[petsc-users] MatFactorNumeric_PASTIX error

Barry Smith bsmith at mcs.anl.gov
Sun Mar 7 11:11:04 CST 2010


    I have fixed a couple bugs in the PETSc PasTix interface.  You  
will need to get an install petsc-dev http://www.mcs.anl.gov/petsc/petsc-as/developers/index.html 
  to use the new code. If there is still a problem then please  
configure with --with-debugging=1and send all the output when it  
crashes.

    Barry

On Mar 7, 2010, at 10:14 AM, Vasia Kalavri wrote:

> Hello again.
>
> > From: bsmith at mcs.anl.gov
> > To: petsc-users at mcs.anl.gov
> > Date: Wed, 3 Mar 2010 08:58:58 -0600
> > Subject: Re: [petsc-users] MatFactorNumeric_PASTIX error
> >
> >
> > Build PETSc and Pastix with debugging turned on.
> >
> > ./configure --with-debugging=1 --download-pastix and see if it
> > still crashes. If it does then in the debugger you can get detailed
> > information about exactly where/why it crashed.
>
> I rebuilt with debugging turned on but unfortunately, it still  
> crashes.
> Where can I find the debugger detailed info?
> I added some prints inside MatFactorNumeric_PaStiX() function
> and found out that the program crashes when calling pastix() function,
> line 441 in pastix.c
>
> > Are you changing the nonzero structure of the matrix?
>
> No I am not.
> In fact, I ran the program without updating the input matrix after  
> the first iteration and it still crashed!
>
> >
> > Barry
> >
> > On Mar 3, 2010, at 1:14 AM, Vasia Kalavri wrote:
> >
> > > Hello.
> > > I use MatGetFactor(), MatCholeskyFactorSymbolic() and
> > > MatCholeskyFactorNumeric() with Pastix solver in an iterative loop
> > > in order to solve a system.
> > > The program behaves as desired when using a matrix of size 118x118
> > > and solves the system after 7 iterations.
> > > When using a matrix of size 1000x1000, the program crashes at the
> > > 2nd iteration just after calling MatCholeskyFactorNumeric().
> > > The message I get is:
> > >
> > > [0]PETSC ERROR: [0] MatFactorNumeric_PASTIX line 347 src/mat/ 
> impls/
> > > aij/mpi/pastix/pastix.c
> > > [0]PETSC ERROR: [0] MatCholeskyFactorNumeric line 2517 src/mat/
> > > interface/matrix.c
> > > [0]PETSC ERROR: --------------------- Error Message
> > > ------------------------------------
> > > [0]PETSC ERROR: Signal received!
> > >
> > > Option -start_in_debugger gives me the following output:
> > >
> > > [1]PETSC ERROR: MPI error 15
> > > [vasia-desktop:15869] *** Process received signal ***
> > > [vasia-desktop:15869] Signal: Aborted (6)
> > > [vasia-desktop:15869] Signal code: (-6)
> > > [vasia-desktop:15869] [ 0] [0xb7eeb440]
> > > [vasia-desktop:15869] [ 1] /lib/tls/i686/cmov/libc.so.6(abort 
> +0x101)
> > > [0xb7b4ea01]
> > > [vasia-desktop:15869] [ 2] ./ex1(Petsc_MPI_DebuggerOnError+0)
> > > [0x843d50e]
> > > [vasia-desktop:15869] [ 3] /usr/lib/libmpi.so.
> > > 0(ompi_errhandler_invoke+0x113) [0xb7e773e3]
> > > [vasia-desktop:15869] [ 4] /usr/lib/libmpi.so.0(MPI_Recv+0x1b5)
> > > [0xb7eac125]
> > > [vasia-desktop:15869] [ 5] ./ex1(Dreceive_one_fanin+0x8e)  
> [0x849aa05]
> > > [vasia-desktop:15869] [ 6] ./ex1(Dwait_contrib_comp_1d+0x43)
> > > [0x84a69f9]
> > > [vasia-desktop:15869] [ 7] ./ex1(Dsopalin_smp+0x2be) [0x84a6d6f]
> > > [vasia-desktop:15869] [ 8] /lib/tls/i686/cmov/libpthread.so.0
> > > [0xb7d6d4fb]
> > > [vasia-desktop:15869] [ 9] /lib/tls/i686/cmov/libc.so.6(clone 
> +0x5e)
> > > [0xb7bf8e5e]
> > > [vasia-desktop:15869] *** End of error message ***
> > > .
> > > .
> > > .
> > > [vasia-desktop:15868] *** Process received signal ***
> > > [vasia-desktop:15868] Signal: Aborted (6)
> > > [vasia-desktop:15868] Signal code: (-6)
> > > [vasia-desktop:15868] [ 0] [0xb7fdd440]
> > > [vasia-desktop:15868] [ 1] /lib/tls/i686/cmov/libc.so.6(abort 
> +0x101)
> > > [0xb7c40a01]
> > > [vasia-desktop:15868] [ 2] ./ex1 [0x848d825]
> > > [vasia-desktop:15868] [ 3] ./ex1(PetscError+0x317) [0x840d5c9]
> > > [vasia-desktop:15868] [ 4] ./ex1(PetscDefaultSignalHandler+0x3b4)
> > > [0x848e32f]
> > > [vasia-desktop:15868] [ 5] ./ex1 [0x848df5b]
> > > [vasia-desktop:15868] [ 6] [0xb7fdd420]
> > > [vasia-desktop:15868] [ 7] ./ex1(sopalin_launch_thread+0x560)
> > > [0x84c8c70]
> > > [vasia-desktop:15868] [ 8] ./ex1(Dsopalin_thread+0x168)  
> [0x84a745e]
> > > [vasia-desktop:15868] [ 9] ./ex1(pastix_task_sopalin+0x52b)
> > > [0x8497f1b]
> > > [vasia-desktop:15868] [10] ./ex1(pastix+0x261) [0x8499804]
> > > [vasia-desktop:15868] [11] ./ex1(MatFactorNumeric_PaStiX+0xff6)
> > > [0x817777c]
> > > [vasia-desktop:15868] [12] ./ex1(MatCholeskyFactorNumeric+0x760)
> > > [0x80957d4]
> > > [vasia-desktop:15868] [13] ./ex1(main+0x3a01) [0x807e5f9]
> > > [vasia-desktop:15868] [14] /lib/tls/i686/cmov/libc.so.
> > > 6(__libc_start_main+0xe0) [0xb7c2a450]
> > > [vasia-desktop:15868] [15] ./ex1 [0x8077291]
> > > [vasia-desktop:15868] *** End of error message ***
> > >
> > > Any suggestions?
> > >
> > > Thanks,
> > > Vasia.
> > > Hotmail: Powerful Free email with security by Microsoft. Get it  
> now.
> >
> Vasia.
>
> Hotmail: Trusted email with Microsoft’s powerful SPAM protection.  
> Sign up now.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20100307/a4fa756e/attachment.htm>


More information about the petsc-users mailing list