[petsc-dev] PETSc release soon, request for input on needed fixes or enhancements

[Barry Smith]

  Please send me a command line option and example (one in the repository?) where this problem happens.

  Barry

> On Mar 16, 2016, at 1:38 PM, Lawrence Mitchell <lawrence.mitchell at imperial.ac.uk> wrote:
> 
> 
> 
> On 16/03/16 18:24, Barry Smith wrote:
>> 
>>> On Mar 16, 2016, at 9:37 AM, Matthew Knepley <knepley at gmail.com> wrote:
>>> 
>>> On Wed, Mar 16, 2016 at 9:34 AM, Lawrence Mitchell <lawrence.mitchell at imperial.ac.uk> wrote:
>>> Barry,
>>> 
>>> On 09/03/16 15:37, Barry Smith wrote:
>>> 
>>> ...
>>> 
>>>> There is no way we can "fix" the management of the formats at this
>>>> point. What value do we need for PETSCVIEWERFORMATPUSHESMAX to get
>>>> everything to work for you?
>>> 
>>> I ran some tests and I can't give you a sensible number here.  So
>>> please leave the current value as is.
>>> 
>>> This is breaking for me as well. I would like 640.
>> 
>>  Please send me a case where you get breakage. It seems inconceivable to me that you are really running with 640 monitors at the same time.
> 
> One need not run with that many monitors concurrently.  One just needs
> (say) KSPMonitorSetFromOptions to be called with the default stdout
> viewer more than PETSCVIEWERFORMATPUSHESMAX.  This could easily happen
> with a multigrid PC (which calls KSPSetFromOptions on each smoother)
> inside a nonlinear solve (PCSetup_MG calls KSPSetFromOptions) when I'm
> monitoring the levels.  Additionally, the push (if no viewer output
> arguments are provided) affects the global viewer obtained from
> PetscViewerASCIIGetStdout, so anything that pushes but doesn't pop
> (any TS, SNES or KSP) onto that stdout viewer will count.
> 
> Cheers,
> 
> Lawrence