> > : I am also upgrading my computer to 6x86. I've read that Linux
> > : works on 6x86 perfectly, doesn't it? So, why bother detecting the
> > : CPU?
> > It does have some points to have the kernel recognize it. As it
> > claims, Cyrix 6x86 should support MMX codes.
> 'Fraid not. Cyrix's M2 chip (in development) is supposed to support MMX
> extensions, but the currently shipping 6x86 does not.
> > If, possibly in the future, some programs need to detect for the
> > CPUID in order to turn the functions on, will it be a bug to regard
> > it as 4x86? 6x86 is actually different from a Intel's Pent...
> I honestly don't know if the Linux kernel could currently use any 6x86-specific
> features if it knew it was running on a 6x86. I do know that the 6x86 handles
> code optimized for the 486 better than it handles code optimized for the Pentium
> (I tried compiling kernel 2.0.7 with each optimization, and the BogoMips value
> was halved when going from 486 to Pentium optimizations, if that means
> anything).
The 6x86 has pentium sized pages (4MB) or 486 size pages (4K ?), it also
has
Cyrix sized pages (apparently varible sized between 4K and 4M ?).
I would think the that the kernel would be at an advantage if it used
the
pentium sized pages ( and maybe still the 486 optimizations?).
The 6x86 can also finish instructions out-of-order. So can the pentium
Pro.
there are region registers that you can disable this for specific
regions on
both the PentiumPro and the 6x86 [like a Memory-mapped IO card]. I
would
presume that using this feature would also be advantageous (sp?) to the
kernel.
My motherboard BIOS correctly recognizes the CPU as a Cyrix-6x86 P166+
, and I also
think it probably disables some of these features. Sooo, I am looking
at
making the linux kernel recognize the 6x86 and use the best mix of
features of it.
Brynn