Ok, so on one process all is fine. Likely an issue with how you handle the viewer in the call to your custom PC
> On Jul 10, 2015, at 2:11 PM, Michele Rosso <[email protected]> wrote: > > Barry, > > I made a mistake naming the files: the messy one is from the file while the > correct one is from command line. I am using a custom PC (dmdarepart). > I attached 4 files: > > ksp_view_cl_1proc.txt: output on command line for 1 process run > ksp_view_cl_8proc.txt: output on command line for 8 processes run > ksp_view_fl_1proc.txt: output on file for 1 process run > ksp_view_fl_8proc.txt: output on file for 8 processes run > > This is what I do in the code: > > CALL KSPSolve( ksp, b, x, ierr ) > CALL PetscViewerASCIIOpen( PETSC_COMM_WORLD, ksp_view_file, viewer, > ierr ) > CALL KSPView( ksp, viewer, ierr ) > CALL PetscViewerDestroy( viewer, ierr ) > > I will try to provide a minimal example ASAP. > > Michele > > > On Fri, 2015-07-10 at 13:43 -0500, Barry Smith wrote: >> Are you sure you named the files correctly and sent the right files. This >> is very confusing. What is in ksp_view_file.txt looks reasonable to me: You >> are running three level multigrid where the coarse solve is two level >> multigrid with a coarse solve of SuperLU_Dist. What is in >> ksp_view_command.txt is a jumbled mess which seems to be missing something. >> >> If you run both on one process what happens? >> >> Barry >> >> >> > On Jul 10, 2015, at 1:14 PM, Michele Rosso <[email protected]> wrote: >> > >> > Hi, >> > >> > I am having a problem with KSPview. If I use the command line option, the >> > output of KSPView is complete. If instead I call KSPView after KSPsolve >> > and I have the ksp infos printed on file, the output misses some >> > parts (see attached files). >> > What can I do to fix this? >> > >> > Thanks, >> > Michele >> > >> > <ksp_view_commandline.txt><ksp_view_file.txt> >> >> >> > > <ksp_view_cl_1proc.txt><ksp_view_cl_8proc.txt><ksp_view_fl_1proc.txt><ksp_view_fl_8proc.txt>
