Re: How to know correct hardware options for compiling 2.4.23 kernel
On Mon, Dec 15, 2003 at 05:00:06PM -0600, Jason Norment wrote:
> Debian installed pretty well without a hitch. (There was a minor sound
> problem, but it was easily remedied.)
> I'm now trying to compile the newest and greatest kernel available
> (that's listed as stable), and I aim to get such features as
> 'automounting' working... (A few other 'wants', such as the capability
> to communicate with my HP calculator -- found that laying in the kernel
> config.) ... and I hope to make it as well optimized as possible.
> Assuming I can get Knoppix up and working, am I correct in thinking that
> a combination of capturing dmesg and the output of lspci should get me
> nearly all of the information that I need to build a well-optimized kernel?
Based greatly on general discussions about optimizing for chips, etc, I
venture a yes to this. If you were going to use this laptop for
molecular mechanics studies or weather simulations, then you wouldn't
be using this machine... ;-)
But if (as I'm guessing) you're headed towards the best (perceived)
response time for normal stuff, then I think that a careful analysis of
1) what you want module-wise (loadable vs. compiled in),
2) service-wise (daemons taking up memory for no immediate purpose,
forcing swapping as other stuff comes in--define and use run levels
3) what can be done with what I think will be a big bottleneck (the
parallel IDE connection, if that's a reality for you)
...will go a long way, after any work on the kernel. My guts tell me
memory management and disk i/o will be big concerns (I think they
probably always are).
The gurus are likely chuckling. I'm quite open to enlightenment!
In a completely rational society, the best of us would aspire to be
_teachers_ and the rest of us would have to settle for something less,
because passing civilization along from one generation to the next
ought to be the highest honor and the highest responsibility anyone
could have. - Lee Iacocca