6502 and CPU licences in general

No, there is no copyright on CPU architectures. (Speaking of the US and Germany) Also, the architecture of the free-ip 6502 is radically different from the architecture of the original 6502, so even if there was such a copyright the 6502 would likely be not infringing. Only the ISA is identically, but there also is no copyright on ISAs.

There can be patents on architectures and on implementation techniques or features, without which a certain ISA might be next to impossible to implement (e.g. ARM has a patent on the barrelshifter/ALU combination) But the original 6502 is to old to have still open patent issues.

Copyright protects only the actual implementation: The HDL, netlists, bitstreams, layouts, etc.

You can reimplement any ISA, but it might happen that the architecture that you create violates some patent. BTW: This could be a patent that was created for a different architecture. I can imagine a 6502 implemenation that violates ARMs ALU patents, etc.

Kolja Sulimma

It also depends on what you intend to do with it. Companies lawyers will protect their patch - IIRC Atmel sued a phone company, that did what sounded like 'license creepage' of their AVR core : ie started designs using the AVR, and then did an ASIC with the same core. Look up the details.

If your target is FPGA, then only ARM would be likely to go after you. (as they have small revenue streams via Actel FPGA).

What sort of core are you looking at doing ? Why not look at the Mico8 and Mico32 from lattice, and contribute to that. That is fully opensource, and legal problem free. You could do a MicoFB, with your pet features... ?

-jg

AVR cores (developed by 3rd parties) are being used in ASICs and Atmel is not (yet) hunting them. but doing AVR-ASIC in large scale could couse some trouble(attempts)

Antti

Thanks, this sounds very good. I've found a C compiler for the 6502, too, and I will post a link to my website, when I have a working project with my latest TREX board.

--
Frank Buss, fb@frank-buss.de
http://www.frank-buss.de, http://www.it4-systems.de

Needs not to be fast, 8 bit registers and 16 bit address space is enough. Should be possible to use internal block RAM and to memory map special locations for accessing hardware, but should be small. It will be used in an FPGA, which is mounted as an extension board to another system, connected by I2C, which has some more VHDL code for accessing some other chips. The core should implement the control logic for interpreting I2C commands for accessing the other chips and providing the result.

I would like to have a C compiler for it, too. The cc65 compiler for the

6502 looks very mature and feature rich and I already know the 6502 instruction set very well (as an old C64 demo programmer :-) , if I need to implement something in inline assembler, so this was the reason why I think it would be a good choice.

I didn't found a C compiler on the mico8 page:

--
Frank Buss, fb@frank-buss.de
http://www.frank-buss.de, http://www.it4-systems.de

Yes, 6502 is mature and well supported. Have you compared the FPGA resource usage of the 6502 core, and the Mico32 ? Also, look up the PicoBlaze C compiler - recently moved I think ?

-jg

Patents is not just about how to implement things but also methods and ideas on how to solve things. You can do patent without ever implementing anything. So just doing a clean-room implementation is not something that will get you free from any patents violations.

I doubt that you can patent ISA but you can patent ideas that prohibit others from doing something that will execute the same machine code. MIPS had their patent on how to handle unaligned memory accesses which was tied into instructions. I think they used it when they went after MIPS clones. ARM has the shadow register handling during interrupts patent which makes it hard to do an usable ARM clone without violating that patent. I think ARM also have some other patents, I never bother to investigate what patent they have and when they expire. There should be some patents from the early 80s when Acorn(ARM) did the first ARM1 core that should expire soon. Since most of the stuff in ARM7 is coming from the early version of ARM cores. But there is nothing to stop them from suing you even if you don't violate any patents since an free ARM clone would be something that they want off the market. Just that you are under a lawsuit would tied you down so you in the end will remove the clone from the market.

I remember meeting one of the 6502 architecture guys in 2000 and he claim to me that he had legal rights to 6502 and that he could sue 6502 clone makers. I have no idea how valid his claim was or if the patents has expired.

In most cases if you do something for yourself and keep it for yourself, I think you are "free" to do anything you like since nobody will bother to sue you. If you start doing products with it and the revenue is starting to come in then I think you better be sure that you not violating anything.

Göran Bilski

eas

you

it

hat

ill

to

rs.

sue

an

I doubt you have much to worry about regarding patents on the 6502 core architecture. First of all, as has already been pointed out, the patent applies only to the implementation, specifically, HDL, netlists, etc. Since they used no hdl back in '74//'75, I believe you're on sound footing. However, the free-IP implementation of 6502 was (a) incomplete and (b) not very much like the physical implementation of the 6502. Further, you'd probably want to implement something that actually implemented the R65C02 instruction set, as that was the richest and most powerful. In order to make it both physically small and fast, you'll probably want to implement it such that it uses the ALU for both data and address arithmetic and uses transparent latches rather than clocked registers. It would use RAM instead of flipflops for those registers, and probably would implement its BCD operations by means of redundant registers rather than a deeper (and therefore slower) ALU. That way you avoid all the delays associated with wide synchronous counters. That way you can focus on an efficient 8-bit ALU rather than wide counters, since the 650x core always allowed two cycles for instructions that incremented the PCH.

Take a close look at that 'C' compiler, BTW, as you'll find it not very simple to implement independently of its console/video-related requirements.

Richard Erlacher Erlacher Associates Denver, CO

on

you can get the latest version of the T65 core from Opencores.

I did a lot of debuging work on the original and still maintain it as the original author has vanished. It is used in a number of the arcade game projects with great success.

/Mike

No, patents have nothing to do with that. All that is covered by copyright. Patents deal with the ideas and concepts.

Copyright issues can easily be avoided but last (virtually) for ever. Patents are difficult to avoid but expire after some time or when the patent fee is not paid.

Kolja Sulimma

possibly, we talked a lot for a while and then .... nothing. He was very interested in Asteroids so I would have expected some response. Possibly the patent lawyers got him ! I should try and get opencores to give me cvs access to the cores, then I could update the ones there. /Mike

"MikeJ" wrote

Mysterious. I hope he's fine.

I know that in my life my personal projects get put on hold when I'm busy with higher priority projects (e.g. buying a home).

Hmm, I think the 6502 has been around for so long it is not a choice for new designs. It has vanished into far-eastern rice cookers but with cores made just out of reach of patent infringements. I can't imagine Zilog pursuing Z80 VHDL models either: any significant commercial market will use full custom Z80 cores like the Rabbit.

So long as we can download your latest models that is the main thing.

Frank,

Now that the truly paranoid have worked it out of their systems, here is a bit of advice:

Like Kolja (I don't count him in the "paranoid category) has cleaqred things up (patents time out, and expire: trademarks last much longer), here is the practical side of things.

Patent holders have better things to do with their time than to start legal proceedings, so it goes like this (a flowchart in pseudo-code for all those software types out there):

if 'someone is potentially using our patents' then if 'revenues of said company > 1 billion $ (US)' then if 'letter sent to ask for license fees rejected' then if 'letter sent back contains demands for licenses from their portfolio is scary' then do: crosslicense() -- no money changes hands end else do; startlegalproceedings() end else do: negociatelicensefees() end else do: sendnastyscaryletter() --company too small end endif --keep looking for a big company with money infringing

Now, I suppose this is an engineer's view of the patent world, but I am sure that it is a pretty good description.

Now, there is a different algorithm for patent trolls, but it only differs in that cross-licensing is not an option.

Austin

Thanks for the algorithm. But what is the content of the nastyscaryletter? If it is a cease and desist letter from a lawyer, where I have to pay some thousands of dollars and have to stop using it in my designs, if I don't want to start legal proceedings, this would be not an option.

--
Frank Buss, fb@frank-buss.de
http://www.frank-buss.de, http://www.it4-systems.de

Frank,

The letter is to try to get you to cease and desist, or to pay a licensing fee (licensing is typically 1 to 3% of "value" -- whatever that is).

If you do not reply, or if you reply with "no way, sue me" then most companies will not waste the time. Since you are less than a billion dollars a year, chances are you don't have the money anyway, or you will likely not be in business in a year or two.

In the cold, hard, real world, legal folks are incredibly practical: only companies that make money, and are likely to be around in ten years are the targets of lawsuits (unless the people involved are just running on ego, and are willing to waste time and money).

So, the patent trolls (and companies with large portfolios) watch and wait: they keep track of how many patents you may be infringing on, and as soon as you announce passing one billion dollars in sales BINGO! the letters start arriving. Standard practice for legal departments, wait until the fish is worth catching before you put the hook in the water.

6502? What a laugh. How much money is there left to be made on this? How much money will a legal department spend? Do the math, then think about it.

There are some companies who when faced with knowingly infringing on someone's patent, consciously choose to do so. Yes, they may have to pay treble damages if they get caught (worst of scenarios) and also cease and desist. But how often do you read about someone losing a patent lawsuit like that? Often, the benefit outweighs the downside, and once brought into court, asking for a jury trial, and having a good legal team may cost you less than licensing it in the first place would have cost. Now, a good legal department won't let anyone do this, as the risk is very large, and the treble damages and shutdown is far too great a risk.

That said, in seems to happen all the time, regardless.

Some products have a six month lifecycle (like a printer). Go into business, build and sell 100,000 printers, and then disappear. Happens all the time. Do they even care about infringement? Nope, they do it again under a new name...

Getting back to ego, there are some who just love to go to court, and are willing to waste huge sums of money doing so. Eventually they run out of money, and end up bitching about their loss until they finally realize they were just stupid, and should have spent their energies more wisely. Or not. They never learn.

And lawyers never turn down a job, as even when they lose for a client, they at least get paid for it.

Austin

Not in the US. "Submarine patents" can be issued many years after the invention was actually made. For instance, the Gilbert Hyatt patent, US

4,942,516, on the microprocessor, based on an original filing in 1970 but granted in 1990. There are other cases where patents issued more than 60 years after application.

TI got a court to invalidate some but not all of the claims of that patent.

I think it is safe to say that it is *unlikely* for any patent issues to arise based on a new straightforward implementation of the 6502 ISA, but it's not impossible.

Submarine US patents are somewhat less likely to be a problem in the future due to the recent changes to US patent law.

Eric

References: