Rob, <br><br>Setting the environment variable works, <br><br>Thanks,<br>Jim<br><br><div><span class="gmail_quote">On 7/11/06, <b class="gmail_sendername">Robert Latham</b> <<a href="mailto:robl@mcs.anl.gov">robl@mcs.anl.gov
</a>> wrote:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">On Tue, Jul 11, 2006 at 12:02:02PM -0600, Jim Edwards wrote:<br>> Hi Rob,
<br>><br>> In order to avoid using the independent mode I am trying to call<br>> nfmpi_put_vara_all with count=0 on some processors. This appears to<br>> work fine under the AIX implementation but causes a malloc error in
<br>> romio on BG/L. Here is a core dump referring to source code in<br>> romio, I don't have that source code but I have some hope that this<br>> will be enough info for you to pinpoint the problem...<br><br>Hi Jim
<br>We talk about this a little bit in our README.bgl. If you set the<br>BGLMPIO_TUNEBLOCKING environment variable to '0', you should no longer<br>get a seg fault<br><br>They just rolled out a new version of IBM's MPI-IO for BGL on our
<br>machine that fixes this problem: with driver version V1R3M0 I no<br>longer see this segfault with our test cases.<br><br>Let us know how those workarounds perform for you.<br><br>==rob<br><br>--<br>Rob Latham<br>Mathematics and Computer Science Division A215 0178 EA2D B059 8CDF
<br>Argonne National Labs, IL USA B29D F333 664A 4280 315B<br></blockquote></div><br>