LAPACK Archives

[Lapack] Mistaken error report of October 16, 2009

Hello William,

Thanks for the bug report.

I have looked briefly over the routine xLASQx.

I0 is INPUT only, isn't it?

Now, you are right, it seems that N0 is INPUT/OUPUT.
(as opposed to what is written in the comments of the routines.)

(This is interesting that we did not catch this problem with the new C 
interface testing .... )

So, I do not know these routines very well. We should definetely do 
something. Is there a N0 = N0IN statement missing at the end of the 
routine to restore the value of N0 in output? Or should we simply have the 
variable INPUT/OUTPUT. If we want the variable to be input only, this is 
not that easy ... since an INPUT variable should not be changed at all 
during the execution of the routine in case another thread wants it ... 
(So the N0IN = N0 // N0 = N0IN would not work.)

Let's go with declaring the variable INOUT.



On Fri, 23 Apr 2010, Gandler, William (NIH/CIT) [E] wrote:



? Please remove my error report of October 16, 2009 of an error in version 
3.2 dlasq2.? My error report was based on the assumption that the DLASQ3 
introductory comments of I0 and N0 as input variables were
correct.? In fact, the DLASQ3 introductory comments are incorrect and I0 and 
N0 are input/output variables.? In fact, dlasq2 works correctly.





<Prev in Thread] Current Thread [Next in Thread>

For additional information you may use the LAPACK/ScaLAPACK Forum.
Or one of the mailing lists, or