On Tue, 20 Feb 1996, David Mosberger wrote:
> As long as it's capable of running Linux, most people probably won't
> have a problem with this, if it means cheaper machines. However, is
> this different behavior *documented* someplace? I'd really hate
> spending hours debugging something that just doesn't work according to
> specs. If the hw ref manuals are accurate all is fine. Otherwise, I
> might just as well buy an x86. Intel has a real good history of not
> documenting chip features (at least not without NDA). I hope Digital
> is not following the same path.
And Erik wrote:
>>Last thing I heard, none of these crippled chips have actually been
>>manufactured yet. Jim Paradis knows all about this stuff though... care
>>to speak up Jim?
>>Erik
The 'crippled' versions have a -Pn suffix (for various integer n) and
the difference in behaviour is documented in the IPR chapter of the
HRM, rev C (EC-QAEQC-TE) and later (I expect it is also documented
elsewhere).
This change is not (nor is it intended to be) a hidden, undocumented
feature. The reason for its introduction is a reflection of the way
that businesses are organised (and therefore revenue channelled)
within Digital. That's beyond the scope of this discussion (no flames,
please).
To repeat: behaviour of -Pn versions is INDETERMINATE if you try to
run the SRM, OpenVMS and Digital Unix. What that means is that it may
work or it may not work. This gives us (Digital) the ability to sell
identical chips or chips that have had the documented features
disabled. Today, the chips are identical. Tomorrow (and particularly
with the introduction of the 21164A) that may not be the case.
The two features that are (maybe) disabled in the -Pn versions are:
1) SPE[0] must always be SET in ICSR (Ibox csr)
2) SP[0] must always be SET in the MCSR (Mbox CSR).
These features control one of the superpage mappings. It is a static
configuration that is performed during initialisation.
My job is to provide factory apps. support for customers who design
with Digital's silicon so I guess, on this one, I speak for Digital.
regards,
Neal.
----------------------------------------------------------------------
Neal Crook
Principal Engineer, European Semiconductor Applications Engineering
Digital Equipment Co. LTD EMAIL neal.crook@reo.mts.dec.com
Mailstop RE02-F/B3 Tel. +44 1734 206297
Digital Park FAX +44 1734 203133
Imperial Way,
Reading, RG2 0TU
ENGLAND
----------------------------------------------------------------------
---------------------------------oOOo---------------------------------
Copyright © 1995-1997 Red Hat Software. Legal notices