[petsc-users] Loosing a flux when using the asm preconditioned for >2 cores

[Barry Smith]

  Force really tight convergence on the nonlinear solver -snes_rtol 1.e-12

  Note also that initial function norm 

  0 SNES Function norm 2.418625311837e+01 
  0 SNES Function norm 3.156023513435e+04

is very different in the two cases. This means either you are solving a different nonlinear problem on 2 and 3 processes or the initial conditions are very different (or both).

  Barry


On Sep 3, 2014, at 2:05 PM, Welland, Michael J. <mwelland at anl.gov> wrote:

> Thanks Barry, I attach the terminal outputs. From my reading, it seems like everything converged alright, no? Usually I get an error message when the linear or snes solver doesn't converge. 
> 
> By the way, I don't think it should matter but for your reference I am doing an operator split method in FEniCS where I am alternating between solving two system until their solution vectors stop changing between iterations. 
> 
> Thanks
> 
> <correct_fluxes.txt><lost_fluxes.txt>