<div dir="ltr">I'm just going to leave this here :-)<div><br></div><div><a href="http://funroll-loops.info/">http://funroll-loops.info/</a><br></div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On Thu, Jun 20, 2013 at 4:05 AM, Greg Folkert <span dir="ltr"><<a href="mailto:greg@gregfolkert.net" target="_blank">greg@gregfolkert.net</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="HOEnZb"><div class="h5">On Wed, 2013-06-19 at 22:52 -0400, Kyle wrote:<br>
> On 06/19/2013 08:44 PM, Adam Tauno Williams wrote:<br>
> > On Wed, 2013-06-19 at 18:11 -0500, Philip Robar wrote:<br>
> >> I’m hoping that someone, like Mike Mol - our resident Gentoo advocate,<br>
> >> might know the answer to this (and save me hours of experimenting):<br>
> >> Will compiling Gentoo specifically for Intel P4/Netburst with GCC,<br>
> >> Intel or LLVM make a noticeable difference on a general purpose<br>
> >> desktop as opposed to using a generically compiled distribution?<br>
> > I know next to nothing about Gentoo... but I'd be willing to wager lunch<br>
> > that the answer is "No way, no difference". I've seen more bugs<br>
> > introduced by eager-to-optimize setups than I have performance<br>
> > improvements. Compiler optimization performance improvements are pretty<br>
> > much limited to code written specifically to demonstrate they make<br>
> > things go faster.<br>
> ><br>
> > Note that you said ****"difference on a general purpose desktop"***.<br>
> > Special cases, especially those involved math / real-data-processing<br>
> > might be different, but a "general purpose desktop"... no way, no<br>
> > difference. You're more likely to break something.<br>
> ><br>
> > [snip]<br>
><br>
><br>
> Can't answer for Netburst, either, but I can say that I run Gentoo and<br>
> haven't seen much of a change in performance with compiling specifically<br>
> for my CPU (an i7) over a generic x86_64 build. However, Gentoo lets<br>
> you compile out all kinds of junk you don't want or use by changing your<br>
> USE flags, and that, for me anyway, makes a significant difference in<br>
> speed and power consumption vs. a generic "cram every feature in, just<br>
> in case it's ever needed" build.<br>
<br>
</div></div>Perception is 90% of the perceived benefit.<br>
<span class="HOEnZb"><font color="#888888"><br>
--<br>
<a href="mailto:greg@gregfolkert.net">greg@gregfolkert.net</a><br>
PGP key 1024D/B524687C 2003-08-05<br>
Fingerprint: E1D3 E3D7 5850 957E FED0 2B3A ED66 6971 B524 687C<br>
"A man of personality can formulate ideals, but only a man of character<br>
can achieve them."<br>
-- Herbert Read<br>
</font></span><br>_______________________________________________<br>
grlug mailing list<br>
<a href="mailto:grlug@grlug.org">grlug@grlug.org</a><br>
<a href="http://shinobu.grlug.org/cgi-bin/mailman/listinfo/grlug" target="_blank">http://shinobu.grlug.org/cgi-bin/mailman/listinfo/grlug</a><br></blockquote></div><br></div>