On Mon, 21 Oct 2002 psyche@computerdatasafe.com.au wrote:
>> Well - having thought about it a bit more ...
>>
>> Removing one 12Mb RPM is quite rediculous when almost EVERY other
>> intel RPM is built for an i386. Even glibc has an i386 version.
>> If you say that you no longer support i386 - then build all the
>> RPM's to i586 and be done with it. If every RPM was at least i586
>> then all intel machines would run a ilttle bit faster.
>
>I think this is not true, that code optimised for the 586 may run slowe
>on Pentium {II,II,IV} and AMD CPUs.
>
>Correct optimisation relies on knowledge of how the target processor
>executes its instructions, and on how the CPU is designed, so that
>each of the CPU's major components can be kept busy.
Yes that is true, but don't confuse instruction set optimization
(-march) with instruction ordering (-mcpu). Optimizing
instruction ordering for the specific CPU is always a gain for
that given CPU type, so long as gcc has code to optimize
instruction scheduling for that CPU, and it works properly.
We optimize instruction scheduling for i686 on all packages
currently, and have done so for as far back as I can remember
anyway.. someone else here at Red Hat would have to guess when
-mcpu was set to something other than i686.
The optimization that people have been debating for ages however,
is the -march one, "instruction set", and that is the one we
claim makes either no difference at all, or at best very little
difference except in very special cases.
>> The argument has been stated before that the majority of performance
>> gain is in using the kernel and glibc that matches your processor -
>> and that all the rest is more effort than worth the gain.
>> However, if they all were already i586 then the effort would be zero
>> to anyone installing to have all to be at least i586
>>
>> Secondly, there is no such thing as a height measurement that puts
>> the lowest pentium above the highest Cyrix 6x86.
>
>I think the 6x86 is an imperfect clone of the Pentium. I don't know what
>its imperfections are, but I noticed that motherboard manufacturers
>are a bit picky about which ones thay say "works with this board."
The Cyrix CPU's do not implement all of Intels "optional"
architectural features. There are some other quirks also.
>> I can think of a lot of reasons why the i386 kernel was not there -
>> but maybe one would be that general RedHat support for older hardware
>> is not as good as MS (RedHat seems to sometimes drop support for old
>> hardware that was supported in the previous release)
>
>The principal point of the i386 kernel is that it will run on anything. Apparently
>that's not so wrt the i586 kernel, even where "compatible" CPUs are used.
The i586 kernel will run on anything that we claim to support.
If a CPU is truely compatible, it should work. If it does not,
then it is very much a bug in the kernel, and should be reported
in bugzilla.
--
Mike A. Harris ftp://people.redhat.com/mharris
OS Systems Engineer
XFree86 maintainer
Red Hat Inc.
--
Psyche-list mailing list
Psyche-list@redhat.com
https://listman.redhat.com/mailman/listinfo/psyche-list
