[petsc-users] Using higher order ILU preconditioner

[Barry Smith]

> On Sep 28, 2017, at 2:27 PM, Rachit Prasad <rachitp at vt.edu> wrote:
> 
> I guess I missed out an important piece of information, due to legacy code issues, I am working with Petsc version 2.3.2. Unfortunately, it doesn't have PCFactorSetMatOrderingType(). Thanks for the advice though. I still don't understand, why would it ask for so much memory, it is almost 16 Petabytes! Any other advice would be greatly appreciated.

  You really, really, really need to change ASAP to the latest PETSc release. That is the only advice I can provide.

> 
> Regards,
> Rachit Prasad
> 
> On Thu, Sep 28, 2017 at 5:13 PM, Kong, Fande <fande.kong at inl.gov> wrote:
> Calling PCFactorSetMatOrderingType() (or command line option: -pc_factor_mat_ordering_type) to use RCM or 1WD usually helps me a lot. My application is based on highly-unstructured meshes.
> 
> Fande,
> 
> On Thu, Sep 28, 2017 at 2:53 PM, Rachit Prasad <rachitp at vt.edu> wrote:
> Hi,
> 
> I am trying to solve a highly ill-conditioned matrix and am unable to get convergence when using ILU(0) or Jacobi and SOR preconditioners. I tried to implement a higher order ILU by increasing the level of fill-in to 1 by calling
> 
> call PCFactorSetLevels(pc,1,ierr)
> 
> However, when I do that I get the following memory error:
> 
> [0]PETSC ERROR: --------------------- Error Message ------------------------------------
> [0]PETSC ERROR: Out of memory. This could be due to allocating
> [0]PETSC ERROR: too large an object or bleeding by not properly
> [0]PETSC ERROR: destroying unneeded objects.
> [0]PETSC ERROR: Memory allocated -2147483648 Memory used by process -2147483648
> [0]PETSC ERROR: Try running with -malloc_dump or -malloc_log for info.
> [0]PETSC ERROR: Memory requested 18446744071863943168!
> 
> While I do understand that increasing the level of fill-in will require higher memory, the memory requested seems to be way too high. Am I implementing the higher order ILU correctly? Is there any other subroutine which I need to call?
> 
> Regards,
> Rachit
> 
> 
>