HP35s Internal Investigations - new processor?

[stefan]

Hi,

I have just purchased a new HP35S and surfing on internet I go through the following thread contained in this forum:

"HP35s Internal Investigations re: Creation of the PH35sx
Message #1 Posted by Jeff O. on 10 Jan 2008, 8:37 a.m."

I was really curios to read it (I'm an HP vintage calculator fan). I read that there was the idea to modify the HP35S (PH35S) and I wanted just to flag this link: http://ex6502.altervista.org/

It is regarding a 32bit version of the 6502 processor (the one used into the HP35S). This site was just mentioning the HP35S as a possible system where to use this processor.

I don't know if it could be interesting for some of you. It seems to fit well in the idea explained by Jeff O.

Cheers
Stefan

[Eric Smith]

As a long time 6502 programmer, I find that interesting. There have been other attempts at 32-bit extensions of the 6502, including the WDC65832/WDC65T32 "Terbium" from Western Design Center, the company that developed the first CMOS 6502 and the 65816, as well as independent projects such as the 65020. There even was one group trying to define a 64-bit extension.

However, I see a few problems with using such a thing in a calculator design.

1. It isn't actually available as a commercial product.

2. If it was available, it would probably use more power than is desirable for a handheld calculator processor.

3. The 6502 architecture, even extended to 16 or 32 bits, isn't really competitive with modern architectures such as the ARM Cortex-M3. The only reason it made sense for HP to use a microcontroller with a 6502 core in some of their calculators is that the masked-ROM version of those microcontrollers is cheaper than dirt. For a hobbyist or small company to develop a calculator, it is much more important to have a part that is readily available and uses flash memory.

[uhmgawa]

Quote:

---

As a long time 6502 programmer, I find that interesting. There have been other attempts at 32-bit extensions of the 6502..

---

The publicized efforts were post MOS Technology after it had
been absorbed into Commodore, the result thereafter disintegrated
and the IP somehow ended up elsewhere.

There was a little known original internal engineering effort at
MOS Technology with the goal of developing a 32-bit processor
called the 65E4 relative to the 68K which was coming into
popularity at that time. The extra digit in the exponent was a
brainstorm which could only have originated in the marketing
department.

Anyway the "merger" with Commodore terminally sunk that project.
The other nearer miss was an MOS Technology second source of the
Motorola MC6809 which I'd desperately wanted to happen.
Unfortunately given the prior legal morass involving Motorola's
MC6800 and the pin compatible MOS6501 (the 6502's ill-fated and
short lived predecessor), Jack Tramiel informed us the only way
we'd have access to MC6809 IP was to open the package and take
pictures.

[Eric Smith]

The part that was pin-compatible with the Motorola MC6800 was the MCS6501. I've got a computer using an MCS6501 from the late but not much lamented company "the Digital Group". The part number for a plastic-packaged part would have used the prefix "MPS", but I haven't seen any evidence that it actually shipped. The more well-known part was originally designated the MCS6502 or MPS6502.

The original second-source vendors, Synertek and Rockwell, used different prefixes and suffixes. Synertek used "SYC", "SYD", and "SYP" for ceramic, CERDIP, and pastic, respectively. Rockwell used an "R" prefix, with a "C" suffix for ceramic and "P" for plastic. Both Synertek and Rockwell often omitted the letter denoting the package variant from the actual labelling printed on the part.

The more interesting story about the MC6809 (IMHO) concerns Hitachi's CMOS version. Hitachi was a licensed second-source for many Motorola parts, and developed CMOS versions of some of them, generally replacing the "68" with "63" to designate CMOS. In the case of the MC6809, they also decided to extend the architecture, adding more registers, and even a few 32-bit operations. Reportedly Motorola was very unhappy about that, and somehow brought pressure to bear on Hitachi which resulted in the architectural extensions of the HD6309 not being described in the public documentation. It was only years later that the details were leaked to a Japanese magazine.

[uhmgawa]

Quote:

---

Reportedly Motorola was very unhappy about that, and somehow brought pressure to bear on Hitachi which resulted in the architectural extensions of the HD6309 not being described in the public documentation.

---

I've always claimed the 6809 was one of the most under
appreciated, overlooked, yet well-designed architectures of
that era. I suppose a similar claim could be made for the
68K relative to the hauntings of Intel although the 68K
did compete fairly well overall.

IIRC the falling out over the 6809 soured the cross-licensing
relationship and was the primary reason Hitachi wasn't allowed
access to the 68K IP. I believe the bidirectional lawsuits
started to flourish about that time.

Years later when encountering the Hitachi SH it was obvious from
where they'd lifted that architecture's programming model. While
it has a few bizarre quirks peculiar to itself, they did toss out
the 68K's Pascal-ism of separate I & D registers which IMHO made
for a much cleaner design.

[Paul Dale]

The 68k is alive and well in the ColdFire range. Lovely processors.

The SH4 is an interesting processor, although the R0 register becomes a serious bottleneck and the dual issue interactions are difficult to handle well. The only CPU I know of that has a 4x4 matrix times 4x1 vector instruction that is fully pipelines -- if you can massage your problem into this form, it flies along at an absurd rate for its clock.

The 32032 series is the one that lost out, lovely processor but almost never used.

- Pauli