[petsc-users] fstack-array issue

[Adrian Jackson]

Hi,

I have tried with -fstack-protector-all and that didn't pick up
anything, but not with valgrind yet.

cheers

adrianj

On 07/02/2016 16:13, Matthew Knepley wrote:
> It seems like that you have exceeded stack space with your local array.
> Did you run under valgrind?
> 
>   Matt
> 
> On Sun, Feb 7, 2016 at 3:51 AM, Adrian Jackson <adrianj at epcc.ed.ac.uk
> <mailto:adrianj at epcc.ed.ac.uk>> wrote:
> 
>     Hi,
> 
>     We're using PETSc in a fortran code and get an error if we compile the
>     code using the -fstack-arrays flag when using the GNU fortran compilers.
>      This is the error we are getting:
> 
>     [22]PETSC ERROR: --------------------- Error Message
>     --------------------------------------------------------------
>     [22]PETSC ERROR: Arguments are incompatible
>     [22]PETSC ERROR: Zero diagonal on row 23003
>     ....
>     [22]PETSC ERROR: #1 MatInvertDiagonal_SeqAIJ() line 1841 in
>     /b/cray-petsc/.cray-build/GNU/49/sandybridge/cray-petsc-base-dynamic/petsc-3.5.2/src/mat/impls/aij/seq/aij.c
>     [22]PETSC ERROR: #2 MatSOR_SeqAIJ() line 1873 in
>     /b/cray-petsc/.cray-build/GNU/49/sandybridge/cray-petsc-base-dynamic/petsc-3.5.2/src/mat/impls/aij/seq/aij.c
>     ....
>     [22]PETSC ERROR: #14 KSPSolve() line 459 in
>     /b/cray-petsc/.cray-build/GNU/49/sandybridge/cray-petsc-base-dynamic/petsc-3.5.2/src/ksp/ksp/interface/itfunc.c
>      WARNING: Failed to converge.
>      PETSc did not converge for matrix solve of: DeltaU
>      Reason for non-convergence is undefined:            0
>      Number of iterations:            0
>      Sending signal to dump and finish
> 
> 
>     I've checked the right hand side vector passed to KSPSolve in both cases
>     and it's the same.  I've not fully verified the ksp context is exactly
>     the same but it doesn't seem to be corrupted when using -fstack-arrays.
>      Is this a problem people have encountered before?
> 
>     thanks
> 
>     adrianj
> 
>     --
>     The University of Edinburgh is a charitable body, registered in
>     Scotland, with registration number SC005336.
> 
> 
> 
> 
> -- 
> What most experimenters take for granted before they begin their
> experiments is infinitely more interesting than any results to which
> their experiments lead.
> -- Norbert Wiener

-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.