But using -mat_fd_type ds vs -mat_fd_type wp does produce (slightly) different convergence histories?
Barry > On Jun 23, 2016, at 9:34 PM, Adrian Croucher <[email protected]> > wrote: > > hi Barry, > > On 24/06/16 14:23, Barry Smith wrote: >> If your fortran function includes petscsnesdef.h does it work ok? > > No, unfortunately. > > (I forgot to say, I'm doing #include <petsc/finclude/petsc.h90> in my code.) > >> Just pass in the string 'ds' it looks like our Fortran include files are not >> correct. > > This does compile, but doesn't look like it is actually doing anything- i.e > it's not giving the same behaviour as if I set "-mat_fd_type ds" on the > command line. It just gives the default wp behaviour. > > - Adrian > -- > Dr Adrian Croucher > Senior Research Fellow > Department of Engineering Science > University of Auckland, New Zealand > email: > [email protected] > > tel: +64 (0)9 923 84611 >
