Refactor SolKz benchmark, add compositional field benchmark, and add active particle benchmark #1539

hlokavarapu · 2017-05-09T19:11:17Z

The log output of SolKz benchmark at a global refinement level of 5:

harsha_lv@godavari:~/ClionProjects/aspect/benchmarks/solkz$ mpirun -np 1 aspect solkz.prm 
-----------------------------------------------------------------------------
-- This is ASPECT, the Advanced Solver for Problems in Earth's ConvecTion.
--     . version 2.0.0-pre
--     . running in DEBUG mode
--     . running with 1 MPI process
--     . using Trilinos
-----------------------------------------------------------------------------

Loading shared library <./libsolkz.so>

Number of active cells: 1,024 (on 6 levels)
Number of degrees of freedom: 13,764 (8,450+1,089+4,225)

*** Timestep 0:  t=0 years
   Rebuilding Stokes preconditioner...
   Solving Stokes system... 47+0 iterations.
      Relative Stokes residual after nonlinear iteration 1: 1
   Solving Stokes system... 0+0 iterations.
      Relative Stokes residual after nonlinear iteration 2: 9.61484e-08

   Postprocessing:
     Errors u_L1, p_L1, u_L2, p_L2: 1.255365e-08, 1.599227e-04, 3.451232e-08, 2.647014e-04
     Writing graphical output:      output/solution/solution-00000

Termination requested by criterion: end time

The log output of SolKz benchmark using compositional fields at a global refinement level of 5:

harsha_lv@godavari:~/ClionProjects/aspect/benchmarks/solkz/compositional_fields$ mpirun -np 1 aspect solkz_compositional_fields.prm 
-----------------------------------------------------------------------------
-- This is ASPECT, the Advanced Solver for Problems in Earth's ConvecTion.
--     . version 2.0.0-pre
--     . running in DEBUG mode
--     . running with 1 MPI process
--     . using Trilinos
-----------------------------------------------------------------------------

Loading shared library <./libsolkz_compositional_fields.so>

Number of active cells: 1,024 (on 6 levels)
Number of degrees of freedom: 34,179 (8,450+3,072+4,225+9,216+9,216)

*** Timestep 0:  t=0 years
   Rebuilding Stokes preconditioner...
   Solving Stokes system... 48+0 iterations.
      Relative Stokes residual after nonlinear iteration 1: 1
   Solving Stokes system... 0+0 iterations.
      Relative Stokes residual after nonlinear iteration 2: 7.95719e-08

   Postprocessing:
     Writing graphical output:      output/solution/solution-00000
     Errors u_L1, p_L1, u_L2, p_L2: 1.263331e-08, 1.961224e-04, 3.505067e-08, 3.422689e-04

Termination requested by criterion: end time

The log output of SolKz benchmark using active particles at a global refinement level of 5:

harsha_lv@godavari:~/ClionProjects/aspect/benchmarks/solkz/compositional_fields$ mpirun -np 1 aspect solkz_particles.prm 
-----------------------------------------------------------------------------
-- This is ASPECT, the Advanced Solver for Problems in Earth's ConvecTion.
--     . version 2.0.0-pre
--     . running in DEBUG mode
--     . running with 1 MPI process
--     . using Trilinos
-----------------------------------------------------------------------------

Loading shared library <./libsolkz_compositional_fields.so>

Number of active cells: 1,024 (on 6 levels)
Number of degrees of freedom: 34,179 (8,450+3,072+4,225+9,216+9,216)

*** Timestep 0:  t=0 years
   Skipping temperature solve because RHS is zero.
   Rebuilding Stokes preconditioner...
   Solving Stokes system... 47+0 iterations.

   Postprocessing:
     Writing particle output:       output/particles/particles-00000
     Writing graphical output:      output/solution/solution-00000
     Errors u_L1, p_L1, u_L2, p_L2: 3.056647e-07, 4.168362e-03, 4.669234e-07, 6.572723e-03

Termination requested by criterion: end time

gassmoeller · 2017-05-10T01:21:46Z

/run-tests

…tive particles

gassmoeller

Looks quite good. The one thing I would like to discuss is the discontinuous vs continuous pressures. Is it important for you to use the discontinuous pressure for this benchmark? I.e. is it important enough to confuse users that the temperature benchmark is set up differently than the composition benchmark?

gassmoeller · 2017-05-10T20:28:12Z

benchmarks/solkz/compositional_fields/solkz_compositional_fields.h

@@ -0,0 +1,38 @@
+#ifndef SOLKZ_SOLKZ_COMPOSITIONAL_FIELDS_H
+#define SOLKZ_SOLKZ_COMPOSITIONAL_FIELDS_H


make this ASPECT_SOLKZ_...

gassmoeller · 2017-05-10T20:31:58Z

benchmarks/solkz/compositional_fields/solkz_compositional_fields.prm

+
+subsection Discretization
+  set Stokes velocity polynomial degree       = 2
+  set Use locally conservative discretization = true


Is there a reason you use discontinous pressures for the compositional benchmark but continuous pressures for the existing solkz benchmark? This will confuse people trying to run this benchmark. It also confused me to see different numbers of degrees of freedom in the results you posted in the pull request. It also makes me wonder if the difference in accuracy is due to this difference in the input file. Can you change this to false and still get similar results? Then please do this. Lets discuss what to do, if that is not the case.

gassmoeller · 2017-05-10T20:32:43Z

benchmarks/solkz/compositional_fields/solkz_particles.prm

+
+subsection Discretization
+  set Stokes velocity polynomial degree       = 2
+  set Use locally conservative discretization = true


Same comment as above

SolKz (base case):

Q2_Q1 h 3 Errors u_L1, p_L1, u_L2, p_L2: 5.710496e-07, 2.769692e-03, 1.071151e-06, 4.210709e-03 4 Errors u_L1, p_L1, u_L2, p_L2: 9.024102e-08, 6.588213e-04, 2.330051e-07, 1.068890e-03 5 Errors u_L1, p_L1, u_L2, p_L2: 1.255368e-08, 1.599227e-04, 3.451230e-08, 2.647015e-04 6 Errors u_L1, p_L1, u_L2, p_L2: 1.603879e-09, 3.975991e-05, 4.499684e-09, 6.568000e-05 7 Errors u_L1, p_L1, u_L2, p_L2: 2.016031e-10, 9.926911e-06, 5.685278e-10, 1.637434e-05

SolKz (refactored):

Q2_Q1 h 3 Errors u_L1, p_L1, u_L2, p_L2: 5.710496e-07, 2.769692e-03, 1.071151e-06, 4.210709e-03 4 Errors u_L1, p_L1, u_L2, p_L2: 9.024102e-08, 6.588213e-04, 2.330051e-07, 1.068890e-03 5 Errors u_L1, p_L1, u_L2, p_L2: 1.255368e-08, 1.599227e-04, 3.451230e-08, 2.647015e-04 6 Errors u_L1, p_L1, u_L2, p_L2: 1.603879e-09, 3.975991e-05, 4.499684e-09, 6.568000e-05 7 Errors u_L1, p_L1, u_L2, p_L2: 2.016031e-10, 9.926911e-06, 5.685278e-10, 1.637434e-05

SolKz (Compositional fields)

Q2_Q1 h 3 Errors u_L1, p_L1, u_L2, p_L2: 5.656321e-07, 2.770715e-03, 1.040702e-06, 4.210785e-03 4 Errors u_L1, p_L1, u_L2, p_L2: 8.922945e-08, 6.610882e-04, 2.291564e-07, 1.066119e-03 5 Errors u_L1, p_L1, u_L2, p_L2: 1.251153e-08, 1.600241e-04, 3.441676e-08, 2.645674e-04 6 Errors u_L1, p_L1, u_L2, p_L2: 1.602676e-09, 3.976684e-05, 4.497021e-09, 6.567566e-05 7 Errors u_L1, p_L1, u_L2, p_L2: 2.015608e-10, 9.927347e-06, 5.684425e-10, 1.637421e-05

SolKz (Active Particles using cell average interpolation)

Q2_Q1_particles h 3 Errors u_L1, p_L1, u_L2, p_L2: 3.924171e-06, 3.080755e-03, 6.435722e-06, 5.349705e-03 4 Errors u_L1, p_L1, u_L2, p_L2: 9.465434e-07, 1.042225e-03, 1.527446e-06, 2.848889e-03 5 Errors u_L1, p_L1, u_L2, p_L2: 2.396696e-07, 3.813762e-04, 3.798516e-07, 1.430330e-03 6 Errors u_L1, p_L1, u_L2, p_L2: 5.976014e-08, 1.148414e-04, 9.487535e-08, 6.125869e-04 7 Errors u_L1, p_L1, u_L2, p_L2: 1.494555e-08, 3.211711e-05, 2.367541e-08, 2.381214e-04

gassmoeller · 2017-05-10T20:34:24Z

benchmarks/solkz/solkz.cc

@@ -1,856 +1,4 @@
-#include <aspect/material_model/simple.h>


Did you just move all of this into the header file? That is ok.

Yeah, pretty much the refactoring was equivalent to the command $ mv solkz.cc solkz.h 😄

…ure element for solkz_particles and solkz_compositional_field benchmark"

egpuckett · 2017-05-10T22:14:50Z

I think this looks good. Is there a manual entry required? If not, how are users going to understand what we have done with these benchmarks. I think we should at least have an entry in the manual that points to the paper, that will presumably be written some day soon.

tjhei · 2017-05-10T23:10:19Z

benchmarks/check.sh

@@ -86,6 +86,8 @@ echo "Please be patient..."

 ( (cd solkz; make_lib && run_all_prms ) || { echo "FAILED"; exit 1; } ) &

+( (cd solkz/compositional_fields; make_lib && run_all_prms ) || { echo "FAILED"; exit 1; } ) &


tjhei · 2017-05-10T23:10:26Z

can you add a changelog entry?

This was referenced May 9, 2017

Compositional field benchmarks #1345

Closed

Active particle benchmarks #1344

Closed

Refactor SolKz, add SolKz benchmark using compositional fields and ac…

d7978ec

…tive particles

gassmoeller reviewed May 10, 2017

View reviewed changes

Rename header guard solkz_compositional_field.h; Use continuous press…

1d3d81b

…ure element for solkz_particles and solkz_compositional_field benchmark"

hlokavarapu force-pushed the refactor_solkz branch from 9a1578f to 1d3d81b Compare May 10, 2017 22:13

tjhei reviewed May 10, 2017

View reviewed changes

Add change log entry to 20170510_egp-cig-reu

286895a

gassmoeller merged commit 220f991 into geodynamics:master May 11, 2017

This was referenced May 12, 2017

Update SolCx and SolKz benchmarks in tests folder #1583

Closed

Fix bug for Inclusion, SolCx, and SolKz benchmarks in check.sh #1664

Merged

gassmoeller mentioned this pull request Aug 12, 2017

Fixed bug preventing particles from working in Solcx benchmark #1884

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Refactor SolKz benchmark, add compositional field benchmark, and add active particle benchmark #1539

Refactor SolKz benchmark, add compositional field benchmark, and add active particle benchmark #1539

hlokavarapu commented May 9, 2017

gassmoeller commented May 10, 2017

gassmoeller left a comment

gassmoeller May 10, 2017

gassmoeller May 10, 2017

gassmoeller May 10, 2017

hlokavarapu May 10, 2017

gassmoeller May 10, 2017

hlokavarapu May 10, 2017

egpuckett commented May 10, 2017

tjhei May 10, 2017

tjhei commented May 10, 2017

		@@ -0,0 +1,38 @@
		#ifndef SOLKZ_SOLKZ_COMPOSITIONAL_FIELDS_H
		#define SOLKZ_SOLKZ_COMPOSITIONAL_FIELDS_H

		@@ -86,6 +86,8 @@ echo "Please be patient..."

		( (cd solkz; make_lib && run_all_prms ) \|\| { echo "FAILED"; exit 1; } ) &

		( (cd solkz/compositional_fields; make_lib && run_all_prms ) \|\| { echo "FAILED"; exit 1; } ) &

Refactor SolKz benchmark, add compositional field benchmark, and add active particle benchmark #1539

Refactor SolKz benchmark, add compositional field benchmark, and add active particle benchmark #1539

Conversation

hlokavarapu commented May 9, 2017

gassmoeller commented May 10, 2017

gassmoeller left a comment

Choose a reason for hiding this comment

gassmoeller May 10, 2017

Choose a reason for hiding this comment

gassmoeller May 10, 2017

Choose a reason for hiding this comment

gassmoeller May 10, 2017

Choose a reason for hiding this comment

hlokavarapu May 10, 2017

Choose a reason for hiding this comment

gassmoeller May 10, 2017

Choose a reason for hiding this comment

hlokavarapu May 10, 2017

Choose a reason for hiding this comment

egpuckett commented May 10, 2017

tjhei May 10, 2017

Choose a reason for hiding this comment

tjhei commented May 10, 2017