OT: FPGA soft-core humor

Dave Pollum schrieb:

I had cold recently too. But I am serious about designing and soft-core that uses 0 slices, 0 luts and 0 FF's :)

Antti PS I an I almost had some experiene with 2 bit processors, dont recall the part type any more but I had some of them in my hands. Well didnt ever make a PCB for them. Was some funky military 2 bit wide bit-slice thing.

And this design has no inputs, outputs, or clocks, either?

-Dave

Dave Pollum schrieb:

Why?

it is a clocked desing with lots of inputs and outputs, no problems with them at all.

it is really doable and fun thing to design.

Antti

Hmmm....I wasn't sure if you were joking when you said "0 slices, 0 luts and 0 FF's". I assumed that there was nothing left. I'm guessing that I/O buffers, clock logic, and interconnect circuitry would be left (I'm faimiliar with CPLDs, but I'm still learning about FPGAs).

-Dave Pollum

Signetics N3002 or something like that. Made the AMD 29xx series of bit slices look as well planned as the IBM 360 or something, by comparison. I had some kind of disk controller that used them, I think.

Jon

Dave Pollum schrieb:

no, I wasnt joking. meaning a design that yields to "0 slice" report those 0 luts and 0 FF in slices.

there aredifferent other resources to use, well depends on target family of course

Antti

what I had was K589 IK 2, see here

its 3002 compatible, so you guessed right :)

the 589 was boring, but some 583 family was in real beatiful rose ceramic with golden leads

Antti

There are still memories and DSP48's if this is V4, or MULTs if many other families. You can make a microcoded state machine with just a BRAM.

To split the difference, you could use Motorola's 14500 1-bit CPU. Yes, this is/was a real product!

Dave Pollum schrieb:

You can't do that with Xilinx. At least in previous version my

0 input design would not run through map because "XC9572XL has not enough input pins available for the design".

It was a design for a CPLD that was reserved for future use but had to set an output enable. Here is the code, released under GPL: output

Ray Andraka schrieb:

2 BRAMs give a pretty complete processor. And then there is that PPC405 block. I wonder whether you could turn that into a CPU.

Kolja Sulimma

Signetics was a second source of the Intel 3001 and 3002 bit slice components.

Yes. Much easier to deal with than the 3001/3002 parts.

The AMD 2901 4-bit part was based on the MMI 5701/6701 bit slice design. Reportedly one of the engineers left MMI to work for AMD, and AMD managed to get the very similar but not 100% compatible 2901 design to market first. Raytheon second-sourced it (as did National Semiconductor later), and it took off for military applications due to the availability of a second source (an absolute requirement back then).

Intel used the 300x parts in their own disk controllers, naturally. For instance, they had the iSBC 201 and iSBC 202 floppy disk controllers for Multibus, which each consisted of a channel board and an interface baord. The channel board had the 300x components and was the brains of the controller as well as a DMA engine.

The 300x were probably also used in Intel's hard drive controllers of that era (back when the hard drives were 8-inch or 14-inch, before the new-fangled 5.25-inch stuff).

I'm not familiar with any other 300x based disk controllers, but there probably were some. There were many vendors with 290x-based disk controllers, and also many using the Signetics 8X300 or 8X305 (10 MHz bipolar microprocessor, but not a bit-slice).

Yes, a one-bit CPU. It contains all of eight flip-flops, and a little bit of combinatorial logic. It does not contain a program counter; the user is expected to provide one, and a stack if desired.

There's a completely untested VHDL model at

It synthesizes to use 14 macrocells of an XC9536. It would probably fit in a 22V10 or maybe even a 20V8.

If anyone needs such a thing commercially, my arm could be twisted into verifying the model against the actual part.