Hi,
sorry for the previous e-mail. I was on the cell phone and the corrector
changed a couple
of words....
In the mean time I found out that the problem is related with the intel
fortran compiler (ifort),
since when compiled with gfortran, the problem disappears, (I am still
interested that it
works with ifort, anyway)
The code would be
$> cat omp_test2.f90
program omp_test
implicit none
write(*,*) "Hello ..."
!$OMP PARALLEL
write(*,*) "Good bye ..."
!$OMP END PARALLEL
end program
This code, if compiled with gfortran (4.8.5) with
gfortran -o omp_test2.x -fopenmp omp_test2.f90
and executed using (coordiantor) and dmtcp_launch works fine
and gives the following output (OMP_NUM_THREADS=3)
Hello ...
Good bye ...
Good bye ...
Good bye ...
On the other hand, when compiled with ifort with
ifort -o omp_test2.x -qopenmp omp_test2.f90
and executed using (coordiantor) and dmtcp_launch only
outputs
Hello ...
and the program never ends (looking at top, it stays there,
using 1 thread with 100% cpu usage). i.e. never enters the parallel region.
*Summarizing, the problems is with OpenMP code generated by ifort*
(I am using *version 18.0.2*).
To further investigate the issue, I have compiled the code with
*ifort version 13.0.2*, and this is the output
Hello ...
[42000] WARNING at dlwrappers.cpp:75 in dlopen; REASON='JWARNING(ret)
failed'
filename = /share/apps/intel/composer_xe_2013_sp1.2.144/compiler/lib/
intel64/libtbbmalloc.so.2
flag = 1
Message: dlopen failed. You may also see a message 'ERROR: ld.so:'
from libdl.so. If this happens only under DMTCP, then consider setting
the environment variable DMTCP_DL_PLUGIN to "0" before 'dmtcp_launch'.
If the problem persists, please write to the DMTCP developers.
Good bye ...
Good bye ...
Good bye ...
I hope this can shed some light on the issue. Apparently, the issue might be
related to the Intel library (I am guessing ...) *libiomp5.so *, since the
same
code (even the one compiled with ifort version 18) when run under the
environment
of version 14, runs (giving also the above warning), and the output of ldd
for
the same code under the different environments are
Environment set for version 14
$> ldd omp_test2.x
linux-vdso.so.1 => (0x00007ffd24a9a000)
libm.so.6 => /lib64/libm.so.6 (0x00007f055bdde000)
libiomp5.so => /share/apps/intel/compilers_and_libraries_2018.2.199/
linux/compiler/lib/intel64/libiomp5.so (0x00007f055ba06000)
libpthread.so.0 => /lib64/libpthread.so.0 (0x00007f055b7ea000)
libc.so.6 => /lib64/libc.so.6 (0x00007f055b427000)
libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00007f055b211000)
libdl.so.2 => /lib64/libdl.so.2 (0x00007f055b00d000)
/lib64/ld-linux-x86-64.so.2 (0x00007f055c0e0000)
Environment set for version 18
$> ldd omp_test2.x
linux-vdso.so.1 => (0x00007ffd24a9a000)
libm.so.6 => /lib64/libm.so.6 (0x00007f055bdde000)
libiomp5.so => /share/apps/intel/compilers_and_libraries_2018.2.199/
linux/compiler/lib/intel64/libiomp5.so (0x00007f055ba06000)
libpthread.so.0 => /lib64/libpthread.so.0 (0x00007f055b7ea000)
libc.so.6 => /lib64/libc.so.6 (0x00007f055b427000)
libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00007f055b211000)
libdl.so.2 => /lib64/libdl.so.2 (0x00007f055b00d000)
/lib64/ld-linux-x86-64.so.2 (0x00007f055c0e0000)
Again, thank you in advance for your help (and for your fast reply)
best regards,
adolfo
2018-05-18 17:00 GMT-03:00 Kapil Arya <kapil.arya...@gmail.com>:
> Hi Adolfo,
>
> Can you also provide instructions to compile this code?
>
> Kapil
>
> On Fri, May 18, 2018 at 3:53 PM ADOLFO JAVIER BANCHIO <
> banc...@famaf.unc.edu.ar> wrote:
>
>>
>>
>> Hi all,
>>
>> After having googled quite a lot without success and also having
>> checked archive posts, I still can not run fortran compiled openmp codes
>> using dmtcp_launch.
>>
>> I have installed on a Rocks 7 (CENTOS 7) cluster dmtcp version 2.5.2
>> (from rpm and also compiled with --enable-openm flag),
>> and I still can not run openmp executables produced by ifort compilded
>> f90 codes.
>>
>> I run:
>>
>> in *shell 1*
>>
>> /export/added_soft/dmtcp/dmtcp-2.5.2/bin/dmtcp_coordinator
>>
>>
>> and in *shell 2*
>>
>> export OMP_NUM_THREADS=3
>>
>> /export/added_soft/dmtcp/dmtcp-2.5.2/bin/dmtcp_launch ./omp_test.x
>>
>>
>> output in *shell 1 *is:
>>
>> $ /export/added_soft/dmtcp/dmtcp-2.5.2/bin/dmtcp_coordinator
>> dmtcp_coordinator starting...
>> Host: bandurria.fis.uncor.edu (0.0.0.0)
>> Port: 7779
>> Checkpoint Interval: disabled (checkpoint manually instead)
>> Exit on last client: 0
>> Type '?' for help.
>>
>> [28865] NOTE at dmtcp_coordinator.cpp:1368 in updateCheckpointInterval;
>> REASON='CheckpointInterval updated (for this computation only)'
>> oldInterval = 0
>> theCheckpointInterval = 0
>> [28865] NOTE at dmtcp_coordinator.cpp:917 in onConnect; REASON='worker
>> connected'
>> hello_remote.from = 1ba5f63f5ba22d27-29111-99b9e2da0f18
>> [28865] NOTE at dmtcp_coordinator.cpp:667 in onData; REASON='Updating
>> process Information after exec()'
>> progname = omp_test.x
>> msg.from = 1ba5f63f5ba22d27-40000-99b9e3d17fe2
>> client->identity() = 1ba5f63f5ba22d27-29111-99b9e2da0f18
>>
>>
>> And* in shell 2*, the code starts (if I do top, it is running with one
>> thread
>> only, using 100% of cpu, but it seems not to spawn the threads, it seems
>> that it gets stuck when it reaches a parallel section (the part of the
>> code previous to parallel block it is actually executed).
>>
>>
>> Thank you in advance for any help.
>> I am new with dmtcp (coming from blcr), so, my apologies if this is
>> a stupid issue ...
>>
>> regards,
>>
>> adolfo
>>
>>
>>
>>
>> P.S.: the code I am using for testing (other real codes fail in the same
>> way)
>> program omp_test
>> implicit none
>> integer(8) :: i,j
>> integer :: nt,tn,omp_get_num_threads,omp_get_thread_num
>>
>> write(*,*) "Hello ..."
>>
>> !nt = omp_get_num_threads()
>> write(*,*) 'num threads = ',nt
>>
>> !$OMP PARALLEL PRIVATE(i,tn,nt)
>> do i = 1, 10**9
>> j = int( sqrt( log( real(i)/real(i**2.4) ) ) )
>> if (mod(i,10**8) == 0) then
>> ! nt = omp_get_num_threads()
>> ! tn = omp_get_thread_num()
>> write(*,*) tn, '/',nt,' -- > ', nint( log(real(i))/log(10.) )
>> endif
>> enddo
>> !$OMP END PARALLEL
>>
>> end program
>>
>>
>> ------------------------------------------------------------
>> ------------------
>> Check out the vibrant tech community on one of the world's most
>> engaging tech sites, Slashdot.org! http://sdm.link/slashdot______
>> _________________________________________
>> Dmtcp-forum mailing list
>> Dmtcp-forum@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/dmtcp-forum
>>
>
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Dmtcp-forum mailing list
Dmtcp-forum@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dmtcp-forum
