I've rewriten the code to use process dictionary, and noticed an interesting thing.<br>Single core cpu results: <br>ets based net takes 300us with smp disabled, and 400us with smp enabled, but still only 1 core in both cases, and 1 scheduler.<br>
process dict based: 200us with smp disabled, and 300us with smp enabled, same as above only 1 core.<br><br>quad core cpu results:<br>ets based net takes 200us with smp disabled, and 350us with smp enabled, 4 cores 4 schedulers.<br>
process dictionary based:150us with smp disabled, 250us with smp enabled, 4 cores 4 schedulers,<br>and 6424us with 4 cores 8 schedulers smp enabled.<br><br>
The only thing that seems to speed up as you add cores is the dict solution, I still don't see why.<br><br>Regards,<br>-Gene<br><br><div class="gmail_quote">On 27 Mar 2008 08:50:03 +0100, Bjorn Gustavsson <<a href="mailto:bjorn@erix.ericsson.se" target="_blank">bjorn@erix.ericsson.se</a>> wrote:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div>"Gene Sher" <<a href="mailto:corticalcomputer@gmail.com" target="_blank">corticalcomputer@gmail.com</a>> writes:<br>
<br>
</div><div>> Bjorn,<br>
><br>
> I see, so then no matter what, the more cpus, the Slower it will be for me?<br>
> Do you suggest that I use the Process Dictionary or Dict (which is about 3<br>
> times slower in my case than ETS, but it does go faster and faster as more<br>
> cpus are added, and this will eventually run on a server/workstation with at<br>
> least 8cores, and most likely 16+).<br>
<br>
</div>For the moment, yes, the process dictionary or dict could be a better choice.<br>
<div><div></div><div><br>
/Bjorn<br>
--<br>
Björn Gustavsson, Erlang/OTP, Ericsson AB<br>
</div></div></blockquote></div><br>