[rescue] The aesthetics of rescue

[Skeezics Boondoggle]

[d'oh!  fatfingered the attribution lines, was in a hurry.  forgive...]

> What Auspex do you have?

a really old one. :-)  i think it's an NS5000, but damned if there aren't
any markings on it anywhere - i got it free, but without the cabinets!  
it was decommissioned by "firing the explosive bolts", so the ac wiring
harness was mostly gone, and i had to rebuild that.  i took an inventory
of the boards; i'll have to pull the machine apart again or find that list 
and post it. :-)

mounted in one of my old dec racks, it features 4 of the big 5.25" drive
shelves with 18 disks (mostly 3gb and 9gb varieties), two 6-channel scsi
procs and two 6-channel ethernet (?) procs, in addition to the host
processor.  the only thing missing is a suitable mbus module for the host
processor so i can boot the silly thing...

at some point i plan to write up a more detailed description of it so
auspex-knowledgable folks could help me identify the beastie and get it
running again.  the power harness works; both halves of the power supply
come up, the disks all spin, all eleven fans kick on, but without a cpu
module (i've tried three or four different ones) it obviously doesn't get 
very far. :-)

i also have six of the 3.5" disk shelves and a separate power supply for
those; i stole a number of the drives out of that and stuck 'em in my
netapp F330, oh the irony (netapp was founded by mostly ex-auspex guys).  
if there are some auspex collectors out there, i could be talked into
letting those shelves go - 42 disk slots with carriers and a random
smattering of 3.5" drives, plus the power supply and DC harness (i
repaired the AC input plug) - but it's bulky, heavy as hell, and the
shipping would be a bitch.  any auspex fanatics in PORTLAND? :-)  i could
use the storage space...


> Do you actually own a perq?  They were vector graphics based, weren't they?

three, actually!  two 1A's (one with "F" rev boards, one with "G") and a
T2.  i have boxes and boxes of documentation, including original
hand-written schematics and hard copy dumps of the pal code; as much
software as i could get for the thing (which really wasn't much, sadly);
some spare parts, like an extra boschert power supply, some disk boards,
etc. - that stuff is so hard to come by these days... (i plan to put up
the web shrine to the perq (and the cs6400, when that arrives :-) once i'm
rich and retired and can indulge my eccentricities full time!  woo!)

the perq is a bitmapped graphics machine, and features 6-way interleaved
dual-ported main memory, from which the screen bitmap is refreshed by the
integrated memory/dma/video controller.  it features a microcoded, custom
bit-sliced 16-bit cpu (built around the amd2910 sequencer) with a 170ns
microcycle time, 16k of writable control store for user-loadable custom
microcode, 256 general purpose 20-bit registers, a 16-level expression
stack, and a rasterop unit to assist in memory moves (i.e., fast
bitblt's).  so although it's a 16-bit machine (alu & memory word size), it
has 20-bit internal address paths so it can handle a megaword of memory;
the rasterop engine can cycle 64 bits every 680ns.  it had a z80 to handle 
low-speed i/o - keyboard, tablet, gpib, serial ports, audio output, but 
separate dma channels for ethernet and disk (a massive, slow 14" shugart).

for 1980, this was a badass little graphics box. :-)  years later i ran it
side-by-side with a sun 3/50 and it still whooped ass in the graphics
department.  7th fastest machine in the world in byte's 1983 "sieve of
eratosthenes" benchmark - so if, in 1983, you wanted to find all the prime
numbers up to 8192, boy howdy the perq was faaast. ;-)  i was pretty
pumped to see the perq listed 7th, after the crays and ibm 3083s, on a 
meaningless benchmark!  ah, youth. :-)

yes, "POS" was aptly named. :-)  but accent - the forerunner of mach - is
really cool.  the first common lisp came out on the perq, called spice
lisp; i think i still have a working copy.  there was also an very
unfortunately named unix system iii port called... "PNX".  generally
pronounced "peenix".  oh, dear.  (pnx, incidentally, is thought to be the 
first unix with an in-kernel windowing system...)

so if it isn't obvious, yes, the perq was the machine that got me hooked.  
hanging out not only at cmu (who had a room full of them called the "spice
rack") and playing with the xerox altos too (the model and inspiration for
the perq, the alto a very cool machine and i'd LOVE to have one to play
with :-), i also knew most of the founders and employees at three rivers
computer through my stepdad, a cmu alum and software engineer doing cae
tools on the perq, so i got to see the perq development offices and
learned pascal (remember pascal?  i loved pascal) and bitmapped graphics
(1024x768 @ 100dpi, 60hz non-interlaced, a big deal in those days) and
ethernet (i have a wire-wrapped "experimental 3mhz ethernet" board from
before the 10 mbit standard was published) and laser printing (on those
awful, smelly 240dpi canon "wet process" lbp-10's) - all in 1981/82...
damn, i wish i could go back... <sweet nostalgic sighs...>

much as i love to rescue older workstations from the late '80s/early '90s,
it really is the perq that defined and created the commercial "engineering
workstation" marketplace - before sun, before apollo, before symbolics or
any of those early vendors - taking the xerox ideas expressed in the alto
and the "d-machines" to market since xerox just couldn't seem to figure it
out.  sadly, three rivers computer took all the arrows in the back... but
at least this rescuer still carries the torch. :-)

cripes, i should probably go to, uh, "work" now!  later, y'all.

-- skeez
the original perq fanatic