX_LOCATION:string=c:/Users/opensubdiv/demo/src/ptex/x64" \

# copy Ptex dependencies (Windows only)
mkdir -p bin/{Debug,Release}
\cp -f c:/Users/opensubdiv/demo/src/zlib-1.2.7/contrib/vstudio/vc10/x64/ZlibDllRelease/zlibwapi.dll bin/Debug/
\cp -f c:/Users/opensubdiv/demo/src/zlib-1.2.7/contrib/vstudio/vc10/x64/ZlibDllRelease/zlibwapi.dll bin/Release/
\cp -f c:/Users/opensubdiv/demo/src/ptex/x64/lib/Ptex.dll bin/Debug/
\cp -f c:/Users/opensubdiv/demo/src/ptex/x64/lib/Ptex.dll bin/Release/


Notice that the following scripts start by recursively removing the ../build/ and ../inst/ directories. Make sure you modify them to suit your build workflow.

Here is a similar script for *Nix-based platforms:

echo "*** Removing build"
cd ..; rm -rf build/ inst/; mkdir build; cd build;
echo "*** Running cmake"
cmake -DPTEX_LOCATION=/home/opensubdiv/dev/opensource/ptex/install \
      -DGLFW_LOCATION=/home/opensubdiv/dev/opensource/glfw/build \
      -DDOXYGEN_EXECUTABLE=/home/opensubdiv/dev/opensource/doxygen/inst/bin/doxygen \
      -DCMAKE_INSTALL_PREFIX=../inst \
      -DCMAKE_BUILD_TYPE=Debug \

Here is a similar script for macOS:

echo "*** Removing build"
cd ..; rm -rf build/ inst/; mkdir build; cd build;
echo "*** Running cmake"
cmake -DOPENGL_INCLUDE_DIR=/Applications/ \
      -DGLFW_LOCATION=/Users/opensubdiv/dev/opensource/glfw/inst \
      -DCMAKE_INSTALL_PREFIX=../inst \
      -DCMAKE_BUILD_TYPE=Debug \

Using Intel's C++ Studio XE

OpenSubdiv can be also be built with Intel's C++ compiler (icc). The default compiler can be overriden in CMake with the following configuration options:

-DCMAKE_CXX_COMPILER=[path to icc executable]
-DCMAKE_C_COMPILER=[path to icc executable]

The installation location of the C++ Studio XE can be overriden with:

-DICC_LOCATION=[path to Intel's C++ Studio XE]

Using Clang

CMake can also be overriden to use the clang compilers by configuring the following options:


Step 3: Building

CMake provides a cross-platform command-line build:

cmake --build . --target install --config Release

Alternatively, you can use native toolkits to launch the build. The steps differ for each OS:

  • Windows :

    launch VC++ with the solution generated by CMake in your build directory.

  • macOS :

    launch Xcode with the xcodeproj generated by CMake in your build directory

  • *Nix :
    run make in your build directory
    - use the clean target to remove previous build results
    - use VERBOSE=1 for verbose build output

Build Targets

Makefile-based builds allow the use of named target. Here are some of the more useful target names:

The core components of the OpenSubdiv libraries

Builds specific code examples by name (glViewer, ptexViewer...)

Builds ReST and doxygen documentation

Builds ReST documentation

Builds Doxygen documentation

Compiling & Linking an OpenSubdiv Application

Here are example commands for building an OpenSubdiv application on several architectures:


g++ -I$OPENSUBDIV/include -c myapp.cpp
g++ myapp.o -L$OPENSUBDIV/lib -losdGPU -losdCPU -o myapp


g++ -I$OPENSUBDIV/include -c myapp.cpp
g++ myapp.o -L$OPENSUBDIV/lib -losdGPU -losdCPU -o myapp
install_name_tool -add_rpath $OPENSUBDIV/lib myapp

(On 64-bit OS-X: add -m64 after each g++.)


cl /nologo /MT /TP /DWIN32 /I"%OPENSUBDIV%\include" -c myapp.cpp
link /nologo /out:myapp.exe /LIBPATH:"%OPENSUBDIV%\lib" libosdGPU.lib libosdCPU.lib myapp.obj


HBR uses the offsetof macro on a templated struct, which appears to spurriously set off a warning in both gcc and Clang. It is recommended to turn the warning off with the -Wno-invalid-offsetof flag.