On Thu, Oct 20, 2022 at 10:27 PM Barry Smith wrote:
>
> Some of the valgrind information does not appear to make sense
>
> PetscMemcpy() is not calling SNESSolve() so I suspect there must be some
> serious corruption of something to this impossible stack trace
>
I ran Valgrind on it, and it
Some of the valgrind information does not appear to make sense
PetscMemcpy() is not calling SNESSolve() so I suspect there must be some
serious corruption of something to this impossible stack trace
==236074==by 0x6FD160F: SNESSolve (snes.c:4569)
==236074==by 0x711917E: PetscMemcpy
On Thu, Oct 20, 2022 at 6:48 PM Sepideh Kavousi wrote:
> Hello,
>
> I want to solve my 5 PDEs based on finite difference method using
> periodic BC in x-direction and non-periodic in y-direction but I run into
> error (Segmentation Violation, probably memory access out of range).
>
> For this,
Hello,
I want to solve my 5 PDEs based on finite difference method using periodic BC
in x-direction and non-periodic in y-direction but I run into error
(Segmentation Violation, probably memory access out of range).
For this, I discretize my equation in FormFunction function. My PDE
I am running a case with periodic boundary condition. It worked fine with PETSc
3.5.1. When PETSc was upgraded to 3.6.0, I got the error below.
[0]PETSC ERROR: - Error Message
--
[0]PETSC ERROR: No support for this
We had to remove the support for DMLocalToGlobalBegin/End() for this case
since it was too difficult to maintain.
Why are you using DMLocalToGlobalBegin/End() here? All it does it discard
the ghost values. Instead of putting values into the wb0_gfpp_loc and then
calling