Gentoo Linux faster app-load than (Mandrake|Fedora)

[Mark Mielke]

On Thu, Oct 09, 2003 at 12:14:40PM -0400, Owen Taylor wrote:
> The initial sentence of the "What can be gleaned from these results" is
>  "For one, the default optimizations of Gentoo Linux 1.4 for Pentium III
> appear to make a significant difference in "real world" application  
>   load-time performance with Mozilla."

The "Gentoo Linux 1.4 for Pentium III" is optimized using the compiler
flags, sure, but it also has the kernel with Pentium III support
compiled into it, as well as hundreds of patches. I haven't looked at
each patch, but many of them are patches that have not made it into
the RedHat tree (NOTE: they used their 'gaming desktop' patch set for
the speed test), and definately not the Linus tree. The Pentium III
focus is more of a category as in "You can build the system any way
you want, but our LiveCD set comes with x86, Pentium, and Pentium
III (or whatever the current set is) pre-compiled, and pre-tuned for
you."

> A claim that I don't see any substantiation of in the rest of the
> article. What is documented is that in the test, Mozilla loaded a lot
> faster, but no tests were done to separate out different possibilities
> for what accounted for the speed difference. 

*nod*

> (It also should be noted that this page has undergone at least some
> changes since I read it a few days ago... unlike an article, you 
> can't really debate what a web page says.)

Perhaps I read the later revision, in which case, I shouldn't be
criticising you until I read the initial revision... :-)

> Well, doing the tests on Gentoo would be more likely to convince Gentoo
> users one way or the other :-). Also, the commonly heard Gentoo claim 
> is that you get significant advantages by compiling the *entire system*
> with architecture-specific optimization flags.

I've always interpretted this to mean that the entire system is compiled
using "i686" or later as a target, which for some packages (especially
including the kernel), includes more than just compiler flags. You are
right, though, that most people likely interpret this to mean compiler flags.

As for which was intended? I guess I can't speak for their motivations. :-)

I believe that RedHat has ventured into this area with the inclusion
of the FUTEX kernel feature, and the -nptl kernel/glibc builds. Components
of these features work more efficiently with a later instruction set.

I guess this issue is important to me because I find it a little
embarassing that my wonderful 'linux desktop' at work takes 3X as long
to start my web browser, and 10X as long to start my word
processor. Much of this isn't RedHat's fault at all -- The Mozilla
people and OpenOffice people are learning how to make their systems
more efficient (or at least appear more efficient), and once they are
done, hopefully the results will be competitive.

Cheers,
mark

-- 
mark at mielke.cc/markm at ncf.ca/markm at nortelnetworks.com __________________________
.  .  _  ._  . .   .__    .  . ._. .__ .   . . .__  | Neighbourhood Coder
|\/| |_| |_| |/    |_     |\/|  |  |_  |   |/  |_   | 
|  | | | | \ | \   |__ .  |  | .|. |__ |__ | \ |__  | Ottawa, Ontario, Canada

  One ring to rule them all, one ring to find them, one ring to bring them all
                       and in the darkness bind them...

                           http://mark.mielke.cc/