I used it on a IA64 platform, so I supposed x86_64 is supported, but
I never use it on an AMD 64. On the mpiP webpage they claim they
support the Cray XT3, which as far as I know are based on AMD Opteron
64 bits. So, there is at least a spark of hope in the dark ...
I decide to give it a try on my x86_64 AMD box (Debian based system).
First problem, my box didn't have the libunwind. Not a big deal, it's
freely available on HP website (http://www.hpl.hp.com/research/linux/
libunwind/download.php4). Few minutes later, the libunwind was
installed in /lib64. Now, time to focus on mpiP ... For some obscure
reason the configure script was unable to detect my g77 compiler
(whatever!!!) nor the installation of libunwind. Moreover, it keep
trying to use the clock_gettime call. Fortunately (which make me
think I'm not the only one having trouble with this), mpiP provide
configure options for all these. The final configure line was: ./
configure --prefix=/opt/ --without-f77 --with-wtime --with-include=-I/
include --with-lib=-L/lib64. Then a quick "make shared" followed by
"make install", complete the work. So, at least mpiP can compile on a
x86_64 box.
Now, I modify the makefile of NetPIPE, and add the "-lmpiP -lunwind",
compile NetPIPE and run it. The mpiP headers showed up, the
application run to completion and my human readable output was there.
@ mpiP
@ Command : ./NPmpi
@ Version : 3.1.0
@ MPIP Build date : Mar 29 2007, 13:35:47
@ Start time : 2007 03 29 13:43:40
@ Stop time : 2007 03 29 13:44:42
@ Timer Used : PMPI_Wtime
@ MPIP env var : [null]
@ Collector Rank : 0
@ Collector PID : 22838
@ Final Output Dir : .
@ Report generation : Single collector task
@ MPI Task Assignment : 0 dancer
@ MPI Task Assignment : 1 dancer
However, I got some strange reading inside the output.
------------------------------------------------------------------------
---
@--- Callsites: 5
---------------------------------------------------------
------------------------------------------------------------------------
---
ID Lev File/Address Line Parent_Funct MPI_Call
1 0 0x0000000000402ffb [unknown] Barrier
2 0 0x0000000000403103 [unknown] Recv
3 0 0x00000000004030ad [unknown] Send
4 0 0x000000000040319f [unknown] Send
5 0 0x00000000004031d5 [unknown] Recv
I didn't dig further to see why. But, this prove that for at least a
basic usage (general statistics gathering) mpiP works on x86_64
platforms.
Have fun,
george.
On Mar 29, 2007, at 11:32 AM, Heywood, Todd wrote:
George,
Any other simple, small, text-based (!) suggestions? mpiP seg
faults on
x86_64, and indeed its web page doesn't list x86_64 Linux as a
supported
platform.
Todd
On 3/28/07 10:39 AM, "George Bosilca" <bosi...@cs.utk.edu> wrote:
Stephen,
There are a huge number of MPI profiling tools out there. My
preference will be something small, fast and where the output is in
human readable text format (and not fancy graphics). The tools I'm
talking about is called mpiP (http://mpip.sourceforge.net/). It's not
Open MPI specific, but it's really simple to use.
george.
On Mar 28, 2007, at 10:10 AM, stephen mulcahy wrote:
Hi,
What is the best way of getting statistics on the size of MPI
messages
being sent/received by my OpenMPI-using application? I'm guessing
MPE is
one route but is there anything built into OpenMPI that will give me
this specific statistic?
Thanks,
-stephen
--
Stephen Mulcahy, Applepie Solutions Ltd, Innovation in Business
Center,
GMIT, Dublin Rd, Galway, Ireland. http://www.aplpi.com
_______________________________________________
users mailing list
us...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/users
_______________________________________________
users mailing list
us...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/users
_______________________________________________
users mailing list
us...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/users
"Half of what I say is meaningless; but I say it so that the other
half may reach you"
Kahlil Gibran
