public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug fortran/18218] New: Miscompare in sixtrack benchmark caused by loss of precision
@ 2004-10-29 12:26 lei at il dot ibm dot com
  2004-11-05 16:56 ` [Bug fortran/18218] " jgrimm2 at us dot ibm dot com
                   ` (8 more replies)
  0 siblings, 9 replies; 10+ messages in thread
From: lei at il dot ibm dot com @ 2004-10-29 12:26 UTC (permalink / raw)
  To: gcc-bugs

I could reproduce this problem only on power4 hardware,
running SUSE linux.

% uname -a
Linux regsuse01 2.6.5-7-pseries64 #1 SMP Thu Apr 8 15:55:41 UTC 2004 ppc64 ppc64
ppc64 GNU/Linux

% ~/gcc/bin/gfortran --version
GNU Fortran 95 (GCC 4.0.0 20041027 (experimental))

Running SIXTRACK spec benchmark results in a miscompare,
even with all optimizations turned off.  The miscompare is caused
by the following place in file eig66.f

      do 40 i=1,ndim2
        if(abs(reval(i)**2+aieval(i)**2 -one).gt.c1m10) then
          write(6,*) ' EIG6: EIGENVALUES OFF THE UNIT CIRCLE!'
          return
        endif

This code checks that the distance from computed complex
eigenvalues to the unit circle is no more than 10**-10.
With g77, the distance is about 10**-14, and I get similar
results on other platforms with gfortran.  On power4+gfortran
the distance is about 10**-8, which triggers the error
message and causes miscompare.

There is some possibility that this problem is related
to 17927.

-- 
           Summary: Miscompare in sixtrack benchmark caused by loss of
                    precision
           Product: gcc
           Version: 4.0.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: fortran
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: lei at il dot ibm dot com
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: powerpc64-unknown-linux-gnu
  GCC host triplet: powerpc64-unknown-linux-gnu
GCC target triplet: powerpc64-unknown-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18218


^ permalink raw reply	[flat|nested] 10+ messages in thread
end of thread, other threads:[~2004-11-15  4:14 UTC | newest]

Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2004-10-29 12:26 [Bug fortran/18218] New: Miscompare in sixtrack benchmark caused by loss of precision lei at il dot ibm dot com
2004-11-05 16:56 ` [Bug fortran/18218] " jgrimm2 at us dot ibm dot com
2004-11-08 14:53 ` lei at il dot ibm dot com
2004-11-08 14:54 ` pinskia at gcc dot gnu dot org
2004-11-09 22:09 ` pbrook at gcc dot gnu dot org
2004-11-09 22:11 ` pbrook at gcc dot gnu dot org
2004-11-10  2:16 ` cvs-commit at gcc dot gnu dot org
2004-11-10  2:19 ` cvs-commit at gcc dot gnu dot org
2004-11-11 14:34 ` pbrook at gcc dot gnu dot org
2004-11-15  4:14 ` pinskia at gcc dot gnu dot org

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).