[petsc-users] petsc4py

[Jed Brown]

2010/9/30 Николай Кутузов <kutuzovnp at gmail.com>:
> ts.setIFunction(ode.evalFunction, f) ---> ts.setIFunction(ode.evalFunction)

This won't work, you need to provide a residual vector, otherwise
PETSc does not know how large your vector is.  Your earlier emails
said that you removed the line that reads

  ts.setIJacobian(ode.evalJacobian, J)

but said nothing about modifying the line

  ts.setIFunction(ode.evalFunction, f)

That was why we couldn't understand what was going wrong for you.  You
*can* skip the ts.setIJacobian line if you use the option -snes_mf, it
will do exactly the same algorithm as if you leave rober.py exactly as
it is and run with -snes_mf.

Jed