capp -- C++ Aspect-Oriented Parallel Programming -- uses aspects in c++ to simplify parallel programming with OpenCL. AspectC++ is used to encapsulate most of the overhead of setting up the parallel programming context into an aspect, allowing users to focus on writing the kernels, rather than setting up context and transferring memory.
Currently there is only support for linux and OSX, but Windows support is intended. I have managed to cross-compile this from Linux to Windows 8, so it should working with Windows if the pre-requists are all installed.
To install AspectC++ on Ubuntu simply type sudo apt-get install aspectc++. Alternatively constult AspecC++.
To install OpenCL you need to download the SDK relevant to the device you have. For any CPU or AMD GPU, the AMD SDK will work, for NVIDIA GPU's the NVIDIA OpenCL SDK and cuda development tools are required.
First get the AMD SDK from AMD.
Extract the file and run sudo sh Install-AMD-APP.sh.
Then install the opencl headers. On Ubuntu this is done by sudo apt-get install opencl-headers. Alternatively they can be downloaded from OpenCL-Headers. Whichever way you get the headers make sure that they are version 1.1 or lower if you are going to be using an NVIDIA GPU at some point. Check the cl.hpp file (usually at /usr/local/CL/cl.hpp or /usr/local/include/OpenCL/cl.hpp) and make sure that it is version 1.1.
Download and install the latest CUDA development kit from CUDA and follow their instructions. Again make sure that cl.hpp is version 1.1 (this is very important for NVIDIA).
Two examples are provided for using capp, one for vector addition and the other for option pricing using the Black-Scholes option pricing model.
The examples can be found under examples/capp in the relevant folder. A Makefile is included to build
the files. You will need to specify the number of elements for computation as a command line argument to the
executable.
Examples are also included for sequential implementations at examples/sequential and for opencl
implementation at examples/opencl (the opencl examples are the Nvidia samples implementation).
The intention of including the additional examples is to illustrate how simple the capp c++ files are for
parallel programming (in the /src directories in the relevant example), as most of the overhead is
perofrmed by the aspect before compilation.
- Please do not change the name, move or delete the
tempdirectory in the capp example folder as it is used by the makefile to remove messy object and aspect files from the aspect compiler.
The Makefile must be in the project root directory (see the capp examples folder). The directory names can be changed appropriately in the Makefile by editing the *_DIR variables in the Makefile.
Additionally, make sure that the environment variables in the Makefile you wish to use are appropriate for your environment.
To compile any program make sure that the files are in correct directories as illustrated by the examples provided in the examples/capp directory (the provided Makefile will not work if the structure is changed but not updated in the Makefile).
Once in the example root directory, issue the
make
command and the executable will reside in the builds directory.