<div class="gmail_extra"><div class="gmail_quote">On Mon, Nov 12, 2012 at 8:56 PM, Jed Brown <span dir="ltr"><<a href="mailto:jedbrown@mcs.anl.gov" target="_blank">jedbrown@mcs.anl.gov</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div class="im"><div class="gmail_extra">On Mon, Nov 12, 2012 at 7:49 PM, John Fettig <span dir="ltr"><<a href="mailto:john.fettig@gmail.com" target="_blank">john.fettig@gmail.com</a>></span> wrote:<br><div class="gmail_quote">
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Then add on top of this the fact that you could simply recompile PETSc, run it natively on the card, and still run it on your CPU's as MPMD. </blockquote></div><br></div></div><div class="gmail_extra">This is a good way to get terrible performance.</div>
</blockquote></div><br>Why? Decompose your domain to take into account the imbalance in computational power. The link between the card and the CPU is going to be faster than going to another node.<br><br>John<br></div>