[Users] Warnings when recovering

Erik Schnetter schnetter at cct.lsu.edu
Mon Jan 16 08:13:00 CST 2012


On Mon, Jan 16, 2012 at 7:49 AM, Ian Hinder <ian.hinder at aei.mpg.de> wrote:
> Hi,
>
> There has been a discussion in the comments of https://trac.einsteintoolkit.org/ticket/550 concerning changing the behaviour of the warning messages that Carpet gives when the process decomposition changes on recovery.  These messages look like this:
>
>> WARNING level 1 in thorn CarpetIOHDF5 processor 21 host c312-313.ls4.tacc.utexas.edu
>>
>> (line 640 of /work/00920/tg459479/Cactus/arrangements/Carpet/CarpetIOHDF5/src/Input.cc):
>> -> Variable AHFINDERDIRECT::ahmask on rl 0 and tl 0 not read completely. Will have to look for it in other files
>>
>
> The process decomposition might be different because of using a different number of processes, using a different version of Carpet with a different algorithm, or because of differences in the way Carpet does processor decomposition during regridding and recovery.
>
> The problem with the current warnings is that they are very verbose, reporting every component of every variable on every refinement level.  This can lead to huge standard output and standard error files.  The current proposal, by Frank, is to replace these warnings with a single message which indicates what is happening (debug-level 3, so ON by default with SimFactory), and suppressing the remaining messages by default (debug-level 4, so OFF by default with SimFactory).  Further, the warning message would be level 1 if the number of processes is the same on checkpoint and recovery, as this likely indicates a real problem.
>
> These messages were apparently requested by a user originally, so we don't want to change things without getting feedback from the wider community.  Does anyone object to Frank's proposal?

I believe Christian D. Ott was one of the people requesting such
warnings, and I asked Frank to contact Christian directly for feedback
to make sure this discussion doesn't pass by him.

-erik

-- 
Erik Schnetter <schnetter at cct.lsu.edu>   http://www.cct.lsu.edu/~eschnett/


More information about the Users mailing list