16 Aug
2004
16 Aug
'04
5:35 p.m.
Nico de Jong wrote:
From: "Tony Duell"
<ard(a)p850ug1.demon.co.uk>
>It is, of course, possible to make a processor where the microcode can be
>changed by the user. Many early graphics workstations (the PERQ, many of
>the Xerox machines, etc) had microcode that was loaded from disk when the
>machine booted. WCS (Writeable Control Store) options were available for
>a few models of PDP11 to allow the user to write microcode.
Actually, intel and AMD x86 chips all have a small WCS for patching bugs
that are discovered after the chips are out in the field. The BIOS
loads these patch words on boot up.
To make this more vintage, I will relate it to something I know and
qualifies as vintage.
The Wang 2200 had boards full of low-density ROMs (at least originally
-- over time they replaced the ROM boards with a single board as the
density of available ROMs went up). Because doing a field upgrade of
70+ mask ROM chips would have been really expensive (no usable EPROM
then), Wang had a very interesting ucode patch mechanism somewhat like
that employed by AMD and Intel.
After a system was out in the field, if a ucode patch had to be
employed, a "patch" board would go out to update the system. The patch
board was a small PLA made from diodes and resistors, the output of
which would (1) assert a signal to disable the ROM boards and (2)
generate an address to a patch ROM that supplied the replacement ucode.
Apparently that was cheaper than retrofitting all the ROMs.
When the 2200 VP model came around (1975 or so) they changed to a WCS
scheme.