> Hello,
>
> Teunis, I would *not* want to see the (fixed) 2.1.39 Cyrix patch go into
> the kernel source, as it stands now.
>
> I have a simple technical reason, and a more "philosophical" reason for
> this.
>
> Technical: no performance gains
> ==========
> Philosophical: kernel pollution vs. user-space utility
> ==============
An excellent commentary. This was one of the arguments I was looking for
actually - I wanted to know why it WASN'T already included.
As things sit:
- the 2.1.6x VGER snapshot has some of the cyrix patch present, so
this will make it into the kernel :)
- Cyrix _DETECTION_ is absolutely necessary, regardless of any other
features.
- For the performance/paging/... changes you're right - this patch
does not address those properly. But it does address them
at all which prepares the way for a formal policy. Not all
of the processor features are possible to enable in userspace
(VSPM + caching changes as examples IIRC)
I could get angry at cyrix for making a nonstandard detection policy, but
there's no point. Their cpu line is compatible with the i586 line and it
would be best if compilers could recognize such a processor to tailor
programs for efficiency.
Not everyone writes/uses software for distribution.
(if that were the case i386 support would be the only thing necessary...
for intel processor support...)
I wonder how many of these things are going to appear again for other CPU
clone manufacturers? (eg: alpha cpu clones)
As a sort of completion : Sorry about the mess on linux-kernel - I didn't
realize when I posted that cyrix support had made it into 2.1.6x vger :)
... Looking forward to the remerger of mainstream kernel and vger again...
G'day, eh? :)
- Teunis