Re: [petsc-dev] WTF

2016-07-02 Thread Mark Adams
On Thu, Jun 30, 2016 at 8:48 PM, Jeff Hammond wrote: > > The default settings often don’t work well for 3D problems. > > Are 2D (1D?) problems really the common case for PDE solvers? Aren't > interesting problems 3D? Shouldn't the defaults be set to optimize for 3D? > Lots of people do 2D in p

Re: [petsc-dev] WTF

2016-06-30 Thread Yang, Ulrike Meier
That might be the way to go ☺ Ulrike From: Jeff Hammond [mailto:jeff.scie...@gmail.com] Sent: Thursday, June 30, 2016 11:49 AM To: Yang, Ulrike Meier Cc: Barry Smith ; Justin Chang ; PETSc ; Falgout Rob (rfalg...@llnl.gov) Subject: Re: [petsc-dev] WTF > The default settings often don’t w

Re: [petsc-dev] WTF

2016-06-30 Thread Jeff Hammond
> The default settings often don’t work well for 3D problems. Are 2D (1D?) problems really the common case for PDE solvers? Aren't interesting problems 3D? Shouldn't the defaults be set to optimize for 3D? Jeff -- Jeff Hammond jeff.scie...@gmail.com http://jeffhammond.github.io/

Re: [petsc-dev] WTF

2016-06-30 Thread Yang, Ulrike Meier
; Jeff Hammond ; PETSc ; Falgout Rob (rfalg...@llnl.gov) Subject: Re: [petsc-dev] WTF > On Jun 30, 2016, at 9:55 AM, Yang, Ulrike Meier wrote: > > Do you know which settings were used for BoomerAMG? > The default settings often don’t work well for 3D problems. Ulrike,

Re: [petsc-dev] WTF

2016-06-30 Thread Barry Smith
> (rfalg...@llnl.gov) > Subject: Re: [petsc-dev] WTF > > That guy's results actually make sense to me. > > I also get poor strong-scaling for the FEM version of the poisson equation > (via firedrake) using HYPRE's boomerAMG. The studies were done on Intel >

Re: [petsc-dev] WTF

2016-06-30 Thread Matthew Knepley
o:jychan...@gmail.com] > *Sent:* Thursday, June 30, 2016 12:40 AM > *To:* Barry Smith > *Cc:* Jeff Hammond ; Yang, Ulrike Meier < > yan...@llnl.gov>; PETSc ; Falgout Rob ( > rfalg...@llnl.gov) > *Subject:* Re: [petsc-dev] WTF > > > > That guy's results actually make

Re: [petsc-dev] WTF

2016-06-30 Thread Yang, Ulrike Meier
...@llnl.gov) Subject: Re: [petsc-dev] WTF That guy's results actually make sense to me. I also get poor strong-scaling for the FEM version of the poisson equation (via firedrake) using HYPRE's boomerAMG. The studies were done on Intel E5-2670 machines and had proper OpenMPI binding

Re: [petsc-dev] WTF

2016-06-30 Thread Jeff Hammond
On Wed, Jun 29, 2016 at 8:18 PM, Barry Smith wrote: > > > On Jun 29, 2016, at 10:06 PM, Jeff Hammond > wrote: > > > > > > > > On Wednesday, June 29, 2016, Barry Smith wrote: > > > >Who are these people and why to they have this webpage? > > > > > > Pop up 2-3 directories and you'll see this

Re: [petsc-dev] WTF

2016-06-30 Thread Andrew McRae
Justin, Note that the Firedrake authors did their own Poisson scaling in the firedrake paper (arXiv ), and they needed to set custom BoomerAMG options to get good performance. See pages 6 and 16-18. If I recall correctly, with the out-of-the-box options, far too

Re: [petsc-dev] WTF

2016-06-30 Thread Justin Chang
That guy's results actually make sense to me. I also get poor strong-scaling for the FEM version of the poisson equation (via firedrake) using HYPRE's boomerAMG. The studies were done on Intel E5-2670 machines and had proper OpenMPI bindings.No HYPRE configure options were set via command line so

Re: [petsc-dev] WTF

2016-06-29 Thread Barry Smith
> On Jun 29, 2016, at 10:06 PM, Jeff Hammond wrote: > > > > On Wednesday, June 29, 2016, Barry Smith wrote: > >Who are these people and why to they have this webpage? > > > Pop up 2-3 directories and you'll see this is a grad student who appears to > be trying to learn applied math. I

Re: [petsc-dev] WTF

2016-06-29 Thread Jeff Hammond
On Wednesday, June 29, 2016, Barry Smith wrote: > >Who are these people and why to they have this webpage? > > Pop up 2-3 directories and you'll see this is a grad student who appears to be trying to learn applied math. Is this really your enemy? Don't you guys have some DOE bigwigs to bash?

Re: [petsc-dev] WTF

2016-06-29 Thread Barry Smith
Who are these people and why to they have this webpage? Almost for sure they are doing no process binding and no proper assignment of processes to memory domains. In addition they are likely filling up all the cores on the first node before adding processes to the second core etc. If the

Re: [petsc-dev] WTF

2016-06-29 Thread Paul T. Bauman
On Wed, Jun 29, 2016 at 8:47 PM, Matthew Knepley wrote: > Why the hell does BoomerAMG have such shitty scaling on the Laplace > equation: > > http://guest.ams.sunysb.edu/~zgao/work/airfoil/scaling.html > > Can we rerun this on something at ANL since I think this cannot be true. > Why are they

[petsc-dev] WTF

2016-06-29 Thread Matthew Knepley
Why the hell does BoomerAMG have such shitty scaling on the Laplace equation: http://guest.ams.sunysb.edu/~zgao/work/airfoil/scaling.html Can we rerun this on something at ANL since I think this cannot be true. Matt -- What most experimenters take for granted before they begin their exper

[petsc-dev] wtf

2011-12-18 Thread Mark F. Adams
On Dec 18, 2011, at 6:44 PM, Tim Gallagher wrote: > >>> I used >>> >>> hg log gamg.c >>> >>> to see that you modified this file. Is there a way to see what you did? >> >> Better to ask someone who knows how to do all that stuff then just me. >> >> Barry > > If you do > > hg annotate -u

[petsc-dev] wtf

2011-12-18 Thread Mark F. Adams
Barry, I just pushed with your code removed ... Mark On Dec 18, 2011, at 6:25 PM, Barry Smith wrote: > > On Dec 18, 2011, at 5:21 PM, Mark F. Adams wrote: > >> Barry, did you add: >> >> #define foo 1 >> #if defined(foo) >> PetscFunctionReturn(0); >> #endif >> >> to gamg.c? > > Yes. I ac

[petsc-dev] wtf

2011-12-18 Thread Tim Gallagher
>> I used >> >> hg log gamg.c >> >> to see that you modified this file. Is there a way to see what you did? > > Better to ask someone who knows how to do all that stuff then just me. > >Barry If you do hg annotate -un it will dump the file to stdout with the username and revision tag ne

[petsc-dev] wtf

2011-12-18 Thread Hong Zhang
> ? ?Why did I add that horrible #if defined() stuff? So I could run tests with > MatMatMultSym() on your matrices without running the more memory using and > time consuming part of the example. I am determined to beat Hong's best on > the symbolic part :-). Have you achieved it? I saw PetscHea

[petsc-dev] wtf

2011-12-18 Thread Sean Farley
> > Yes. I accidently pushed the #define foo 1 line, you can remove that one > line, sorry about that. I may have removed it in a later push. > If you have, then you haven't pushed it. > Why did I add that horrible #if defined() stuff? So I could run tests with > MatMatMultSym() on your matrices

[petsc-dev] wtf

2011-12-18 Thread Barry Smith
On Dec 18, 2011, at 5:21 PM, Mark F. Adams wrote: > Barry, did you add: > > #define foo 1 > #if defined(foo) >PetscFunctionReturn(0); > #endif > > to gamg.c? Yes. I accidently pushed the #define foo 1 line, you can remove that one line, sorry about that. I may have removed it in a late