MAPS is a header-only C++ CUDA template library for automatic optimization of GPU kernels and transparent partitioning of multi-GPU tasks. It uses memory access patterns to provide near-optimal performance while maintaining code simplicity.
For more information, see the framework website at: http://www.cs.huji.ac.il/project/maps/
CUDA 7.0 or higher.
gflags (for command-line arguments in samples): https://github.com/gflags/gflags
Google Test (for unit tests): https://github.com/google/googletest
To compile code with MAPS, use the includes under the "include" directory.
It is generally recommended to include MAPS using the all-inclusive header (from .cu files only):
#include <maps/maps.cuh>Code samples are available under the "samples" directory. To compile, either use Visual Studio on Windows or CMake on other platforms (http://www.cmake.org/)