Runtime I/O error in neb.F (neb.f90)

abf123 · Post by **abf123** » Fri Jul 06, 2007 6:03 pm

I have been experiencing a problem at runtime, independent of compiler (it is exhibited in the latest Pathscale, Portland, and the latest Intel) pointing to a formatted/unformatted file conflict. Reverting to an older version of the VTST tools, the problem no longer presents itself - it comes up in the most recent release (2.03 I believe) downloaded within the last week. The cluster on which this is running is an Intel Core 2 platform running RHEL4 and using mpich 1.2.7. The problem occurs on other similar platforms (Opteron w/ Infiniband) as well, so it appears to be platform and MPI independent as well. Is this something that has been previously observed? Please let me know if any further information is needed.

Below is the output from the most informative compilation (pgf90):

1 F= -.92727089E+03 E0= -.92728740E+03 d E =-.927271E+03 mag= 27.4336
BRION: g(F)= 0.187E+00 g(S)= 0.000E+00
bond charge predicted
N E dE d eps ncg rms
rms(c)
RMM: 1 -0.232085386940E+03 -0.23209E+03 -0.12561E+01 4833 0.603E+00
0.127E+00
RMM: 2 -0.231972688443E+03 0.11270E+00 -0.36162E-01 4874 0.763E-01
0.129E+00
RMM: 3 -0.231971369887E+03 0.13186E-02 -0.21210E-02 4849 0.307E-01
0.234E+00
RMM: 4 -0.231944647764E+03 0.26722E-01 -0.12066E-02 4857 0.287E-01
0.106E+00
RMM: 5 -0.231959827324E+03 -0.15180E-01 -0.10017E-02 4831 0.260E-01
0.100E+00
RMM: 6 PGFIO-F-215/unformatted write/unit=8/formatted/unformatted file confli
ct.
File name = 01/OUTCAR formatted, sequential access record = 2705
In source file neb.f90, at line number 254
PGFIO-F-215/unformatted write/unit=8/formatted/unformatted file conflict.
File name = 04/OUTCAR formatted, sequential access record = 2705
In source file neb.f90, at line number 254
PGFIO-F-215/unformatted write/unit=8/formatted/unformatted file conflict.
File name = 03/OUTCAR formatted, sequential access record = 2705
In source file neb.f90, at line number 254
PGFIO-F-215/unformatted write/unit=8/formatted/unformatted file conflict.
File name = 02/OUTCAR formatted, sequential access record = 2705
In source file neb.f90, at line number 254
p11_2080: p4_error: net_recv read: probable EOF on socket: 1
rm_l_11_2101: (5040.644531) net_send: could not write to fd=5, errno = 32
[...]

abf123 · Post by **abf123** » Fri Jul 06, 2007 6:08 pm

Also,
Here's the block of code containing line 254:

!======================================================================
! find climbing image
!======================================================================

IF (lclimb) THEN
lclimb_local = .TRUE.
! use energy of first image if specified in the INCAR
IF(energy_first.NE.0) THEN
max_energy_image = 0
max_energy = energy_first
ELSE
max_energy_image = 1
max_energy = toten_all(1)
ENDIF
DO ni=1,images
IF (toten_all(ni).GT.max_energy) THEN
max_energy = toten_all(ni)
max_energy_image = ni
ENDIF
ENDDO
IF (energy_last.NE.0 .AND. energy_last.GT.max_energy) THEN
max_energy_image = images+1
max_energy = energy_last
ENDIF

IF (max_energy_image > 0 .AND. max_energy_image < images+1) THEN
climbing_image = max_energy_image
ELSE
lclimb_local = .FALSE.
**LINE254** IF (iu6>=0) WRITE(iu6) 'NEB: no climbing image found'
ENDIF
ENDIF

abf123 · Post by **abf123** » Fri Jul 06, 2007 7:50 pm

After a bit of coffee, I noticed this can be fixed by changing line 210 in neb.F from:

IF (iu6>=0) WRITE(iu6) 'NEB: no climbing image found'

to:

IF (iu6>=0) WRITE(iu6,*) 'NEB: no climbing image found'

Just a tiny syntax error ...

andri · Post by **andri** » Sat Jul 07, 2007 11:06 am

Thanks for pointing this out. I guess it was a pretty obvious, and embarrassing, error.