Thank you both for your speedy suggestions!
Unfortunately I need NetCDF-4 for this, because the F77 NetCDF interface won’t
allow REAL otherwise. However, I will try what combinations I can manage on
Monday (screen sharing over a VPN is not ideal for *nix tinkering).
Cheers
Harry
On 23 Jun 2017, at 21:38, dmh@xxxxxxxx<mailto:dmh@xxxxxxxx> wrote:
All good suggestions.
The goal is to try to isolate the problem
to see if the error is associated with HDF5, or mpi, or
the core netcdf-c library.
On 6/23/2017 2:17 PM, Gus Correa wrote:
Hi Ian, Dennis
May I suggest a workaround?
If your files are netCDF "Classic" or "64 bit offset" ('ncdump -k' will tell)
they don't need
NetCDF-4 or HDF5 support to be processed.
Am I right?
If so, you could compile the netCDF 3.6.3 library and link your program to it
instead.
Alternatively, you could compile netCDF 4.4.1 disabling netCDF-4 (run configure
without --enable-netcdf4).
Either one doesn't require HDF5, and *may* not have these errors.
Other than this, perhaps lowering the compiler optimization level,
or turning off OpenMP, if it is on.
I hope this helps,
Gus Correa
On 06/23/2017 03:32 PM, dmh@xxxxxxxx<mailto:dmh@xxxxxxxx> wrote:
Is there any chance that you can run this using valgrind?
My speculation is that it is a memory leak of some kind.
=Dennis Heimbigner
Unidata
On 6/23/2017 7:38 AM, Ian Harris (ENV) wrote:
Hi,
I have NetCDF installed (netcdf library version 4.4.1 per ncdump) on a fairly
powerful iMac (4GHz i7, 16GB). I use Fortran 77 (yes, I know, but it’s very
fast).
I’m processing a lot of data, inputs and outputs all NetCDF, in loops, and it’s
dying at predictable and reproducible points.
One example is a simple prog I knocked up; it opens a nc file, then continually
reads the first time value in a loop until it dies.
The last successful loop is 16777187, which may sound familiar, as it’s been
here before:
http://www.unidata.ucar.edu/mailing_lists/archives/netcdfgroup/2014/msg00173.html
The error is usually 'NetCDF: HDF error', though occasionally 'NetCDF: Error
initializing for parallel access'.
I should note that a much more complicated program, which reads in three files
and writes one, also dies like this after a fixed number of runs through the
loops (taking over an hour).
The simple test program is below.
Please can anyone hazard a guess as to what’s happening?
Cheers
Harry
Here’s the entire program:
program netcdftestloop
c NETCDFTESTLOOP Testing iteration counts before it dies
implicit none
include 'netcdf.inc'
integer nits,st,ncid,tvarid
real rday
nits = 0
st = nf_open('Prec_daily_WFD_CRU_grid_198605.nc',0,ncid)
if (st.ne.nf_noerr) call handle_err(st,'openff')
st = nf_inq_varid(ncid,'time',tvarid)
if (st.ne.nf_noerr) call handle_err(st,'inqtvi')
do
nits=nits+1
if (mod(nits,100000).eq.0) write(*,'(i10)')nits ! keep
track
st = nf_get_var1_real(ncid,tvarid,1,rday) ! get
day 1 time value
if (st.ne.nf_noerr) then
write(*,'("N. iterations = ",i10)')nits
call handle_err(st,'getday')
endif
enddo
st = nf_close(ncid) !
totally unnecessary close file
if (st.ne.nf_noerr) call handle_err(st,'closef')
stop
end
c
***********************************************************************************************
subroutine handle_err(errcode,code)
implicit none
include 'netcdf.inc'
integer errcode
character code*6
print *, 'Error: ', code, ': ', nf_strerror(errcode)
stop
end
c
***********************************************************************************************
In a more complicated program, I get a different error, ('')
Ian "Harry” Harris
NCAS-Climate
Climatic Research Unit
School of Environmental Sciences
University of East Anglia
Norwich NR4 7TJ
United Kingdom
IMPORTANT NOTICE - This email is intended for the named recipient only. It may
contain privileged and confidential information. If you are not the intended
recipient, notify the sender immediately and destroy this email. You must not
copy, distribute or take action in reliance upon it. Whilst all efforts are
made to safeguard emails, The School of Environmental Sciences cannot guarantee
that attachments are virus free or compatible with your systems and does not
accept liability in respect of viruses or computer problems experienced.
