Issue 68 in pencil-code: pc_run relies on a line which is optimized away to detect MPI runs

1 view
Skip to first unread message

penci...@googlecode.com

unread,
Aug 16, 2013, 6:28:30 AM8/16/13
to pencil-co...@googlegroups.com
Status: New
Owner: ----
Labels: Type-Defect Priority-Medium

New issue 68 by mcnal...@gmail.com: pc_run relies on a line which is
optimized away to detect MPI runs
http://code.google.com/p/pencil-code/issues/detail?id=68

With gfortran-4.8 -O3 pc_run fails to detect that start.x uses MPI.

It does at line
https://code.google.com/p/pencil-code/source/browse/trunk/bin/pc_run#657

$settings{'mpi'}
= `fgrep -c 'initialize_mpicomm: MPICOMM neighbors' src/start.x` + 0;


This relies on the string from:

https://code.google.com/p/pencil-code/source/browse/trunk/src/mpicomm.f90#495

if (ldebug_mpi) write(6,'(A,I4,"(",3I4,"): ",8I4)') &
'initialize_mpicomm: MPICOMM neighbors ', &
iproc,ipx,ipy,ipz, &
ylneigh,llcorn,zlneigh,ulcorn,yuneigh,uucorn,zuneigh,lucorn

showing up in the binary.

If you compile with ldebug_mpi=.false. this line can be optimized away, so
the strong does not show up in the binary start.x

pc_run needs to do the detection in some other way - perhaps some useful
default output message can be added to MPICOMM to tag the binary.

--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings
Reply all
Reply to author
Forward
0 new messages