[petsc-users] MUMPS memory allocation problem

[Lukasz Kaczmarczyk]

Thanks, you right.


On 13 Sep 2012, at 22:27, Jed Brown wrote:

> No point wasting time debugging such archaeological relics. There have been numerous improvements and bug fixes to both packages since then. You should upgrade to petsc-3.3 which also has the latest MUMPS.
> 
> On Thu, Sep 13, 2012 at 4:24 PM, Lukasz Kaczmarczyk <Lukasz.Kaczmarczyk at glasgow.ac.uk> wrote:
> Hello,
> 
> I use petsc-3.1-p3 and MUMPS_4.9.2.  I hope that this will help, it is tricky to track the error, usually is for relatively large problems,  I have to check if it happens always at the same time, or is random error.
> 
> CYCLE 19 ref_mesh 0
> Vol = 1.6749e+08 ave_Vol = 4.2698e+03 min_quali = 1.7609e-01
> Nodes Ordering
> Elemnets Ordering
> 
> arc length lambda 4.4581e-01
> 
> arc length lambda value 5.6751e+02
> 
>   0 SNES Function norm 7.929158005697e+01
>   0 SNES (MY) norms Fint 6.7783e+01, arc length res_lambda 0.0000e+00 lambda 4.4581e-01
>     0 KSP (MY) F 9.0999e-01 Res 1.0000e+00
>     1 KSP (MY) F 3.8659e-15 Res 4.4892e-15
>     PC ArcLength norm dF_dLambda 5.8658e+03 norm dD_dF_dLambda 3.29e+01 b_dD_dF_dLambda 0.00e+00
>     0 KSP (MY) norms F 6.7783e+01 Res 7.9292e+01 res arc length lambda 0.0000e+00 dlambda 0.0000e+00
>     1 KSP (MY) norms F 9.1576e-14 Res 9.9498e-14 res arc length lambda 0.0000e+00 dlambda 0.0000e+00
>   1 SNES Function norm 2.355295496832e-02
>   1 SNES (MY) norms Fint 2.0118e-02, arc length res_lambda 0.0000e+00 lambda 4.4581e-01
> At line 593 of file dmumps_part3.F
> Fortran runtime error: Attempt to DEALLOCATE unallocated 'perm'
> --------------------------------------------------------------------------
> mpirun has exited due to process rank 3 with PID 25487 on
> node rdb-srv1 exiting without calling "finalize". This may
> have caused other processes in the application to be
> terminated by signals sent by mpirun (as reported here).
> 
> 
> On 13 Sep 2012, at 21:23, Hong Zhang wrote:
> 
> > Lukasz :
> > Hello,
> >
> > I solve a strongly nonlinear problem using SNES. When I use MUMS with petsc form time to time I get following error,
> >
> > Fortran runtime error: Attempt to DEALLOCATE unallocated 'perm'
> >
> > I've never seen report from our users about this error.
> > Can you provide more info about it, e.g., run with a debugger and find
> > error stack showing which routine and the line number that the error occurs?
> > Which version of petsc and mumps do you use?
> > Does it work on petsc examples, e.g.
> > petsc-3.3/src/ksp/ksp/examples/tutorials>mpiexec -n 4 ./ex2 -pc_type lu -pc_factor_mat_solver_package mumps
> > Norm of error 2.71948e-15 iterations 1
> >
> > Hong
> > --------------------------------------------------------------------------
> > mpirun has exited due to process rank 3 with PID 25487 on
> > node rdb-srv1 exiting without calling "finalize". This may
> > have caused other processes in the application to be
> > terminated by signals sent by mpirun (as reported here).
> >
> > Other solvers (f.e. SuperLU Dist) working without problems. Is this know issue with MUMPS and similar problem where reported in past?
> >
> > Lukasz
> >
> 
>