b.ne599.n500.np8000.sh), I get
> > MatPtAP 3.6224e+01 (nonscalable for small mats,
> scalable for larger ones)
> > scalable MatPtAP 4.6129e+01
> > hypre1.9389e+02
> >
> > I'm merging this work to next, then to master
x27;t support TSX but the programming guide states that TSX users
> must query cpuid to confirm support.
>
> Jeff
>
> On Wed, Feb 15, 2017 at 1:38 PM, Mark Adams wrote:
>
>> Oh, this is because git does not work on KNL. Maybe I should do
>> with-batch=1,
>>
>>
ad-fblaslapack=1 on cray? Shouldn't you be using
> > one of the blas impls provied by cray? [libsci?]
> >
> > Satish
> >
> > On Fri, 17 Feb 2017, Mark Adams wrote:
> >
> > > p4est seems to work now, but I am getting a BLAS failure. But I
> dow
and I had some numerical problem with MKL, so download my own
now.
>
> Satish
>
> On Fri, 17 Feb 2017, Mark Adams wrote:
>
> > p4est seems to work now, but I am getting a BLAS failure. But I download
> > blas
> >
> > On Fri, Feb 17, 2017 at 12:48 PM, Satish
Hong, do you do this for KNL? -C knl
On Thu, Feb 16, 2017 at 6:29 PM, Mark Adams wrote:
>
>
> On Thu, Feb 16, 2017 at 4:49 PM, Zhang, Hong wrote:
>
>> On cori, I tried to compile PETSc on a compute node in the interactive
>> mode, but it did not work. What worked for m
e binary to the queue: srun -n 1 -c 1 ./conftest-xxx
> 3. run the reconfigure-xxx script after the submitted job is done
>
> Hong (Mr.)
>
> On Feb 15, 2017, at 3:38 PM, Mark Adams wrote:
>
> Oh, this is because git does not work on KNL. Maybe I should do
> with-batch=1,
>
I just grab this and put it in git.p4est?
> The "bootstrap" step is only called if `configure` isn't present,
> which it should be in the tarball.
>
> Cheers,
> Toby
>
> >
> > Satish
> >
> > On Wed, 15 Feb 2017, Mark Adams wrote:
>
NUPackage.__init__(self, framework)
> > self.gitcommit = 'v2.0'
> > -self.download = ['git://https://github.com/
> cburstedde/p4est']
> > +self.download = ['git://https://github.com/
> cburstedde/p4est','https://
Oh, this is because git does not work on KNL. Maybe I should do
with-batch=1,
On Wed, Feb 15, 2017 at 4:36 PM, Mark Adams wrote:
> I get this error on the KNL partition at NERSC. P4est works on the
> Haswell partition and other downloads seem to work on the KNL partition.
> Any ideas?
>
>
>
> > Its not a coincidence that it is equal to the FunctionEval time. That is
> the main cost.
>
>Mark,
>
> When Newton is in the quadratic basin then it will do one Function
> evaluation and one Jacobian evaluation per Newton step. The one function
> evaluation takes place inside the li
I am getting a lot of time in line search. see below. This is what I see
with view:
SNESLineSearch Object: 1 MPI processes
type: basic
maxstep=1.00e+08, minlambda=1.00e-12
tolerances: relative=1.00e-08, absolute=1.00e-15,
lambda=1.00e-08
maximum iterations=1
mmit+' in
> repository: '+self.packageDir+'.\n If its a remote branch- use:
> origin/'+self.gitcommit)
>
> Please do
>
> cd $PETSC_DIR/$PETSC_ARCH/externalpackages/git.p4est
>
> git rev-parse v2.0
>
>
OK, but that may make mathematical sense ...
On Sat, Feb 11, 2017 at 2:28 PM, Barry Smith wrote:
>
> Simple damping line search.
>
>
> The default damping is one.
>
>
> > On Feb 11, 2017, at 1:21 PM, Mark Adams wrote:
> >
> >
> >
> >
On Sat, Feb 11, 2017 at 1:17 PM, Barry Smith wrote:
>
> > On Feb 11, 2017, at 12:12 AM, Mark Adams wrote:
> >
> > The web page for linear search seems to be wrong for basic, right?
>
> What is wrong with this?
>
> http://www.mcs.anl.gov/petsc/pe
The web page for linear search seems to be wrong for basic, right? basic
turns line search off. I hope.
Mark
> 2) We do NOT have scalable implementation for MatPtAP() yet.
> We have non-scalable PtAP and interface to Hypre's PtAP. Comparing the
> two,
> Petsc MatPtAP() is approx 3x faster than Hypre's.
>
> I'm writing a scalable MatPtAP() now.
>
> Hong
&
Can you get -info to print on just one processor (0 is fine).
>
>
> I'll check if we can avoid this non-scalable approach.
>
Thanks,
> Meanwhile, can you try option '-matrap false' to use P^T*A*P?
>
>>
>>
Treb can try this.
On Thu, Feb 2, 2017 at 11:13 AM, Hong wrote:
> Mark:
> Try '-matmatmult_via scalable' first. If this works, should we set it as
> default?
>
If it is robust I would say yes unless it is noticeably slower (say >20%)
small scale problems.
ble
Thanks,
>
> https://www.nersc.gov/users/computational-systems/edison/configuration/
> 64GB/node => 5.3GB/process for tasks-per-node=12 [and 2.6GB/node for
> tasks-per-node=24]
>
> Satish
>
> On Wed, 1 Feb 2017, Mark Adams wrote:
>
> > We are getting this error on Edis
We are getting this error on Edison. This problem has run before. This is
using maint from last week.
I recall there are low memory options for mpiptap. I don't think we are
using them.
Thanks,
3404]PETSC ERROR: - Error Message
On Thu, Jan 26, 2017 at 10:53 AM, Hong wrote:
> Mark:
> I see MatTransposeMatMult() is called by PCGAMGCoarsen_AGG().
>
Oh, right. I was thinking of RAP.
> When Gmat1 is created, if you know it is symmetric, then set
> Gmat1->symmetric = PETSC_TRUE;
>
> I can add following:
> +if (Gmat1->
>
>
>>>
> MatTransposeMatMult() tends to use much larger ctable than MatMatMult.
> Is your matrix symmetric? If so, MatMatMult() will use smaller ctable and
> require less communication.
>
>
Humm, that is good to know.
How would I add this to GAMG? Does the user tag the matrix as symmetric and
th
-
> > and see if that works?
> >
> > Satish
> >
> > On Wed, 25 Jan 2017, Mark Adams wrote:
> >
> > > We are still having problems with the large hash value. This should be
> with
> > > ma
We are still having problems with the large hash value. This should be with
maint from earlier this week: v3.7.4-2584-g7175242.
[2998]PETSC ERROR: - Error Message
--
[2998]PETSC ERROR: Argument out of range
[2998]PETSC
ed recently.
>
> Ok, then I am confused. Maybe master fixes the problem?
>
>
It looks like master still has 'aN=A->cmap->N' and
ierr = PetscTableCreate(aN,aN,&ta);CHKERRQ(ierr);
>
> >
> > Fande,
> >
> > On Wed, Jan 11, 2017 at 1:31 PM, Ma
size. That doesn't look right.
>
> https://bitbucket.org/petsc/petsc/commits/c373ccc6#chg-
> src/mat/impls/aij/mpi/mpimatmatmult.c
>
> I see Hong had previously updated this code. Will have to recheck..
>
> Satish
>
> On Wed, 11 Jan 2017, Mark Adams wrote:
>
> >
Oh, I See Fande's thread now. So you extended this.
On Wed, Jan 11, 2017 at 2:52 PM, Mark Adams wrote:
>
>> Must be the same place Fande encountered this error.
>>
>>
> Satish, here is our stack trace. Do you have any thoughts vis-a-vis
> Fande's experien
>
>
> Must be the same place Fande encountered this error.
>
>
Satish, here is our stack trace. Do you have any thoughts vis-a-vis Fande's
experience what the problem is here? My best guess is that the local
matrix size is huge, but this code can have some very large stencils.
Treb: about how man
>
>
>
> Must be the same place Fande encountered this error.
>
> MatSetUpMultiply_MPIAIJ() does:
>
> ierr = PetscTableCreate(aij->B->rmap->n,mat->cmap->N+1,&gid1_lid1);
> CHKERRQ(ierr);
>
> So the local matrix size (n) must be: 4294967296
>
>
Ah, that is a good hint.
Treb: is it possible that so
) to get a stack trace. And
can you tell me how many global equations you have here.
> Satish
>
> On Tue, 10 Jan 2017, Kong, Fande wrote:
>
> > On Tue, Jan 10, 2017 at 1:47 PM, Mark Adams wrote:
> >
> > > We have a code that is failing on large scale runs on Edison
-with-mpiexec=srun LIBS=-lstdc++ --with-64-bit-indices
--prefix=/global/homes/m/madams/petsc_install/petscv3.7-x[16527]PETSC
ERROR:
using:
PETSC_DIR=/global/homes/m/madams/petsc_install/petscv3.7-xc30-dbg64-intel/
maybe go back to opt version?
On 12/22/2016 03:22 PM, Mark Adams wrote:
>
>
>
>
> >
> >
> >
> >
> >Note that Picard
> >
> >Solves the equation A(x) x = b(x) via the defect correction
> algorithm A(x^{n}) (x^{n+1} - x^{n}) = b(x^{n}) - A(x^{n})x^{n}
> >
> >The functions you need to provide a
ep in the form A(x) x = b(x). This may
> require additions to TS, I don't know.
>
>Barry
>
>
>
> > On Jan 1, 2017, at 8:56 AM, Mark Adams wrote:
> >
> > I tried using Picard inside of TS, and SNES diverged. Is this supposed
> to be possible? I t
I tried using Picard inside of TS, and SNES diverged. Is this supposed to
be possible? I tried to do do this by wrapping my operator in a SNES
version and simply adding it like this.
Mark
ierr = TSSetIFunction(ts,NULL,LandIFunction,ctx);CHKERRQ(ierr);
ierr = TSSetIJacobian(ts,J,J,LandIJacobi
>
>
> BTW: Why use --download-fblaslapack on cray? [and not system blas - like
> cray-libsci]
>
Excellent question. Same answer as how did I spend a week preintf debugging
in November. The LAPACK invert method in MKL (at NERSC) used in Plex gaves
differences (the ones that I saw were small) that r
--with-batch=0 seems to fix the problem.
On Thu, Dec 29, 2016 at 4:40 AM, Mark Adams wrote:
> BTW, I cloned this (working) config file and changed to single precision
> and 32 bit ints, and added with-batch=1.
>
> On Wed, Dec 28, 2016 at 11:20 PM, Satish Balay wrote:
>
>>
;
> >I suggest try removing all options except the
> --download-fblaslapack=1 and the compiler names and the --with-batch and
> see if it goes through. Do not cut and past but type each argument; it
> could be some strange character being pass in you cannot see.
> >
> >Bar
OK it works. I added:
ierr = DMViewFromOptions(ctx->dm,NULL,"-dm_view");CHKERRQ(ierr);
Before the solve. There are problems, but nothing fatal:
1) It only plots the first field. I am just debugging so this is not fatal.
2) The current version of Visit v2.11 fails to load it (bad format message)
>
>
>> #PETSc Option Table entries:
>> -dm_view hdf5:res.h5
>> -snes_monitor_residual hdf5:res.h5::append
>>
>> ierr = DMViewFromOptions(ctx->dm,NULL,"-dm_view");CHKERRQ(ierr);
>>
>
> 1) You need this in the code. Its in all my examples.
>
> 2) Did you put it before the vec views? Otherwise, it nee
>
>
> That will stash vectors in the HDF5 file, but you still need to put in the
> DM as well
>
> -dm_view hdf5:ex1.h5
> -snes_monitor_residual hdf5:ex1.h5::append
> -snes_monitor_solution hdf5:ex1.h5::append
>
I tried:
#PETSc Option Table entries:
-dm_view hdf5:res.h5
-snes_monitor_residual
;
> > On Dec 21, 2016, at 1:14 PM, Matthew Knepley wrote:
> >
> > On Wed, Dec 21, 2016 at 12:56 PM, Mark Adams wrote:
> > I ran with:
> >
> > -snes_monitor_residual hdf5:res.h5
> > -snes_monitor_solution hdf5:sol.h5
> >
> > and got h5 file
I ran with:
-snes_monitor_residual hdf5:res.h5
-snes_monitor_solution hdf5:sol.h5
and got h5 files that look OK:
-rw-r--r-- 1 markadams staff 19760 Dec 21 13:46 res.h5
-rw-r--r-- 1 markadams staff 19760 Dec 21 13:46 sol.h5
but then got this. Am I doing this wrong?
13:46 multi-species *
, thanks again for getting me setup on cg.
Mark
On Thu, Dec 15, 2016 at 9:06 PM, Mark Adams wrote:
> Damn, it runs pretty clean, and the code runs, even with optimized. This
> one warning goes away with -malloc_debug -malloc_dump (appended). This is
> the source of the valgrind warning. Th
k analysis, rerun with: --leak-check=full
==26477==
==26477== For counts of detected and suppressed errors, rerun with: -v
==26477== ERROR SUMMARY: 12 errors from 1 contexts (suppressed: 0 from 0)
On Thu, Dec 15, 2016 at 8:10 PM, Mark Adams wrote:
> Great, thanks, all installed and configuri
login then you need to do it at
> accounts.mcs.anl.gov (note that you cannot ssh via a password you need to
> set the ssh key).
>
> Barry
>
> > On Dec 15, 2016, at 5:32 PM, Mark Adams wrote:
> >
> > OK, that was useful. I don't have access to a Linux machine right n
or_if_not_converged -ksp_error_if_not_converged
> to try to get it to output as soon as a problem is detected.
>
>Barry
>
>
> > On Dec 15, 2016, at 4:56 PM, Mark Adams wrote:
> >
> > I have a code that work on my Mac but it fails on both a Cray XC30 and a
>
I have a code that work on my Mac but it fails on both a Cray XC30 and a
KNL, unless the code is build with debug. I get this error message + -info
output. I am using -snes_monitor but get no output. This code was working
and I added a new feature. It does seem to fail when this new feature is
used
time", seqnum, (PetscScalar) seqval,
viewer);CHKERRQ(ierr); */
I'm not sure what the problem is here. Something in the DMSequenceView_HDF5
and/or petsc_gen_xdmf.py. I tried this with ParaView and got an error
message so it looks like it is not a Visit problem.
On Sat, Oct 22, 2016 at 1:51 PM,
/src/dm/impls/forest/p4est/pforest.c:4332
#11 0x000100317a13 in VecView (vec=0x104b3c860, viewer=0x104137a60) at
/Users/markadams/Codes/petsc/src/vec/vec/interface/vector.c:619
#12 0x0001000216a1 in viewDMVec ()
#13 0x00010001a074 in main ()
On Sat, Oct 22, 2016 at 1:51 PM, Mark Adams
I have tried a million things but I have a strange error that I am not
fixing and wanted to see if anyone has anh ideas.
I am printing an 8 field cell centered 3D field (960 variables total). I
initialize the field and print it. It looks perfect. I then do a TSSolve
and it looks fine except all th
Of the 5 (the last one that I checked of course), CEPHES has just raw
source, like LAPACK, and its in C. That is much better as we will
eventually need to vectorize this code, it is a kernel.
On Sat, Oct 22, 2016 at 10:24 AM, Mark Adams wrote:
>
>
> On Sat, Oct 22, 2016 at 8:07 AM
On Sat, Oct 22, 2016 at 8:07 AM, Matthew Knepley wrote:
> Boost is an unmitigated nightmare. What exactly do you need?
>
7.13.3 Legendre Form of Complete Elliptic IntegralsFunction: *double*
*gsl_sf_ellint_Kcomp* *(double k, gsl_mode_t mode)*Function: *int*
*gsl_sf_ellint_Kcomp_e* *(double k, gs
GSL is a GNU math library. It seems to have exactly what we need.
On Fri, Oct 21, 2016 at 11:30 PM, Oxberry, Geoffrey Malcolm <
oxber...@llnl.gov> wrote:
> Do you mean GSL?
>
> From: on behalf of Mark Adams <
> mfad...@lbl.gov>
> Date: Friday, October 21, 2016 at 8:2
Not really a PETSc issue but I need elliptic functions and I tried
downloading GLP but I am getting error with dynamic libs. Matlab and
Mathematica have them, maybe that is a route in PETSc. Anyone have any
advice on better options to get elliptic functions in a code?
Mark
Thanks, rolled back to python 2.7.
On Fri, Oct 21, 2016 at 7:24 AM, Lawrence Mitchell <
lawrence.mitch...@imperial.ac.uk> wrote:
>
> > On 21 Oct 2016, at 15:10, Mark Adams wrote:
> >
> > petsc_gen_xdmf.py works fine on my Mac but I get this error on a Cray.
> It is
petsc_gen_xdmf.py works fine on my Mac but I get this error on a Cray. It
is looking for "file" but I give it a file name. On my Mac I have pything
v2.7.11 and on the Cray I have Python 3.4.3 :: Anaconda 2.2.0 (64-bit).
Any ideas?
-rw-r--r-- 1 madams madams206176 Oct 21 07:03 ex3.h5
07:04 nid
FYI, I pushed a fix for snes/ex56 into knepley/feature-plex-bdfunc so this
should fix next.
On Fri, Oct 14, 2016 at 12:25 PM, Mark Adams wrote:
> Note, I have a pull request in (for weeks) that adds a convergence study
> to snes/ex56 so this might conflict Stefano is doing.
>
> A
Note, I have a pull request in (for weeks) that adds a convergence study to
snes/ex56 so this might conflict Stefano is doing.
Also, next is broken because of what I think was an API change from Matt
et.al. We are getting a bit tied up here.
On Fri, Oct 14, 2016 at 11:05 AM, Mark Adams wrote
Stefano, There is a bug, I think, that was discussed on an earlier thread
and it did not get resolved AFAIK. If you feel like folding this into this
project (if you page it back in) then that would be nice. Here is a quote:
My pull request for snes/ex56 has:
ierr = VecNorm(xx,NORM_INFINITY,&mdisp
Excellent question!
-- Forwarded message --
From: Xiaoye S. Li
Date: Wed, Oct 5, 2016 at 10:44 AM
Subject: Re: reference to GAMG in PETSc
To: Mark Adams
Why Hong Zhang appears twice, are they really 2 persons?
Sherry
On Wed, Oct 5, 2016 at 7:38 AM, Mark Adams wrote:
> A
suggests using CRAIG in the underdetermined
> case, but I don't know if CRAIG is implemented
> in PETSc.
>
> Other norms are more difficult to obtain, but can
> be done. One and infinity norms are recast as
> linear programming problems. Matrix norms are
> equality constrained
I thought least squares was for tall skinny (overdetermined) solves? I have
a short fat (5 x ~100) matrix to solve.
On Wed, Sep 21, 2016 at 4:24 PM, Stefano Zampini
wrote:
> Mark,
>
> You can use KSPLSQR
>
> Stefano
>
> Il 21 set 2016 11:21 PM, "Mark Adams" ha sc
I want to solve for w in V' w = b, where V is tall and skinny. So a short
fat matrix "solve". This is underdetermined. I would like to minimize the
two norm (or any norm) of w. This looks like an optimization problem, would
TAO do this?
Mark
On Thu, Sep 15, 2016 at 5:23 AM, wrote:
>
>
> Dear developer,
>
> I am using a machine whose architecture is not Intel.
> The preconditioner 'gamg' fails in the setup phase, but there is no error
> information output.
>
Please send the output and if there is no output try running with -info.
>
Garth is getting a segv when he uses -pc_gamg_asm_use_agg and an overlap of
1.
This seems like a pretty hard thing to do right and I would not be
surprised if it was not done with complete generality. MatGetSubMatrix is
find to get any submatrix but enlarging a remote subdomain seems very
tricky.
ov 4, 2014 at 4:20 PM, Barry Smith wrote:
>
> You need to be in the correct directory to run the make. and likely you
> also need the LOC=${PETSC_DIR}
>
> > On Nov 4, 2014, at 3:15 PM, Mark Adams wrote:
> >
> > Thanks Barry, it does not seem to pick changes in
);CHKERRQ(ierr);
> if (viamatmatmatmult) {
> ierr = MatMatMatMult(Pt,A,P,scall,fill,C);CHKERRQ(ierr);
> ierr = PetscInfo(*C,"MatPtAP via MatMatMatMult\n");CHKERRQ(ierr);
> } else {
> Mat AP;
> ierr = MatMatMult(A,P,MAT_INITIAL_MATRIX,fill,
>
>
>
> "Give me the branch and exact command line to run."
>
>
12:50 mark/dm-bs-bug= ~/Codes/petsc/src/snes/examples/tutorials$ make ex56
12:50 mark/dm-bs-bug= ~/Codes/petsc/src/snes/examples/tutorials$ make
runex56
> [2]PETSC ERROR: [3]PETSC ERROR: - Error Message
;Look at the source code for MatPtAP. It depends on the approach to how
> it is computed what logging takes place. I am not sure the rational for
> this exact logging is, it would seem better to always log it so perhaps
> that could be changed.
>
> Barry
>
>
> > On
On Sun, Sep 4, 2016 at 8:35 PM, Matthew Knepley wrote:
> On Sun, Sep 4, 2016 at 7:07 PM, Mark Adams wrote:
>
>> On Sat, Sep 3, 2016 at 11:05 PM, Matthew Knepley
>> wrote:
>>
>>> On Sat, Sep 3, 2016 at 11:12 AM, Mark Adams wrote:
>>>
>>>>
I don't seem to be seeing RAP (MatPtAP, or whatever) profile data. Did this
get dumped?
I see that there are two timers in GAMG that I can scrap but if RAP is not
timed then I will add that to GAMG profile data.
Mark
On Sat, Sep 3, 2016 at 11:05 PM, Matthew Knepley wrote:
> On Sat, Sep 3, 2016 at 11:12 AM, Mark Adams wrote:
>
>> The DM in snes/ex56 does elasticity but the block size does not get into
>> vectors. It does something like this:
>>
>> PetscFECreateDefault(dm,
The DM in snes/ex56 does elasticity but the block size does not get into
vectors. It does something like this:
PetscFECreateDefault(dm, dim, dim, PETSC_FALSE, NULL, 3,
&fe);CHKERRQ(ierr); /* elasticity */
DMGetDS(dm, &prob);CHKERRQ(ierr);
PetscDSSetDiscretization(prob, 0, (PetscObject) fe);CHKERRQ
Just an update for the list, Steve seems to have narrowed it down to a
compiler bug in the array assignment as Matt suggested.
Mark
Quick update: just checking ierr isn’t enough to avoid the compiler bug,
but adding zero to n1 is.
Compiling the code as sent:
145, Loop not fused: function c
>
>
>
> Yes, it does. However, I think its more likely that the compiler plays
> with those array copies.
>
>
Matt's first girlfriend dumped him for a Fortran programmer. Sad.
>
>
>
>
> I agree completely, they should be there. I would be much more comfortable
> if just putting them in didn’t solve the error I put them in to look for..
>
>
>
Yes, we have to figure this out, if for no other reason this is a silent
error.
>
>
>
>
> The troubling bit about this, which caused me to suggest that it was a
> race condition, was that if I modify the code a little bit I can get the
> correct result from the scatter. For example, we weren’t always checking
> the returned error flag.
>
If there was an error then you should
if was not updated it should not be
zero. Is that correct or could a_par be zero coming into this routine?
If need be I could add print statements in PETSc to dig into this.
On Thu, Aug 25, 2016 at 5:04 PM, Mark Adams wrote:
> Matt is getting confused with Fortran, I just talked with him and
;>
>>
>>
>> On Aug 25, 2016, at 3:16 PM, Matthew Knepley wrote:
>>
>> On Thu, Aug 25, 2016 at 1:59 PM, Mark Adams wrote:
>>
>>> We have the subroutine below that scatters three vectors. We have used
>>> this code on many machines a
>
>
>> ! end
>> call VecScatterEnd( a_ts%from_petsc(0),a_XX,xxvec
>> (0),INSERT_VALUES,SCATTER_FORWARD,ierr)
>> a_n1 = n1
>>
>
> Why is the line above in here?
>
a_n1 is the return value of the subroutine. I just did this to keep the
types clear, real kind=8 vs PetscScalar.
We have the subroutine below that scatters three vectors. We have used
this code on many machines and it works fine but on one machine data does
not get scattered correctly. The first scatter looks OK, but it looks like
the other two are missing data.
Am I using this correctly? Should we use Vec
Thanks, I keep forgetting that --help works for configure.
On Mon, Aug 22, 2016 at 7:08 PM, Jed Brown wrote:
> Mark Adams writes:
>
> > I am trying to set the memory alignment and added PETSC-MEMALIGN=64, but
> it
> > did not seem to work.
>
> Where did
c-3.6.4
> > >
> > > Executing: mpicc -show
> > > stdout: gcc -fPIC -g3 -I/home/mantis/petsc/petsc-3.7.2/linux-gnu/include
> -L/home/mantis/petsc/petsc-3.7.2/linux-gnu/lib -Wl,-rpath
> -Wl,/home/mantis/petsc/petsc-3.7.2/linux-gnu/lib -Wl,--enable-new-dtags
> -lmp
>
> Okay, there is
>
> -snes_fd_color -snes_fd_color_use_mat -mat_coloring_type greedy
>
> which takes a properly preallocated matrix.
>
Which Mat will it use? Our operators matrix is a shell so it must be the PC ...
If it uses the PC Mat we could take the transpose of the PC Mat, scale
it by
We are pushing the bounds of -snes_fd and I wanted to understand what
we can expect. We do not use a DM. I assume FD has to apply the
operator N times. I don't see how you can color unless we give SNES a
graph (in a DM). Am I missing something?
If not, I could construct a symmetric matrix that
We have a question about the RHS Jacobian (G) for TS. The manual says:
• Jacobian Gu
If using a fully implicit method and the function G() is provided, one
also can provide an appropriate (approximate) Jacobian matrix of G().
If we use, say, backward Euler and provide a G Jacobian, does PETSc
us
thanks,
On Mon, Aug 1, 2016 at 2:04 PM, Satish Balay wrote:
> can you do 'rm -rf arch-titan-opt64-pgi' - and rerun configure?
>
> Satish
>
> On Mon, 1 Aug 2016, Mark Adams wrote:
>
> > make and configure log attached:
> >
> > PGC-S-0055-Illegal oper
Cool, so just --with-openmp and we are off. I was hoping (and sort of
claimed) it would easy,
Thanks,
On Thu, Jul 28, 2016 at 3:08 PM, Satish Balay wrote:
>
> https://bitbucket.org/petsc/petsc/commits/e872ca4eda51751516e9741c8758494774519449
>
> satish
>
> On Thu, 28 Jul 2016
On Thu, Jul 28, 2016 at 3:07 PM, Matthew Knepley wrote:
> Have they lost their fucking collective mind?
>
We will get some data. Good.
>
>Matt
>
> On Thu, Jul 28, 2016 at 12:04 PM, Mark Adams wrote:
>
>> I have a project (XGC) that has heard that hypre suppo
I have a project (XGC) that has heard that hypre supports threads and they
are interested in using it. They would call PETSc outside of a threaded
section, of course, but can hypre pick up the threads in some way?
Mark
nverged in 7 iterations to the relative tolerance.
>
> -Ari
>
> - Original Message -
> From: "Mark Adams" >
> To: "Ari Rappaport" >, "For users of the
> development version of PETSc" >
> Sent: Wednesday, July 27, 2016 3:26:03 PM
&
>>> No, although since we mirror storage you could get an idea if you broke
>>> out the GPU vecs into a separate stage.
>>>
>>> Matt
>>>
>>> On Wed, Jul 27, 2016 at 9:05 AM, Mark Adams wrote:
>>> Would/could PETSc memory usage m
"Linear solve did not converge due to DIVERGED_INDEFINITE_PC
> iterations 2" in the output, could this be causing the problem? It only
> appears to be going for 2 iterations now.
>
> -Ari
>
> - Original Message -
> From: "Mark Adams" >
> To: &
Would/could PETSc memory usage methods (PetscMemoryGetMaximumUsage) pick up
GPU memory usage?
l 26, 2016 at 6:58 PM, Mark Adams wrote:
> So remove one of the -mg_levels_pc_type jacobi and add -mg_coarse_ksp_type
> preonly, then verify that this works on one proc and then try two procs.
>
> On Tue, Jul 26, 2016 at 6:56 PM, Mark Adams wrote:
>
>> Oh, actually this wo
So remove one of the -mg_levels_pc_type jacobi and add -mg_coarse_ksp_type
preonly, then verify that this works on one proc and then try two procs.
On Tue, Jul 26, 2016 at 6:56 PM, Mark Adams wrote:
> Oh, actually this worked. You have this ...pc_type jacobi in there twice,
> so one o
Oh, actually this worked. You have this ...pc_type jacobi in there twice,
so one of them was "unused".
Try this with 2 processors now.
On Tue, Jul 26, 2016 at 6:42 PM, Mark Adams wrote:
>
>
> On Tue, Jul 26, 2016 at 6:24 PM, Ari Rappaport wrote:
>
>> So I comment
-mg_coarse_ksp_type preonly
>
> -Ari
>
> ----- Original Message -
> From: "Mark Adams"
> To: "Ari Rappaport" , "For users of the development
> version of PETSc"
> Sent: Tuesday, July 26, 2016 4:03:03 PM
> Subject: Re: [petsc-de
At the end of this you have:
#PETSc Option Table entries:
-ksp_view
-mg_levels_pc_type jacobi
-options_left
#End of PETSc Option Table entries
There is one unused database option. It is:
Option left: name:-mg_levels_pc_type jacobi (no value)
So this jacobi parameter is not being used.
Do you cal
701 - 800 of 1122 matches
Mail list logo
