[petsc-dev] FAS indentation

[Dener, Alp]

Hi Matt,

Some notes about this issue:

  1.  Correction to my previous email: the XYZMonitor() routines were not touched/modified in this feature branch. I misspoke about that. We modified the XYZView() only, and indeed these are the ones that appear to have some different indentations in the output files.

  2.  I’m seeing incorrect tab levels in only two ex12 tests: ex12_tri_p1_adapt_0 and ex12_tri_p1_adapt_1. Turns out SNESReasonView() did not have the earlier tab level changes correctly propagated into it. I’ve attached a git format-patch that fixes this (I didn’t wanna push directly to next without approval). But the remaining tab level changes are consistent with the XYZView() changes we’ve made. So I would recommend updating the output files after applying the patch (again, I didn’t want to touch the outputs without your approval first).

  3.  Your first email on this mentioned FAS solver errors, but all the ex12_fas_% tests are passing for me locally. Can you confirm that these are working for you as well? If not, which test machine should I run this on to reproduce your errors?

  4.  I’ve got some number of the ex12 tests crashing on my machine. ex12_restart-b reports HDF5 errors, while ex12_mg_newton_coarse_0, ex12_mg_newton_coarse_1 and ex12_mg_newton_coarse_2 report “inconsistent data” with regards to dof mapping. These just might have to do with my local configuration but I still wanted to put it on your radar just in case.

Thanks,

—
Alp Dener


On April 3, 2018 at 6:24:42 PM, Matthew Knepley (knepley at gmail.com<mailto:knepley at gmail.com>) wrote:

On Tue, Apr 3, 2018 at 7:16 PM, Dener, Alp <adener at anl.gov<mailto:adener at anl.gov>> wrote:
Sure thing!

Matt, just to give you a bit of detail, the Monitor() calls to ASCII viewers were changed to respect the tab levels for the parent objects instead of manually pushing and popping tabs on the printouts. The code also tries to preserve the existing tab levels of the viewer though — that is, we first save the existing tab level, then indent monitor printouts from the root (0 tab) using the object’s tab level, and then set the viewer back to whatever tab level it was at before the monitor is called. So in theory, it shouldn’t be changing the behavior of the viewer outside of the monitor. It is entirely possible I introduced a bug somewhere and that’s what I’m going to look for right now, but in the meantime please let me know if there’s a use-case or a test I can use for debugging as well.

The easiest thing to do is run all the SNES ex12 tests in next and look at the indentation errors. I am now pushing fixes for all other errors, so
by the time you run there should be nothing else.

  Thanks,

     Matt

Thanks,
—
Alp Dener


On April 3, 2018 at 6:11:12 PM, Smith, Barry F. (bsmith at mcs.anl.gov<mailto:bsmith at mcs.anl.gov>) wrote:

Alp,

Can you please take a look at this?

Thanks

Barry


> On Apr 3, 2018, at 4:46 PM, Matthew Knepley <knepley at gmail.com<mailto:knepley at gmail.com>> wrote:
>
> The recent indentation fix (I think) has broken the FAS solver indentation.
>
> Matt
>
> --
> What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.
> -- Norbert Wiener
>
> https://www.cse.buffalo.edu/~knepley/




--
What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.
-- Norbert Wiener

https://www.cse.buffalo.edu/~knepley/<http://www.caam.rice.edu/~mk51/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20180404/d95a4a81/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-SNESReasonView-fixed-to-adopt-earlier-tab-level-chan.patch
Type: application/octet-stream
Size: 1715 bytes
Desc: 0001-SNESReasonView-fixed-to-adopt-earlier-tab-level-chan.patch
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20180404/d95a4a81/attachment.obj>