<div class="gmail_quote"><div class="gmail_quote"><div>On Mon, Dec 19, 2011 at 10:08 PM, Erik Schnetter <span dir="ltr"><<a href="mailto:schnetter@cct.lsu.edu" target="_blank">schnetter@cct.lsu.edu</a>></span> wrote:<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div>There was a discussion regarding where to place checkpoint files. I originally intended them to reside within the corresponding output-NNNN directories, and this required setting the checkpoint directory to $parfile, same as the output directory. However, since then others made the point that they rather wanted all checkpoint files of a simulation in the same directory, because this enables Cactus to remove the checkpoint files of previous restarts -- this makes restarts less independent, but can save disk space. This requires setting the checkpoint directory to "../checkpoints".</div>
<div><br></div><div>I don't recall what changes were necessary in Simfactory to support this, but there may have been some.</div></blockquote><div><br></div></div><div>In the end, I don't think any changes were made in order to support this. The only changes were in the parfiles.</div>
<div>
<div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div>However, Simfactory always expected all output directories to be set to $parfile, or to subdirectories of $parfile. It may be that things worked without doing so, but I never used or tested this myself.</div>
</blockquote><div><br></div></div><div>I guess it must have always been the case that this didn't work. It does seem somewhat counter-intuitive that by default Cactus outputs checpoints to a different location than where it looks for them.</div>
<span><font color="#888888">
<div><br></div><div>Barry</div></font></span></div>
</div><br>