keaton (03/28/83)
I would be truly awestruck if the iAPX-286 behaved
as Dave Taylor says; however, he is slightly mistaken.
In REAL address mode, the 286 behaves EXACTLY like
the 86 and 186 -- this means that segment registers are
shifted left four bits instead of fourteen and therefore
yield a one megabyte address space, not sixteen megabytes.
In protected mode, the sixteen bit segment registers are
shifted left fourteen bits, yielding a thirty bit virtual
address -- one gigabyte, not one terabyte. It is the
iAPX-432 which has a one terabyte virtual address space
(forty bits of address). The place where the sixteen
megabytes come into play is when the operating system
(in protected mode) has that much real memory in which to
store part of the virtual memory of each process -- i.e.
the CPU chip has 24 address lines coming out of it but
the top four can only be used in protected mode.
Of course, this is still impressive, especially for
a microprocessor.
David Keaton
ucbvax!unmvax!nmtvax!keaton
P.S. The compatibility shown is about right. Programs
written for the iAPX-86,88 (the 86 and 88 are identical
in software) will run on the iAPX-186 and programs which
run on the 186 will run on the iAPX-286 at least in real
address mode, and usually in protected mode. Any 8080/8085
programs must be translated at the source code level (a
simple task for which Intel has developed an automatic
tool) to run on an iAPX-*86.tihor (03/30/83)
#R:nmtvax:-27100:cmcl2:7300004:000:200 cmcl2!tihor Mar 29 19:32:00 1983 Note however that the 386 and 486 are, based on Intel's current designs, intended to continue expanding the address spaces and are "scheduled" at one year intervals. Only time will tell of course...