g77: Different results when compiling with/without -fbounds-check ?!

[Richard Ems <r dot ems dot mtg at gmx dot net>]

Hi all,

After 2 days searching for errors and comparing my test case result's with different option combinations, I found
that compiling with -fbounds-check and without it returns (for my test case) different results.

Attached are my two files, the main program file and the subroutine one.

After running:

clear
set -x
for OPT in O0 O3
do
  g77 -$OPT -o testit-$OPT testit.f zwwe.f && time ./testit-$OPT
  mv -f fort.77 fort-$OPT.77
done
diff fort-O0.77 fort-O3.77
set +x

(the same in one line:   clear ; set -x ; for OPT in O0 O3 ; do g77 -$OPT -o testit-$OPT testit.f zwwe.f && time ./testit-$OPT && mv -f fort.77 fort-$OPT.77 ; done && diff fort-O0.77 fort-O3.77 ; set +x   )

several times, I get no differences from the diff command. As you can see, I'm compiling WITHOUT -fbounds-check.

Running the same, also several times, but now compiling WITH -fbounds-check, I get different results for the executable compiled with -O0 and the one compiled with -O3.

( The same as above WITH -fbaound-check :
clear ; set -x ; for OPT in O0 O3 ; do g77 -fbounds-check -$OPT -o testit-$OPT testit.f zwwe.f && time ./testit-$OPT && mv -f fort.77 fort-$OPT.77 ; done && diff fort-O0.77 fort-O3.77 ; set +x   )

Is this supposed to behave like this? I suppose not!

I'm runing on SuSE Linux 6.4, kernel 2.2.16
g77 --version: GNU Fortran 0.5.25 19991030 (prerelease)
gcc --version: 2.95.2
as --version:GNU assembler 2.9.5
 

Many thanks, Richard Ems
 
 

-- 
   Richard Ems
   ... e-mail: r.ems@gmx.net
   ... Computer Science, University of Hamburg

   Unix IS user friendly. It's just selective about who its friends are.

 
 
 
 
 
 
 

      program testit
c
      implicit none
c
      integer i, mynx
      parameter (mynx=100)
      real a, myx, myxu, mydx, zwwe
      real myf(mynx)
c
      myx = acos(-1.)
      myxu = 1.0
      mydx = 1.0
c
      do 1 i=1,mynx,1
         myf(i) = real(i)/100000*rand(i)
 1    continue 
c
      do 2 i=1,10000,1
         myx = myx*(rand(i)+0.6)
         write (77,*) 'in testit: vor zwwe: ', myx, myxu, mydx, mynx
         a = zwwe(myx, myxu, mydx, mynx, myf)
         write (77,*) 'in testit: a = ', a
 2    continue
c
      end
c



      REAL FUNCTION ZWWE (X, XU, DX, NX, F)
C                   ====
C
C KURZBESCHREIBUNG: LINEARE INTERPOLATION IN AEQUIDISTANTEM FELD VON
C                   STUETZWERTEN
C
C AUTOR(EN):
C
C DATUM DER ERSTEN VERSION:
C
C-----------------------------------------------------------------------
C AENDERUNGEN:                   BESCHREIBUNG DER AENDERUNG
C
C GEAENDERT VON:  R. PEREIRA
C DATUM:          25.09.90
C
C                 R. PEREIRA     FALL K1 = NX MITBERUECKSICHTIGT
C                 13.09.94
C
C                 R. Pereira     Kein Abbruch bei Stuetzwert ausserhalb
C                 21.06.2000     des zulaessiges Bereiches.
C
C-----------------------------------------------------------------------
C AUFRUF: ZWWE (X, XU, DX, NX, F)
C
C BESCHREIBUNG DER PARAMETER:
C
C      EINGABE: X  - INTERPOLATIONSSTELLE
C               XU - KLEINSTER WERT DER STUETZPUNKTKOORDINATEN
C               DX - SCHRITTWEITE DER STUETZPUNKTKOORDINATEN
C               NX - ANZAHL DER STUETZPUNKTE
C               F  - FELD DER STUETZWERTE
C
C      AUSGABE: ZWWE - INTERPOLATIONSWERT (LOESUNG)
C
C-----------------------------------------------------------------------
C LISTE DER COMMON BLOCKS:
C
C      KANAL
C
C LISTE DER VARIABLEN:
C
C      VARIABLEN
C
C
C      VEKTOREN UND MATRIZEN
C
C
C-----------------------------------------------------------------------
C AUFGERUFENE SUBROUTINES UND FUNCTIONS:
C
C             KEINE
C
C BEMERKUNGEN:  LIEGT X AUSSERHALB DES DURCH XU, DX UND NX BESTIMMTEN
C               BEREICHES, wird der Grenzwert gesetzt.
C
C LITERATUR:
C
C-----------------------------------------------------------------------
      IMPLICIT NONE
C
C
C Fuer Ultrix
c      INCLUDE '../source/cmn/kanal.cmn'
C Fuer Open/VMS
C      INCLUDE '[PEREIRA.SIMBEL.BACKUP.SOURCE]KANAL.CMN'
C
      INTEGER K1, NX
      REAL X, XU, DX, X1
      REAL F(NX)
C
C.......................................................................
      write (77,*) 'X, XU, DX, NX = ', X, XU, DX, NX
C
      K1=INT((X-XU)/DX+0.001)
      write (77,*) 'k1 = ', k1
C
      IF (K1 .GE. NX-1) THEN
         ZWWE = F(NX)
         write (77,*) '1 - zwwe = ', zwwe
         RETURN
      ELSEIF (K1 .LE. 0) THEN
         ZWWE = F(1)
         write (77,*) '2 - zwwe = ', zwwe
         RETURN
      ELSE
         X1 = XU+K1*DX
         K1 = K1+1
         write (77,*) 'x1, k1, dx, f(k1+1), f(k1) = ',
     *                 x1, k1, dx, f(k1+1), f(k1)
         ZWWE = (X-X1)/DX*(F(K1+1)-F(K1))+F(K1)
         write (77,*) '3 - zwwe = ', zwwe
      ENDIF
C
      RETURN
      END
C ENDE -ZWWE-
C***********************************************************************
C***********************************************************************