Rogers and BitTorrent: another datapoint

[Ian Petersen]

>   For a true apples-to-apples comparison, what about "optimizing VM"
> versus optimizing compiler?  I'm talking...
> CFLAGS="-O2 -pipe -fomit-frame-pointer -march=athlon -m3dnow -mmmx -msse -msse2 -mfpmath=sse"
> and putting mmx, sse, and sse2 into "USE" (under Gentoo).

If you google "dynamo dynamic optimization", you can find a PDF that
answers this question in great detail.  The summary is that dynamic
optimization has a bigger effect (percentage wise) the higher you set
the optimizations in the static compiler.  Also, dynamic optimization
is not an alternative to static optimization--it's a complement.  If
you want the best possible performance, you should do _both_ (assuming
your run times are at least 2 minutes--you have to let the dynamic
optimizer "warm up").

The example in the PDF I'm referring to compared static optimization
equivalent to -O plus dynamic optimization to static optimization
equivalent to either -O2 or -O3 without dynamic optimization.  The
dynamically optimized version was slightly faster on average.  Dynamic
plus -O2/3 is even faster.  (I say "static optimization equivalent to
..." because they wrote Dynamo for HP's Unix and they were using HP's
static compiler, not gcc.)

Ian

-- 
Tired of pop-ups, security holes, and spyware?
Try Firefox: http://www.getfirefox.com
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists