Conference turris::fortran

1298.0. "Compilation does not complete using -O5 switch" by NNTPD::"[email protected]" (Donna McLaughlin) Wed May 14 1997 09:26

Hi, 

I have a piece of code which compiles fine using -O4, -fast, etc. But if I try

to use the -O5 switch the compilation never completes (I can compile it using
-O5 -notransform_loops).

e.g.
>f90 -v -O5 -c pyggam.F
>/lib/cpp -DLANGUAGE_FORTRAN -DLANGUAGE_FORTRAN_90 -D__LANGUAGE_FORTRAN__
-D__LANGUAGE_FORTRAN_90__ -Dunix -D__unix__ -D__osf__ -D__alpha pyggam.F >
pyggam.i 
>/usr/lib/cmplrs/fort90/decfort90 -platinum -O5 -I/usr/lib/cmplrs/hpfrtl -o
pyggam.o pyggam.i 

The same thing happens using f77 - V4.1

The code is not long so I have included it below.
Thanks
Donna

 
C*********************************************************************
 
C...The following routines are adapted from
C...SaSgam - parton distributions of the photon
C...by Gerhard A. Schuler and Torbjorn Sjostrand
C...For further information see CERN-TH/95-62.
C...The version found here is NOT suitable for standalone usage.
 
      SUBROUTINE PYGGAM(ISET,X,Q2,P2,F2GM,XPDFGM)
C...Purpose: to construct the F2 and parton distributions of the photon
C...by summing homogeneous (VMD) and inhomogeneous (anomalous) terms.
C...For F2, c and b are included by the Bethe-Heitler formula;
C...in the 'MSbar' scheme additionally a Cgamma term is added.
      DIMENSION XPDFGM(-6:6)
      COMMON/PYINT8/XPVMD(-6:6),XPANL(-6:6),XPANH(-6:6),XPBEH(-6:6),
     &XPDIR(-6:6)
      SAVE /PYINT8/
 
C...Temporary array.
      DIMENSION XPGA(-6:6)
C...Charm and bottom masses (low to compensate for J/psi etc.).
      DATA PMC/1.3/, PMB/4.6/
C...alpha_em and alpha_em/(2*pi).
      DATA AEM/0.007297/, AEM2PI/0.0011614/
C...Lambda value for 4 flavours.
      DATA ALAM/0.20/
C...Mixture u/(u+d), = 0.5 for incoherent and = 0.8 for coherent sum.
      DATA FRACU/0.8/
C...VMD couplings f_V**2/(4*pi).
      DATA FRHO/2.20/, FOMEGA/23.6/, FPHI/18.4/
C...Masses for rho (=omega) and phi.
      DATA PMRHO/0.770/, PMPHI/1.020/
 
C...Reset output.
      F2GM=0.
      DO 100 KFL=-6,6
      XPDFGM(KFL)=0.
      XPVMD(KFL)=0.
      XPANL(KFL)=0.
      XPANH(KFL)=0.
      XPBEH(KFL)=0.
      XPDIR(KFL)=0.
  100 CONTINUE
 
C...Set k0 cut-off parameter as function of set used.
      IF(ISET.LE.2) THEN
        AK0=0.6
      ELSE
        AK0=2.
      ENDIF
 
C...Call VMD parametrization for d quark and use to give rho+omega+ phi.
C...Note scale choice and dipole dampening for off-shell photon.
      P2MX=MAX(P2,AK0**2)
      CALL PYGVMD(ISET,1,X,Q2,P2MX,ALAM,XPGA)
      XFVAL=XPGA(1)-XPGA(2)
      XPGA(1)=XPGA(2)
      XPGA(-1)=XPGA(-2)
      FACUD=AEM*(1./FRHO+1./FOMEGA)*(PMRHO**2/(PMRHO**2+P2))**2
      FACS=AEM*(1./FPHI)*(PMPHI**2/(PMPHI**2+P2))**2
      DO 110 KFL=-5,5
      XPVMD(KFL)=(FACUD+FACS)*XPGA(KFL)
  110 CONTINUE
      XPVMD(1)=XPVMD(1)+(1.-FRACU)*FACUD*XFVAL
      XPVMD(2)=XPVMD(2)+FRACU*FACUD*XFVAL
      XPVMD(3)=XPVMD(3)+FACS*XFVAL
      XPVMD(-1)=XPVMD(-1)+(1.-FRACU)*FACUD*XFVAL
      XPVMD(-2)=XPVMD(-2)+FRACU*FACUD*XFVAL
      XPVMD(-3)=XPVMD(-3)+FACS*XFVAL
 
C...Call anomalous parametrization for d + u + s.
      CALL PYGANO(-3,X,Q2,P2MX,ALAM,XPGA)
      DO 120 KFL=-5,5
      XPANL(KFL)=XPGA(KFL)
  120 CONTINUE
 
C...Call anomalous parametrization for c and b.
      CALL PYGANO(4,X,Q2,P2MX,ALAM,XPGA)
      DO 130 KFL=-5,5
      XPANH(KFL)=XPGA(KFL)
  130 CONTINUE
      CALL PYGANO(5,X,Q2,P2MX,ALAM,XPGA)
      DO 140 KFL=-5,5
      XPANH(KFL)=XPANH(KFL)+XPGA(KFL)
  140 CONTINUE
 
C...Call Bethe-Heitler term expression for charm and bottom.
      CALL PYGBEH(4,X,Q2,P2,PMC**2,XPBH)
      XPBEH(4)=XPBH
      XPBEH(-4)=XPBH
      CALL PYGBEH(5,X,Q2,P2,PMB**2,XPBH)
      XPBEH(5)=XPBH
      XPBEH(-5)=XPBH
 
C...For MSbar subtraction call C^gamma term expression for d, u, s.
      IF(ISET.EQ.2.OR.ISET.EQ.4) THEN
        CALL PYGDIR(X,Q2,P2,AK0,XPGA)
        DO 150 KFL=-5,5
        XPDIR(KFL)=XPGA(KFL)
  150   CONTINUE
      ENDIF
 
C...Store result in output array.
      DO 160 KFL=-5,5
      CHSQ=1./9.
      IF(IABS(KFL).EQ.2.OR.IABS(KFL).EQ.4) CHSQ=4./9.
      XPF2=XPVMD(KFL)+XPANL(KFL)+XPBEH(KFL)+XPDIR(KFL)
      IF(KFL.NE.0) F2GM=F2GM+CHSQ*XPF2
      XPDFGM(KFL)=XPVMD(KFL)+XPANL(KFL)+XPANH(KFL)
  160 CONTINUE
 
      RETURN
      END

[Posted by WWW Notes gateway]

This is still failing in the lastest compiler.
It's a GEM bug.

This will be fixed in the next ECO kit, to be announced very soon.  It'll
appear in the low-numbered notes of this conference when it's ready.  We've
been promising it for some time now and are trying to finish it up... it'll
happen when it does, not before, so don't ask.

/Stan