CUDA library used to assess performance of CUDA GPGPU against popular multi precission arithmetic library GMP.
Concole application to make measurements comparing GMP and cuGMP performance.
- Download and install CUDA toolkit from nvidia.com.
- Download MPIR (GMP fork for Windows/VS) from mpir.org (Tested with version 2.7.2)
- Extract MPIR to repository directory (or anywhere else, but then edit paths in TestApp project properties - Additional Library Directories and Additional Include Directories accordingly)
- In MPIR directory, open build.vc12\mpir.sln (for VS2013) and build lib_mpir_cxx and lib_mpir_gc in your desired configuration (or use batch build and build debug and release configurations)
Now you should be able to start using cuGMP.sln (recommended project configuration - x64)
- Install cuda toolkit
- Install GMP library from your distribution repository or build from sources available at gmplib.org
- Open makefile and check all include paths (primarily CCCUDAINCL variable)
make builds cuGMP library and TestApp
make cuGMP.a builds only cuGMP library