Re: Will a compiled kernel run faster/better?
From: Jean-David Beyer (jdbeyer_at_exit109.com)
Date: 07/17/04
- Next message: Michael Heiming: "Re: Will a compiled kernel run faster/better?"
- Previous message: Al C.: "Will a compiled kernel run faster/better?"
- In reply to: Al C.: "Will a compiled kernel run faster/better?"
- Next in thread: Michael Heiming: "Re: Will a compiled kernel run faster/better?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Sat, 17 Jul 2004 13:56:34 -0400
Al C. wrote:
> Is there a measurable difference in speed between using the stock
> kernel you get with your distro, and one that you compile yourself?
>
> By speed I mean how long it takes windows (KDE) to pop, apps to load and run
> (like FireFox).
>
> What is the payback for doing a kernel compile besides getting some kind of
> enhancement (like Win4Lin).
>
> I know the first thing a lot of you guys do after you've installed a distro
> is to compile your own kernel. Why?
>
You can figure a lot out about this without trying it.
How long does it take a KDE window to come up? Well, first, the
application that will ultimately want to pop a window has to realize that
is what is to be done and to do it. Sooner or later it will send a message
to X. So message passing speed might be important (I doubt it). Then X may
create a blank window. Then the application would have to find out what
the window is to look like and what its contents should be. Mostly disk
IO, I would imagine. So what is your disk performance? There are tools to
measure that.
My guess is that most windowing stuff is disk limited. On my machine
almost everything is disk limited even though I have four 10,000 rpm
Ultra/320 LVD SCSI hard drives. Still disk limited (more precisely, seek
limited). Another problem I have is that my two 3.06GHz processors are
cache limited. I have 4096 Megabytes RAM, but if I am running a normally
IO limited database application and I run compute-limited seti@home
processes at the same time at nice 19, the seti@home processes slow down
the database stuff running at normal priority. My interpretation is that
whenever the database pauses for a disk IO, seti@home gets the
processor(s) and dirty up the 1 Megabyte cache on each processor so
thoroughly that the database application is effectively running from the
relatively slow RAM (533MHz is slow compared to 3060 MHz processor)
instead of the high speed caches on the processors.
So for me, recompiling the kernel would make no noticeable improvement. My
guess is that the same applies to you.
-- .~. Jean-David Beyer Registered Linux User 85642. /V\ Registered Machine 241939. /( )\ Shrewsbury, New Jersey http://counter.li.org ^^-^^ 13:45:00 up 23:25, 4 users, load average: 11.70, 11.51, 10.81
- Next message: Michael Heiming: "Re: Will a compiled kernel run faster/better?"
- Previous message: Al C.: "Will a compiled kernel run faster/better?"
- In reply to: Al C.: "Will a compiled kernel run faster/better?"
- Next in thread: Michael Heiming: "Re: Will a compiled kernel run faster/better?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
|
