This small code base proposes an implementation of a reasonably fast and flexible tasking system. Note that a lengthy description may be found in tasking.hpp. Some useful comments in particular about the implementation may be found in tasking.cpp
The code base uses CMake to build. CMake can be downloaded at www.cmake.org The classical CMake profiles are supported (Debug, Release, RelWithDebInfo, MinSizeRel).
- On Windows, use basically the user interface to generate the VS solutions.
- On Linux, you can use ccmake to get the similar user interface.
- The code should work on MacOS but I did not have a Mac to test it.
The code was tested and compiled on:
- Linux 64 bits. You need a relatively new compiler since the code use unordered_map
- Windows 32 bits with VS2008 (but memory debugger which uses unordered_map is not suppored)
- Windows 32/64 bits with VS2010
Beyond the build mode, you may choose to have:
- a memory debugger (it will slow down the system considerably since it locks malloc/free) by setting the variable PF_DEBUG_MEMORY with CMake
- a blob which compiles the program with one big cpp file (use PF_USE_BLOB)
In tasking.hpp, you have some options to configure the tasking system.
Also, note that parts of the code (related to mutexes/conditions/multi-platform abstraction) were directly taken from Intel Embree project: http://software.intel.com/en-us/articles/embree-photo-realistic-ray-tracing-kernels/
The code only includes stress tests you may find in utests.cpp
Benjamin Segovia (segovia.benjamin@gmail.com)