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

Re: [petsc-dev] Slides on Barcelona "Mont Blanc" project
by Matthew Knepley 31 Jul '12

31 Jul '12

On Sun, Jul 29, 2012 at 3:56 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: The problem with the programming model they propose, namely unroll the entire task graph and schedule, is that it has been proposed at least 20 times before by smart, well-funded people. These slides just tell us how the system is supposed to work, which is bullshit. They need to tell us how it is different from other similar systems which failed, and why that will make it succeed. Matt -- 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

1 0

Re: [petsc-dev] Semismooth Newton
by Todd Munson 31 Jul '12

31 Jul '12

On Jul 31, 2012, at 2:36 AM, Jed Brown wrote: > On Mon, Jul 30, 2012 at 5:40 PM, Todd Munson <tmunson(a)mcs.anl.gov> wrote: > > 1. change SNESVI to support general complementarity constraints because only doing box constraints on state variables is lame > > Okay; and how do you want to do this? You need to write out the corresponding KT system > and then you just have a box constrained problem. One can make it easier to write the > KT system, but then you have to precondition the KT system. > > Yes, but if the user created the augmented system, they have to inform the linear solver of that structure. I think it is much more convenient for the user to not have to manage those things. Knowing the constraint structure and forming the augmented system is fine and can be helpful information to know when determining the active set. I would concentrate on linear constraints and inequalities though; they are nicer to deal with and you do not have to be concerned with constraint qualifications. Nonlinear constraints are possible, but you need to satisfy a constraint qualification and you need second derivatives of the constraints for the augmented system. (In optimization terms, the Hessian of the Lagrangian.) > There are some reformulations for polyhedral constraints, but they are, in my opinion, > a bit unwieldy. > > Why unwieldy? The standard semismooth formulation using the Fischer-Burmeister function and its cousins works only for bound constraints. Designing a semismooth reformulation for general polyhedral constraints is difficult. It can be done, but its really not easy. You can define the normal map using projections, but then you need to be able to do the projections onto the constraint set. Of course, the augmented system results in a box constrained problem and the standard formulations work. One thing we should try is adding slacks in the semismooth algorithm for box constraints and then applying the reformulation to the redefined system. There is some evidence that this will work better. > Note that changing the size of the constrained size is also important, and box constraints strike me as even more confusing in that context. I do not follow this at all. Are you talking about something like a time dependent problem and the constraints in the VI are a function of time? That becomes a somewhat odd problem and I'd have to see an example. Normally the constraint set is fixed for all time and just the activities change. Cheers, Todd.

1 0

Re: [petsc-dev] Semismooth Newton
by Jed Brown 31 Jul '12

31 Jul '12

On Mon, Jul 30, 2012 at 5:40 PM, Todd Munson <tmunson(a)mcs.anl.gov> wrote: > > 1. change SNESVI to support general complementarity constraints because > only doing box constraints on state variables is lame > > Okay; and how do you want to do this? You need to write out the > corresponding KT system > and then you just have a box constrained problem. One can make it easier > to write the > KT system, but then you have to precondition the KT system. > Yes, but if the user created the augmented system, they have to inform the linear solver of that structure. I think it is much more convenient for the user to not have to manage those things. > > There are some reformulations for polyhedral constraints, but they are, in > my opinion, > a bit unwieldy. > Why unwieldy? Note that changing the size of the constrained size is also important, and box constraints strike me as even more confusing in that context. > > > 2. have a mode to solve the system in the full space, instead of > eliminating the semi-smooth variables (which produces the bad conditioning) > > There are all these options in the TAO methods.

1 0

Re: [petsc-dev] Jack has produced a scalable sparse direct solver
by Hong Zhang 30 Jul '12

30 Jul '12

Jack, Thanks. Xuan is reading and working on it :-) Hong On Mon, Jul 30, 2012 at 3:55 PM, Jack Poulson <jack.poulson(a)gmail.com>wrote: > Hi Hong, > > As promised, there is significantly more documentation here now: > http://poulson.github.com/Clique/ > > Also, here is an example of solving several right-hand sides with the > high-level interface: > https://github.com/poulson/Clique/blob/master/tests/MultiVectorSolve.cpp > > Jack > > > On Mon, Jul 23, 2012 at 11:53 AM, Hong Zhang <hzhang(a)mcs.anl.gov> wrote: > >> Jack: >> >>> >>> I'm getting started on creating documentation right now; it should end >>> up looking very similar to Elemental's. >>> >> >> The detailed instruction will be very helpful, especially >> the sparse matrix data structures. Notify us when the documentation is >> out. >> >

1 0

Re: [petsc-dev] Semismooth Newton
by Todd Munson 30 Jul '12

30 Jul '12

> 1. change SNESVI to support general complementarity constraints because only doing box constraints on state variables is lame Okay; and how do you want to do this? You need to write out the corresponding KT system and then you just have a box constrained problem. One can make it easier to write the KT system, but then you have to precondition the KT system. There are some reformulations for polyhedral constraints, but they are, in my opinion, a bit unwieldy. > 2. have a mode to solve the system in the full space, instead of eliminating the semi-smooth variables (which produces the bad conditioning) There are all these options in the TAO methods. Cheers, Todd.

1 0

Semismooth Newton
by Jed Brown 30 Jul '12

30 Jul '12

I think we should: 1. change SNESVI to support general complementarity constraints because only doing box constraints on state variables is lame 2. have a mode to solve the system in the full space, instead of eliminating the semi-smooth variables (which produces the bad conditioning) 3. eventually allow the dimension of the constraint space to change on each SNES iteration (would be fantastic for implicit contact)

1 0

Re: [petsc-dev] Jack has produced a scalable sparse direct solver
by Jack Poulson 30 Jul '12

30 Jul '12

Hi Hong, As promised, there is significantly more documentation here now: http://poulson.github.com/Clique/ Also, here is an example of solving several right-hand sides with the high-level interface: https://github.com/poulson/Clique/blob/master/tests/MultiVectorSolve.cpp Jack On Mon, Jul 23, 2012 at 11:53 AM, Hong Zhang <hzhang(a)mcs.anl.gov> wrote: > Jack: > >> >> I'm getting started on creating documentation right now; it should end up >> looking very similar to Elemental's. >> > > The detailed instruction will be very helpful, especially > the sparse matrix data structures. Notify us when the documentation is out. >

1 0

Re: [petsc-dev] Complex Cholesky
by Hong Zhang 30 Jul '12

30 Jul '12

Jose : > > Thanks. I think the same error should go to MatGetFactor_seqaij_petsc > otherwise my second sample run gives wrong results: > Done for seqaij and seqbaij. Hong

1 0

Re: [petsc-dev] Slides on Barcelona "Mont Blanc" project
by Chris Eldred 29 Jul '12

29 Jul '12

Reminds me of the (unfortunately now defunct) green flash project http://www.lbl.gov/cs/html/greenflash.html On Jul 29, 2012 4:00 PM, "Barry Smith" <bsmith(a)mcs.anl.gov> wrote: > > > > > > > _______________________________________________ > > >

1 0

Slides on Barcelona "Mont Blanc" project
by Barry Smith 29 Jul '12

29 Jul '12

> > _______________________________________________

1 0