January 2017 - petsc-dev - mailman3.cels.anl.gov

Re: [petsc-dev] Bump MOAB version?
by Vijay S. Mahadevan 31 Jan '17

31 Jan '17

Toby, thanks for the update. Yes, I recently updated the "petsc" branch in MOAB to make sure that v4.9.2 is merged. Some of these changes are needed for one of my upcoming PR's to improve DMMoab support. I however have been on some international travel and haven't found enough time to take care of this. Iulian, can you create a tarball that corresponds to the HEAD in petsc branch (if not already available from the ftp site) ? Vijay PS: The macports errors are related to ESMF. They have their custom version of MOAB (4.6.x) and so I'm unsure whether the port needs the latest MOAB as its dependency. On Wed, Feb 1, 2017 at 3:07 AM, Tobin Isaac <tisaac(a)uchicago.edu> wrote: > > The commit of MOAB that we use in MAOB.py doesn't compile in gcc6 > due to c++14 compatibility issues [1]. I checked out the latest petsc > branch from their repo and it seems to compile. Can we ask them to > host an updated tarball, too? > > [1] https://trac.macports.org/ticket/52427#comment:2 > > Toby

1 0

Bump MOAB version?
by Tobin Isaac 31 Jan '17

31 Jan '17

The commit of MOAB that we use in MAOB.py doesn't compile in gcc6 due to c++14 compatibility issues [1]. I checked out the latest petsc branch from their repo and it seems to compile. Can we ask them to host an updated tarball, too? [1] https://trac.macports.org/ticket/52427#comment:2 Toby

1 0

Re: [petsc-dev] Matrix-free preconditioners
by Chris Eldred 30 Jan '17

30 Jan '17

Hey Jed, That's what I thought, thanks for the clarification! Regards, Chris Eldred On Sat, Jan 28, 2017 at 2:58 PM, Jed Brown <jed(a)jedbrown.org> wrote: > That's pretty much it. You can probably easily apply the local part of > the operator so you can do a block Jacobi by running an iterative method > on subdomains. Depending on the preconditioner, you might be able to > apply Neumann subdomain operators -- a research topic would be to set up > a BDDC-type method without explicit subdomain matrices. > > Chris Eldred <chris.eldred(a)gmail.com> writes: > >> Hey Petsc Dev, >> >> I had some questions about preconditioners, and I was hoping the list >> could help me out. Specifically, if I have a MatShell with the ability >> to compute the diagonal of the matrix; and a geometric multigrid >> hierarchy with appropriate operators defined; what are my >> preconditioner options? >> >> From reading the manual and the doc pages, it appears that my options are >> 1) Jacobi (and block variants) >> 2) Multigrid >> plus the ability to use Fieldsplit for block matrices. Then within >> multigrid, any iterative KSP (ie one that requires only matrix-vector >> products) can be used as a smoother, with the basic ones being >> Chebyshev and Richardson using Jacobi as a preconditioner. >> >> Is this correct? Is there anything else? >> >> Regards, >> Chris Eldred >> >> -- >> Chris Eldred >> https://www.math.univ-paris13.fr/~eldred/ >> Postdoctoral Fellow, LAGA, University of Paris 13 >> PhD, Atmospheric Science, Colorado State University, 2015 >> DOE Computational Science Graduate Fellow (Alumni) >> B.S. Applied Computational Physics, Carnegie Mellon University, 2009 >> chris.eldred(a)gmail.com -- Chris Eldred https://www.math.univ-paris13.fr/~eldred/ Postdoctoral Fellow, LAGA, University of Paris 13 PhD, Atmospheric Science, Colorado State University, 2015 DOE Computational Science Graduate Fellow (Alumni) B.S. Applied Computational Physics, Carnegie Mellon University, 2009 chris.eldred(a)gmail.com

1 0

Re: [petsc-dev] Matrix-free preconditioners
by Jed Brown 28 Jan '17

28 Jan '17

That's pretty much it. You can probably easily apply the local part of the operator so you can do a block Jacobi by running an iterative method on subdomains. Depending on the preconditioner, you might be able to apply Neumann subdomain operators -- a research topic would be to set up a BDDC-type method without explicit subdomain matrices. Chris Eldred <chris.eldred(a)gmail.com> writes: > Hey Petsc Dev, > > I had some questions about preconditioners, and I was hoping the list > could help me out. Specifically, if I have a MatShell with the ability > to compute the diagonal of the matrix; and a geometric multigrid > hierarchy with appropriate operators defined; what are my > preconditioner options? > > From reading the manual and the doc pages, it appears that my options are > 1) Jacobi (and block variants) > 2) Multigrid > plus the ability to use Fieldsplit for block matrices. Then within > multigrid, any iterative KSP (ie one that requires only matrix-vector > products) can be used as a smoother, with the basic ones being > Chebyshev and Richardson using Jacobi as a preconditioner. > > Is this correct? Is there anything else? > > Regards, > Chris Eldred > > -- > Chris Eldred > https://www.math.univ-paris13.fr/~eldred/ > Postdoctoral Fellow, LAGA, University of Paris 13 > PhD, Atmospheric Science, Colorado State University, 2015 > DOE Computational Science Graduate Fellow (Alumni) > B.S. Applied Computational Physics, Carnegie Mellon University, 2009 > chris.eldred(a)gmail.com

1 0

Matrix-free preconditioners
by Chris Eldred 28 Jan '17

28 Jan '17

Hey Petsc Dev, I had some questions about preconditioners, and I was hoping the list could help me out. Specifically, if I have a MatShell with the ability to compute the diagonal of the matrix; and a geometric multigrid hierarchy with appropriate operators defined; what are my preconditioner options? >From reading the manual and the doc pages, it appears that my options are 1) Jacobi (and block variants) 2) Multigrid plus the ability to use Fieldsplit for block matrices. Then within multigrid, any iterative KSP (ie one that requires only matrix-vector products) can be used as a smoother, with the basic ones being Chebyshev and Richardson using Jacobi as a preconditioner. Is this correct? Is there anything else? Regards, Chris Eldred -- Chris Eldred https://www.math.univ-paris13.fr/~eldred/ Postdoctoral Fellow, LAGA, University of Paris 13 PhD, Atmospheric Science, Colorado State University, 2015 DOE Computational Science Graduate Fellow (Alumni) B.S. Applied Computational Physics, Carnegie Mellon University, 2009 chris.eldred(a)gmail.com

1 0

Re: [petsc-dev] new test suite stops on example compile error
by Satish Balay 27 Jan '17

27 Jan '17

Scott, The issue I was refering to was : >>>>>>>> arch-linux-pkgs-valgrind/tests/src/dm/impls/plex/examples/tests/ex15.o: In function `main': /sandbox/petsc/petsc.clone-4/src/dm/impls/plex/examples/tests/ex15.c:84: undefined reference to `PetscViewerHDF5Open' /sandbox/petsc/petsc.clone-4/src/dm/impls/plex/examples/tests/ex15.c:108: undefined reference to `PetscViewerHDF5Open' collect2: error: ld returned 1 exit status gmake[2]: *** [arch-linux-pkgs-valgrind/tests/src/dm/impls/plex/examples/tests/ex15] Error 1 gmake[2]: *** Waiting for unfinished jobs.... gmake[1]: [allgtest] Error 1 (ignored) <<<<< Satish On Fri, 27 Jan 2017, Scott Kruger wrote: > > > This appears to be the legacy builds, not the new system: > > Possible problem with ex115_5, diffs above > ============================== > ******* Testing: testexamples_SUPERLU_DIST ******* > /usr/bin/sh: line 1: 1090 Killed ( cd $dir; if [ nofoo = > dotfoo ]; then > printf "."; > fi; if [ nofoo = foo -o nofoo = yesfoo -o nofoo = dirfoo ]; then > echo testexamples_SUPERLU_DIST in: `pwd`; > fi; /usr/bin/gmake --no-print-directory tree ACTION=testexamples_SUPERLU_DIST > PETSC_ARCH=arch-linux-pkgs-valgrind LOC= DATAFILESPATH=/home/petsc/datafiles > BASE_DIR= PRINT_PROGRESS=no ) > ****************************************************** > Finished Build on el7.novalocal arch-linux-pkgs-valgrind Fri Jan 27 08:35:12 > CST 2017 > X******************************************************X > > > > On 1/27/17 12:59 PM, Satish Balay wrote: > > Jed, > > > > I see the test build stops on encountering errors. This is fine for > > library build - but for tests - I think its best if it skips this test > > and continues with the remaining tests. Is this possible? > > > > http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2017/01/27/examples_ne… > > > > Satish > > > >

1 0

Re: [petsc-dev] new test suite stops on example compile error
by Satish Balay 27 Jan '17

27 Jan '17

Thanks! pushed this change. Satish On Fri, 27 Jan 2017, Jed Brown wrote: > make -k > > Satish Balay <balay(a)mcs.anl.gov> writes: > > > Jed, > > > > I see the test build stops on encountering errors. This is fine for > > library build - but for tests - I think its best if it skips this test > > and continues with the remaining tests. Is this possible? > > > > http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2017/01/27/examples_ne… > > > > Satish >

1 0

Re: [petsc-dev] new test suite stops on example compile error
by Jed Brown 27 Jan '17

27 Jan '17

make -k Satish Balay <balay(a)mcs.anl.gov> writes: > Jed, > > I see the test build stops on encountering errors. This is fine for > library build - but for tests - I think its best if it skips this test > and continues with the remaining tests. Is this possible? > > http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2017/01/27/examples_ne… > > Satish

1 0

Re: [petsc-dev] new test suite stops on example compile error
by Scott Kruger 27 Jan '17

27 Jan '17

This appears to be the legacy builds, not the new system: Possible problem with ex115_5, diffs above ============================== ******* Testing: testexamples_SUPERLU_DIST ******* /usr/bin/sh: line 1: 1090 Killed ( cd $dir; if [ nofoo = dotfoo ]; then printf "."; fi; if [ nofoo = foo -o nofoo = yesfoo -o nofoo = dirfoo ]; then echo testexamples_SUPERLU_DIST in: `pwd`; fi; /usr/bin/gmake --no-print-directory tree ACTION=testexamples_SUPERLU_DIST PETSC_ARCH=arch-linux-pkgs-valgrind LOC= DATAFILESPATH=/home/petsc/datafiles BASE_DIR= PRINT_PROGRESS=no ) ****************************************************** Finished Build on el7.novalocal arch-linux-pkgs-valgrind Fri Jan 27 08:35:12 CST 2017 X******************************************************X On 1/27/17 12:59 PM, Satish Balay wrote: > Jed, > > I see the test build stops on encountering errors. This is fine for > library build - but for tests - I think its best if it skips this test > and continues with the remaining tests. Is this possible? > > http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2017/01/27/examples_ne… > > Satish > -- Tech-X Corporation kruger(a)txcorp.com 5621 Arapahoe Ave, Suite A Phone: (720) 974-1841 Boulder, CO 80303 Fax: (303) 448-7756

1 0

new test suite stops on example compile error
by Satish Balay 27 Jan '17

27 Jan '17

Jed, I see the test build stops on encountering errors. This is fine for library build - but for tests - I think its best if it skips this test and continues with the remaining tests. Is this possible? http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2017/01/27/examples_ne… Satish

1 0