6502.org

It's not like we could get a MOS7529 these days, or at least easily. It was a "calculator chip" from MOS used in several calculators, providing advanced mathematical functions.

And you could make it work with a 6502

Part 1: https://archive.org/details/byte-magazi ... 7/mode/2up
Part 2: https://archive.org/details/byte-magazi ... 5/mode/2up

Even if you could make it work with your 6502, then that's the crux - it would work with your 6502 and highly unlikely to work with someone elses )-:

I suspect the actual interfacing might not make it worthwhile either -
Also, I suspect that today an ATmega or PIC can run just as fast, if not faster, so interfacing in a way to poke 2 x 4-byte FP numbers into it, a function then read back a 4-byte result... well, someones just used latches to do a similar thing with an ATmega, so ...

I do use an ATmega in my Ruby BCPL system for floating point work. I know my interface has a relatively high latency, but it's much faster than the fastest BASIC (BBC Basic), so must be doing something right...

-Gordon

I was wondering if hooking up my Teensy 4.1 to my 6502 would let me use it as a floating point processor, perhaps emulating similar functionality as the AM9511.

I also came up with a integer multiplier in hardware that uses shift registers and counters. If I'm right it could multiply two 8 bit numbers in about 8 clock cycles. Which I think after having to load/read the registers puts it on par with some of the faster multiplication functions.

At the end of the day I think the process of loading/storing the registers add a lot of extra clock cycles that could kill it's performance, IDK.

I haven't built it yet though, it's just a working prototype in Logisim.

We did mention the MCS7529 in a previous thread:
viewtopic.php?f=4&t=6956

I like the idea of any kind of coprocessor, and back in the day wiring up a calculator chip was an interesting kind of project, even though the performance was unlikely to be world-beating - because it was in your own computer!

As for getting the floating point numbers in and out, that can be a bottleneck, which is why it pays to have floating point registers on the far side, inside the copro, so that any chained or complicated operations can proceed without being transfer-limited.

Edit: I see Neil (barnacle) wired something up back in the day:

(Yes, but sometimes I wake up screaming...)

Neil

While I can't tell if that implementation is done efficiently, I found another site where they interface the chip with a KIM-1 and it seems they didn't needed that many chips.

https://archive.org/details/73-magazine-1978-08/page/100/mode/2up