LAPACK Archives

[Lapack] patches for scalapack to reduce use of uninitialized variables

Ake and Robert
Thank you very much for the patches.
We will have a ScaLAPACK release very soon and of course I will  
include your corrections.
Just let me know when you have the corrections for pxlahqr ready.
For information we are planning to "detach" LAPACK from ScaLAPACK. It  
means that the user should have LAPACK installed before installing  
ScaLAPACK, and thus has to provide the LAPACK library for the linking.
Let me know if you need anything from me
Thanks again for your contribution
Regards,
Julie langou
On Mar 19, 2007, at 5:51 AM, ?ke Sandgren wrote:

Hi!

The attached set of patches for scalapack 1.7.4 does two things.
1 - reduce the usage of uninitialized variables
2 - fix a couple of incorrect calls to blacs (bad LDA)

The gehdrv patch is just the complete patch related to
https://icl.cs.utk.edu/lapack-forum/viewtopic.php?p=1153#1153

pxsepinfo doesnt initialize THRESH when INFO != 0.

I'm not sure if the pxlahrd patch is the best. Maybe something  
should be
done in pxlarfg instead since alpha isn't set in all cases there, like
myrow != ixrow for row distribution and likewise for column
distribution.

pxlasmsub destroys irow1/icol1 in the "find some norm of the local H"
part.

pxrot used incorrect LDA values for buff in several places, not  
sure if
the intention was to have buff Mx1 or 1xM but it shouldn't really  
matter
should it?

PBLAS/pxscal must not test ALPHA unless it is really going to be used
since scalapack routines sometimes call pxscal with ALPHA  
uninitialized
when myrow != Xrow/mycol != Xcol.

pxstein must initialize ONENRM since it isn't always initialized in  
the
"IF( NBLK.EQ.IBLOCK( NEXT-1 ) .AND. NBLK.NE.OLNBLK ) THEN" case before
being used in the "IF( TMPFAC.GT.ODM18 ) THEN" case. Maybe setting to
ZERO is wrong but its not worse then the original code.

pxtrevc and pxevcdriver are just incorrect LDA param to blacs  
routines.

All these was found with pathscale compiler with -trapuv -O0 -g which
initialized everything to NaN and turns FPE traps on.

Upcoming is a patch to pxlahqr to fix at least some of the  
uninitialized
variables and some other problems.

-- 
Ake Sandgren, HPC2N, Umea University, S-90187 Umea, Sweden
Internet: ake@Domain.Removed   Phone: +46 90 7866134 Fax: +46 90 7866126
Mobile: +46 70 7716134 WWW: http://www.hpc2n.umu.se
<evcdriver.patch>
<gehdrv.patch>
<gsepreq.patch>
<lahrd.patch>
<lasmsub.patch>
<rot.patch>
<scal.patch>
<stein.patch>
<trevc.patch>
_______________________________________________
Lapack mailing list
Lapack@Domain.Removed
http://lists.cs.utk.edu/listinfo/lapack

******************************
Julie Langou; Innovative Computing Laboratory; Computer Science Dept;
University of Tennessee from Denver, Colorado ;-)
julie@Domain.Removed; http://www.cs.utk.edu/~julie/




-------------- next part --------------
An HTML attachment was scrubbed...
URL: 
http://lists.cs.utk.edu/private/lapack/attachments/20070320/89c824d5/attachment.html
 

<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