Thanks, Mark, Now, the total compute time using GAMG is competitive with ASM. Looks like I could not use something like: "-mg_level_1_ksp_type gmres" because this option makes the compute time much worse.
Fande, On Thu, Apr 13, 2017 at 9:14 AM, Mark Adams <mfad...@lbl.gov> wrote: > > > On Wed, Apr 12, 2017 at 7:04 PM, Kong, Fande <fande.k...@inl.gov> wrote: > >> >> >> On Sun, Apr 9, 2017 at 6:04 AM, Mark Adams <mfad...@lbl.gov> wrote: >> >>> You seem to have two levels here and 3M eqs on the fine grid and 37 on >>> the coarse grid. I don't understand that. >>> >>> You are also calling the AMG setup a lot, but not spending much time >>> in it. Try running with -info and grep on "GAMG". >>> >> >> I got the following output: >> >> [0] PCSetUp_GAMG(): level 0) N=3020875, n data rows=1, n data cols=1, >> nnz/row (ave)=71, np=384 >> [0] PCGAMGFilterGraph(): 100.% nnz after filtering, with threshold >> 0., 73.6364 nnz ave. (N=3020875) >> [0] PCGAMGCoarsen_AGG(): Square Graph on level 1 of 1 to square >> [0] PCGAMGProlongator_AGG(): New grid 18162 nodes >> [0] PCGAMGOptProlongator_AGG(): Smooth P0: max eigen=1.978702e+00 >> min=2.559747e-02 PC=jacobi >> [0] PCGAMGCreateLevel_GAMG(): Aggregate processors noop: new_size=384, >> neq(loc)=40 >> [0] PCSetUp_GAMG(): 1) N=18162, n data cols=1, nnz/row (ave)=94, 384 >> active pes >> [0] PCSetUp_GAMG(): 2 levels, grid complexity = 1.00795 >> [0] PCSetUp_GAMG(): level 0) N=3020875, n data rows=1, n data cols=1, >> nnz/row (ave)=71, np=384 >> [0] PCGAMGFilterGraph(): 100.% nnz after filtering, with threshold >> 0., 73.6364 nnz ave. (N=3020875) >> [0] PCGAMGCoarsen_AGG(): Square Graph on level 1 of 1 to square >> [0] PCGAMGProlongator_AGG(): New grid 18145 nodes >> [0] PCGAMGOptProlongator_AGG(): Smooth P0: max eigen=1.978584e+00 >> min=2.557887e-02 PC=jacobi >> [0] PCGAMGCreateLevel_GAMG(): Aggregate processors noop: new_size=384, >> neq(loc)=37 >> [0] PCSetUp_GAMG(): 1) N=18145, n data cols=1, nnz/row (ave)=94, 384 >> active pes >> > > You are still doing two levels. Just use the parameters that I told you > and you should see that 1) this coarsest (last) grid has "1 active pes" and > 2) the overall solve time and overall convergence rate is much better. > > >> [0] PCSetUp_GAMG(): 2 levels, grid complexity = 1.00792 >> GAMG specific options >> PCGAMGGraph_AGG 40 1.0 8.0759e+00 1.0 3.56e+07 2.3 1.6e+06 1.9e+04 >> 7.6e+02 2 0 2 4 2 2 0 2 4 2 1170 >> PCGAMGCoarse_AGG 40 1.0 7.1698e+01 1.0 4.05e+09 2.3 4.0e+06 5.1e+04 >> 1.2e+03 18 37 5 27 3 18 37 5 27 3 14632 >> PCGAMGProl_AGG 40 1.0 9.2650e-01 1.2 0.00e+00 0.0 9.8e+05 2.9e+03 >> 9.6e+02 0 0 1 0 2 0 0 1 0 2 0 >> PCGAMGPOpt_AGG 40 1.0 2.4484e+00 1.0 4.72e+08 2.3 3.1e+06 2.3e+03 >> 1.9e+03 1 4 4 1 4 1 4 4 1 4 51328 >> GAMG: createProl 40 1.0 8.3786e+01 1.0 4.56e+09 2.3 9.6e+06 2.5e+04 >> 4.8e+03 21 42 12 32 10 21 42 12 32 10 14134 >> GAMG: partLevel 40 1.0 6.7755e+00 1.1 2.59e+08 2.3 2.9e+06 2.5e+03 >> 1.5e+03 2 2 4 1 3 2 2 4 1 3 9431 >> >> >> >> >> >> >> >> >>> >>> >>> On Fri, Apr 7, 2017 at 5:29 PM, Kong, Fande <fande.k...@inl.gov> wrote: >>> > Thanks, Barry. >>> > >>> > It works. >>> > >>> > GAMG is three times better than ASM in terms of the number of linear >>> > iterations, but it is five times slower than ASM. Any suggestions to >>> improve >>> > the performance of GAMG? Log files are attached. >>> > >>> > Fande, >>> > >>> > On Thu, Apr 6, 2017 at 3:39 PM, Barry Smith <bsm...@mcs.anl.gov> >>> wrote: >>> >> >>> >> >>> >> > On Apr 6, 2017, at 9:39 AM, Kong, Fande <fande.k...@inl.gov> wrote: >>> >> > >>> >> > Thanks, Mark and Barry, >>> >> > >>> >> > It works pretty wells in terms of the number of linear iterations >>> (using >>> >> > "-pc_gamg_sym_graph true"), but it is horrible in the compute time. >>> I am >>> >> > using the two-level method via "-pc_mg_levels 2". The reason why >>> the compute >>> >> > time is larger than other preconditioning options is that a matrix >>> free >>> >> > method is used in the fine level and in my particular problem the >>> function >>> >> > evaluation is expensive. >>> >> > >>> >> > I am using "-snes_mf_operator 1" to turn on the Jacobian-free >>> Newton, >>> >> > but I do not think I want to make the preconditioning part >>> matrix-free. Do >>> >> > you guys know how to turn off the matrix-free method for GAMG? >>> >> >>> >> -pc_use_amat false >>> >> >>> >> > >>> >> > Here is the detailed solver: >>> >> > >>> >> > SNES Object: 384 MPI processes >>> >> > type: newtonls >>> >> > maximum iterations=200, maximum function evaluations=10000 >>> >> > tolerances: relative=1e-08, absolute=1e-08, solution=1e-50 >>> >> > total number of linear solver iterations=20 >>> >> > total number of function evaluations=166 >>> >> > norm schedule ALWAYS >>> >> > SNESLineSearch Object: 384 MPI processes >>> >> > type: bt >>> >> > interpolation: cubic >>> >> > alpha=1.000000e-04 >>> >> > maxstep=1.000000e+08, minlambda=1.000000e-12 >>> >> > tolerances: relative=1.000000e-08, absolute=1.000000e-15, >>> >> > lambda=1.000000e-08 >>> >> > maximum iterations=40 >>> >> > KSP Object: 384 MPI processes >>> >> > type: gmres >>> >> > GMRES: restart=100, using Classical (unmodified) Gram-Schmidt >>> >> > Orthogonalization with no iterative refinement >>> >> > GMRES: happy breakdown tolerance 1e-30 >>> >> > maximum iterations=100, initial guess is zero >>> >> > tolerances: relative=0.001, absolute=1e-50, divergence=10000. >>> >> > right preconditioning >>> >> > using UNPRECONDITIONED norm type for convergence test >>> >> > PC Object: 384 MPI processes >>> >> > type: gamg >>> >> > MG: type is MULTIPLICATIVE, levels=2 cycles=v >>> >> > Cycles per PCApply=1 >>> >> > Using Galerkin computed coarse grid matrices >>> >> > GAMG specific options >>> >> > Threshold for dropping small values from graph 0. >>> >> > AGG specific options >>> >> > Symmetric graph true >>> >> > Coarse grid solver -- level ------------------------------- >>> >> > KSP Object: (mg_coarse_) 384 MPI processes >>> >> > type: preonly >>> >> > maximum iterations=10000, initial guess is zero >>> >> > tolerances: relative=1e-05, absolute=1e-50, >>> divergence=10000. >>> >> > left preconditioning >>> >> > using NONE norm type for convergence test >>> >> > PC Object: (mg_coarse_) 384 MPI processes >>> >> > type: bjacobi >>> >> > block Jacobi: number of blocks = 384 >>> >> > Local solve is same for all blocks, in the following KSP >>> and >>> >> > PC objects: >>> >> > KSP Object: (mg_coarse_sub_) 1 MPI processes >>> >> > type: preonly >>> >> > maximum iterations=1, initial guess is zero >>> >> > tolerances: relative=1e-05, absolute=1e-50, >>> divergence=10000. >>> >> > left preconditioning >>> >> > using NONE norm type for convergence test >>> >> > PC Object: (mg_coarse_sub_) 1 MPI processes >>> >> > type: lu >>> >> > LU: out-of-place factorization >>> >> > tolerance for zero pivot 2.22045e-14 >>> >> > using diagonal shift on blocks to prevent zero pivot >>> >> > [INBLOCKS] >>> >> > matrix ordering: nd >>> >> > factor fill ratio given 5., needed 1.31367 >>> >> > Factored matrix follows: >>> >> > Mat Object: 1 MPI processes >>> >> > type: seqaij >>> >> > rows=37, cols=37 >>> >> > package used to perform factorization: petsc >>> >> > total: nonzeros=913, allocated nonzeros=913 >>> >> > total number of mallocs used during MatSetValues >>> calls >>> >> > =0 >>> >> > not using I-node routines >>> >> > linear system matrix = precond matrix: >>> >> > Mat Object: 1 MPI processes >>> >> > type: seqaij >>> >> > rows=37, cols=37 >>> >> > total: nonzeros=695, allocated nonzeros=695 >>> >> > total number of mallocs used during MatSetValues calls >>> =0 >>> >> > not using I-node routines >>> >> > linear system matrix = precond matrix: >>> >> > Mat Object: 384 MPI processes >>> >> > type: mpiaij >>> >> > rows=18145, cols=18145 >>> >> > total: nonzeros=1709115, allocated nonzeros=1709115 >>> >> > total number of mallocs used during MatSetValues calls =0 >>> >> > not using I-node (on process 0) routines >>> >> > Down solver (pre-smoother) on level 1 >>> >> > ------------------------------- >>> >> > KSP Object: (mg_levels_1_) 384 MPI processes >>> >> > type: chebyshev >>> >> > Chebyshev: eigenvalue estimates: min = 0.133339, max = >>> >> > 1.46673 >>> >> > Chebyshev: eigenvalues estimated using gmres with >>> translations >>> >> > [0. 0.1; 0. 1.1] >>> >> > KSP Object: (mg_levels_1_esteig_) 384 >>> 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=10, initial guess is zero >>> >> > tolerances: relative=1e-12, absolute=1e-50, >>> >> > divergence=10000. >>> >> > left preconditioning >>> >> > using PRECONDITIONED norm type for convergence test >>> >> > maximum iterations=2 >>> >> > tolerances: relative=1e-05, absolute=1e-50, >>> divergence=10000. >>> >> > left preconditioning >>> >> > using nonzero initial guess >>> >> > using NONE norm type for convergence test >>> >> > PC Object: (mg_levels_1_) 384 MPI processes >>> >> > type: sor >>> >> > SOR: type = local_symmetric, iterations = 1, local >>> iterations >>> >> > = 1, omega = 1. >>> >> > linear system matrix followed by preconditioner matrix: >>> >> > Mat Object: 384 MPI processes >>> >> > type: mffd >>> >> > rows=3020875, cols=3020875 >>> >> > Matrix-free approximation: >>> >> > err=1.49012e-08 (relative error in function >>> evaluation) >>> >> > Using wp compute h routine >>> >> > Does not compute normU >>> >> > Mat Object: () 384 MPI processes >>> >> > type: mpiaij >>> >> > rows=3020875, cols=3020875 >>> >> > total: nonzeros=215671710, allocated nonzeros=241731750 >>> >> > total number of mallocs used during MatSetValues calls =0 >>> >> > not using I-node (on process 0) routines >>> >> > Up solver (post-smoother) same as down solver (pre-smoother) >>> >> > linear system matrix followed by preconditioner matrix: >>> >> > Mat Object: 384 MPI processes >>> >> > type: mffd >>> >> > rows=3020875, cols=3020875 >>> >> > Matrix-free approximation: >>> >> > err=1.49012e-08 (relative error in function evaluation) >>> >> > Using wp compute h routine >>> >> > Does not compute normU >>> >> > Mat Object: () 384 MPI processes >>> >> > type: mpiaij >>> >> > rows=3020875, cols=3020875 >>> >> > total: nonzeros=215671710, allocated nonzeros=241731750 >>> >> > total number of mallocs used during MatSetValues calls =0 >>> >> > not using I-node (on process 0) routines >>> >> > >>> >> > >>> >> > Fande, >>> >> > >>> >> > On Thu, Apr 6, 2017 at 8:27 AM, Mark Adams <mfad...@lbl.gov> wrote: >>> >> > On Tue, Apr 4, 2017 at 10:10 AM, Barry Smith <bsm...@mcs.anl.gov> >>> wrote: >>> >> > > >>> >> > >> Does this mean that GAMG works for the symmetrical matrix only? >>> >> > > >>> >> > > No, it means that for non symmetric nonzero structure you need >>> the >>> >> > > extra flag. So use the extra flag. The reason we don't always use >>> the flag >>> >> > > is because it adds extra cost and isn't needed if the matrix >>> already has a >>> >> > > symmetric nonzero structure. >>> >> > >>> >> > BTW, if you have symmetric non-zero structure you can just set >>> >> > -pc_gamg_threshold -1.0', note the "or" in the message. >>> >> > >>> >> > If you want to mess with the threshold then you need to use the >>> >> > symmetrized flag. >>> >> > >>> >> >>> > >>> >> >> >