CookMem is header-only, so building is only needed to run the tests, examples and performance tests, or to generate the documentation. To use CookMem in your own project, see Installation.

Prerequisites

CMake 3.14 or later.
A C++11 compiler, such as GCC, Clang or Visual Studio.
For code coverage (cov builds on Linux): gcov, lcov and genhtml.
For documentation: Doxygen and Graphviz (for the class diagrams).

On Ubuntu, for example:

sudo apt-get install g++ cmake lcov doxygen graphviz

Linux

./make.sh [type]

type can be one of the following. The default is debug. Each type builds in a directory of the same name.

debug - builds with debug information and runs the tests. The internal assertions are enabled (COOKMEM_DEBUG).
cov - like debug, but with code coverage. The performance tests are skipped. The report is generated at cov/test_coverage/index.html.
release - builds with optimization and runs the tests, including the performance tests.
profiler - like release, but built with -pg for profiling with gprof.
docs - generates the Doxygen documentation into the docs directory.
help - prints the usage.

To run a single test, use ctest in the build directory, for example cd debug && ctest -R test_rbtree.

Windows

make.cmd [type]

type can be debug (the default), release or profiler. It builds with CMake's default generator, normally Visual Studio, and runs the tests. The build directory is named after the type with a win- prefix, such as win-debug, so that it does not clash with the directories of make.sh when the same checkout is built from both Windows and WSL.

Code coverage, documentation and the performance tests are only available on Linux.

Using CMake Directly

The scripts are wrappers around CMake. For example:

mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=Release ..
cmake --build .
ctest

Pass -DCOVERAGE=1 for code coverage (with a Debug build) or -DBUILD_DOC=1 for documentation.

Performance Tests

The performance tests compare CookMem with dlmalloc, using both the red-black tree and the AVL tree. They are built by the release type and only run on Linux.

./make.sh release

./perftest.sh [build directory] [report file]

perftest.sh runs each test 5 times (set RUNS to change this), drops the best and the worst time of each allocator, and averages the rest. It prints a markdown table that can be pasted into README.md. It also writes the results to performances/perfdata.js, which the performance chart displays.