[petsc-dev] TS not counting linear iterations properly?

[Barry Smith]

  Beats the heck out of me, too hard to run myself. Requires all kinds of data files and other crapola I don't have.



> On Dec 3, 2014, at 8:11 PM, Emil Constantinescu <emconsta at mcs.anl.gov> wrote:
> 
> On 12/3/14 11:21 AM, Barry Smith wrote:
>> 
>>   Jed and Emil,
>> 
>>   type: arkimex
>>   total number of nonlinear solver iterations=1522
>>   total number of nonlinear solve failures=158
>>   total number of linear solver iterations=4871
>>   total number of rejected steps=158
>> 
>> KSPGMRESOrthog   1601671 1.0 9.8384e+03 1.1 6.96e+12 1.1 0.0e+00 0.0e+00
>> 1.6e+06  8  9  0  0 14   8  9  0  0 14  2777
>> KSPSetUp            6721 1.0 2.0089e-02 1.6 0.00e+00 0.0 0.0e+00 0.0e+00
>> 2.0e+01  0  0  0  0  0   0  0  0  0  0     0
>> KSPSolve            1680 1.0 1.1547e+05 1.0 7.63e+13 1.1 8.6e+07 3.5e+02
>> 1.1e+07 99100100 96100  99100100 96100  2593
>> PCSetUp             5040 1.0 5.0914e+02 1.0 1.01e+09 1.0 8.0e+01 2.8e+04
>> 1.4e+04  0  0  0  0  0   0  0  0  0  0     8
>> PCApply          1584871 1.0 9.0297e+04 1.0 5.83e+13 1.1 7.6e+07 3.8e+02
>> 8.0e+06 77 76 89 93 71  77 76 89 93 71  2535
>> 
>>    Seems to be missing counting some of the linear solves.
>> 
>>    What's up with this?
> 
> What are the counts if you run with -ts_dt 1e-12 -ts_max_snes_failures 0 -ts_max_steps 1 ?
> 
> Emil