July 2012 - petsc-dev - mailman3.cels.anl.gov

Re: [petsc-dev] DMGetNamedGlobalVector() implementation?
by Barry Smith 19 Jul '12

19 Jul '12

On Jul 19, 2012, at 1:25 PM, Jed Brown wrote: > On Thu, Jul 19, 2012 at 1:19 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > Why not implement it using PetscObjectQuery()/PetscObjectCompose()? and not have to introduce a whole new datatype DMNamedVecLink linked list and other crap? > > The only reason this thing has to exist, and it really has to exist in some form, is that we have a reference loop. For example, the SNES Vec X is obtained from the DM (thus X references the DM), but must be made available to the KSP callback that uses the DM. If we attached X to the DM, then we would have a loop X -> DM -> X. Same for the TS vector needed in SNESComputeFunction. > > The general form of PetscObjectCompose allowing reference loops is terribly dangerous and imprecise (you end up implementing a sort of limited garbage collector that periodically looks for loops, but that is super dangerous in parallel because you need determinism in destruction). Is there a way to fix the reference loop problem in PetscObjectCompose()? (Since you have a magic fix for DMGetNamedGlobalVector() can we use that everywhere) You have to admit that introducing in a bunch of places separate constructs for containment rather than using a single universal construct is not good software development. Barry

1 0

Re: [petsc-dev] DMGetNamedGlobalVector() implementation?
by Jed Brown 19 Jul '12

19 Jul '12

On this topic, we should become much more adamant, at least in documentation and maybe also in naming about which arguments are scale-free and which are scale-dependent. Internally, we (Peter and I) have a system that is being followed carefully, including in the examples that we write. Things compose well when these unwritten guidelines are followed. On Thu, Jul 19, 2012 at 1:25 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > On Thu, Jul 19, 2012 at 1:19 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > >> Why not implement it using PetscObjectQuery()/PetscObjectCompose()? and >> not have to introduce a whole new datatype DMNamedVecLink linked list and >> other crap? >> > > The only reason this thing has to exist, and it really has to exist in > some form, is that we have a reference loop. For example, the SNES Vec X is > obtained from the DM (thus X references the DM), but must be made available > to the KSP callback that uses the DM. If we attached X to the DM, then we > would have a loop X -> DM -> X. Same for the TS vector needed in > SNESComputeFunction. > > The general form of PetscObjectCompose allowing reference loops is > terribly dangerous and imprecise (you end up implementing a sort of limited > garbage collector that periodically looks for loops, but that is super > dangerous in parallel because you need determinism in destruction). >

1 0

Re: [petsc-dev] fortran binding for DMGetNamedGlobalVector and DMRestoreNamedGlobalVector
by Satish Balay 19 Jul '12

19 Jul '12

ok pushed now http://petsc.cs.iit.edu/petsc/releases/petsc-3.3/rev/3c63675dd01c Satish On Thu, 19 Jul 2012, Satish Balay wrote: > There are bugs in the code. I'll fix and push. > > Satish > > On Thu, 19 Jul 2012, Barry Smith wrote: > > > > > Looks ok to me. If you can push go ahead and do it or Satish can do it. > > > > Barry > > > > On Jul 19, 2012, at 10:17 AM, Blaise Bourdin wrote: > > > > > Hi, > > > > > > I am attaching a patch adding fortran bindings for DMGetNamedGlobalVector and DMRestoreNamedGlobalVector. I think I got the file names right, this time. > > > > > > Is it possible to push this patch to petsc-3.3 and petsc-dev? > > > > > > Blaise > > > > > > <DMGetRestoreNamedGlobalVector.patch> > > > -- > > > Department of Mathematics and Center for Computation & Technology > > > Louisiana State University, Baton Rouge, LA 70803, USA > > > Tel. +1 (225) 578 1612, Fax +1 (225) 578 4276 http://www.math.lsu.edu/~bourdin > > > > > > > > > > > > > > > > > > > > > > > > > > >

1 0

Re: [petsc-dev] DMGetNamedGlobalVector() implementation?
by Jed Brown 19 Jul '12

19 Jul '12

On Thu, Jul 19, 2012 at 1:19 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > Why not implement it using PetscObjectQuery()/PetscObjectCompose()? and > not have to introduce a whole new datatype DMNamedVecLink linked list and > other crap? > The only reason this thing has to exist, and it really has to exist in some form, is that we have a reference loop. For example, the SNES Vec X is obtained from the DM (thus X references the DM), but must be made available to the KSP callback that uses the DM. If we attached X to the DM, then we would have a loop X -> DM -> X. Same for the TS vector needed in SNESComputeFunction. The general form of PetscObjectCompose allowing reference loops is terribly dangerous and imprecise (you end up implementing a sort of limited garbage collector that periodically looks for loops, but that is super dangerous in parallel because you need determinism in destruction).

1 0

Re: [petsc-dev] fortran binding for DMGetNamedGlobalVector and DMRestoreNamedGlobalVector
by Satish Balay 19 Jul '12

19 Jul '12

There are bugs in the code. I'll fix and push. Satish On Thu, 19 Jul 2012, Barry Smith wrote: > > Looks ok to me. If you can push go ahead and do it or Satish can do it. > > Barry > > On Jul 19, 2012, at 10:17 AM, Blaise Bourdin wrote: > > > Hi, > > > > I am attaching a patch adding fortran bindings for DMGetNamedGlobalVector and DMRestoreNamedGlobalVector. I think I got the file names right, this time. > > > > Is it possible to push this patch to petsc-3.3 and petsc-dev? > > > > Blaise > > > > <DMGetRestoreNamedGlobalVector.patch> > > -- > > Department of Mathematics and Center for Computation & Technology > > Louisiana State University, Baton Rouge, LA 70803, USA > > Tel. +1 (225) 578 1612, Fax +1 (225) 578 4276 http://www.math.lsu.edu/~bourdin > > > > > > > > > > > > > > > >

1 0

Re: [petsc-dev] fortran binding for DMGetNamedGlobalVector and DMRestoreNamedGlobalVector
by Barry Smith 19 Jul '12

19 Jul '12

Looks ok to me. If you can push go ahead and do it or Satish can do it. Barry On Jul 19, 2012, at 10:17 AM, Blaise Bourdin wrote: > Hi, > > I am attaching a patch adding fortran bindings for DMGetNamedGlobalVector and DMRestoreNamedGlobalVector. I think I got the file names right, this time. > > Is it possible to push this patch to petsc-3.3 and petsc-dev? > > Blaise > > <DMGetRestoreNamedGlobalVector.patch> > -- > Department of Mathematics and Center for Computation & Technology > Louisiana State University, Baton Rouge, LA 70803, USA > Tel. +1 (225) 578 1612, Fax +1 (225) 578 4276 http://www.math.lsu.edu/~bourdin > > > > > > >

1 0

DMGetNamedGlobalVector() implementation?
by Barry Smith 19 Jul '12

19 Jul '12

Why not implement it using PetscObjectQuery()/PetscObjectCompose()? and not have to introduce a whole new datatype DMNamedVecLink linked list and other crap? Code reuse is a good thing. Barry

1 0

A small bug
by Adam Rose 19 Jul '12

19 Jul '12

To whom it may concern, I believe that I've found a small error in some example code located in $PETSC_DIR/src/ksp/ksp/examples/tutorials/ex45.c, and I just wanted to bring it to your attention. It's a Laplace equation equation example with Dirichlet boundary conditions u=1 on the square [0,1]x[0,1], but on running the program I noticed that the boundary condition wasn't consistent with the solution. This seems due to line 119 which sets the diagonal values of our discretization matrix in the case of boundary points. It currently reads "v[0] = 2.0*(HxHydHz + HxHzdHy + HyHzdHx);" which is the same as the diagonal away from the boundary, but a quick change to "v[0] = Hx*Hy*Hz;" produces the desired result. Thanks for reading, Adam Rose (Intern, Tech-X Corporation)

1 0

Re: [petsc-dev] downtime for petsc.cs.iit.edu
by Satish Balay 19 Jul '12

19 Jul '12

Its up now. Satish On Thu, 19 Jul 2012, Jed Brown wrote: > Still down. Does anyone know how long the power outage was expected to take? > > On Thu, Jul 19, 2012 at 12:06 AM, Satish Balay <balay(a)mcs.anl.gov> wrote: > > > There is a scheduled power outage at IIT for early morning [5am] of > > 19th. I believe some of the labs are switching off hardware in > > anticipation of this outage. > > > > petsc.cs.iit.edu is unreachable now. I expect it to come back up in > > the morning. [when power is restored after the outage] > > > > Satish > > > > > > >

1 0

fortran binding for DMGetNamedGlobalVector and DMRestoreNamedGlobalVector
by Blaise Bourdin 19 Jul '12

19 Jul '12

Hi, I am attaching a patch adding fortran bindings for DMGetNamedGlobalVector and DMRestoreNamedGlobalVector. I think I got the file names right, this time. Is it possible to push this patch to petsc-3.3 and petsc-dev? Blaise -- Department of Mathematics and Center for Computation & Technology Louisiana State University, Baton Rouge, LA 70803, USA Tel. +1 (225) 578 1612, Fax +1 (225) 578 4276 http://www.math.lsu.edu/~bourdin

1 0