Measuring FEP time

[jinca]

Hello dear MOOSE group,

I found documentation of PETSc with regards to the performance in the output file.
See here.

However, it seems quite outdated to me. Could you please enlighten me on this?
Having the following Performance Graph:

---------------------------------------------------------------------------------------------------------------------------------------------------------------
|                  Section                  | Calls |   Self(s)  |   Avg(s)   |    %   | Children(s) |   Avg(s)   |    %   |  Total(s)  |   Avg(s)   |    %   |
---------------------------------------------------------------------------------------------------------------------------------------------------------------
| pookaTestApp (main)                       |     1 |      0.015 |      0.015 |   0.00 |    1145.325 |   1145.325 | 100.00 |   1145.340 |   1145.340 | 100.00 |
|   FEProblem::outputStep                   |     2 |      0.448 |      0.224 |   0.04 |       0.505 |      0.252 |   0.04 |      0.953 |      0.476 |   0.08 |
|   Steady::PicardSolve                     |     1 |      0.000 |      0.000 |   0.00 |    1061.416 |   1061.416 |  92.67 |   1061.416 |   1061.416 |  92.67 |
|     FEProblem::solve                      |     1 |   1000.687 |   1000.687 |  87.37 |      59.227 |     59.227 |   5.17 |   1059.914 |   1059.914 |  92.54 |
|       FEProblem::computeResidualInternal  |     4 |      0.000 |      0.000 |   0.00 |       9.506 |      2.376 |   0.83 |      9.506 |      2.376 |   0.83 |
|       FEProblem::computeJacobianInternal  |     2 |      0.000 |      0.000 |   0.00 |      49.711 |     24.856 |   4.34 |     49.711 |     24.856 |   4.34 |
|     FEProblem::outputStep                 |     1 |      0.205 |      0.205 |   0.02 |       0.004 |      0.004 |   0.00 |      0.209 |      0.209 |   0.02 |
|     AuxiliarySystem::computeElementalVars |     1 |      1.002 |      1.002 |   0.09 |       0.291 |      0.291 |   0.03 |      1.293 |      1.293 |   0.11 |
|   Steady::final                           |     1 |      0.001 |      0.001 |   0.00 |       0.049 |      0.049 |   0.00 |      0.051 |      0.051 |   0.00 |
|     FEProblem::outputStep                 |     1 |      0.043 |      0.043 |   0.00 |       0.006 |      0.006 |   0.00 |      0.049 |      0.049 |   0.00 |
---------------------------------------------------------------------------------------------------------------------------------------------------------------

1- Is it OK to consider the 'Avg(s)' column (penultimate column) of the raw 'FEProblem::solve' to plot my benchmarking?
2- There is another block that determine performance and timing:

Summary of Stages:   ----- Time ------  ----- Flop ------  --- Messages ---  -- Message Lengths --  -- Reductions --
                        Avg     %Total     Avg     %Total    Count   %Total     Avg         %Total    Count   %Total
 0:      Main Stage: 1.1462e+03 100.0%  2.8356e+12 100.0%  5.721e+05 100.0%  9.959e+04      100.0%  8.080e+02  99.1%

------------------------------------------------------------------------------------------------------------------------

Is the column after 'Main Stage:' the total time of my simulation?

Thanks in advance for the help!
Julita

[GiudGiud]

Hello

This is the MOOSE perf graph, not the PETSc one. Documentation is here:
https://mooseframework.inl.gov/source/outputs/PerfGraphOutput.html

1- yes, since it's only called once for your case. Otherwise Total (s) is usually the better one.

2- this does like a different perf graph. Is this the PETSc one? I've never seen that one.
I reckon since you want to look at moose performance, you should stick to the moose one.

Guillaume