[ET Trac] #1805: triggers statement does not include reductions output

Roland Haas trac-noreply at einsteintoolkit.org
Mon Mar 20 14:35:38 CDT 2023


#1805: triggers statement does not include reductions output

 Reporter: Wolfgang Kastaun
   Status: new
Milestone: 
  Version: ET_2014_05
     Type: bug
 Priority: minor
Component: Other

Comment (by Roland Haas):

I note that computing variables whose reductions are being computed \(for output or otherwise\) only every once in a while and not at every timestep is very firmly in the “[give you rope to hang yourself with, and then some more](https://motd.ambians.com/quotes.php/name/linux_computers/toc_id/1-1-4/s/881#882)” territory. Technically should be supported though \(and I thought there are the required `CCTJ_TriggerFoo` calls in `CarpetIOScalar`\).

The only way to get away with it somewhat safely is to compute \(and output\) only every coarse step \(if not computing every single step of course\). Everything else stands a danger of using stale / invalid data on past timelevels if and when interpolation in time happens. This also implies \(well mostly\) that the computed quantity must be computable in a pointwise manner, not involving any stencil operations \(to avoid having to restrict and prolongate\) to avoid different results obtained when computing every step and output every coarse or computing and outputting every coarse.

--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/1805/triggers-statement-does-not-include
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.einsteintoolkit.org/pipermail/trac/attachments/20230320/166ea861/attachment.html 


More information about the Trac mailing list