[petsc-dev] Test Problem Recommendations

[Jed Brown]

On Fri, Jan 20, 2012 at 16:03, Nystrom, William D <wdn at lanl.gov> wrote:

> I'm getting ready to use petsc-dev as part of our test suite for doing
> stability testing of some new hardware platforms with gpus.  I'm
> interested in testing both with and without using gpus.  So far, I have
> not extensively explored the various example problems that come with
> PETSc.  Mainly, I have run src/ksp/ksp/examples/tutorials/ex2.c
>

This is a horrible example because the parallel decomposition is 1D. You
can run essentially the same model with a better decomposition (and inside
a nonlinear problem) with src/snes/examples/tutorials/ex5.c (which can also
do geometric multigrid, but you don't have to use it).


> which is a 2d Laplace's equation problem.  I'm looking for recommendations
> of PETSc example problems that I should try, especially ones that might
> be more challenging numerically than ex2.c or more realistic of real
> world problems.  Does anyone reading this email have a favorite PETSc
> example problem that you would suggest I consider and try?
>

Here are some options and profiling output (on a BG/P) for a 3D ice flow
problem (src/snes/examples/tutorials/ex48.c):

https://github.com/jedbrown/tme-ice/tree/master/shaheen/b


src/ksp/ksp/examples/tutorials/ex43.c (2D Stokes, preconditioned using
fieldsplit inside multigrid)

mpiexec -n 8 ./ex43 -stokes_ksp_type gcr -stokes_ksp_gcr_restart 60
-stokes_ksp_norm_type unpreconditioned -stokes_ksp_rtol 1e-8 -c_str 3
-sinker_eta0 1.0 -sinker_eta1 100 -sinker_dx 0.4 -sinker_dy 0.3 -mx 1024
-my 1024 -stokes_ksp_monitor -o sc1024.dat -stokes_pc_type mg
-stokes_mg_levels_pc_type fieldsplit -stokes_pc_mg_galerkin
-stokes_mg_levels_pc_fieldsplit_block_size 3
-stokes_mg_levels_pc_fieldsplit_0_fields 0,1
-stokes_mg_levels_pc_fieldsplit_1_fields 2
-stokes_mg_levels_fieldsplit_0_pc_type sor -stokes_mg_levels_ksp_type
chebychev -stokes_mg_levels_ksp_max_it 1
-stokes_mg_levels_ksp_chebychev_estimate_eigenvalues 0,0.2,0,1.1
-stokes_pc_mg_levels 5 -log_summary


src/ksp/ksp/examples/tutorials/ex56.c (3D elasticity, process count should
be 8^n)
mpiexec -n 64 ./ex56 -ne 255 -alpha 1.e-3 -ksp_monitor_short -ksp_type cg
-pc_gamg_type sa  -ksp_norm_type unpreconditioned -pc_gamg_verbose
-pc_gamg_repartition -pc_gamg_process_eq_limit 10 -pc_gamg_coarse_eq_limit
20 -pc_gamg_mat_partitioning_type ptscotch -log_summary
-pc_gamg_mat_partitioning_ptscotch_strategy scalability
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20120120/f08093e0/attachment.html>