Re: kernel-{image,headers} package bloat
On Sat, Apr 21, 2001 at 10:09:41PM +1000, Herbert Xu wrote:
> Rahul Jain <rahul@rice.edu> wrote:
>
> > Hmm... I was under the impression that the 486 required the most unique
> > optimizations of all the x86 line. Why not get rid of the k6, p2, and p3
> > kernels and have a 486 one instead?
>
> When you compile the kernel for a particular CPU, it's doing more than just
> passing a different flag to the compiler. It affects things like TSC, APIC,
> hand-tuned checksum routines and more. For all the gory details, hav a
> look at arch/i386/config.in.
Interesting. Thanks for the pointer. A few comments I have after looking at
that file:
It seems that most of the chips are the same as ppro. I think this supports my
idea above to remove the k6, p2, and p3 optimiations. Also, I think the p3-smp
optimization should be changed to ppro-smp. One problem I noted was that there
is no way to get 3dnow enabled with a k6 (I have a k6-3, which fits this
featureset). I don't think APIC should be enabled on the default kernels,
anyway, so that shouldn't be an issue. TSC vs. non-TSC seems to be the major
division here, along with ppro checksum (is that only used for raid5?).
So, I propose the following list of kernels:
386, 486, 586, 586tsc, 686, k7, ppro-smp.
Of course, people will need to figure out what each of these mean and which one
is best for their CPU, at which point they could just compile a kernel for
their exact CPU type and get the best optimizations possible.
--
-> -/- - Rahul Jain - -\- <-
-> -\- http://linux.rice.edu/~rahul -=- mailto:rahul-jain@usa.net -/- <-
-> -/- "I never could get the hang of Thursdays." - HHGTTG by DNA -\- <-
|--|--------|--------------|----|-------------|------|---------|-----|-|
Version 11.423.999.220020101.23.50110101.042
(c)1996-2000, All rights reserved. Disclaimer available upon request.
Reply to: