C/C++ library to provide a unified API for high-performance convolutions.
- boost (unit test framework, multi-array)
- cmake (to build it)
- c/c++ compiler (notably gcc)
- OpenCL (optional)
$ git clone https://github.com/psteinb/anyfold
$ cd anyfold
$ mkdir build
$ cd build
$ cmake ..
$ make
$ ctestThe following cmake flags are supported:
CMAKE_INSTALL_PREFIXto provide a custom installation directoryBUILD_OPENCL_ANYFOLDto build anyfold with OpenCL support
As this is an educational project (until stable), we target Linux primarily using regular x86 instructions. The ultimate goal is to provide all functionality based on OpenCL (and potentially CUDA).
- CPU functionality is provided in namespace
anyfold::cpu - OpenCL functionality is provided in namespace
anyfold::opencl