You can check out the example file titled testblas.d to see how you can use OpenBLAS for low-level matrix computations. Note a couple of things, however.
It’s probably not going to speed things up in most cases. It seems natural to think that using OpenBLAS will lead to better performance for matrix multiplication and other such operations. It will if you’re reusing a matrix/vector. If you already have all vectors and matrices allocated, these functions will be much faster. If you need to allocate a new matrix/vector to hold the result, there’s nothing to gain, since the matrix operations already call into BLAS to do the calculations. The R functions will, in most cases, allocate a new matrix/vector to hold the results. You can avoid that by using BLAS.
Not everything has been demonstrated in that file. I just showed how to use OpenBLAS. I generally don’t have a need to go this low-level, so I have not created a full library.