<div class="gmail_quote">On Tue, Dec 13, 2011 at 13:16, Bogdan Dita <span dir="ltr"><<a href="mailto:bogdan@lmn.pub.ro">bogdan@lmn.pub.ro</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div id=":x9">So I reconfigured PETSc as advised, the result was that the time obtained<br>
from the new version of PETSc Optimised version(with COPTFLAGS) was about<br>
the same as the time from the Debug version(+/- 2 sec).</div></blockquote><div><br></div><div>You have to send output from -log_summary in both cases if you want us to help explain observed performance.</div><div> </div>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div id=":x9"> Although is not<br>
what I expected this is still an intriguing result for me: on one hand<br>
there was definitely something about the flags that the optimised PETSc<br>
didn't like, exactly how Mr. J. Brown predicted, but on the other hand<br>
this raises new questions.<br>
<br>
This is the output from 'mpicc -show'.<br>
for the cluster:<br>
gcc -m64 -O2 -fPIC -Wl,-z,noexecstack -I/usr/include/mpich2-x86_64<br>
-L/usr/lib64/mpich2/lib -lmpich<br></div></blockquote><div><br></div><div>This one is optimized by default.</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div id=":x9">
<br>
and from my laptop:<br>
gcc -I/usr/lib64/mpich2/include -L/usr/lib64/mpich2/lib<br>
-Wl,-rpath,/usr/lib64/mpich2/lib -lmpich -lopa -lmpl -lrt -lpthread</div></blockquote></div><br><div>This one is not.</div><div><br></div><div><br></div><div>PETSc folks, should we fight MPICH2 by explicitly adding -O0 for debug builds in which CFLAGS are not passed by the user?</div>