Hi,
moving from IPP 7.0 to IPP 2017, I have to replace the discontinued small-matrix functions ippmQRDecomp_m_64f() and ippmQRBackSubst_mv_64f() with suitable functionality from the MKL (as per Intel suggestion). I am using these functions to solve min(||A*x - b||_2) problems.
With IPP 7.0, I have (a) called ippmQRDecomp_m_64f() to QR-decompose the matrix A, (b) stored the resulting decomposition, and (c) repeatedly used the decomposition afterwards (calling ippmQRBackSubst_mv_64f()) to solve for specific vectors b.
For performance reasons, I want to maintain this approach of doing the decomposition only once, and using it repeatedly afterwards, so probably calling the ?gels driver routine is suboptimal. The MKL manpage for ?geqrf suggests calling the MKL functions LAPACKE_dgeqrf, LAPACKE_dormqr and cblas_dtrsm to solve linear least-squares problems.
I haven't used LAPACK before and didn't find a sample for this on the net. Is there sample code showing how to do this?
Thanks for your help,
Daniel