PETSc examples: assemble matrix analytically instead of coloring

[jedbrown]

This currently requires development PETSc to correctly handle negative indices through the LocalToGlobalMapping.

[jeremylt]

I think my change to the solids example should be ok, but there is something strange in your PETSc branch:

# +[0]PETSC ERROR: --------------------- Error Message --------------------------------------------------------------
# +[0]PETSC ERROR:
# +[0]PETSC ERROR: Invalid row index -6! Must be in [0,24)
# +[0]PETSC ERROR: See https://petsc.org/release/faq/ for trouble shooting.
# +[0]PETSC ERROR: Petsc Development GIT revision: v3.15.4-739-g1701aa0a1f  GIT Date: 2021-09-09 14:14:31 -0600
# +[0]PETSC ERROR: build/solids-elasticity on a arch-linux-c-debug named sinensis by jeremy Tue Sep 14 14:28:25 2021
# +[0]PETSC ERROR: Configure options --download-exodusii --download-hdf5 --download-hypre --download-metis --download-ml --download-netcdf --download-parmetis --download-pnetcdf --download-zlib --download-openmpi --download-fblaslapack
# +[0]PETSC ERROR: #1 MatSetPreallocationCOO() at /home/jeremy/Dev/petsc_dev/src/mat/utils/gcreate.c:562
# +[0]PETSC ERROR: #2 main() at /home/jeremy/Dev/libCEED/examples/solids/elasticity.c:533

I think it should be a quick fix though

[jedbrown]

Hmm, I fixed your uninitialized variable (pushed) and it's running:

 build/solids-elasticity -E 10 -nu .3 -degree 3 -mesh ~/meshes/holes.msh -problem FSCurrent-NH2 -num_steps 3 -bc_clamp 1 -bc_traction 2 -bc_traction_2 0,1,0 -snes_monitor

-- Elasticity Example - libCEED + PETSc --
  MPI:
    Hostname                           : kichatna
    Total ranks                        : 1
  libCEED:
    libCEED Backend                    : /cpu/self/xsmm/blocked
    libCEED Backend MemType            : host
  PETSc:
    PETSc Vec Type                     : seq
  Problem:
    Problem Name                       : Hyperelasticity finite strain Current configuration Neo-Hookean w/ dXref_dxcurr, tau, constant storage
    Forcing Function                   : None
  Mesh:
    File                               : /home/jed/meshes/holes.msh
    Number of 1D Basis Nodes (p)       : 4
    Number of 1D Quadrature Points (q) : 4
    Global nodes                       : 85530
    Owned nodes                        : 85530
    DoF per node                       : 3
  Multigrid:
    Type                               : P-multigrid, logarithmic coarsening
    Number of Levels                   : 3
    Level 0 (coarse):
      Number of 1D Basis Nodes (p)     : 2
      Global Nodes                     : 3710
      Owned Nodes                      : 3710
    Level 2 (fine):
      Number of 1D Basis Nodes (p)     : 4
      Global Nodes                     : 85530
      Owned Nodes                      : 85530
0 Load Increment
  0 SNES Function norm 2.999573779527e-02
  1 SNES Function norm 3.001944084220e-02
  2 SNES Function norm 1.149946571373e-03
  3 SNES Function norm 4.848972179332e-05
[...]

[jeremylt]

Still not working on my local machine. What do you get with make prove search=sol -j? I don't have the mesh you are using on my current machine.

[jedbrown]

$ make prove search=sol
make: 'lib' with optional backends: /cpu/self/memcheck/serial /cpu/self/memcheck/blocked /cpu/self/avx/serial /cpu/self/avx/blocked /cpu/self/xsmm/serial /cpu/self/xsmm/blocked /gpu/cuda/ref /gpu/cuda/shared /gpu/cuda/gen
Testing backends: /cpu/self/ref/serial /cpu/self/ref/blocked /cpu/self/opt/serial /cpu/self/opt/blocked /cpu/self/memcheck/serial /cpu/self/memcheck/blocked /cpu/self/avx/serial /cpu/self/avx/blocked /cpu/self/xsmm/serial /cpu/self/xsmm/blocked /gpu/cuda/ref /gpu/cuda/shared /gpu/cuda/gen
prove -j 16 --exec 'tests/tap.sh' solids-elasticity
solids-elasticity .. ok
All tests successful.
Files=1, Tests=117, 118 wallclock secs ( 0.07 usr  0.00 sys + 106.64 cusr  8.75 csys = 115.46 CPU)
Result: PASS

[jeremylt]

Strange. I'm on jed/plex-mat-coo for PETSc.

[jedbrown]

Very nice on the CPU profile now, thanks.

$ mpiexec -n 32 build/solids-elasticity -E 10 -nu .3 -degree 3 -mesh ~/meshes/holes.msh -problem FSCurrent-NH2 -num_steps 3 -bc_clamp 1 -bc_traction 2 -bc_traction_2 0,1,0 -snes_monitor -log_view :petsc.flame:ascii_flamegraph

[jeremylt]

Passes locally when rebased with main.