I would like to also suggest TAU (Tuning Analysis and Utilities) from
The University of Oregon.
http://www.cs.uoregon.edu/research/paracomp/tau/tautools
Not only is it free, but I have worked closely with the developers and
they are some of the most responsive I have ever dealt with. If you
have questions/problems they will certainly respond quickly and offer
you more assistance than you would expect.
Dave.
Jeff Squyres wrote:
> Two obvious choices include:
>
> - Use the MPE package from Argonne national lab. It produces data files
> after a run that can be used with a java GUI to display a pictorial
> depiction of your message passing patterns.
>
> - Insert your own profiling function calls (e.g., override MPI_Send) to
> gather your own statistics and then turn around and call the "real"
> MPI_Send by calling PMPI_Send. Check out the Profiling chapter in the
> MPI-1 standard for details.
>
> There's other commercial tools out there for this kind of stuff, but I
> am guessing that you'll want to start with the free stuff first.
>
>
>
> On May 10, 2005, at 10:49 AM, Julio Monetti wrote:
>
>> Hello list,
>>
>> I'm just deploying a little system compiled with LAM-MPI.
>> I've finished my work with good results, but, i like to start
>> analysing possible improvements about the performance
>> of the whole application.
>> So my question is: how can i start to insert instrumentation
>> in my code, (and perhaps inside mpi inner routines).
>>
>> Could you helpme with it?
>>
>> Thanks in advanced.
>>
>>
>> Julio_______________________________________________
>> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
>
--
Dr. David Cronk, Ph.D. phone: (865) 974-3735
Research Leader fax: (865) 974-8296
Innovative Computing Lab http://www.cs.utk.edu/~cronk
University of Tennessee, Knoxville
|
