[MPICH] MPE loggin question

[Wolfram Brenig]

Thanks! That did it.

(Gropp/Lusk/Skjellum's 2nd ed., chaps. 3.7.3 - 6,
was just providing a 'slightly' different view on this ...)

Wolfram

> See the section "VI. a) CUSTOMIZING LOGFILES" under mpich2/src/mpe2/README

> You only put in MPE_Init_log()/MPE_Finish_log() if you don't want
> automatic MPI logging, i.e compiled your code with "mpicc -mpe=log".
> If automatic MPI logging is needed, you do NOT need to put in
> MPE_Init_log()/MPE_Finish_log() in your code, i.e. compiled with
> "mpicc -mpe=mpilog".
> 
> PS.  "mpicc -mpe=xxx" feature is only available in mpich2-1.0.3.
> do "mpicc -mpe=help" for more options.


> 
> A.Chan
> 
> 
> On Fri, 9 Dec 2005, Wolfram Brenig wrote:
> 
> 
>>Hello,
>>
>>trying to use MPE-loggin, I'm am experiencing
>>the problem that even the most elementary
>>'code' - shown below - fails for me
>>
>>#include <mpi.h>
>>#include <mpe.h>
>>int main(int argc, char* argv[]) {
>>  int p,r;
>>  MPI_Init(&argc,&argv);
>>  MPI_Comm_size(MPI_COMM_WORLD,&p);
>>  MPI_Comm_rank(MPI_COMM_WORLD,&r);
>>
>>  MPE_Init_log();                   // *1*
>>
>>  MPE_Finish_log("tessimessi");     // *2*
>>
>>  MPI_Finalize();
>>  return 0;
>>}
>>
>>This 'program' compiles with no warning/error,
>>but on execution causes the output:
>>
>>Enabling the synchronization of the clocks...
>>rank 2 in job 11  dell1_1130   caused collective abort of all ranks
>>   exit status of rank 2: killed by signal 11
>>rank 0 in job 11  dell1_1130   caused collective abort of all ranks
>>   exit status of rank 0: killed by signal 9
>>
>>Commenting away lines *1* and *2*, i.e. leaving out the MPE,
>>the problem disappears.
>>
>>The MPI version I have at hand is:
>>~> mpich2version
>>Version:           1.0.2
>>Device:            ch3:sock
>>Configure Options: --enable-sharedlibs=gcc CFLAGS=-O3 -march=pentium4m
>>CXXFLAGS=-O3 -march=pentium4m
>>
>>The MPE came with the package.
>>
>>Any help would be most welcome.
>>Wolfram
>>
>>