Yes, you need to build a new PETSC_ARCH for 32 with optimization turned on.
export PETSC_ARCH=arch-something-appropriate ./configure --with-debugging=0 all the options you previously used Barry > On Jun 25, 2019, at 7:20 AM, José Lorenzo via petsc-users > <petsc-users@mcs.anl.gov> wrote: > > Matthew, > > thank you for your useful advice. By running the code with -log_view I have > found the following message which might be related to my performance troubles > and that only appears when using PETSC with 32 bit integers: > > ########################################################## > # # > # WARNING!!! # > # # > # This code was compiled with a debugging option. # > # To get timing results run ./configure # > # using --with-debugging=no, the performance will # > # be generally two or three times faster. # > # # > ########################################################## > > > > El mar., 25 jun. 2019 a las 12:32, Matthew Knepley (<knep...@gmail.com>) > escribió: > For any performance question, we need to see the output of -log_view for each > case compared. > > Thanks, > > Matt > > On Tue, Jun 25, 2019 at 5:55 AM José Lorenzo via petsc-users > <petsc-users@mcs.anl.gov> wrote: > I have access to two PETSC installations: one using 64-bit integers and > another one with 32-bit integers. For some reason, when I run the exactly > same FORTRAN code (except for the size of integer type) using both > installations, the 32-bit one exhibits much lower performance in terms of > computation time. This is a problem because I need the 32-bit version in > order to be able to use the LU factorization provided in MUMPS. > > I would appreciate any hint helping me understand why this may happen. > > > -- > What most experimenters take for granted before they begin their experiments > is infinitely more interesting than any results to which their experiments > lead. > -- Norbert Wiener > > https://www.cse.buffalo.edu/~knepley/