On Thursday, December 13, 2012 15:57:18 John Eblen wrote:
> > 1. The hint "don't add the expressions Debug_.* and Release_.* to the
> > exclusion list for ptp synchronization" should be added to the
> > documentation. Not sure whether FAQ is an appropriate place for this.
>
> Can you explain this a little more? Given that these are build
> directories, one might actually want to filter them.
John,
this is precisely why I first filtered them.
Just to be on the safe side, I scratched everything, i.e.
* rm -rf ../eclipse .project .settings .cproject Debug_*
* Then opened the work space again
(which recreated an empty workspace ../eclipse )
* Projects->New->Synchronized C/C++ project
-> executable -> empty MPI C project
* pointed locally to the existing source file
* pointed to a remote host that allowed Linux gcc
(and thus avoiding the known bug with xlc)
* excluded Debug_.* from synchronization.
* compiled locally. This went fine.
* compiled remotely (same tool chain).
This got me the following message.
20:03:52 **** Incremental Build of configuration Debug_remote for project MPI_Output ****
make all
/bin/bash: line 0: cd: /homec/ibm/pospiech/bench/DKRZ/MPI_Output/Debug_remote: No such file or directory
20:03:53 Build Finished (took 1s.318ms)
Then I removed Debug_.* from the filter and added the expressions Debug_.*\/MPI_Output and Debug_.*\/.*\.o to it and retried.
20:16:38 **** Incremental Build of configuration Debug_remote for project MPI_Output ****
make all
Building file: ../helloa.c
Invoking: GCC C Compiler
mpicc -O0 -g3 -Wall -c -fmessage-length=0 -MMD -MP -MF"helloa.d" -MT"helloa.d" -o "helloa.o" "../helloa.c"
/opt/ibmcmp/vac/bg/12.1/bin/.orig/bgxlc_r: 1501-289 (W) Option -Wall was incorrectly specified. The option will be ignored.
/opt/ibmcmp/vac/bg/12.1/bin/.orig/bgxlc_r: 1501-289 (W) Option -Wall was incorrectly specified. The option will be ignored.
/opt/ibmcmp/vac/bg/12.1/bin/.orig/bgxlc_r: 1501-216 (W) command option -fmessage-length=0 is not recognized - passed to ld
/opt/ibmcmp/vac/bg/12.1/bin/.orig/bgxlc_r: 1501-208 (S) command option D is missing a subargument
make: *** [helloa.o] Error 40
20:16:41 Build Finished (took 2s.958ms)
Ooops, this is a BlueGene - mpicc does not wrap gcc but xlc, not really a Linux gcc tool chain. I have to tamper the options and then it will compile. But I think you see the difference.
It appears that Debug_remote is generated locally (given the tool chain and other settings), then replicated to the remote host. Then the compilation is started via ssh. If you exclude Debug_.* from replication, make doesn't find anything to make.
Is this the level of detail that you expected ? If so, what should we do with it ? Put this into some wiki ?
--
Mit freundlichen Grüßen / Kind regards
Dr. Christoph Pospiech
High Performance & Parallel Computing
Phone: +49-351 86269826
Mobile: +49-171-765 5871
E-Mail: christoph.pospiech@xxxxxxxxxx
-------------------------------------------------------------------------------------------------------------------------------------------
IBM Deutschland GmbH / Vorsitzender des Aufsichtsrats: Martin Jetter
Geschäftsführung: Martina Koederitz (Vorsitzende), Reinhard Reschke, Dieter Scholz, Gregor Pillen, Joachim Heel, Christian Noll
Sitz der Gesellschaft: Ehningen / Registergericht: Amtsgericht Stuttgart, HRB 14562 / WEEE-Reg.-Nr. DE 99369940
|