These are chat archives for elemental/chat

14th
Feb 2017
Aidan Dang
@AidanGG
Feb 14 2017 00:30
Hello again Jack, my issues with this matrix occur on the default 4 process grid (2x2 after my recent pull) if that is any help.
Jack Poulson
@poulson
Feb 14 2017 06:19
does the problem occur on a single process?
Aidan Dang
@AidanGG
Feb 14 2017 07:55
mpirun -np 1 appears to work fine, both with useQR true or false.
Aidan Dang
@AidanGG
Feb 14 2017 08:45
I've checked with a modified version of the SVD test driver, and the results are good when I run on one process
Jack Poulson
@poulson
Feb 14 2017 08:45
I found the problem
it is still there when there is a single process
it has to do with when to force a zero shift in the bidiagonal QR algorithm
essentially I was purposefully avoiding a check if the tolerance was only requested to be relative to the maximal singular value
but it seems the zero shift still needs to be carried out if the conditioning of the subbidiagonal is sufficiently bad
Aidan Dang
@AidanGG
Feb 14 2017 08:47
I apologise, but I don't know too much about the actual details on the algorithm.
Jack Poulson
@poulson
Feb 14 2017 08:47
no worries, it's something for me to fix, but thankfully it will just be a few lines
I'm just trying not to unnecessarily sacrifice more performance
Aidan Dang
@AidanGG
Feb 14 2017 08:50
It will be a great help to me to get this fixed, so I can drop the Scalapack need and go to 64-bit ints.
Jack Poulson
@poulson
Feb 14 2017 08:52
hopefully I can submit the fix to master in the next ten minutes and then cherrypick to the 0.87 branch
thank you for providing this example
FWIW, you can enable the bidiagSVDCtrl.progress = true option to see how the bidiagonal convergence was proceding
the issues were towards the end
Aidan Dang
@AidanGG
Feb 14 2017 08:53
I'm glad to have done my part. I will be looking to test the SVD further.
Jack Poulson
@poulson
Feb 14 2017 09:08
the trivial patch has been submitted
Aidan Dang
@AidanGG
Feb 14 2017 09:08
Great, thank you. I'll check it out.
Jack Poulson
@poulson
Feb 14 2017 09:09
please let me know if that does or does not fix the issue!
(I assume that it will)
Aidan Dang
@AidanGG
Feb 14 2017 09:54
Preliminary testing seems good.