Has anyone noticed a difference in DVI overflow behavior on the PDP-8/I EAE
versus the PDP-8/E EAE? The 8/E EAE claims to be 8/I compatible in Mode A,
and I think I agree, for the most part. At least, it's compatible for the
parts that matter.
When a DVI instruction results in overflow, the EAE immediately returns
with the link set. The results in AC and MQ seem to have no relevance, but
they appear to differ between the 8/E and 8/I.
For instance, running the 8/I MUY/DVI diagnostics under SimH fails due to
the following:
sim> lo maindec/maindec-8i-d0ba-pb.bin
sim> d sr 40
sim> g 201
DIVERR L C(AC) C(MQ) C(MB)
PROB 0 000000000000 000000000000 000000000000
GOOD 1 111111111111 000000000000 000000000000
BAD 1 000000000000 000000000001 000000000000
SCA 000000000000
HALT instruction, PC: 01512 (JMP I 1506)
The link is set, but obviously MQ and AC do not match.
Running the same diagnostic on an 8/I works fine.
I can't imagine a scenario outside of diagnostics where this behavior would
impact the software, but it does seem curious nevertheless that the DVI
approach to handling overflow differs slightly between EAEs on the 8/I and
8/E.
Kyle
I've got rights to a fairly nice system located in St. Louis.? It has
working streaming tapes as well as half inch, all working.
It is on till this coming weekend.
The full system is a single bay, I've been told is 7' tall on casters.?
I won't let it be scrapped if possible, but I'd like it to go off the
floor directly to someone interested and not have to use favors to get
help having it moved out.
It will be skinned of an addin UPS but otherwise disconnected and put to
one side till it can be picked up.
Told the location has dock high, but no word on how that is accessible
or what type it is.? Might be able to move dock high to dock high anyway.
Let me know if there's interest.? I will have to have possession of the
drives, but will make sure the hardware that goes with them is kept.? I
hope I can zero them and pass it along.
Cabling will be boxed as appropriate and will be included.
Let me know if you are interested, and pass it along.? I know it's a
dual processor, but don't have other info right handy here.
thanks
Jim
That appears to be an earlier model of a similar system we had at UBC
which could crunch arrays of FP numbers at 10 Mflops. Had it
connected to an 11/44 and just recall doing some frantic programming
mainly involving using minimal code as had to use memory management
to allocate memory pages to get data into array processor and then
fetch results. Realized at that time that a 56 Kb memory space was a
bit limited for this type of work. Did FFT far faster than 11/23
(which took 1 second for 1024 points using DEC's code that shipped
with MINC) but still had to do overnight runs to analyze a lot of our
data. Likely have bad memories of that part of my programming career
as we were under some rather tight deadlines to analyze data to get a
few papers published and I much preferred writing in PDP11 assembler
as very rarely had to deal with running out of memory issues with
data acquisition code.
Out of curiousity, decided to benchmark one of my old, really cheap
PC laptops that got in 2010 and it managed 30 Mflops using double
precision arithmetic. 10 Mflop performance no longer as impressive
as it used to be.
>I picked this up a number of years ago for reasons that entirely escape
>me. It's certainly neat, but I don't see myself ever actually using it and
>it's large and heavy.
>
>Documented here:
>http://bitsavers.trailing-edge.com/pdf/fps/7259-02_AP-120B_procHbk.pdf
>
>Mine appears to have a DEC-style interface but I'm unsure what it talks to
>on the DEC side of things.
>
>I can take pictures if there's interest, but it's fairly nondescript, just
>a large white box with rack-mount ears and a small panel with some switches
>on it.
>
>It's in the Seattle area if anyone wants it, and it's free! Shipping is...
>not something I really want to think about right now.
>
>- Josh
RLX Technologies pioneered the blade server concept between 1999 and 2005
(when they got acquired by HP). I have two of their early RLX 24 blade
enclosures, one fully populated with 24 transmeta-based processor blades,
and the other with 19 blades.
Julf