>
> $ git checkout -b madams/sr-driver3 knepley/feature-plex-refine-3d
> $ git merge knepley/fix-plex-ghost-cells
> $ git cherry-pick [sha1 for "start of segmental refinement"]
I did:
575 git checkout -b madams/sr-driver4 origin/knepley/feature-plex-refine-3d
576 git merge knepley/fix-ple
[amend to add important part of error message]
I am trying to rebase -i, where I squashed one commit message, and get this
error:
12:31 madams/sr-driver2 ~/Codes/petsc$ git pull
Merge made by the 'recursive' strategy.
12:31 madams/sr-driver2 ~/Codes/petsc$ git rebase -i HEAD~10
[detached HEAD 31
I am trying to rebase -i, where I squashed one commit message, and get this
error:
12:31 madams/sr-driver2 ~/Codes/petsc$ git pull
Merge made by the 'recursive' strategy.
12:31 madams/sr-driver2 ~/Codes/petsc$ git rebase -i HEAD~10
[detached HEAD 319baff] TS ex32: Added a parallel test
Author: M
http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/DM/DMPlexConstructGhostCells.html
This page is all mashed up.
Not a big deal, thanks,
Mark
On Oct 15, 2013, at 3:00 PM, Matthew Knepley wrote:
> On Tue, Oct 15, 2013 at 1:42 PM, Mark F. Adams wrote:
> Visit gives me this error message when I try to open the sol.vtk output from
> src/dm/impls/plex/examples/tutorials/ex1. This is from 'ne
Visit gives me this error message when I try to open the sol.vtk output from
src/dm/impls/plex/examples/tutorials/ex1. This is from 'next'.
Mark
localhost:~ markadams$ cd /Users/markadams;
/Applications/VisIt.app/Contents/Resources/2.6.3/../bin/visit -cli -v 2.6.3
-reverse_launch -psn_0_15617
a lot more processor than
what you might expect from a good partitioning. You definitely need to
increase your eq/proc by at least 100x.
On Oct 9, 2013, at 6:50 PM, Matthew Knepley wrote:
> On Wed, Oct 9, 2013 at 5:47 PM, Mark F. Adams wrote:
> >
> > (referring to your previou
>
> (referring to your previous email about the kind of discretization) I'm
> using P_2 finite elements on tetras. The mesh is so small that it might as
> well be partitioned lexicographically (but in this case it was done by
> Metis)
>
Well, This seems fine. Higher order elements can confuse A
gt; > Oh, I think the problem is that GAMG sets the coarse grid solver
> > explicitly as a block jacobi with LU local. GAMG insures that all
> > equation are on one PE for the coarsest grid. ML uses redundant. You
> > should be able to use redundant in GAMG, it is just not the
igh level information.
>>>
>>
>> Oh, I think the problem is that GAMG sets the coarse grid solver
>> explicitly as a block jacobi with LU local. GAMG insures that all
>> equation are on one PE for the coarsest grid. ML uses redundant. You
>> should be able
rry
>
>
> On Oct 8, 2013, at 6:50 PM, "Mark F. Adams" wrote:
>
>> Something is going terrible wrong with the setup in hypre and ML. hypre's
>> default parameters are not setup well for 3D. I use:
>>
>> -pc_hypre
e very light weight (not do excessive
> communication) and provide very high level information.
>
>
> Barry
>
>
> On Oct 8, 2013, at 6:50 PM, "Mark F. Adams" wrote:
>
>> Something is going terrible wrong with the setup in hypre and ML
>
> An IS can start at any number,
Can it start with a negative number in PETSc?
>
>> This is a sort of sensible number.
>
> Nonsense! PETSc and C both use 0-based indexing consistently. This is
> not a Fortran interface.
I meant it is a valid number. An IS can start at 1.
>>
>> No idea. Do you want me to find out? What should a general IS return
>> as 'start'?
>
> ISStrideGetInfo does not dispatch through dynamically registered functionals.
>
> PetscErrorCode ISStrideGetInfo(IS is,PetscInt *first,PetscInt *step)
> {
> IS_Stride *sub;
>
> PetscFunctionBegin
On Sep 26, 2013, at 9:17 PM, Matthew Knepley wrote:
> On Thu, Sep 26, 2013 at 1:15 PM, Mark F. Adams wrote:
> >
> > We have ghost cells in TS ex11, and I think we do them the correct way :)
> > We have a function you register that fills them up, and we
> >
>
> We have ghost cells in TS ex11, and I think we do them the correct way :) We
> have a function you register that fills them up, and we
> loop over them explicitly.
>
Well damn Matt, this looks like the right way to do FV in PETSc, regardless of
the nice BC abstraction, but am I going to g
>
> Here is the magic
> PetscErrorCode VecGetArray2d(Vec x,PetscInt m,PetscInt n,PetscInt
> mstart,PetscInt nstart,PetscScalar **a[])
>
> ierr = VecGetLocalSize(x,&N);CHKERRQ(ierr);
> if (m*n != N) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_ARG_INCOMP,"Local array
> size %D does not match 2d array
at last cell
> force the average of that slot and its neighbor to equal the Dirichlet value.
>
>
> Barry
>
> On Sep 26, 2013, at 11:28 AM, Matthew Knepley wrote:
>
>> On Thu, Sep 26, 2013 at 9:25 AM, Mark F. Adams wrote:
>> I am interested in making a cell cente
I am interested in making a cell centered test problem with Dirichlet BCs. I
want to go high(ish) order and doing explicit stencils (preferred) looks rather
messy and I am interest in using ghost cells. I assume this would entail using
DMDA_BOUNDARY_GHOSTED. I do not see any examples or docum
gt; AlsoOn Thu, 26 Sep 2013, Satish Balay wrote:
>
>> I see Matt pushed a fix for this.
>>
>> Satish
>>
>> On Thu, 26 Sep 2013, Mark F. Adams wrote:
>>
>>> I thought I got my machine working last week but now I seem to get a
>>> Triangle error. I did nuke everything before I ran this.
>>> Mark
>>>
>>
>>
>
>
>> SOR with multis dof?
>
> ex43 and snes ex19 do that in multigrid and there are other examples
> that test more sizes of SOR.
>
In that case I can not think of anything that these test are testing other than
gamg.
So it sounds like you are saying that tests -- tutorial tests -- should not
>>
>
> Something else seems to be amiss. I just did as Mark did (with the
> same gmake in my PATH) and it went through fine. Mark, perhaps you
> could update Xcode to 5.0 and also command-line tools?
I download 5.0 but can not see how to get the command-line tools. Google and
even the Help men
ll gmake
> [and not have macports gmake at-all]
OK, that worked. Thanks.
>
> Satish
>
> On Sat, 21 Sep 2013, Mark F. Adams wrote:
>
>> I'd like to fix gmake. I use macports and see:
>>
>> 19:04 master ~/Codes/petsc$ which gmake
>> /opt/local/bi
later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
Do you have any advice?
On Sep 21, 2013, at 7:07 PM, Satish Balay wrote:
> On Sat, 21 Sep 2013, Mark F. Adams wrote:
>
On Sep 21, 2013, at 5:25 PM, Jed Brown wrote:
> "Mark F. Adams" writes:
>
>>>>
>>>>
>>>> $ make -f gmakefile PETSC_ARCH=arch-macosx-gnu-O
>>>>
>>
>> OK, I'm up and running. Do I need to use this -f flag all
chine.
>
> Parallel build would be: make -f gmakefile PETSC_ARCH=arch-macosx-gnu-O -j 8
>
This works.
> Satish
>
> On Sat, 21 Sep 2013, Mark F. Adams wrote:
>
>> Satish, sorry threads have been getting mixed up. Jed suggested this:
>>
>> $ make -f gma
>>
>>
>> $ make -f gmakefile PETSC_ARCH=arch-macosx-gnu-O
>>
OK, I'm up and running. Do I need to use this -f flag all the time or can I
configure something?
>
> What methods? What are your symptoms?
I believe I sent the full log:
CC arch-macosx-gnu-O/obj/src/ts/impls/implicit/theta/ftn-auto/thetaf.o
FC arch-macosx-gnu-O/obj/src/ts/f90-mod/petsctsmod.o
CLINKER /Users/markadams/Codes/petsc/arch-macosx-gnu-O/lib/libpetsc.dylib
> For some reason gmake is misbehaving on your machine? Previously you had
> a gmake error in --download-mpich part of the build.
>
> Perhaps you can try:
>
> --with-make=make
>
> Satish
>
> On Sat, 21 Sep 2013, Mark F. Adams wrote:
>
>
> [NON-Text Body part not included]
>
>
On Sep 19, 2013, at 7:05 PM, Barry Smith wrote:
>
> On Sep 19, 2013, at 4:28 PM, "Mark F. Adams" wrote:
>
>>
>> On Sep 19, 2013, at 3:53 PM, Barry Smith wrote:
>>
>>>
>>> It is essentially 3 Laplacians
>>
>> If y
On Sep 19, 2013, at 3:53 PM, Barry Smith wrote:
>
> It is essentially 3 Laplacians
If you mean literally 3 scalar laplacians packed in one matrix for some reason
then the one constant vector is fine. Is this a block diagonal matrix with 3
big blocks, if you order it correctly of course?
>>>
>>> results in two levels with the coarse problem being 13*13. Wouldn't
>>> it be natural for the coarse problem to be of size that's an integer
>>> multiple of 3, though?
Yes, something is wrong here. And as Jed said if you have different Mats for
the operator and preconditioner the you ha
I have a field split solver and am using:
-vec_view :x.m:ascii_matlab
I get a zero length file x.m produced with this. Am I using this correctly?
Mark
I seemed to have messed up in trying to build on top of an outstanding pull
request. My branch does not have older material that I thought had been merged
in. I noticed that master is more up to date than the branch I'm working on.
So how should I create a branch to build on a merge of master
When I do SOR I record the state, with the intent of checking that state in
MatResidual to see if we are good to use the cached Ux. But the state of the
vector is incremented after SOR so it can not be the same. So I moved the
increment before the call to MatSOR. Fine.
But I still have a str
Jed, where you going to give me a PetscObjectGetId?
petsc/petsc seems to have disappeared at bitbucket.org - did I miss a memo?
I have changed the architecture of residual calculations in PETSc, in
preparation for some residual optimizations, and have tested with some AMG
tests. This change touched many matrix types other than AIJ and my AMG tests
do not test these nor solvers that use DMs, etc. So my code coverage is
CATED("Use MatResidualDefault()") PETSC_STATIC_INLINE PetscErrorCode
PCMGResidualDefault(Mat A,Vec b,Vec x,Vec r) {
return PCMGResidualDefault(A,b,x,r);
}
?
On Sep 12, 2013, at 5:46 PM, Barry Smith wrote:
>
> On Sep 12, 2013, at 4:39 PM, "Mark F. Adams" wrote:
>
>> I'
Default to using
MatResidualDefault. This will keep from breaking code that uses this to set
the Mat also. And should this be called PCMGSetResidualAndMat?
Mark
On Sep 9, 2013, at 9:25 PM, Barry Smith wrote:
>
> On Sep 9, 2013, at 8:03 PM, "Mark F. Adams" wrote:
>
>
OK, so its ground hog day. The plan?:
1) I should add a residual method to Mat, make a MatResidualDefault, and have
all (yuck) SetUp_MatXXX set this default. I guess there is no base class SetUp
…. then make a MatResidual_SeqAIJ that just does a dumb residual for now, with
testing logic if J
On Sep 8, 2013, at 9:11 PM, Jed Brown wrote:
> "Mark F. Adams" writes:
>
>> So I should add a residual method to Mat, get rid of PCMGResidualDefault,
>> make a MatResidualDefault, and have SetUp_MatAIJ set this default.
>>
>> I guess PCMGSetResidua
[ignore last message from me, here is a revised plan]
>
> PCMGResidualDefault would just call MatResidual instead MatMult, then VecAYPX.
> PetscObjectStateQuery() and PetscObjectGetId() (need to add the latter
> function, but the information is at h->id).
OK, is this the plan:
1) I should add
>
> PetscObjectStateQuery() and PetscObjectGetId() (need to add the latter
> function, but the information is at h->id).
OK, is this the plan:
1) I should add a residual method to Mat, get rid of PCMGResidualDefault, make
a MatResidualDefault, and have all (yuck) SetUp_MatXXX set this default.
So do we have a consensus? use h->id plus a state thing to test the vector and
the state thing to test the this Mat. I thought there was already a 'state
thing' someplace. But it sounds like you are designing it here.
Mark
On Sep 8, 2013, at 5:40 PM, Jed Brown wrote:
> Barry Smith writes:
computation. This way supports the optimization
> automatically without the user or the MG code needing to have any special
> checks or functions to call to use the optimization.
>
> Barry
>
>
> On Sep 8, 2013, at 11:16 AM, "Mark F. Adams" wrote:
>
>
>>
>> 3) How are we going to intercept the residual call? We only use this for MG
>> so we could modify the V-cycle to call a special residual method or check
>> the matrix to see if has the upper part stashed…
>
>MGSetComputeResidual() already exists so this is how you would prorovide
>
On Sep 5, 2013, at 3:39 PM, Satish Balay wrote:
> On Thu, 5 Sep 2013, Mark F. Adams wrote:
>
>>
>> On Sep 5, 2013, at 3:29 PM, Satish Balay wrote:
>>
>>> On Thu, 5 Sep 2013, Mark F. Adams wrote:
>>>
>>>>
>>>> On Sep 5, 2013
On Sep 5, 2013, at 12:39 PM, Matthew Knepley wrote:
> On Thu, Sep 5, 2013 at 11:35 AM, Mark F. Adams wrote:
>
> On Sep 5, 2013, at 6:16 AM, Matthew Knepley wrote:
>
>> On Thu, Sep 5, 2013 at 1:19 AM, Mark F. Adams wrote:
>> In
>> http://www.mcs.anl.gov/petsc/
On Sep 5, 2013, at 12:46 PM, Satish Balay wrote:
> On Thu, 5 Sep 2013, Satish Balay wrote:
>
>> On Thu, 5 Sep 2013, Mark F. Adams wrote:
>
>>> I wanted to not have files get checked and added -skip_petscrc to
>>> the command line but this was not sufficient.
On Sep 5, 2013, at 3:29 PM, Satish Balay wrote:
> On Thu, 5 Sep 2013, Mark F. Adams wrote:
>
>>
>> On Sep 5, 2013, at 12:46 PM, Satish Balay wrote:
>>
>>> On Thu, 5 Sep 2013, Satish Balay wrote:
>>>
>>>> On Thu, 5 Sep 2013, Mark F.
On Sep 5, 2013, at 6:16 AM, Matthew Knepley wrote:
> On Thu, Sep 5, 2013 at 1:19 AM, Mark F. Adams wrote:
> In
> http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Sys/PetscInitialize.html
> I see:
>
> file - [optional] PETSc database file, also checks ~use
>
> 1) ~username/.petscrc
> 2) cwd/.petscrc
> 3) "code specific file", whose name is passed into PetscIntitialize()
>
> If you pass in NULL to PetscInitialize(), it will not check option 3). If you
> pass in a real name to PetscIntiialize(), and
> in the "code specific file" you put -skpi_pet
In
http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Sys/PetscInitialize.html
I see:
file- [optional] PETSc database file, also checks ~username/.petscrc and
.petscrc use NULL to not check for code specific file. Use -skip_petscrc in the
code specific file to skip the .petscrc fi
I was able to run on Edison, 64 bit int w/ hypre, and my user that had the
problems, and was running at large scale, has moved on. GAMG works as well as
hypre. I just wanted to give him options.
On Sep 4, 2013, at 2:19 PM, Satish Balay wrote:
> On Wed, 4 Sep 2013, Satish Balay wrote:
>
>>
I'm rebuilding and trying to reproduce the error, that my user reported. Crays
take forever.
On Sep 4, 2013, at 2:19 PM, Satish Balay wrote:
> On Wed, 4 Sep 2013, Satish Balay wrote:
>
>>
>> I attempted a build on my linuxbox - and I see hypre gets correctly built
>> with --enable-bigint
>
und like you want me to dive into .py land and do this. Should I
do something?
Mark
> I haven't heard back from you.
>
> Barry
>
> Do they have valgrind on this machine?
>
> On Aug 25, 2013, at 1:04 PM, Mark F. Adams wrote:
>
>> I'm getting these memo
>
> if (sml_mype/sml_pe_per_plane .gt. 0) then ! not first, no monitor
> call SNESMonitorCancel(solver%snes,ierr)
> CHKERRQ(ierr)
> call SNESGetKSP(solver%snes,ksp,ierr)
> CHKERRQ(ierr)
> call KSPMonitorCancel(ksp,ierr)
> CHKERRQ(ierr)
> endif
>
> First, a quick
>
> Perhaps you also want SNESSetLagPreconditioner
>
> - lag - -1 indicates NEVER rebuild, 1 means rebuild every time the Jacobian
> is computed within a single nonlinear solve, 2 means every second time
> the Jacobian is built etc. -2 indicates rebuild preconditioner at
> next chance
I have a code that has several concurrent solves (with their own communicators)
and I want to turn off monitors for all but one. My setup code looks something
like this:
call DMSetFromOptions(solver%dalam,ierr)
call DMCompositeAddDM(solver%da,solver%dalam,ierr)
call DMSetFromOptions(solve
I have a SNES solver that runs with a linear operator at times. I use
-ksp_ksponly. The KSPSetOperator seems to get called every solve because the
PC setup code gets called every solve. Should this be happening?
Mark
/autofs/na3_home1/adams/epsi/XGC1_3/./poisson.F90:795: undefined reference to
`snesmonitorcancel_'
0 0100 1700
>>
>> Mark
>>
>> On Aug 23, 2013, at 3:21 PM, Barry Smith wrote:
>>
>>>
>>> Run a bit bigger problem. The results shouldn't be that noisy; unless you
>>> are playing some game in another window :-).
>>>
>>&g
, Barry Smith wrote:
>
> Run a bit bigger problem. The results shouldn't be that noisy; unless you are
> playing some game in another window :-).
>
> Barry
>
> On Aug 23, 2013, at 11:52 AM, Mark F. Adams wrote:
>
>> This code looks funny in MatSOR_SeqAIJ:
history.
Mark
On Aug 23, 2013, at 3:21 PM, Barry Smith wrote:
>
> Run a bit bigger problem. The results shouldn't be that noisy; unless you
> are playing some game in another window :-).
>
> Barry
>
> On Aug 23, 2013, at 11:52 AM, Mark F. Adams wrote:
>
I'm getting these memory error in hypre on the new Cray XC30 at NERSC (Edison)
with 64 bit ints. I found two different stacks in looking around processors.
Any ideas?
Mark
#10 KSPSolve () (at 0x012f121a)
#9 KSPSetUp () (at 0x012f5abb)
#8 PCSetUp () (at 0x011f7f0e)
#7 PC
ular
> storage good for triangular solve and pretty good for multiply.
>
> Barry
>
>
> Barry
>
>> (unless you add an Eisenstat optimization).
>>
>> On Aug 20, 2013 4:26 PM, "Mark F. Adams" wrote:
>> Barry,
>>
>> T
g can be done using those accumulated values at a cost of .5 work
> units instead of 1 unit.
>
> Barry
>
> Each approach has its handful of tricks to shave off work.
>
>
>
> On Aug 13, 2013, at 1:19 PM, Jed Brown wrote:
>
>> "Mark F. Adams
>
>>
>> MatMult2 1.0 1.1801e-02 1.0 1.16e+07 1.0 0.0e+00 0.0e+00
>> 0.0e+00 0 10 0 0 0 0 10 0 0 0 981
>> MatSOR 3 1.0 4.6818e-02 1.0 1.78e+07 1.0 0.0e+00 0.0e+00
>> 0.0e+00 0 16 0 0 0 0 16 0 0 0 380
>>
>> Thus we see that we save all of the
the speed of plain old SOR(3) etc a great
> deal. At the expensive of even more convoluted code :-).
>
> Note also the Eisentat introduces several extra flops per grid point in
> vector operations so for the 5 point stencil the work units are a poor
> measure of actual flops performed.
>
This is happening when the prolongation matrix is repartitioned. The number of
active processors is being reduced. MatGetSubmatrix is called just a few lines
before this with a square matrix, doing the same basic repartitioning. So this
MatGetSubmatrix results in empty local matrices (active
On May 25, 2013, at 8:25 PM, Jed Brown wrote:
> "Mark F. Adams" writes:
>
>> I just updated petsc and rebuilt and am now getting this error:
>
> Have you reconfigured since the commit below?
Yes, I reconfigured after my pull today.
> What do you have for
&
I just updated petsc and rebuilt and am now getting this error:
m/madams> cd BISICLES/code/exec2D/newSolverTest/
exec2D/newSolverTest> aprun -n 64
../driver2d.Linux.64.CC.ftn.OPTHIGH.MPI.PETSC.ex
inputs.ant.solverTest.PETSC.jfnk.steph
[0]PETSC ERROR: - Error Message
I ran a code that once worked with the current version and get this error.
This is a one field solve. I do have a two field version of this code that
uses FieldSplit. Perhaps I polluted to old code in some way but I don't see
how.
[31]PETSC ERROR: - Error Message
tational and Statistical Sciences Division
> Los Alamos National Laboratory
> MS D413, P.O. Box 1663
> Los Alamos, NM 87545
>
> E-mail: christon at lanl.gov
> Phone: (505) 663-5124
> Mobile: (505) 695-5649 (voice mail)
>
> International Journal for Numerical Methods
take a long time and does not seem
to have errors.
> Satish
>
> On Sat, 13 Apr 2013, Mark F. Adams wrote:
>
>> I tried making alldoc1 but got these errors. I tried removing these targets
>> w/o rules. The tried to rung the old make command (not sure what I was
>>
%man+manualpages/%
> ${LOC}/docs/manualpages/manualpages.cit > ${LOC}/docs/manualpages/htmlmap
>- at cat ${PETSC_DIR}/src/docs/mpi.www.index >>
> ${LOC}/docs/manualpages/htmlmap
>-cd src/docs/tex/manual; ${OMAKE} manual.pdf LOC=${LOC}
>-cd src/docs/tex
I'm trying to add a GAMG section to the manual and am having problems building
it:
13:15 2 madams/manual-gamg ~/Codes/petsc/src/docs/tex/manual$ make manual.pdf
LOC=/var/tmp
Traceback (most recent call last):
File "../../../../bin/maint/mapnameslatex.py", line 204, in
raise Ru
> (d) Where do I put binary matrix files that these tests rely upon?
>
> Put them somewhere and we'll add them here:
>
> ftp://ftp.mcs.anl.gov/pub/petsc/matrices/
>
> (We could consider nicer ways to maintain these.)
>
There is currently a 'data' directory in ksp/ksp/examples/tutorials that
On Mar 21, 2013, at 10:07 PM, Jed Brown wrote:
>
> On Thu, Mar 21, 2013 at 3:31 PM, Barry Smith wrote:
> > Why isn't this `PetscOptionsList()` so that it shows up in `-help`?
>
>My mistake, I cut and pasted the PetscOptionsGetInt() from the line above
> and figure that it mustn't be in a
No, not intentionally, we don't touch threads.
On Mar 21, 2013, at 7:45 PM, Jed Brown wrote:
> Did you enable threadcomm? I better finish this work I started and get my
> branch out there...
>
>
> On Thu, Mar 21, 2013 at 6:42 PM, Mark F. Adams
> wrote:
> I'
I'm getting segv in at the end of runs on the Cray XC30. in optimized builds
it segv with a stack like
PetscFinalize
PetscViewerASCIIGetStdout
PetscCommDuplicate
PetscThreadCommGetRank
In debug builds it does not segv but there are a bunch of un freeded memory in
ThreadComm stuff.
Mark
Dave,
I would work from this book:
http://git-scm.com/book/en
it is well done and up to date.
It sounds like branching is what you will want to use.
On Mar 12, 2013, at 7:14 PM, "Nystrom, William D" wrote:
> I have a git project that I call petsc_dev that contains my scripts and
> infrastr
This looks OK, it looks like BuildSystem has been assimilated.
On Mar 12, 2013, at 1:14 PM, "Bacon, Diana H" wrote:
> So I'm assuming that the instructions to build the new petsc-hg repository
> should read:
>
> Once Mercurial is installed obtain petsc-hg with the following:
>
>
>
> * hg clo
On Mar 11, 2013, at 11:08 PM, Richard Tran Mills wrote:
> 1) Jed, what did you do to Barry? =)
> 2) I am am going to interpret the move from Mercurial to Git as a sign that
> the End Times are indeed here.
>
> OK, seriously, I am OK with using Git or Mercurial: I think they are actually
> fai
>
> Richard,
>
> Completely understood. This is why we didn't do what Jed wanted to do
> which was just switch to git completely Sunday at Midnight.
Wow, Barry is the the good cop.
On Mar 2, 2013, at 12:45 PM, Barry Smith wrote:
>
> On Mar 2, 2013, at 11:16 AM, "Mark F. Adams"
> wrote:
>
>>>
>>> An alternative description for the method in the paper is "two-level
>>> unsmoothed aggregation applied to the ASM-
>
> An alternative description for the method in the paper is "two-level
> unsmoothed aggregation applied to the ASM-preconditioned operator".
>
Isn't this paper just doing two level unsmoothed aggregation with a V(0,k)
cycle, where k (and the smoother) is K in the paper?
It looks to me like
I have a SNES solver and use -mat_view ::ascii_matlab.
I do two SNES iterations and get three matrices output. My formJAcobean
explicitly forms AIJ matrices. After the 2nd two matrix printouts I see:
Matrix-free approximation:
err=1.49012e-08 (relative error in function evaluation)
U
>
> This is clearly wrong in the release too. I'm inclined to delete the entire
> useblockis code path since it seems like it will need to be implemented
> completely differently (if it matters for performance).
Yes it should be removed, I just deleted the whole section and tested it.
We could
On Feb 18, 2013, at 4:00 PM, Jed Brown wrote:
> Hmm, this debugging check was added by Mark here.
>
> https://bitbucket.org/petsc/petsc-dev/commits/fb7e4343253b73d3fb40857991a8c6d5d0c7c56b#Lsrc/mat/impls/aij/mpi/mmaij.cT126
>
The logic here is messed up. Valid input could result in a non-col
Do you have more of a stack in the error output?
The ML interface has had some additions to make it more consistent with blocked
matrices. Do you block all of your matrices?
On Feb 18, 2013, at 3:35 PM, Garth N. Wells wrote:
> I have some code using ML as a preconditioner, and it works fine
I pulled today and did a deep clean but get this error:
[ 37%] Building C object CMakeFiles/petsc.dir/src/vec/vec/impls/mpi/pmap.c.o
/Users/markadams/Codes/petsc-dev/src/vec/vec/impls/mpi/pdvec.c: In function
'VecView_MPI_ASCII':
/Users/markadams/Codes/petsc-dev/src/vec/vec/impls/mpi/pdvec.c:144:
f0b0 T _VecNestGetSubVecs_Nest
> 0001d758 S _VecNestGetSubVecs_Nest.eh
> e680 t _VecNestGetSubVecs_Private
> 0001d6c8 s _VecNestGetSubVecs_Private.eh
>
>
>
> On Jan 25, 2013, at 2:49 PM, Jed Brown wrote:
>
>>
>> On Fri, Jan 25, 2013 at 11:13
>
> Can you compare what you've got to snes ex5f90t.F?
This is what I cloned. I just got a debug version working so I'll do a deep
cleaning.
.eh
On Jan 25, 2013, at 2:49 PM, Jed Brown wrote:
>
> On Fri, Jan 25, 2013 at 11:13 AM, Jed Brown wrote:
> On Fri, Jan 25, 2013 at 11:11 AM, Mark F. Adams
> wrote:
> And, when I link I get some missing stuff:
>
> Undefined symbols for architecture x86_64:
> "_
If I nm the .a files I see things like:
_VecNestGetSubVec
But the link errors are for
_vecnestgetsubvec_
It has an extra subscript ? but it is finding other Vec methods.
On Jan 25, 2013, at 12:13 PM, Jed Brown wrote:
>
> On Fri, Jan 25, 2013 at 11:11 AM, Mark F. Adams
> wro
>
> I think you're just missing "use petscdmda".
That did not help:
program main
#include
#include
use petscdm
use petscdmda
use petscsnes
use petsc_kkt_solver_module
use petsc_kkt_solver_moduleinterfaces
! implicit none
And I have this problem also:
>
> And, when I link I get some missing stuff:
>
> Undefined symbols for architecture x86_64:
> "_dmsetoptionsprefix_", referenced from:
> _MAIN__ in petsc_kkt_solver.o
> "_matnestgetsubmat_", referenced from:
> _formjacobian_ in petsc_kkt_solver.o
>
1 - 100 of 353 matches
Mail list logo