• Please review our updated Terms and Rules here

Floating Point Systems AP-120B Array Processor

M

mpickering

Experienced Member
Joined
Sep 8, 2006
Messages
70
Location
Potomac, MD
Hi all,

I have the opportunity to acquire a partial (no PSU, no front panel) Floating Point System AP-120B array processor. For those of you who don't know what that is, it is essentially a mini-supercomputer geared to fast floating point calculations. It was hooked up to another computer (mini or mainframe) to augment its processing power. It required the host computer to feed it instructions and data. This model was used in the late 70s and early 80s as a cheap alternative to a Cray. A blistering 6 MIPS and vector processing capabilities thrown in. At one time, five of these units were yoked to a S/370 to form one of the National Centers for Supercomputing systems (of which five were created by the NSF and the only site that didn't use a conventional super at its core).

The cost is reasonable (I can probably get it for $50 with a little patience). I have some technical documentation (thank you Bitsavers!) but I really need to know if this is worth getting. I like it for the sheer novelty factor but I would really like to get it operational. I have only read about the AP-120B and this is the first one I have ever seen. I suspect they are extremely rare nowadays. I think a complete one would be worthy of a musuem display next to a VAX or System/370.

I don't mind getting it for the sheer sake of having it. It is probably the closest I'll get to a Cray. :) And with a PDP-11 I am getting next year, it might be an interesting project if I can find the bits needed to get it up and running.

Anyone have any experience with these things? Worthy project?

Matt
 
Ow .. nice piece of hardware, but without full technical specs it's going to be pretty much a paperweight.

This sounds like the original math co-processor ... outboard and on an I/O channel.
 
Did u buy it? I worked for FPS in the UK as a support engineer and the AP120b is a big part of my history. If you are needing any bits and pieces I may be able to source what you need.

The biggest problem you'll have is the software. i think the last release was on those horrible little tapes that went in a microvax - tapes plural that is!!
Prior to that it was tape reels and 12Mb removable disks!!

Fount of Knowledge I'm not but I can supply you loads of info if you wish!!
 
Even if you decide not to buy it for yourself, **PLEASE** don't let it go for scrap! It is something worthy of saving for posterity. If you don't want it, there are places which will gladly adopt it (e.g. Digibarn, etc). Put the word out that it's available, and someone will grab it up.

--T
 
AP120 software

AP120 software

> The biggest problem you'll have is the software.

Do you have any of the software for the AP120 or FPS100?

Something to watch out for is the interface is built into the unit.
The most commonly found AP120's will be out of GE CAT scanners
and will have a DG Nova interface.
 
Al Kossow said:
> The biggest problem you'll have is the software.

Do you have any of the software for the AP120 or FPS100?

Something to watch out for is the interface is built into the unit.
The most commonly found AP120's will be out of GE CAT scanners
and will have a DG Nova interface.
Click to expand...

I don't have any of the software but if somebody SERIOUSLY needs it I'll make some enquiries.

On both models the interface was on a single PCB that could be changed according to the computer. I think there was also a format board that might have needed changing for certain interfaces. There were interfaces for Prime, IBM, PDP, Vax, DG, Norsk Data plus a few more.
 
Floating Point Systems AP-120B Array Processor

Terry Yager said:
**PLEASE** don't let it go for scrap! It is something worthy of saving for posterity.
--T
Click to expand...
I was looking at older computers I've used and ran across this great site and thread. I'll 2nd the sentiment to save the machine, although it sounds like you only have the insides. I've always thought an AP120B should have been donated to the Smithsonian. They made high end number crunching possible at a 'reasonable' price during the early days (12 MFLOPS + 6 MIPS).

I used FPS machines quite a bit in the late 70s through early 80s and was the Secretary/VP/Pres. of the FPS User's Group. I think the FPS UG disbanded shortly after my term - testament to my leadership abilities!

FPS was a fantastic company until they turned management over to a 'professional'. They originally licensed an integer only version (AP120) from Culler-Harrison then added floating point capability to create the AP120B.

We typically had 120Bs hooked to a SEL mini because the SEL had a very fast/synchronous backplane and a fast interface to the 120B. We used the SEL/120B configuration for satellite image processing and real time analysis for the B2 bomber. A lot of other companies used them for medical image processing and oil exploration.

As far as I can recall, the comments earlier are correct. Each minicomputer required a unique I/O board (installed in the host minicomputer?). Software was assembled on the host then downloaded to the AP120B. Software development and debug for the AP120B was exceptionally difficult as it used a 64-bit (?) microcoded instruction (e.g., a floating point operation required at least 3 machine cycles) and there was limited visibility into the machine at run time. Although difficult, there was some remarkably efficient code written for the 120B.

So, you'll need a host board for your minicomputer and software both for the AP120B (I don't remember what ran inside the AP120B, but it wasn't much if anything at all) and the development/download/debug/runtime tools mini, plus the scientific subroutine library (forgot it's name) for the AP-120B.

I have intermittent contact with one or two ex-FPS managers in Beaverton who might know where the software might be located if you're still interested.

Best of luck,

Doug
 
Last edited:
Oh, do I remember the AP120B's and FPS-5000 series that were used on Gould/SEL Concept 32's by the aerospace industry in their flight simulation laboratories. There was a hardware connection on the SEL system, a ribbon cable to the FPS box (you had to watch out that the wires were not pinched at the plug), and an interface board on the FPS box where dip switch settings were made.

The software setup was pretty much the same for all of Floating Point's array processors; FPS-100, AP120B, FPS-5000 Series. Files were downloaded off of 9-track reel tape onto the host (Gould/SEL, DEC (VMS, RT-11, etc.), IBM (MVS and VM), Hewlett Packard, Data General, PR1ME, etc.) and built there. OS configuration changes (sysgen tables or whatever) were made so that the host would recognize the new device. In some cases the FPS software would also have to be configured, SEL in particular. Then diagnostics were run from the host to ensure that the new box could be communicated with and that it was functioning properly.

The downloaded files consisted of necessary communications software and a library of standard math routines. Additional math libraries were available for specific applications.
 
3.5 years later and no word as to whether he bought the darn thing.
Not sure that I have anything of substance to add.
I did work as an AP120B and FPS100 interface board test and field service tech ~1980-4.
I would guess that clients like IBM and Schlumberger would have disposed of their machines ages ago... and the people using them long since moved on?
 
Software

Software

Im very late to this however for some more background...
I worked writing the machine code software for the AP120B and later the 5000 model in the 1980's and 1990's. These were installed at the Flight Systems and Measurement Laboratory, Cranfield Institute of Technology (Now Cranfield University)
We used them for processing vibration data acquired on MoD trials aircraft (and other vehicles/devices some now in the Science Museum), with the main use being the FFT library and associated code.
The AP120B was attached to a CTL (later ITL) 8040 mini computer, and we interfaced with an in-house written interpreter.
The 5000 was attached to a DEC MicroVax 3400, with Fortran being the host language.
We used the digital input interface (it was similar to the HP GPIO interface) along with an in house developed Analogue to Digital converter system to acquire the data. Using 12bit ADCs the 120B could handle 8 simultaneous channels and expanded to 16 channels on the 5000. Code name ORAC Online Repetitive data Acquisition Controller (reference to Blakes7 TV series!)
I still have print out pdf scans of the 120 code listings and source files for the 5000 code (and associated Fortran), but no copy of the FPS library software or manuals.
I seem to recall the 5V PSU and the interconnect ribbon cable to the host were the items most likely to fail on the devices.

All the boxes have long gone in the skip, its a kind of shame we didn't appreciate at the time they would be significant pieces of history so soon.
 
John,

Can I suggest that you let Al Kossow have a copy of what you have for storage on the bitsavers website for posterity.

That way future researchers, maintainers and emulator writers may have access to the material if required.

Dave
 
You must log in or register to reply here.
Back
Top