Re: Re: dgelsd_ / Parameter number 4 had an illegal value

[dderigo]

Dear John, 

I just went into this thread which seems to be the latest one discussing the
issue on DLASCL parameter number 4.

A simple code to reproduce the message (tested on GNU Octave 3.2 and 4.0,
both under GNU/Linux) is:




Instead, the code:



appears to loop indefinitely.



Should be it of any use, potentially related reports might be found at 

http://savannah.gnu.org/bugs/?36211 <http://savannah.gnu.org/bugs/?36211>   

and 

https://lists.gnu.org/archive/html/octave-maintainers/2012-04/msg00067.html
<https://lists.gnu.org/archive/html/octave-maintainers/2012-04/msg00067.html>  


If the problem is related to what is reported e.g. in 
https://software.intel.com/en-us/node/520866
<https://software.intel.com/en-us/node/520866>   ( "LAPACK routines assume
that input matrices do not contain IEEE 754 special values such as INF or
NaN values. Using these special values may cause LAPACK to return unexpected
results or become unstable." ), then a possible workaround might exploit a
semantic check on the values of <M> in the operation:



In particular, it might consist in limiting +/- inf values in <M> to +/-
realmax .


Just for testing purposes, something equivalent to: 





appears to provide reasonable results. E.g.:





--
View this message in context: 
http://octave.1599824.n4.nabble.com/dgelsd-Parameter-number-4-had-an-illegal-value-tp4667802p4671147.html
Sent from the Octave - General mailing list archive at Nabble.com.