<div dir="ltr"><span style="font-size:small;text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">On Fri, Jun 8, 2018 at 2:30 AM, Jeff Hammond<span> </span></span><span dir="ltr" style="font-size:small;text-decoration-style:initial;text-decoration-color:initial"><<a href="mailto:jeff.science@gmail.com" target="_blank" style="color:rgb(17,85,204)">jeff.science@gmail.com</a>></span><span style="font-size:small;text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>wrote:</span><br style="font-size:small;text-decoration-style:initial;text-decoration-color:initial"><blockquote class="gmail_quote" style="font-size:small;text-decoration-style:initial;text-decoration-color:initial;margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr">It spins because that is optimal for latency and how the shared-memory protocols work.</div></blockquote><div class="gmail_extra" style="font-size:small;text-decoration-style:initial;text-decoration-color:initial"><br><div class="gmail_quote">On Fri, Jun 8, 2018 at 4:22 AM, Jeff Hammond<span> </span><span dir="ltr"><<a href="mailto:jeff.science@gmail.com" target="_blank" style="color:rgb(17,85,204)">jeff.science@gmail.com</a>></span><span> </span>wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_extra"><div class="gmail_quote"><span class="gmail-"><div>As blocking poll is largely incompatible with low-latency and shared-memory protocols, I don't think there is any implementation that is going to do a good job at this, since it would not be very appealing to the majority of MPI users.</div></span></div></div></div></blockquote></div></div><br><div class="gmail_extra"><br><div class="gmail_quote">On Fri, Jun 8, 2018 at 4:48 AM, Jed Brown <span dir="ltr"><<a href="mailto:jed@jedbrown.org" target="_blank">jed@jedbrown.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On the flipside, MPI would be more attractive to some<br>
(latency-insensitive) users if the implementations were not so committed<br>
to spinning.<br></blockquote><div><br></div><div>I'm trying to have a bit contribution on preparing MPI for next generation of computing systems (Exascale). It seems current MPI implementations cares about latency and performance more than power. I believe future MPI implementations must consider power as a high priority goal so i'm doing something much similar to what @Jeff linked it as my MS.c thesis.</div><div><span style="font-size:small;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">Thanks all,</span><br></div><div><span style="font-size:small;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><br></span></div><div><span style="font-size:small;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">______________________</span></div><div><span style="font-size:small;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">Researcher</span></div><div><span style="background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><div>School of Computer Engineering</div><div>Iran University of Science and Technology</div><div><a href="http://webpages.iust.ac.ir/msharifi/dslab.html">http://webpages.iust.ac.ir/msharifi/dslab.html</a><br></div></span></div></div></div></div>