I have no idea how this e-mail slipped by me -- I'm sorry for the
delay in replying.
Can you run your code through a memory-checking debugger such as
valgrind and see if it turns up any ancillary memory problems?
On Dec 1, 2006, at 1:52 PM, Cupp, Matthew R wrote:
Hi,
In my program I am calling MPI_Barrier(MPI_COMM_WORLD), but it
seems to cause an error on one node. The node changes depending on
how many total nodes I have (it could be 4 or 2). I’m thinking
it’s MPI_Barrier because I put print statements around it and
that’s where the program terminates. This is the error message
that I get:
Signal:11 info.si_errno:0(Success) si_code:1(SEGV_MAPERR)
Failing at addr:0xc900000002
[0] func:/opt/openmpi/st/lib/libopal.so.0 [0x2aaaab04dbc8]
[1] func:/lib64/libpthread.so.0 [0x3be4f0c530]
[2] func:/opt/openmpi/st/lib/openmpi/mca_pml_ob1.so(mca_pml_ob1_recv
+0x2a3) [0x2aaaacf26f33]
[3] func:/opt/openmpi/st/lib/openmpi/mca_coll_tuned.so
(ompi_coll_tuned_barrier_intra_recursivedoubling+0x14a)
[0x2aaaad8699ba]
[4] func:/opt/openmpi/st/lib/libmpi.so.0(PMPI_Barrier+0xa4)
[0x2aaaaad87294]
[5] func:Debug/BioRiskAssessmentMpiLibTest
(_ZN2BL14CMpiTestRunner11SynchronizeEv+0xe) [0x490846]
[6] func:Debug/BioRiskAssessmentMpiLibTest
(_ZN2BL8CMpiTest12FinishedTestEi+0x3c) [0x490884]
[7] func:Debug/BioRiskAssessmentMpiLibTest
(_ZN15CMpiProcessTest8RunTestsEv+0x269) [0x490297]
[8] func:Debug/BioRiskAssessmentMpiLibTest
(_ZN29CMpiConsequenceCalculatorTest3RunEP19ompi_communicator_t
+0xdf) [0x45a8e7]
[9] func:Debug/BioRiskAssessmentMpiLibTest
(_ZN2BL14CMpiTestRunner3RunEv+0x60) [0x4909ba]
[10] func:Debug/BioRiskAssessmentMpiLibTest(main+0x42) [0x44558a]
[11] func:/lib64/libc.so.6(__libc_start_main+0xef) [0x3be481c40f]
[12] func:Debug/BioRiskAssessmentMpiLibTest(__gxx_personality_v0
+0x99) [0x4454b9]
*** End of error message ***
I’m using version 1.1.2. Not sure if it matters, but before I call
the MPI_Barrier I create a comm subset (which in this case happens
to be all of the same processes that are in MPI_COMM_WORLD).
Does anybody have an idea what might be my problem? Or what I
should do to get more information?
Thanks!
Matt
______________________________
Matt Cupp
Battelle Memorial Institute
Statistics and Information Analysis
_______________________________________________
users mailing list
us...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/users
--
Jeff Squyres
Server Virtualization Business Unit
Cisco Systems