[Octave-bug-tracker] [bug #45008] control: Unit tests for ltimodels and @lit/minreal fail

[Lukas Reichlin]

Follow-up Comment #2, bug #45008 (project octave):

Yes, this is a known "error". The examples used in these tests are from the
SLICOT library. The SLICOT guys strongly recommend the use of reference
BLAS/LAPACK and so do I. One could argue that these tests should be removed,
but I prefer to keep them so users are aware of the fact that they are getting
different results which may be wrong.


octave:5> help test_control 
'test_control' is a script from the file
/Users/lukas/control/inst/test_control.m

 -- Script File: test_control
     Execute all available tests at once.  The Octave control package is
     based on the SLICOT (http://www.slicot.org) library.  SLICOT needs
     BLAS and LAPACK libraries which are also prerequisites for Octave
     itself.  In case of failing tests, it is highly recommended to use
     Netlib's reference BLAS (http://www.netlib.org/blas/) and LAPACK
     (http://www.netlib.org/lapack/) for building Octave.  Using ATLAS
     may lead to sign changes in some entries of the state-space
     matrices.  In general, these sign changes are not 'wrong' and can
     be regarded as the result of state transformations.  Such state
     transformations (but not input/output transformations) have no
     influence on the input-output behaviour of the system.  For better
     numerics, the control package uses such transformations by default
     when calculating the frequency responses and a few other things.
     However, arguments like the Hankel singular Values (HSV) must not
     change.  Differing HSVs and failing algorithms are known for using
     Framework Accelerate from Mac OS X 10.7.




    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?45008>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/