[ET Trac] [Einstein Toolkit] #1102: when using the file reader, mark variables not requested fro reading as fully read

Einstein Toolkit trac-noreply at einsteintoolkit.org
Sat Oct 27 19:48:13 CDT 2012


#1102: when using the file reader, mark variables not requested fro reading as
fully read
---------------------+------------------------------------------------------
  Reporter:  rhaas   |       Owner:  eschnett
      Type:  defect  |      Status:  review  
  Priority:  minor   |   Milestone:          
 Component:  Carpet  |     Version:          
Resolution:          |    Keywords:          
---------------------+------------------------------------------------------

Comment (by rhaas):

 Apparently I never actually posted my explanation to comment:2. Oh well.
 This is what I wanted to say:

 I'll gladly remove the continue check since it just clutters the code.

 As for the last hunk. This is an odd piece of code. My understanding was
 that the idea is to synchronize all variables that were read. When
 recovering from a checkpoint, then all variables are attempted to read so
 all might need synchronization (this is the != FILEREADER_DATA case). If
 on the other hand the file reader was used, then a variable is either to
 be read if either all variables are to be read (in which case
 ioUtilGH->do_inVars is NULL) or if do_inVars[vindex] is not zero (it can
 be positive or negative but for the file-reader should really be negative
 or one). This last logic is explained in line 623 of the code. So it seems
 as if the piece of code in line 802 is required to avoid sync'ing grid
 functions that were excluded from being read (and whose read_completely
 entry was set to true at the beginning of the read process). A more
 readable test might be:

 {{{#!C
 if (not (called_from == FILEREADER_DATA and
          (ioUtilGH->do_inVars and not ioUtilGH->do_inVars[vindex])) )
 {
   if (read_completely(...
 }}}

-- 
Ticket URL: <https://trac.einsteintoolkit.org/ticket/1102#comment:3>
Einstein Toolkit <http://einsteintoolkit.org>
The Einstein Toolkit


More information about the Trac mailing list