[wplug] what is optimized by -march in gcc?

[Bill Moran]

terry mcintyre wrote:
> It would be interesting to extend that test
> of optimizations. Split out the kernel and
> other sources, to determine how much each 
> matters; try newer hardware also.

There are a lot of interesting things that could be
done with those tests.  The ones you mentioned are
pretty interesting to me, but I can think of even
more.

Unfortunately, I only have so much time in a day,
and only so many available computers to test on
(and most of them are older machines, newer machines
get _real_ work to do!)

Perhaps I should take the phk approach and soliticit
donations to reasearch this:
http://ezine.daemonnews.org/200404/dadvocate.html


> --- "Vanco, Don" <don.vanco at agilysys.com> wrote:
> 
> 
>>>Probably.  See:
>>>http://www.potentialtech.com/wmoran/source.php
>>
>>	Interesting page - it might be kind of cool to take
>>a distro
>>like Gentoo and install on a "like system" (of P4
>>class) twice, once
>>optimized for i386 and once optimized for i686 and
>>compare things like
>>boot time, browser performance, OpenOffice
>>run-times, etc.  I recall
>>when Mandrake made the leap from an i386 to an i586
>>base compile and the
>>overall performance increase was notable.
>>	It's also interesting to note that this flexibility
>>(in open
>>source) has not been taken advantage of by Red Hat -
>>for whatever reason
>>they've not made the leap to at least i586 on the
>>Enterprise products -
>>even though the minimum hardware requirements exceed
>>that i386 compile.
>>I have tried to re-compile the entire distro in the
>>2.1 days - but there
>>were not that many source packages that even had
>>optimizations for
>>586/686.
>>
>>Don

-- 
Bill Moran
Potential Technologies
http://www.potentialtech.com