slow QR decomposition

[Anton Kostiuk]

Hi guys!

First of all thanks a lot for a great job!

I've got a small remark/desire :)

Current implementation of the QR decomposition seems to be very slow.

On my machine it take quite the same time as fullSVD routine, while numerically speaking it should be faster.

I've checked out the corresponding source file.

It seems, that the SimpleBlas.ormqr routine makes it so slow.

Is there a way (without going into native BLAS) to avoid calling this routine and still get Q and R matrices as an outcome of QR decomposition?

If not, that'd be nice to have a fast alternative for the next release :) 

Thanks in advance!

Anton

p.s. sorry if the last sentence is inappropriate, I'm quite new to such a forums.

[Mikio Braun]

Hi Anton,

LAPACK doesn't compute Q explicitly but rather some intermediate
representation which it can then use to compute the result of
multiplying with Q (that's what ormqr does).

So I'm using ormqr to multiply with an identity matrix to explicitly
compute Q. Maybe that's not the most efficient way to do that... .

-M

> --
> You received this message because you are subscribed to the Google Groups
> "jblas-users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to jblas-users...@googlegroups.com.
> To post to this group, send email to jblas...@googlegroups.com.
> Visit this group at http://groups.google.com/group/jblas-users.
> For more options, visit https://groups.google.com/d/optout.



--
Mikio Braun - http://blog.mikiobraun.de, http://twitter.com/mikiobraun