Re: [petsc-users] gamg failure with petsc-dev
On 21/03/14 04:24, Jed Brown wrote:
Stephan Kramer s.kra...@imperial.ac.uk writes:
We have been having some problems with GAMG on petsc-dev (master) for
cases that worked fine on petsc 3.4. We're solving a Stokes equation
(just the velocity block) for a simple convection in a square box
(isoviscous). The problem only occurs if we supply a near null space
(via MatSetNearNullSpace) where we supply the usual (1,0) (0,1) and
(-y,x) (near) null space vectors. If we supply those, the smoother
complains that the diagonal of the A matrix at the first coarsened
level contains a zero. If I dump out the prolongator from the finest
to the first coarsened level it indeed contains a zero column at that
same index. We're pretty confident that the fine level A matrix is
correct (it solves fine with LU). I've briefly spoken to Matt about
this and he suggested trying to run with -pc_gamg_agg_nsmooths 0 (as
the default changed from 3.4 - dev) but that didn't make any
difference, the dumped out prolongator still has zero columns, and it
crashes in the same way. Do you have any further suggestions what to
try and how to further debug this?
Do you set the block size? Can you reproduce by modifying
src/ksp/ksp/examples/tutorials/ex49.c (plane strain elasticity)?
I don't set a block size, no. About ex49: Ah great, with master (just updated
now) I get:
[skramer@stommel]{/data/stephan/git/petsc/src/ksp/ksp/examples/tutorials}$
./ex49 -elas_pc_type gamg -mx 100 -my 100 -mat_no_inode
[0]PETSC ERROR: - Error Message
--
[0]PETSC ERROR: Arguments are incompatible
[0]PETSC ERROR: Zero diagonal on row 1
[0]PETSC ERROR: See http://http://www.mcs.anl.gov/petsc/documentation/faq.html
for trouble shooting.
[0]PETSC ERROR: Petsc Development GIT revision: v3.4.4-3671-gbb161d1 GIT Date:
2014-03-21 01:14:15 +
[0]PETSC ERROR: ./ex49 on a linux-gnu-c-opt named stommel by skramer Fri Mar 21
11:25:55 2014
[0]PETSC ERROR: Configure options --download-fblaslapack=1 --download-blacs=1
--download-scalapack=1 --download-ptscotch=1 --download-mumps=1
--download-hypre=1 --download-suitesparse=1 --download-ml=1
[0]PETSC ERROR: #1 MatInvertDiagonal_SeqAIJ() line 1728 in
/data/stephan/git/petsc/src/mat/impls/aij/seq/aij.c
[0]PETSC ERROR: #2 MatSOR_SeqAIJ() line 1760 in
/data/stephan/git/petsc/src/mat/impls/aij/seq/aij.c
[0]PETSC ERROR: #3 MatSOR() line 3734 in
/data/stephan/git/petsc/src/mat/interface/matrix.c
[0]PETSC ERROR: #4 PCApply_SOR() line 35 in
/data/stephan/git/petsc/src/ksp/pc/impls/sor/sor.c
[0]PETSC ERROR: #5 PCApply() line 440 in
/data/stephan/git/petsc/src/ksp/pc/interface/precon.c
[0]PETSC ERROR: #6 KSP_PCApply() line 227 in
/data/stephan/git/petsc/include/petsc-private/kspimpl.h
[0]PETSC ERROR: #7 KSPSolve_Chebyshev() line 456 in
/data/stephan/git/petsc/src/ksp/ksp/impls/cheby/cheby.c
[0]PETSC ERROR: #8 KSPSolve() line 458 in
/data/stephan/git/petsc/src/ksp/ksp/interface/itfunc.c
[0]PETSC ERROR: #9 PCMGMCycle_Private() line 19 in
/data/stephan/git/petsc/src/ksp/pc/impls/mg/mg.c
[0]PETSC ERROR: #10 PCMGMCycle_Private() line 48 in
/data/stephan/git/petsc/src/ksp/pc/impls/mg/mg.c
[0]PETSC ERROR: #11 PCApply_MG() line 330 in
/data/stephan/git/petsc/src/ksp/pc/impls/mg/mg.c
[0]PETSC ERROR: #12 PCApply() line 440 in
/data/stephan/git/petsc/src/ksp/pc/interface/precon.c
[0]PETSC ERROR: #13 KSP_PCApply() line 227 in
/data/stephan/git/petsc/include/petsc-private/kspimpl.h
[0]PETSC ERROR: #14 KSPInitialResidual() line 63 in
/data/stephan/git/petsc/src/ksp/ksp/interface/itres.c
[0]PETSC ERROR: #15 KSPSolve_GMRES() line 234 in
/data/stephan/git/petsc/src/ksp/ksp/impls/gmres/gmres.c
[0]PETSC ERROR: #16 KSPSolve() line 458 in
/data/stephan/git/petsc/src/ksp/ksp/interface/itfunc.c
[0]PETSC ERROR: #17 solve_elasticity_2d() line 1053 in
/data/stephan/git/petsc/src/ksp/ksp/examples/tutorials/ex49.c
[0]PETSC ERROR: #18 main() line 1104 in
/data/stephan/git/petsc/src/ksp/ksp/examples/tutorials/ex49.c
[0]PETSC ERROR: End of Error Message ---send entire error
message to petsc-ma...@mcs.anl.gov--
Which is the same error we were getting on our problem
Cheers
Stephan
[petsc-users] Preallocation Memory of Finite Element Method's Sparse Matrices
Your faithfully: Last e-mail has some literal error, sorry~ program src/ksp/ksp/examples/tutorials/ex3.c.html is about Bilinear elements on the unit square for Laplacian. After preallocation using ierr = MatMPIAIJSetPreallocation(A,9,NULL,5,NULL);CHKERRQ(ierr); /* More than necessary */, Results of commands of mpiexec -n 2 ./ex3 and mpiexec -n 3 ./ex3 are Norm of error 2.22327e-06 Iterations 6 and Norm of error 3.12849e-07 Iterations 8. Both results are good! However, if I use mpiexec -n 4 ./ex3 or 5,6,7...precesses, error [2]PETSC ERROR: New nonzero at (4,29) (here is for process 4, other positions for different processes) caused a malloc! appear!. For me, this error is unbelievable, because first, the preallocation is more than necessary,how can the new malloc appear? Second, the global number 4 point originally has no neighbor vertices whose global number is 29! This error has tortured me for a long time. This error seems meaningless, however, my recent 3d finite element method cannot be caculated by more processes owing to the new nonzero malloc error! And this is why I want to use 4 or much more processes to compute ex3.c. Thank you for all previous assistence and hope you have a good life! your sincerely LV CHAO 2014/3/21
Re: [petsc-users] 2 level schur
Is there a way to now what the new numbering is? I am assuming that in y example since there are two fields only the numbers associated with them are 0 and 1 hence I tried: -fieldsplit_0_fieldsplit_Field_2_fields 1 -fieldsplit_0_fieldsplit_Field_3_fields 0 which did not work. As mentioned earlier, the following does not work either: -fieldsplit_0_fieldsplit_Field_2_fields 3 -fieldsplit_0_fieldsplit_Field_3_fields 2 and without too much expectation I also passed the following -fieldsplit_0_fieldsplit_Field_2_fields Field_3 -fieldsplit_0_fieldsplit_Field_3_fields Field_2 to no avail. By the way I attached the output from -ksp_view in case I might be doing something wrong? Best, Luc On 03/20/2014 09:01 PM, Matthew Knepley wrote: On Thu, Mar 20, 2014 at 6:20 PM, Luc Berger-Vergiat lb2...@columbia.edu mailto:lb2...@columbia.edu wrote: Hi all, I am solving a four field problem using two Schur complements. Here are the arguments that I usually pass to PETSc to do it: -ksp_type gmres -pc_type fieldsplit -pc_fieldsplit_type schur -pc_fieldsplit_schur_factorization_type full -pc_fieldsplit_schur_precondition selfp -pc_fieldsplit_0_fields 2,3 -pc_fieldsplit_1_fields 0,1 -fieldsplit_0_ksp_type preonly -fieldsplit_0_pc_type fieldsplit -fieldsplit_0_pc_fieldsplit_type schur -fieldsplit_0_pc_fieldsplit_schur_factorization_type full -fieldsplit_0_pc_fieldsplit_schur_precondition selfp -fieldsplit_0_fieldsplit_Field_2_fields 2 -fieldsplit_0_fieldsplit_Field_3_fields 3 -fieldsplit_0_fieldsplit_Field_2_ksp_type preonly -fieldsplit_0_fieldsplit_Field_2_pc_type ilu -fieldsplit_0_fieldsplit_Field_3_ksp_type preonly -fieldsplit_0_fieldsplit_Field_3_pc_type jacobi -fieldsplit_1_ksp_type preonly -fieldsplit_1_pc_type lu -malloc_log mlog -log_summary time.log One issue with this is that when I change -fieldsplit_0_fieldsplit_Field_2_fields 2 to -fieldsplit_0_fieldsplit_Field_2_fields 3 it is ineffective, as if PETSc automatically assign IS 2 to Field 2 even though it is not what I want. Is there a way to pass the arguments correctly so that PETSc goes about switching the IS set of -fieldsplit_0_fieldsplit_Field_2 and -fieldsplit_0_fieldsplit_Field_3? This is crucial to me since I am using the selfp option and the matrix associated to IS 3 is diagonal. By assigning the fields correctly I can get an exact Schur preconditioner and hence very fast convergence. Right now my convergence is not optimal because of this. I believe the inner Schur field statements should not be using the original numbering, but the inner numbering, after they have been reordered. Matt Thanks! Best, Luc -- 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 KSP Object: 1 MPI processes type: gmres GMRES: restart=30, using Classical (unmodified) Gram-Schmidt Orthogonalization with no iterative refinement GMRES: happy breakdown tolerance 1e-30 maximum iterations=1, initial guess is zero tolerances: relative=1e-08, absolute=1e-16, divergence=1e+16 left preconditioning using PRECONDITIONED norm type for convergence test PC Object: 1 MPI processes type: fieldsplit FieldSplit with Schur preconditioner, factorization FULL Preconditioner for the Schur complement formed from Sp, an assembled approximation to S, which uses A00's diagonal's inverse Split info: Split number 0 Defined by IS Split number 1 Defined by IS KSP solver for A00 block KSP Object: (fieldsplit_0_) 1 MPI processes type: preonly maximum iterations=1, initial guess is zero tolerances: relative=1e-05, absolute=1e-50, divergence=1 left preconditioning using NONE norm type for convergence test PC Object: (fieldsplit_0_) 1 MPI processes type: fieldsplit FieldSplit with Schur preconditioner, factorization FULL Preconditioner for the Schur complement formed from Sp, an assembled approximation to S, which uses A00's diagonal's inverse Split info: Split number 0 Defined by IS Split number 1 Defined by IS KSP solver for A00 block KSP Object:(fieldsplit_0_fieldsplit_Field_2_) 1 MPI processes type: preonly maximum iterations=1, initial guess is zero tolerances: relative=1e-05, absolute=1e-50, divergence=1 left preconditioning using NONE norm type for convergence test PC Object:(fieldsplit_0_fieldsplit_Field_2_) 1 MPI processes type: jacobi linear system
[petsc-users] A modified ex12.c
Does anyone know if the DMPlex solver can be run on GPU? Martin
Re: [petsc-users] Building PETSc with Intel mpi
Did you follow the directions here: http://www.mcs.anl.gov/petsc/documentation/faq.html#libimf Did it make any difference? On Mar 21, 2014, at 9:45 AM, Qin Lu lu_qin_2...@yahoo.com wrote: Hello, I was trying to build PETSc-3.4.2 with Intel MPI using Intel-2013 compilers in Linux, but got the error below. The configure.log is attached. *** UNABLE to EXECUTE BINARIES for ./configure --- Cannot run executables created with FC. If this machine uses a batch system to submit jobs you will need to configure using ./configure with the additional option --with-batch. Otherwise there is problem with the compilers. Can you compile and run code with your C/C++ (and maybe Fortran) compilers? See http://www.mcs.anl.gov/petsc/documentation/faq.html#libimf *** Thanks a lot for any suggestions abut the problem, Regards, Qin configure.log
[petsc-users] MatCreateMPIAdj and mat/examples/tutorials/ex11.c
Hi! In the documentation of MatCreateMPIAdj it says that the fifth argument j should be sorted for each row: http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Mat/MatCreateMPIAdj.html The same page links to an example: http://www.mcs.anl.gov/petsc/petsc-current/src/mat/examples/tutorials/ex11.c.html but in that example the entries in jj, on lines 40 and 42, do not seem to be sorted for each row. On rank 0, the column indices are 0, 1 and 2, which are sorted. But the second row correspond to the column indices 1, 3, 2, which are not given in increasing order. The same goes for the indices given in jj for rank 1 on line 42, corresponding to the second row on rank 1. Doesn't that conflict with the documentation? Best regards Torquil Sørensen
Re: [petsc-users] 2 level schur
On Fri, Mar 21, 2014 at 9:37 AM, Luc Berger-Vergiat lb2...@columbia.eduwrote: Is there a way to now what the new numbering is? I am assuming that in y example since there are two fields only the numbers associated with them are 0 and 1 hence I tried: -fieldsplit_0_fieldsplit_Field_2_fields 1 -fieldsplit_0_fieldsplit_Field_3_fields 0 If its an inner fieldsplit, the numbering for options starts over again -fieldsplit_0_fieldsplit_Field_0_fields 1 -fieldsplit_0_fieldsplit_Field_1_fields 0 Thanks, Matt which did not work. As mentioned earlier, the following does not work either: -fieldsplit_0_fieldsplit_Field_2_fields 3 -fieldsplit_0_fieldsplit_Field_3_fields 2 and without too much expectation I also passed the following -fieldsplit_0_fieldsplit_Field_2_fields Field_3 -fieldsplit_0_fieldsplit_Field_3_fields Field_2 to no avail. By the way I attached the output from -ksp_view in case I might be doing something wrong? Best, Luc On 03/20/2014 09:01 PM, Matthew Knepley wrote: On Thu, Mar 20, 2014 at 6:20 PM, Luc Berger-Vergiat lb2...@columbia.eduwrote: Hi all, I am solving a four field problem using two Schur complements. Here are the arguments that I usually pass to PETSc to do it: -ksp_type gmres -pc_type fieldsplit -pc_fieldsplit_type schur -pc_fieldsplit_schur_factorization_type full -pc_fieldsplit_schur_precondition selfp -pc_fieldsplit_0_fields 2,3 -pc_fieldsplit_1_fields 0,1 -fieldsplit_0_ksp_type preonly -fieldsplit_0_pc_type fieldsplit -fieldsplit_0_pc_fieldsplit_type schur -fieldsplit_0_pc_fieldsplit_schur_factorization_type full -fieldsplit_0_pc_fieldsplit_schur_precondition selfp -fieldsplit_0_fieldsplit_Field_2_fields 2 -fieldsplit_0_fieldsplit_Field_3_fields 3 -fieldsplit_0_fieldsplit_Field_2_ksp_type preonly -fieldsplit_0_fieldsplit_Field_2_pc_type ilu -fieldsplit_0_fieldsplit_Field_3_ksp_type preonly -fieldsplit_0_fieldsplit_Field_3_pc_type jacobi -fieldsplit_1_ksp_type preonly -fieldsplit_1_pc_type lu -malloc_log mlog -log_summary time.log One issue with this is that when I change -fieldsplit_0_fieldsplit_Field_2_fields 2 to -fieldsplit_0_fieldsplit_Field_2_fields 3 it is ineffective, as if PETSc automatically assign IS 2 to Field 2 even though it is not what I want. Is there a way to pass the arguments correctly so that PETSc goes about switching the IS set of -fieldsplit_0_fieldsplit_Field_2 and -fieldsplit_0_fieldsplit_Field_3? This is crucial to me since I am using the selfp option and the matrix associated to IS 3 is diagonal. By assigning the fields correctly I can get an exact Schur preconditioner and hence very fast convergence. Right now my convergence is not optimal because of this. I believe the inner Schur field statements should not be using the original numbering, but the inner numbering, after they have been reordered. Matt Thanks! Best, Luc -- 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 -- 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
Re: [petsc-users] MatCreateMPIAdj and mat/examples/tutorials/ex11.c
Thanks. Now fixed in master. Barry On Mar 21, 2014, at 10:51 AM, Torquil Macdonald Sørensen torq...@gmail.com wrote: Hi! In the documentation of MatCreateMPIAdj it says that the fifth argument j should be sorted for each row: http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Mat/MatCreateMPIAdj.html The same page links to an example: http://www.mcs.anl.gov/petsc/petsc-current/src/mat/examples/tutorials/ex11.c.html but in that example the entries in jj, on lines 40 and 42, do not seem to be sorted for each row. On rank 0, the column indices are 0, 1 and 2, which are sorted. But the second row correspond to the column indices 1, 3, 2, which are not given in increasing order. The same goes for the indices given in jj for rank 1 on line 42, corresponding to the second row on rank 1. Doesn't that conflict with the documentation? Best regards Torquil Sørensen
Re: [petsc-users] Preallocation Memory of Finite Element Method's Sparse Matrices
Thank you for reporting this. It was our error. In fact 4 is not enough under certain circumstances; consider where each process has only a single degree of freedom (vertex) then it is coupled to 8 other vertices ALL on other processes. Thus we really need to use 8 instead of 4 as the maximum number of off process coupling. I have fixed this in master so it now runs on any number of processes. Barry On Mar 21, 2014, at 9:11 AM, 吕超 luc...@mail.iggcas.ac.cn wrote: Your faithfully: Last e-mail has some literal error, sorry~ program src/ksp/ksp/examples/tutorials/ex3.c.html is about Bilinear elements on the unit square for Laplacian. After preallocation using ierr = MatMPIAIJSetPreallocation(A,9,NULL,5,NULL);CHKERRQ(ierr); /* More than necessary */, Results of commands of mpiexec -n 2 ./ex3 and mpiexec -n 3 ./ex3 are Norm of error 2.22327e-06 Iterations 6 and Norm of error 3.12849e-07 Iterations 8. Both results are good! However, if I use mpiexec -n 4 ./ex3 or 5,6,7...precesses, error [2]PETSC ERROR: New nonzero at (4,29) (here is for process 4, other positions for different processes) caused a malloc! appear!. For me, this error is unbelievable, because first, the preallocation is more than necessary,how can the new malloc appear? Second, the global number 4 point originally has no neighbor vertices whose global number is 29! This error has tortured me for a long time. This error seems meaningless, however, my recent 3d finite element method cannot be caculated by more processes owing to the new nonzero malloc error! And this is why I want to use 4 or much more processes to compute ex3.c. Thank you for all previous assistence and hope you have a good life! your sin cerely LV CHAO 2014/3/21
Re: [petsc-users] 2 level schur
I hear you though that is not what petsc does. When I name the fields as you suggest: -fieldsplit_0_fieldsplit_Field_0_fields 1 -fieldsplit_0_fieldsplit_Field_1_fields 0 petsc ignores it and still call the fields -fieldsplit_0_fieldsplit_Field_2 -fieldsplit_0_fieldsplit_Field_3 But the automatic naming scheme is not really the issue. It would just be nice to be able to switch the two fields. I will try to change the order in which I pass the IS to the DM and see if I can go around the problem that way. Best, Luc On 03/21/2014 12:13 PM, Matthew Knepley wrote: On Fri, Mar 21, 2014 at 9:37 AM, Luc Berger-Vergiat lb2...@columbia.edu mailto:lb2...@columbia.edu wrote: Is there a way to now what the new numbering is? I am assuming that in y example since there are two fields only the numbers associated with them are 0 and 1 hence I tried: -fieldsplit_0_fieldsplit_Field_2_fields 1 -fieldsplit_0_fieldsplit_Field_3_fields 0 If its an inner fieldsplit, the numbering for options starts over again -fieldsplit_0_fieldsplit_Field_0_fields 1 -fieldsplit_0_fieldsplit_Field_1_fields 0 Thanks, Matt which did not work. As mentioned earlier, the following does not work either: -fieldsplit_0_fieldsplit_Field_2_fields 3 -fieldsplit_0_fieldsplit_Field_3_fields 2 and without too much expectation I also passed the following -fieldsplit_0_fieldsplit_Field_2_fields Field_3 -fieldsplit_0_fieldsplit_Field_3_fields Field_2 to no avail. By the way I attached the output from -ksp_view in case I might be doing something wrong? Best, Luc On 03/20/2014 09:01 PM, Matthew Knepley wrote: On Thu, Mar 20, 2014 at 6:20 PM, Luc Berger-Vergiat lb2...@columbia.edu mailto:lb2...@columbia.edu wrote: Hi all, I am solving a four field problem using two Schur complements. Here are the arguments that I usually pass to PETSc to do it: -ksp_type gmres -pc_type fieldsplit -pc_fieldsplit_type schur -pc_fieldsplit_schur_factorization_type full -pc_fieldsplit_schur_precondition selfp -pc_fieldsplit_0_fields 2,3 -pc_fieldsplit_1_fields 0,1 -fieldsplit_0_ksp_type preonly -fieldsplit_0_pc_type fieldsplit -fieldsplit_0_pc_fieldsplit_type schur -fieldsplit_0_pc_fieldsplit_schur_factorization_type full -fieldsplit_0_pc_fieldsplit_schur_precondition selfp -fieldsplit_0_fieldsplit_Field_2_fields 2 -fieldsplit_0_fieldsplit_Field_3_fields 3 -fieldsplit_0_fieldsplit_Field_2_ksp_type preonly -fieldsplit_0_fieldsplit_Field_2_pc_type ilu -fieldsplit_0_fieldsplit_Field_3_ksp_type preonly -fieldsplit_0_fieldsplit_Field_3_pc_type jacobi -fieldsplit_1_ksp_type preonly -fieldsplit_1_pc_type lu -malloc_log mlog -log_summary time.log One issue with this is that when I change -fieldsplit_0_fieldsplit_Field_2_fields 2 to -fieldsplit_0_fieldsplit_Field_2_fields 3 it is ineffective, as if PETSc automatically assign IS 2 to Field 2 even though it is not what I want. Is there a way to pass the arguments correctly so that PETSc goes about switching the IS set of -fieldsplit_0_fieldsplit_Field_2 and -fieldsplit_0_fieldsplit_Field_3? This is crucial to me since I am using the selfp option and the matrix associated to IS 3 is diagonal. By assigning the fields correctly I can get an exact Schur preconditioner and hence very fast convergence. Right now my convergence is not optimal because of this. I believe the inner Schur field statements should not be using the original numbering, but the inner numbering, after they have been reordered. Matt Thanks! Best, Luc -- 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 -- 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
Re: [petsc-users] Preallocation Memory of Finite Element Method's Sparse Matrices
Barry Smith bsm...@mcs.anl.gov writes: Thank you for reporting this. It was our error. In fact 4 is not enough under certain circumstances; consider where each process has only a single degree of freedom (vertex) then it is coupled to 8 other vertices ALL on other processes. Thus we really need to use 8 instead of 4 as the maximum number of off process coupling. Note that _your_ code should generally not have this problem because you should use a non-pathological partition. pgpWf1NEACFhu.pgp Description: PGP signature
[petsc-users] The same code randomly produce different residual norms
Hi, I run the code src/snes/examples/tutorials/ex5.c with options: * mpirun -n 8 ./ex5 -pc_type mg \ -ksp_monitor\ -pc_mg_levels 3 \ -pc_mg_galerkin \ -da_grid_x 17 \ -da_grid_y 17 \ -mg_levels_ksp_norm_type unpreconditioned \ -snes_monitor \ -mg_levels_ksp_chebyshev_estimate_eigenvalues 0.5,1.1 \ -mg_levels_pc_type sor \ -pc_mg_type \* I run this script several times, but at each time, the residual norms have some tiny differences ( these should not happen). For example: Case 1: * 0 SNES Function norm 1.188788066192e+00 0 KSP Residual norm 1.573384253521e+00 1 KSP Residual norm 3.616321708396e-02 2 KSP Residual norm 2.780221563755e-04 3 KSP Residual norm 2.194662354037e-06 1 SNES Function norm 5.125240595190e-03 0 KSP Residual norm 1.217284338266e-01 1 KSP Residual norm 1.774247017346e-04 2 KSP Residual norm 2.557118591292e-06 3 KSP Residual norm 1.622173269367e-08 2 SNES Function norm 3.922335995111e-05 0 KSP Residual norm 9.960745924876e-04 1 KSP Residual norm 1.273916336665e-06 2 KSP Residual norm 1.571259383270e-08 3 KSP Residual norm 1.250266145356e-10 3 SNES Function norm 2.662898279023e-09 * Case 2: * 0 SNES Function norm 1.188788066192e+00 0 KSP Residual norm 1.573384253521e+00 1 KSP Residual norm 3.616321708396e-02 2 KSP Residual norm 2.780221563755e-04 3 KSP Residual norm 2.194662354037e-06 1 SNES Function norm 5.125240595190e-03 0 KSP Residual norm 1.217284338266e-01 1 KSP Residual norm 1.774247017347e-04 2 KSP Residual norm 2.557118591292e-06 3 KSP Residual norm 1.622173269367e-08 2 SNES Function norm 3.922335995108e-05 0 KSP Residual norm 9.960745924862e-04 1 KSP Residual norm 1.273916336654e-06 2 KSP Residual norm 1.571259383257e-08 3 KSP Residual norm 1.250266145346e-10 3 SNES Function norm 2.662898285038e-09 * These differences are marked by red color. So here, I was wondering there are any explanations why these happen? Thanks, Fande,
Re: [petsc-users] The same code randomly produce different residual norms
Fande Kong fd.k...@siat.ac.cn writes: * mpirun -n 8 ./ex5 -pc_type mg \ -ksp_monitor\ -pc_mg_levels 3 \ -pc_mg_galerkin \ -da_grid_x 17 \ -da_grid_y 17 \ -mg_levels_ksp_norm_type unpreconditioned \ -snes_monitor \ -mg_levels_ksp_chebyshev_estimate_eigenvalues 0.5,1.1 \ -mg_levels_pc_type sor \ -pc_mg_type \* I run this script several times, but at each time, the residual norms have some tiny differences ( these should not happen). For example: Some messages are unpacked in the order in which they are received, rather than doing a moderately expensive sorting. There are some options to make VecScatter and other components reproducible, but this configuration might use a feature for which such an option has not been implemented (-vecscatter_reproduce helps a little in my test, but is not sufficient). The difference you're seeing is harmless, but if you have an application for which this is required (either to reproduce rare events for re-analysis of an unstable dynamical system or due to programmatic requirements that misunderstand error analysis of finite-precision arithmetic), we can hunt down any such places and provide an option to make it reproducible. pgpkI40cDJhSv.pgp Description: PGP signature
Re: [petsc-users] Building PETSc with Intel mpi
--with-mpi-dir=/apps/compilers/intel_2013/impi/4.1.0.024/intel64 --with-mpi-compilers=0 Does this mpi not come with mpicc/mpif90 wrappers? If they do - its best to use them. If not - its best to look at the docs for this compiler - and specify it with the appropriate --with-mpi-include --with-mpi-lib options, [instead of the above] Satish On Fri, 21 Mar 2014, Qin Lu wrote: Sourcing the .csh files of the compiler fixed the problem. Thanks! However, later it got another error (see the attached configure.log for details): *** UNABLE to CONFIGURE with GIVEN OPTIONS (see configure.log for details): --- Fortran error! mpi_init() could not be located! *** It seems the configure did not link the Intel MPI libs. I used --with-mpi-dir to specify the MPI directory, can configure get the correct Intel MPI lib names? If I have to specify the lib names (using --with-mpi-lib?), which libs should I specify? I saw a lot of libs under the directory, such as libmpi.a, libmpi_ipl64.a, libmpi_mt.a, etc. Thanks a lot, Qin From: Barry Smith bsm...@mcs.anl.gov To: Qin Lu lu_qin_2...@yahoo.com Cc: petsc-users petsc-users@mcs.anl.gov Sent: Friday, March 21, 2014 10:11 AM Subject: Re: [petsc-users] Building PETSc with Intel mpi Did you follow the directions here: http://www.mcs.anl.gov/petsc/documentation/faq.html#libimf Did it make any difference? On Mar 21, 2014, at 9:45 AM, Qin Lu lu_qin_2...@yahoo.com wrote: Hello, I was trying to build PETSc-3.4.2 with Intel MPI using Intel-2013 compilers in Linux, but got the error below. The configure.log is attached. *** UNABLE to EXECUTE BINARIES for ./configure --- Cannot run executables created with FC. If this machine uses a batch system to submit jobs you will need to configure using ./configure with the additional option --with-batch. Otherwise there is problem with the compilers. Can you compile and run code with your C/C++ (and maybe Fortran) compilers? See http://www.mcs.anl.gov/petsc/documentation/faq.html#libimf *** Thanks a lot for any suggestions abut the problem, Regards, Qin configure.log
