[Bug fortran/23318] New: program works correctly with -g option but fails with -O option on LINUX

[Bug fortran/23318] New: program works correctly with -g option but fails with -O option on LINUX

[dir at lanl dot gov]

This program run Ok on the Macintosh or on LINUX with the -g option, but fails
with the -O option on LINUX.

dir/junk2> gfortran -w -O -o timefun timefun.f
dir/junk2> timefun < timefun.in
1t i m e   f u n c t i o n   d a t a

     number of time functions               (ntfn) =    1

     max number of points in time functions (nptm) =    2

 time function number   =    1
 number of time points  =    2

     time value      function

        0.00000    0.0000000E+00
        1.00000    0.1000000E+01
 ""  error   time is larger than in the time function
           2           2           1           1   1.00000000000000       
1.00000000000000
STOP 0
dir/junk2> gfortran -w -g -o timefun timefun.f
dir/junk2> timefun < timefun.in
1t i m e   f u n c t i o n   d a t a

     number of time functions               (ntfn) =    1

     max number of points in time functions (nptm) =    2

 time function number   =    1
 number of time points  =    2

     time value      function

        0.00000    0.0000000E+00
        1.00000    0.1000000E+01
STOP 0
dir/junk2> cat timefun.in
    1    2
    1    2
        0.        0.        1.        1.
dir/junk2> cat timefun.f
      program main
      implicit real*8 (a-h,o-z)
      save
      common /sol/ numnp,neq,nwk,nwm,nwc,numest,midest,maxest,nste,ma
      common/const/ dt,dta,acoef(21),dtod,iope
      common a(1000)

      dt=0.1d0
      dta=0.1d0
      nste=10
      itwo=2


      read (5,1010) ntfn,nptm
      if (idatwr.le.1) write (6,2250) ntfn,nptm
c
      if (ntfn.eq.0) go to 15
      m2=n1 + ntfn
      m3=m2 + ntfn*nptm*itwo
      m4=m3 + ntfn*nptm*itwo
      m5=m4 + ntfn*nste*itwo
      m6=m5 + ntfn*itwo - 1
c
      call timfun (a(m5),a(n1),a(m2),a(m3),a(m4),ntfn,nptm)
   15 continue
      stop

 1010 format (16i5)
 2250 format (1h1,35ht i m e   f u n c t i o n   d a t a   //4x,
     148h number of time functions               (ntfn) =,i5//4x,
     248h max number of points in time functions (nptm) =,i5)
      end


      subroutine timfun (rgst,ipnt,timv,rv,rg,ntfn,nptm)
c
c . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
c .                                                                   .
c .   subroutine to calculate time function values at all time points .
c .   the time function values are stored in rg                       .
c .                                                                   .
c . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
c
      implicit real*8 (a-h,o-z)
      save
c
      common /isubst/ isub,nsubst,nsub,ntuse,negls,negnls,numnps,
     1                nodcon,nodret,idofs(6),ndofs,neqs,nwks,maxes,
     2                mas,nstape,iloa(9),krsizm,neqc
      common /sol/ numnp,neq,nwk,nwm,nwc,numest,midest,maxest,nste,ma
      common /var/ ng,modex,iupdt,kstep,itemax,ieqref,ite,kpri,
     1             iref,iequit,ipri,kplotn,kplote
      common/const/ dt,dta,acoef(21),dtod,iope
      common /adinai/ opvar(7),tstart,irint,istote
      common /prcon/ idatwr,ipric,npb,idc,ivc,iac,ipc,ipnode(3,15)
c
      dimension rg(ntfn,1),timv(nptm,1),rv(nptm,1),ipnt(1),rgst(1)
c
c     write(6,*)tstart,dt,dta,nptm,ntfn,nste
      do 100 l=1,ntfn
      read (5,1000) ll,npts
      if (ll - l) 80,90,80
   80 write (6,2000)
      stop
c
   90 if (idatwr.le.1) write (6,2002) l,npts
      ipnt(ll)=npts
      read  (5,1020) (timv(i,ll),rv(i,ll),i=1,npts)
      if (idatwr.gt.1) go to 95
      write (6,2004) (timv(i,ll),rv(i,ll),i=1,npts)
   95 if (npts.le.nptm) go to 100
      write (6,2100) l,npts,nptm
      stop
  100 continue
c
      nt=13
      if (nsubst.gt.0) nt=15
      rewind nt
      do 200 l=1,ntfn
      rgst(l)=rv(1,l)
      npts=ipnt(l)
      time=tstart + dt
      timep=tstart + dta
      i=0
      k=1
  120 i=i + 1
      if (i-npts) 190,130,130
  130 write (6,2010)
      write(6,*)i,npts,ntfn,l,time,timep
      stop
c
  190 ddr=rv(i+1,l) - rv(i,l)
      ddt=timv(i+1,l) - timv(i,l)
      if (ddt) 110,120,150
  110 write (6,2020)
      stop
  150 slope=ddr/ddt
  180 if (timv(i+1,l)-time) 120,140,140
  140 rg(l,k)=rv(i,l) + slope*(timep-timv(i,l))
      timep=time + dta
      time=time + dt
      k=k + 1
      if (nste-k) 195,180,180
  195 write (nt) rgst(l),(rg(l,k),k=1,nste),npts,
     1           (rv(j,l),timv(j,l),j=1,npts)
  200 continue
c
      return
c
 1000 format (2i5)
 1020 format (8f10.0)
 2000 format (43h ""  error   time functions out of order   )
 2002 format (/25h time function number   =,i5/
     1           25h number of time points  =,i5//4x,
     2           25h time value      function/)
 2004 format (3x,f12.5,2x,e15.7)
 2010 format (53h ""  error   time is larger than in the time function)
 2020 format (42h ""  error   time points are out of order  )
 2100 format (///28h *** i n p u t   e r r o r -//
     1        30h detected by subroutine timfun/
     2        30h while reading time functions //
     3     5x,23h time function number =,i5/
     4     5x,36h number of points in this function =,i5,
     5        17h  is greater than/
     6     5x,36h the max number of points requested=,i5,
     7        49h  as specified on the time function control card. //
     4        12h *** s t o p)
c
      end

-- 
           Summary: program works correctly with -g option but fails with -O
                    option on LINUX
           Product: gcc
           Version: 4.0.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: fortran
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: dir at lanl dot gov
                CC: gcc-bugs at gcc dot gnu dot org
  GCC host triplet: Suse Linux professional 9.3


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23318

[Bug fortran/23318] program works correctly with -g option but fails with -O option on LINUX

[kargl at gcc dot gnu dot org]

------- Additional Comments From kargl at gcc dot gnu dot org  2005-08-10 20:19 -------
I suggest running the code through ftnchek and fixing potential
problems.  These looks suspicious

Warning in module MAIN in file show_bug.f: Variables used before set
    IDATWR used at line 15 file show_bug.f; never set
    N1 used at line 18 file show_bug.f; never set

Warning: Common block ISUBST Elements used but never set:
    NSUBST

Warning: Common block ADINAI Elements used but never set:
    TSTART

Warning: Subprogram TIMFUN argument data type mismatch at position 2:
    Dummy arg IPNT in module TIMFUN line 35 file show_bug.f is type intg
    Actual arg A(N1) in module MAIN line 24 file show_bug.f is type real*8

Warning: Subprogram TIMFUN argument arrayness mismatch at position 1:
    Dummy arg RGST in module TIMFUN line 35 file show_bug.f is whole array
    Actual arg A(M5) in module MAIN line 24 file show_bug.f is array element
  and at position 2:
    Dummy arg IPNT in module TIMFUN line 35 file show_bug.f is whole array
    Actual arg A(N1) in module MAIN line 24 file show_bug.f is array element
  and at position 3:
    Dummy arg TIMV in module TIMFUN line 35 file show_bug.f is whole array
    Actual arg A(M2) in module MAIN line 24 file show_bug.f is array element

Warning: Subprogram TIMFUN argument usage mismatch at position 1:
    Dummy arg RGST in module TIMFUN line 35 file show_bug.f is modified
    Actual arg A(M5) in module MAIN line 24 file show_bug.f may be same as arg 
 5: A(M4)
  and at position 2:
    Dummy arg IPNT in module TIMFUN line 35 file show_bug.f is modified
    Actual arg A(N1) in module MAIN line 24 file show_bug.f may be same as arg 
 1: A(M5)
  and at position 3:
    Dummy arg TIMV in module TIMFUN line 35 file show_bug.f is modified
    Actual arg A(M2) in module MAIN line 24 file show_bug.f may be same as arg 
 2: A(N1)



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23318

[Bug fortran/23318] program works correctly with -g option but fails with -O option on LINUX

[dir at lanl dot gov]

------- Additional Comments From dir at lanl dot gov  2005-08-11 13:41 -------
In pulling the routines out, I forgot to set n1=1 and zero the others, but it
does not affect the results. The rest of the warning messages indicate that
ftnchek does not know how to correctly handle a program that does its own
dynamic memory management. In this case, there is enough information for ftnchek
to tell that the program is correct - if it analyzed the program and did the
math symbolically. I suspect that it is impossible to make ftnchek happy with
this type of program.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23318

[Bug fortran/23318] program works correctly with -g option but fails with -O option on LINUX

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-11 13:43 -------
Hmm, -fno-tree-loop-im fixes it for some reason

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23318

[Bug tree-optimization/23318] [4.1 Regression] program works correctly with -g option but fails with -O option on LINUX

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-11 13:48 -------
Hmm, this worked in 4.0.0.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|fortran                     |tree-optimization
   GCC host triplet|Suse Linux professional 9.3 |
 GCC target triplet|                            |i686-linux
           Keywords|                            |alias, wrong-code
            Summary|program works correctly with|[4.1 Regression] program
                   |-g option but fails with -O |works correctly with -g
                   |option on LINUX             |option but fails with -O
                   |                            |option on LINUX
   Target Milestone|---                         |4.1.0
            Version|4.0.0                       |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23318

[Bug tree-optimization/23318] [4.1 Regression] program works correctly with -g option but fails with -O option on LINUX

[dir at lanl dot gov]

------- Additional Comments From dir at lanl dot gov  2005-08-11 19:54 -------
   Here is the slightly revised version that I tried on 9 compilers on four
different system - "gfortran -O" on linux is the only place that it failed.

      program main
      implicit real*8 (a-h,o-z)
      save
      common /prcon/ idatwr,ipric,npb,idc,ivc,iac,ipc,ipnode(3,15)
      common /sol/ numnp,neq,nwk,nwm,nwc,numest,midest,maxest,nste,ma
      common/const/ dt,dta,acoef(21),dtod,iope
      common a(1000)
      n1=1        
      dt=0.1d0
      dta=0.1d0
      nste=10
      itwo=2
      idatwr=0 
                     
      read (5,1010) ntfn,nptm
      write (6,2250) ntfn,nptm
c
      if (ntfn.eq.0) go to 15
      m2=n1 + ntfn
      m3=m2 + ntfn*nptm*itwo
      m4=m3 + ntfn*nptm*itwo
      m5=m4 + ntfn*nste*itwo
      m6=m5 + ntfn*itwo - 1
c
      call timfun (a(m5),a(n1),a(m2),a(m3),a(m4),ntfn,nptm)
   15 continue      
      stop
      
 1010 format (16i5)
 2250 format (1h1,35ht i m e   f u n c t i o n   d a t a   /4x,
     148h number of time functions               (ntfn) =,i5/4x,
     248h max number of points in time functions (nptm) =,i5)
      end
                
                     
      subroutine timfun (rgst,ipnt,timv,rv,rg,ntfn,nptm)
c
c . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
c .                                                                   .
c .   subroutine to calculate time function values at all time points .
c .   the time function values are stored in rg                       .
c .                                                                   .
c . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
c
      implicit real*8 (a-h,o-z)
      save
c
      common /isubst/ isub,nsubst,nsub,ntuse,negls,negnls,numnps,
     1                nodcon,nodret,idofs(6),ndofs,neqs,nwks,maxes,
     2                mas,nstape,iloa(9),krsizm,neqc
      common /sol/ numnp,neq,nwk,nwm,nwc,numest,midest,maxest,nste,ma
      common /var/ ng,modex,iupdt,kstep,itemax,ieqref,ite,kpri,
     1             iref,iequit,ipri,kplotn,kplote
      common/const/ dt,dta,acoef(21),dtod,iope
      common /adinai/ opvar(7),tstart,irint,istote
      common /prcon/ idatwr,ipric,npb,idc,ivc,iac,ipc,ipnode(3,15)
c
      dimension rg(ntfn,1),timv(nptm,1),rv(nptm,1),ipnt(1),rgst(1)
c
c     write(6,*)tstart,dt,dta,nptm,ntfn,nste
      do 100 l=1,ntfn
      read (5,1000) ll,npts
      if (ll - l) 80,90,80
   80 write (6,2000)
      stop
c
   90 if (idatwr.le.1) write (6,2002) l,npts
      ipnt(ll)=npts
      read  (5,1020) (timv(i,ll),rv(i,ll),i=1,npts)
      if (idatwr.gt.1) go to 95
      write (6,2004) (timv(i,ll),rv(i,ll),i=1,npts)
   95 if (npts.le.nptm) go to 100
      write (6,2100) l,npts,nptm
      stop
  100 continue
c
      nt=13
      rewind nt
      do 200 l=1,ntfn
      rgst(l)=rv(1,l)
      npts=ipnt(l)
      time=tstart + dt
      timep=tstart + dta
      i=0
      k=1
  120 i=i + 1
      if (i-npts) 190,130,130
  130 write (6,2010)
      write(6,*)i,npts,ntfn,l,time,timep
      stop
c
  190 ddr=rv(i+1,l) - rv(i,l)
      ddt=timv(i+1,l) - timv(i,l)
      if (ddt) 110,120,150
  110 write (6,2020)
      stop
  150 slope=ddr/ddt
  180 if (timv(i+1,l)-time) 120,140,140
  140 rg(l,k)=rv(i,l) + slope*(timep-timv(i,l))
      timep=time + dta
      time=time + dt
      k=k + 1
      if (nste-k) 195,180,180
  195 write (nt) rgst(l),(rg(l,k),k=1,nste),npts,
     1           (rv(j,l),timv(j,l),j=1,npts)
  200 continue
c
      return
c
 1000 format (2i5)
 1020 format (8f10.0)
 2000 format (43h ""  error   time functions out of order   )
 2002 format (/25h time function number   =,i5/
     1           25h number of time points  =,i5/4x,
     2           25h time value      function)
 2004 format (3x,f12.5,2x,e15.7)
 2010 format (53h ""  error   time is larger than in the time function)
 2020 format (42h ""  error   time points are out of order  )
 2100 format (///28h *** i n p u t   e r r o r -//
     1        30h detected by subroutine timfun/
     2        30h while reading time functions //
     3     5x,23h time function number =,i5/
     4     5x,36h number of points in this function =,i5,
     5        17h  is greater than/
     6     5x,36h the max number of points requested=,i5,
     7        49h  as specified on the time function control card. //
     4        12h *** s t o p)
c
      end
    
        
        

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23318

[Bug tree-optimization/23318] [4.1 Regression] program works correctly with -g option but fails with -O option on LINUX

[kargl at gcc dot gnu dot org]

------- Additional Comments From kargl at gcc dot gnu dot org  2005-08-11 22:33 -------
Dale, I've compiled your new code with both 4.0 and 4.1 with -O0, -O1, -O2, 
and -O3.  It runs and a comparison of the output shows the results that
you expect.  How old is your version(s) of gfortran?  Can you add 
-fdump-tree-all to the command line for the case that fails, and 
send me the *.original and *.optimized files?

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |kargl at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23318

[Bug tree-optimization/23318] [4.1 Regression] program works correctly with -g option but fails with -O option on LINUX

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-11 22:38 -------
(In reply to comment #6)
> How old is your version(s) of gfortran?  
Mine is last night's compiler: gcc version 4.1.0 20050811 (experimental) and I can reproduce Dale's 
issue and found it to be lim which is causing it but how I have not looked.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rakdver at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23318

[Bug tree-optimization/23318] [4.1 Regression] program works correctly with -g option but fails with -O option on LINUX

[sgk at troutmask dot apl dot washington dot edu]

------- Additional Comments From sgk at troutmask dot apl dot washington dot edu  2005-08-11 22:50 -------
Subject: Re:  [4.1 Regression] program works correctly with -g option but fails with -O option on LINUX

On Thu, Aug 11, 2005 at 10:38:59PM -0000, pinskia at gcc dot gnu dot org wrote:
> 
> ------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-11 22:38 -------
> (In reply to comment #6)
> > How old is your version(s) of gfortran?  
> Mine is last night's compiler: gcc version 4.1.0 20050811 (experimental)
> and I can reproduce Dale's issue and found it to be lim which is causing
> it but how I have not looked.
> 

OK.  This may be a regression.  I accidently used the wrong gfortran
compilers.  I need to find my newest builds and run some more tests.

troutmask:sgk[242] gfc41 --version
GNU Fortran 95 (GCC 4.1.0 20050610 (experimental))

troutmask:sgk[243] gfc40 --version
GNU Fortran 95 (GCC 4.0.1 20050527 (prerelease))



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23318

[Bug tree-optimization/23318] [4.1 Regression] program works correctly with -g option but fails with -O option on LINUX

[rakdver at gcc dot gnu dot org]

------- Additional Comments From rakdver at gcc dot gnu dot org  2005-08-16 10:55 -------
I don't see anything suspicious after lim.  But without lim (store motion),
almost no other loop optimizers are able to do anything on this testcase
(everything is a global variable), so perhaps the problem is somewhere else.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23318

[Bug tree-optimization/23318] [4.1 Regression] program works correctly with -g option but fails with -O option on LINUX

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-16 13:21 -------
Ok, then it is a dup of 323 then.

*** This bug has been marked as a duplicate of 323 ***

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |DUPLICATE


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23318

[Bug tree-optimization/23318] [4.1 Regression] program works correctly with -g option but fails with -O option on LINUX

[dir at lanl dot gov]

------- Additional Comments From dir at lanl dot gov  2005-08-16 12:59 -------
I found that using the -ffloat-store option fixed the problem.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23318