How to extend the 6502 for a bigger memory space, and some forth
I believe we are fortunate to have the author of that article, @Dr_Jefyll, among us on this forum.
I would love to hear anything from them that didn’t go into the article they still would like to discuss.
Thanks to all three of you for your interest. Are you curious about anything in particular that didn’t go into the article? Warning: if you let me, I will talk about my crazy contraption all day! LOL
One thing I always worried wasn’t adequately explained is how nimble (IMO) the 24-bit addressing is. I admit it’s clunky in the sense that you typically need two instructions to do what the (at the time unknown to me) 65816 can do in one instruction.
But you can’t complain that it’s slow. The first instruction – the one that loads the bank register you intend to use – takes as little as two cycles! I always worried that folks would assume that loading a bank register will take dozens of cycles. After all, that’s what they’ve been conditioned to expect from schemes that break the 64K barrier.
Forgive me, as a lot of this goes over my head so if You’ve already explained point to that instead.
Can this work with the 65C816?
How much wouldit cost? Given products like the raspberry pi make a majority of people just headscratch at anything less capible that costs more, I’m trying to figure out how to market something like this i na way that aims at the retr ocommunity even though… well this aims into another question.
Could this be put on a card to slot into an apple ii/c/GS? Fine fine software would need to be written that uses it ,because it isn’t enough for the hardware to exist software must know how to put it to use but… Can it be done?
I have a variety of ideas for using external logic to upgrade the 65C816, but the goals (and the means of achieving them) would be different than for my KK Computer (KimKlone upgraded 65C02). For example, the 'C02 address space is limited to 64K. But breaking the 64K barrier is not a goal on the '816. Also, on the 'C02 I was able to take advantage of the large number of opcodes that are undefined, but on the '816 this resource is extremely limited (only one undefined opcode), so other measures are required.
A great deal is possible, but I fear the realm of what might be profitable is small to non-existent. As you say, it’d be necessary to have a body of users willing to use the software that exploits the new hardware. This limits the potential audience. And there’s probably not a lot of money being spent by Retro enthusiasts in the first place.
Industry is different – the folks there have real money to spend, and as a freelance contractor I have had occasion to modify computers used on the factory floor. But my creativity in regard to 65C02 and 65C816 is likely to remain hobby based.
Honestly I’m more excited about extending these old workhorse computers from the earliest decadeish or so of personal computing tha nI am for ‘oh hey intel is switching to femtometer scale now’ (that WOULD be exciting since goign that small means they cracked the problem of electron tunneling but that is neither here nor there (; anyhoo…)
Plus, and this is anticdotal so please feel free to correct me of the notion if wrong, a 6502 is a heck of a lot simpler to produce and probably more fault tolerant along with the sheer number of videos i’ve seen of people either popping these in and out of zif sockets vs the utter lack of anyone withotu specialist equipment dealing with a modern rpocessor that wasn’t explicitely made to be unsocketed.
I just like the idea of pushing the old hardware as far as it can go since part of me does see a need to scale back compared to what we have at the bleeding edge.
So your kimklone? Take a 6502 and give it some much needed expanded ram (seriously when i saw 16 meg my first thought was ‘oooh i misunderstood this is for the thing that went in the iiGS.’ Nope, ye olden 6502. And that, frankly, is pretty damned cool since i could see this living on a ram card for a ii/e/c with applications written to do just that little bit more than the stock system would allow.
Pity you’d need a dedication encryption/decryption bit of hardware for TLS support otherwise i"d float the idea of a gemini client. Ah well, mercury (gemini sans the security layer) is still doable, and i’m pretty sure someone’s working on a decoupled client that has the router or other hardware between client and wider internet do the security dance.
Again, off topic.
Keep us posted Dr_Jefyll. I might not comprehend a lot of the nitty gritty details, but i’ve had my head in the clouds and hands on keyboard since the days of the apple ii/e. i live and breath for this kind of ‘hey look at this cool thing I made the old workhorse do!’ stuff that this project is the embodiment of.