[Bug libfortran/41157] New: dtime not consistent in times reported

[Bug libfortran/41157] New: dtime not consistent in times reported

[fkrogh#gcc at mathalacarte dot com]

The following code was compiled with 
gfortran -g -o test_dtime test_dtime.f

The code is given by (very similar to that given in the info file for gfortran

      program test_dtime
      integer(8) :: i, j, k
      real, dimension(2) :: tarray
      real :: result
      real :: accum
      call dtime(tarray, result)
      print '(''tarray,result:'', 1P, 3E20.11)', tarray, result
      do i=1,100000000          ! Just a delay
        j = i * i - i
      end do
      call dtime(tarray, result)
      print '(''tarray,result:'', 1P, 3E20.11)', tarray, result
      accum = 0.e0    ! Should accumulate 10 times previous tarray(1)
      call dtime(tarray, result)
      do k=1, 10
        do i=1,100000000        ! Just a delay
          j = i * i - i
        end do
        call dtime(tarray, result)
        accum = accum + tarray(1)
      end do
      print '(''accum ='', 1P, E20.11 )', accum
      end program test_dtime

The resulting output is 

tarray,result:   9.99000039883E-04   9.99000039883E-04   1.99800007977E-03
tarray,result:   3.86941999197E-01   0.00000000000E+00   3.86941999197E-01
accum =   1.35349416733E+01

Note that accum should be close to 10 times the value the tarray(1) above, that
is, close to 3.8.  I have no idea where the 13.5... came from.


-- 
           Summary: dtime not consistent in times reported
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libfortran
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: fkrogh#gcc at mathalacarte dot com
 GCC build triplet: x86_64-pc-linux-gnu-4.4.1
  GCC host triplet: x86_64-pc-linux-gnu-4.4.1
GCC target triplet: x86_64-pc-linux-gnu-4.4.1


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41157

[Bug libfortran/41157] dtime not consistent in times reported

[kargl at gcc dot gnu dot org]

------- Comment #1 from kargl at gcc dot gnu dot org  2009-08-24 17:19 -------
(In reply to comment #0)
> The following code was compiled with 
> gfortran -g -o test_dtime test_dtime.f
> 
> The code is given by (very similar to that given in the info file for gfortran
> 
>       program test_dtime
>       integer(8) :: i, j, k
>       real, dimension(2) :: tarray
>       real :: result
>       real :: accum
>       call dtime(tarray, result)
>       print '(''tarray,result:'', 1P, 3E20.11)', tarray, result
>       do i=1,100000000          ! Just a delay
>         j = i * i - i
>       end do
>       call dtime(tarray, result)
>       print '(''tarray,result:'', 1P, 3E20.11)', tarray, result
>       accum = 0.e0    ! Should accumulate 10 times previous tarray(1)
>       call dtime(tarray, result)
>       do k=1, 10
>         do i=1,100000000        ! Just a delay
>           j = i * i - i
>         end do
>         call dtime(tarray, result)
>         accum = accum + tarray(1)
>       end do
>       print '(''accum ='', 1P, E20.11 )', accum
>       end program test_dtime
> 
> The resulting output is 
> 
> tarray,result:   9.99000039883E-04   9.99000039883E-04   1.99800007977E-03
> tarray,result:   3.86941999197E-01   0.00000000000E+00   3.86941999197E-01
> accum =   1.35349416733E+01
> 
> Note that accum should be close to 10 times the value the tarray(1) above, that
> is, close to 3.8.  I have no idea where the 13.5... came from.
>

Why should it be close to 3.8?

dtime returns the time since the start of the process?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41157

[Bug libfortran/41157] dtime not consistent in times reported

[fkrogh#gcc at mathalacarte dot com]

------- Comment #2 from fkrogh#gcc at mathalacarte dot com  2009-08-24 17:34 -------
I had thought that was the case as well.  But when I started getting some
negative times by subtracting the previous value from the current one in the
first location of the array I looked at the info file and found

    Subsequent invocations of `DTIME' return values accumulated since
     the previous invocation.

I modified the code to compute the times between invocations.  The code now
looks like 

      program test_dtime
      integer(8) :: i, j, k
      real, dimension(2) :: tarray
      real :: result
      real :: accum, tarray1
      real :: ptimes(10)
      call dtime(tarray, result)
      print '(''tarray,result:'', 1P, 3E20.11)', tarray, result
      do i=1,100000000          ! Just a delay
        j = i * i - i
      end do
      call dtime(tarray, result)
      print '(''tarray,result:'', 1P, 3E20.11)', tarray, result
      accum = 0.e0    ! Should accumulate 10 times previous tarray(1)
      call dtime(tarray, result)
      tarray1 = tarray(1)
      do k=1, 10
        do i=1,100000000        ! Just a delay
          j = i * i - i
        end do
        call dtime(tarray, result)
        ptimes(k) = tarray(1) - tarray1
        tarray1 = tarray(1)
        accum = accum + tarray(1)
      end do
      print '(''accum ='', 1P, E20.11 )', accum
      print '(''ptimes ='', 1P, (3E20.11))', ptimes
      end program test_dtime

And the output is

tarray,result:   0.00000000000E+00   9.99000039883E-04   9.99000039883E-04
tarray,result:   3.80941987038E-01   3.00000002608E-03   3.83941978216E-01
accum =   1.34219570160E+01
ptimes =   7.64882981777E-01  -3.78940939903E-01   7.60881841183E-01
  -3.77939820290E-01   7.63881921768E-01  -3.80940914154E-01
   7.63882875443E-01  -3.78940820694E-01   7.61881709099E-01
  -3.78939628601E-01

There is a pattern here, but why?  If dtime is getting overflows in times less
than a few seconds, there is a problem as well.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41157

[Bug libfortran/41157] dtime not consistent in times reported

[dominiq at lps dot ens dot fr]

------- Comment #3 from dominiq at lps dot ens dot fr  2009-08-24 17:45 -------
> dtime returns the time since the start of the process?

The manual says:

Subsequent invocations of DTIME return values accumulated since the previous
invocation.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41157

[Bug libfortran/41157] dtime not consistent in times reported

[dominiq at lps dot ens dot fr]

------- Comment #4 from dominiq at lps dot ens dot fr  2009-08-24 18:16 -------
If I am not mistaken, the dtime behavior change between 4.2 and 4.3.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41157

[Bug libfortran/41157] dtime not consistent in times reported

[kargl at gcc dot gnu dot org]

-- 

kargl at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |kargl at gcc dot gnu dot org
                   |dot org                     |
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2009-08-24 19:40:51         |2009-08-24 19:41:06
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41157

[Bug libfortran/41157] dtime not consistent in times reported

[kargl at gcc dot gnu dot org]

------- Comment #5 from kargl at gcc dot gnu dot org  2009-08-24 19:40 -------
Confirmed.  

I have a patch.


-- 

kargl at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2009-08-24 19:40:51
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41157

[Bug libfortran/41157] dtime not consistent in times reported

[kargl at gcc dot gnu dot org]

------- Comment #6 from kargl at gcc dot gnu dot org  2009-08-24 20:53 -------
Patch is here:

http://gcc.gnu.org/ml/gcc-patches/2009-08/msg01311.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41157

[Bug libfortran/41157] dtime not consistent in times reported

[kargl at gcc dot gnu dot org]

------- Comment #7 from kargl at gcc dot gnu dot org  2009-08-25 01:47 -------
Subject: Bug 41157

Author: kargl
Date: Tue Aug 25 01:47:23 2009
New Revision: 151072

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=151072
Log:
2009-08-24  Steven G. Kargl  <kargl@gcc.gnu.org>

    PR fortran/41157
    * dtime.c (dtime_sub): Fix computing time increment.
    * time_1.h: Add <sys/types.h> header.  Use RUSAGE_SELF macro instead
    of a hardcoded 0.

Modified:
    trunk/libgfortran/ChangeLog
    trunk/libgfortran/intrinsics/dtime.c
    trunk/libgfortran/intrinsics/time_1.h


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41157

[Bug libfortran/41157] dtime not consistent in times reported

[kargl at gcc dot gnu dot org]

------- Comment #8 from kargl at gcc dot gnu dot org  2009-08-25 03:30 -------
Subject: Bug 41157

Author: kargl
Date: Tue Aug 25 03:30:25 2009
New Revision: 151073

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=151073
Log:
2009-08-24  Steven G. Kargl  <kargl@gcc.gnu.org>

    PR fortran/41157
    * dtime.c (dtime_sub): Fix computing time increment.
    * time_1.h: Add <sys/types.h> header.  Use RUSAGE_SELF macro instead
    of a hardcoded 0.


Modified:
    branches/gcc-4_4-branch/libgfortran/ChangeLog
    branches/gcc-4_4-branch/libgfortran/intrinsics/dtime.c
    branches/gcc-4_4-branch/libgfortran/intrinsics/time_1.h


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41157

[Bug libfortran/41157] dtime not consistent in times reported

[kargl at gcc dot gnu dot org]

------- Comment #9 from kargl at gcc dot gnu dot org  2009-08-25 03:33 -------
Fixed on trunk and 4.4.
Thanks for the bug report.


-- 

kargl at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.4.3


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41157