Page 1 of 1

question to QR decomposition routine culaDgeqrf

PostPosted: Thu Feb 25, 2016 5:25 am
by jessi

I have a question to the QR decomposition routine culaDgeqrf.
From the documentation of culaDgeqrf, I know that Householder's QR decomposition is used. It also specifies how the output looks like.
But I would like to understand roughly how the QR decomposition is performed in culaDgeqrf.
Is it basically the same algorithm as used in LAPACK's Dgeqrf using block Householder?

Would it be possible to send me some material describing the algorithm basically, giving an idea about why it is suitable for the implementation on GPUs and explaining roughly how parallelization is achieved?
I would also like to know which routines are included in this algorithm.

Thank you very much for your help,