<br><tt><font size=2>robl@mcs.anl.gov (Robert Latham) wrote on 02/06/2008
03:05:55 PM:<br>
<br>
> Mostly for my own notes: For "upstream" romio, we've
found an<br>
> astonishing number of stat variants. ROMIO needs (but does not
have)<br>
> a generalized stat routine that can be called from different places
in<br>
> the code. Also, this blocksize should be tunable with an MPI-IO
hint. <br>
> <br>
</font></tt>
<br><tt><font size=2>Yes, understand the stat issues. Since this
was BGL specific, we just </font></tt>
<br><tt><font size=2>went with stat64() which we knew would work.</font></tt>
<br>
<br><tt><font size=2>We actually discussed a hint and decided not to do
it. We allow the user</font></tt>
<br><tt><font size=2>to enable/disable this tuning/optimizing code. But
if it is enabled, we</font></tt>
<br><tt><font size=2>decided that it was important to match the underlying
file system blocksize. </font></tt>
<br>
<br><tt><font size=2>Bob Cernohous: (T/L 553) 253-6093</font></tt>
<br>
<br><tt><font size=2>BobC@us.ibm.com</font></tt>
<br><tt><font size=2>IBM Rochester, Building 030-2(C335), Department 61L</font></tt>
<br><tt><font size=2>3605 Hwy 52 North, Rochester, MN 55901-7829</font></tt>
<br>
<br><tt><font size=2>> Chaos reigns within.</font></tt>
<br><tt><font size=2>> Reflect, repent, and reboot.</font></tt>
<br><tt><font size=2>> Order shall return.</font></tt>