compilation architecture

[Eric Sandeen]

Jakub 'Livio' Rusinek wrote:
> 2008/1/12, Patrice Dumas <pertusus at free.fr <mailto:pertusus at free.fr>>:
> 
>     On Sat, Jan 12, 2008 at 10:34:38PM +0100, Jakub 'Livio' Rusinek wrote:
>     > >
>     >
>     > So where Fedora just... sucks? Don't tell me "everything in Fedora is
>     > perfect". Fedora is slow. It's a fact.
> 
>     If you want to know, search. And avoid coming with ideas, but with
>     facts.

> You're funny...


No, serious.  And I think Patrice is willing to believe you when you say
Fedora feels slower than OpenSuSE.  But, if it's real, and you're
motivated to see it change, stick with it, and *really* find out why it
is so.  First prove it: find a representative test, measure it
objectively, and if Fedora is slower, find out why by further
measurement and testing.  You can be part of the solution.

For example - the Linux Battery Life Toolkit (BLTK) has 6 representative
workloads - Idle, Reader, Office, DVD Player, SW Developer, and
3D-Gamer.  Maybe that could be a place to start: tweak it to measure the
time for one run of an interesting workload on a fresh boot & login on
both platforms.  (This may not be the right test; "feels faster" is
probably more of a latency/responsiveness thing than a total runtime
thing, but perhaps this gets you thinking in the right direction).

Show the numbers that prove your impression, and people will probably
get much more interested.

One thing I think SuSE does is more aggressive preloading of shared
libraries:

http://sourceware.org/ml/libc-alpha/2004-07/msg00135.html

Like anything, this has tradeoffs of course.  But it's an example of one
difference worth investigating.  (Incidentally, in that thread, Ulrich
asks for hard numbers to show the performance gain, if any.  Detect a
theme here?) :)

-Eric