On Dec 22, 2010, at 6:09 PM, jim s wrote:
On 12/22/2010 5:29 PM, Chuck Guzis wrote:
On 22 Dec 2010 at 16:25, Fred Cisin wrote:
On Wed, 22 Dec 2010, Chuck Guzis wrote:
Himself BillG even gave it some happy words (but
then, what products
of Intel didn't he give happy words to?).
He called the 80286 "brain
dead"!
I remember Gordon Letwin (Microsoft OS/2) describing switching back
and forth between real and protected mode as being "like having to
turn off your engine to switch gears on the freeway".
I assume this was
before the LOADALL instruction was discovered. I
wonder if Letwin's statement was designed to confuse the competition,
since you"d think that Microsoft would be on the Intel developer
distribution list...
--Chuck
Developers didn't know about it. A friend of mine found and documented it.
It was not even in the Red covers for the processor.
There was a separate document that described it.
Some other friends at Micro 5, who were writing a clone of the 286 bios for their clone
and the emm driver found that microsoft knew about it by discovering it in their code,
which could do protected mode entry / exits and data transfers as a result "too
fast" They disassembled and reverse engineered the method and included it in their
driver, which when discovered caused a big fuss because intel thought they had violated
IP, (they were not nondisclosed).
Several of us at IBM were doing "interesting" things with the loadall
instruction on the 286. The 386 also had a variation of the loadall instruction but with
the introduction of the vm86 mode, the major need to use (outside of Intel) the loadall
instruction was obviated.
And it was brain dead to do that. By the way, the loadall was in there at all to support
the bondout package version used in ICE. Up to that time, INtel never had done a bondout
chip special for any of their ice products, but used the same die with more signals for
that purpose. Other vendors of ICE had to do elaborate things to support the ICE function
if it was doable at all.
The loadall instruction was a "test" instruction (Intel does similar things on
most of their parts). It allowed them to load all of the state of the processor at once.
It simplified their testing of the part.
The ICE parts were special bondouts/fuse versions of the part (Intel still does that
today...but for different SKUs...no more ICE). When you start getting into high frequency
buses it's just not feasible. Looking at the FSB analyzers for Core & Core 2
parts it was amazing that they worked at all (the analyzers). How to you probe every
signal on a high pin count 1GHz+ bus? It's not easy and the equipment is $$$$$$$.
The ICE functionality has been replaced with a JTAG like interface/protocol.
TTFN - Guy