Skip to content

Various documents and code related to proposals for WG21

License

Notifications You must be signed in to change notification settings

BobSteagall/wg21

Repository files navigation

wg21

CMake Build

Work on revision R7 of linear algebra proposal P1385 is currently underway in the r7y branch. We are targeting C++26.

Support

This reference implementation requires several C++20 capabilities which are only available in newer compiler releases.

Current compiler support is as follows, but will change as new compilers become available and catch up with implementing new language features for C++20. Builds and unit testing have been successful on the following combinations of operating system and compiler:

  • Windows 10

    • Visual Studio 2019, Version 16.5.4 and higher
  • Linux (written and tested on RHEL 8.X)

    • GCC 12.x

Cloning the Repo

After ensuring that git is in your executable path, open a terminal window on Linux, or a Command Prompt on Windows 10, and then:

cd <repo_root>
git clone --recurse-submodules https://github.com/BobSteagall/wg21.git <project_root>

If you do not specify <project_root>, it will default to wg21.

Building with the Visual Studio Solution

Open the lin_alg_test.sln solution file in the <repo_root>/<project_root> directory. Click Build..Build Solution to build the unit test program. (Note: This build method uses the NuGet package manager to download and install Google Test in the <project_root>/packages directory, so you'll need to ensure NuGet is installed.)

Building Manually Via CMake on Windows 10

The unit test program can be built via CMake on Windows 10 by proceeding as follows. First, generate the Visual Studio project files:

cd <project_root>
mkdir build-win
cd build-win
cmake -G "Visual Studio 16 2019" ..

Next, build the Debug configuration and run the unit test:

cmake --build . --config Debug
ctest -C Debug

Finally, build the Release configuration and run the unit test:

cmake --build . --config Release
ctest -C Release

Note: CMake uses a multi-configuration generator for Visual Studio, so there is no need to specify a CMake build type; in fact, CMake will ignore CMAKE_BUILD_TYPE when using the Visual Studio generator.

The following CMake configuration options are available on Windows:

Name Possible Values Description Default Value
BUILD_TESTING ON, OFF Build the test suite ON
LA_BUILD_USING_PCH ON, OFF Build using precompiled headers OFF

Building Manually Via CMake on Linux

The project can be built via CMake on Linux as follows:

cd <project_root>
mkdir build-dir
cd build-dir
cmake [-G <generator>] [configuration options] ..
cmake --build . [--config <build_type>]
ctest [--config <build_type>]

For example, to build and test the Debug configuration, you could do the following:

cd <project_root>
mkdir build-debug
cd build-debug
cmake -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=Debug ..
cmake --build .
ctest

Likewise, for the Release configuration, you could do this:

cd <project_root>
mkdir build-release
cd build-release
cmake -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=Release ..
cmake --build .
ctest

Note: CMake uses a single-configuration generator when generating makefiles, so there is no need to use the --config flag with CMake or the -C flag with CTest. When using CMake on Linux, a good practice is to create a separate build directory for each CMake build type.

The following configuration options are available on Linux:

Name Possible Values Description Default Value
BUILD_TESTING ON, OFF Build the test suite ON
CMAKE_BUILD_TYPE Debug, Release, RelWithDebInfo Build type Debug
LA_BUILD_USING_PCH ON, OFF Build using precompiled headers OFF
LA_ENABLE_SANITIZERS ON, OFF Build with address and UB sanitizers OFF
LA_VERBOSE_TEST_OUTPUT ON, OFF Write verbose test results OFF

Installing Via CMake

Installing the project can be performed as follows:

cd <project root>
mkdir build
cd build

cmake [-G <generator>] [configuration options] -DCMAKE_INSTALL_PREFIX=<install_dir> ..
cmake --build . --target install

Packages

The Linear Algebra library is available for integration into your own project via our favorite package manager: Conan.

Getting the Conan package

To add the linear_algebra library to your project as a dependency, you need to add a remote to Conan that points to the location of the package:

pip install conan
conan remote add linear_algebra https://twonington.jfrog.io/artifactory/api/conan/conan-oss

Once this is set, you can add the linear_algebra dependency to you project via the following signature:

linear_algebra/0.7.0@conan-oss/testing

Available versions of the Linear Algebra package can be search via Conan:

conan search linear_algebra

Building the Conan package

The linear_algebra project and package can be built locally via the Conan as such:

cd <project_root>
pip install conan
conan create [conan options] <project_root> [--test-folder None]