[petsc-dev] SEGV in hypre_BoomerAMGCycleT

[Barry Smith]

   Jed,

    Sorry I've never used the transpose stuff in hypre. I don't see  
why one would need to do any additional setup to apply the transpose,  
but perhaps there is. If we know what it is, we can add it to the  
PETSc calls to hypre.

    This is more a petsc-maint then petsc-dev issue, but no big deal  
either way,


     Barry

On Jan 9, 2010, at 1:34 PM, Jed Brown wrote:

> Useing Hypre-2.4.0b through PETSc's interface, I get the following
>
> $ cd petsc/src/ksp/ksp/examples/tutorials
> $ gdb -ex r --args ./ex2 -ksp_type cgne -pc_type hypre
> No symbol table is loaded.  Use the "file" command.
> Breakpoint 1 (PetscError) pending.
> Reading symbols from /home/jed/petsc/src/ksp/ksp/examples/tutorials/ 
> ex2...done.
> Starting program: /home/jed/petsc/src/ksp/ksp/examples/tutorials/ex2  
> -ksp_type cgne -pc_type hypre
> [Thread debugging using libthread_db enabled]
>
> Program received signal SIGSEGV, Segmentation fault.
> 0x00007ffff783b7b8 in hypre_BoomerAMGCycleT (amg_vdata=0x851950,  
> F_array=0x859b00, U_array=0x859bd0) at par_amg_solveT.c:432
> 432              relax_points =   grid_relax_points[cycle_param][j];
> (gdb) p grid_relax_points
> $1 = (int **) 0x0
> (gdb) bt
> #0  0x00007ffff783b7b8 in hypre_BoomerAMGCycleT (amg_vdata=0x851950,  
> F_array=0x859b00, U_array=0x859bd0) at par_amg_solveT.c:432
> #1  0x00007ffff783b1c7 in hypre_BoomerAMGSolveT (amg_vdata=0x851950,  
> A=0x850790, f=0x85b860, u=0x8589e0) at par_amg_solveT.c:203
> #2  0x00007ffff782d263 in HYPRE_BoomerAMGSolveT (solver=0x851950,  
> A=0x850790, b=0x85b860, x=0x8589e0) at HYPRE_parcsr_amg.c:94
> #3  0x00007ffff772b5ff in PCApplyTranspose_HYPRE_BoomerAMG  
> (pc=0x84f390, b=0x856f90, x=0x854150) at hypre.c:354
> #4  0x00007ffff77c8540 in PCApplyTranspose (pc=0x84f390, x=0x856f90,  
> y=0x854150) at precon.c:480
> #5  0x00007ffff76c8f61 in KSPSolve_CGNE (ksp=0x8408e0) at cgne.c:113
> #6  0x00007ffff776408f in KSPSolve (ksp=0x8408e0, b=0x7f8110,  
> x=0x834d10) at itfunc.c:396
> #7  0x0000000000402829 in main (argc=5, args=0x7fffffffd048) at  
> ex2.c:196
>
>
> Are we expected to do something special so that
> amg_data->grid_relax_points is allocated?
>
> Jed