Hi all,
I have a weird bug with the CUDA compiler nvcc (v. 4.1, Linux) and Boost::uBLAS (1.48.0). I need to copy data out of uBLAS vectors/matrices into GPU space, hence I need to compile a file containing Boost::ublas structures as well as CUDA code -- hence compiling Boost::uBLAS with nvcc.
I am including ublas/operation.hpp, ublas/matrix.hpp, ublas/matrix_sparse.hpp, ublas/matrix_proxy.hpp, ublas/vector.hpp, and ublas/io.hpp.
The underlying version of GCC is 4.3.2 on a Suse Linux platform. If I comment out all the CUDA code and compile with raw g++, then everything is fine. However, if I compile with nvcc, then I get many errors like the following:
/home/stats/stsiab/include/boost/numeric/ublas/vector_sparse.hpp: In member function ‘const T* boost::numeric::ublas::mapped_vector