Cupertino copied processor pipelining claims Wisconsin U

The patent management arm of the University of Wisconsin-Madison has taken aim at Cupertino over patent infringement in Apple's A7 chip. Its complaint here alleges infringement of a 1998 patent covering processor pipelining. The Wisconsin Alumni Research Foundation – WARF for short – took on Intel back in 2008 over the same …

[Anonymous Coward]

"Wisconsin's complaint notes that Apple has cited the '752 patent in some of its own patent applications (meaning that Cupertino was aware of the patent), but rebuffed WARF's approaches seeking a licensing deal."

That seems to be an on-going theme with Apple. Hopefully Wisconsin University won't let Apple get away with it and sees this to the end. They could easily get triple damages as Apple knew they were infringing and thus it was willful. Since it is also not FRAND, the sky is the limit for licensing. They could make the A7 and any successor to it quite expensive. Even better, when Apple loses, they could get a court order barring Samsung from selling the CPU to Apple until the fines and all licensing fees are paid.

I would bet that the PA Semi guys at Apple are working on a new chip that doesn't violate that patent but we all know, it won't be coming out tomorrow.

[Steve Todd]

They don't even know for sure

that the A7 is using a design that infringes the patent. Read the complaint. They only suspect that it does.

If Apple knew about the patent and COULD work around it don't you think they'd have done that already? You're making contradictory arguments.

[Anonymous Coward]

Re: They don't even know for sure

If Apple knew about the patent and COULD work around it don't you think they'd have done that already? You're making contradictory arguments.

Why are they contradicting themselves, how about....they could work around it, but thought it easier / cheaper to just ignore the patent and go ahead anyway and worry about the consequences (if any) later.

[John Smith 19]

@Lost all faith...

"Why are they contradicting themselves, how about....they could work around it, but thought it easier / cheaper to just ignore the patent and go ahead anyway and worry about the consequences (if any) later."

Hard to believe?

Well that's pretty much what Kodak did with the Polaroid patents on instant photography.

And Ford over the intermittent windscreen wiper.

Didn't end well for either of them.

[Anonymous Coward]

Re: They don't even know for sure

Wrong, I'm not making contradictory statements. If Apple didn't infringe, why did they list that patent? They can't argue they didn't know when they listed that patent and didn't pay the royalties for it. The sky is the limit for damages as the licensing costs could be claimed to be $929 for chip, which covers the price of the most expensive iPad that contains that chip. Then they could get triple damages as well. That is a lot of money.

[busycoder99]

Just another case of

The rich man slaughering the poor man's one sheep and refusing to pay, even though he has a fold full of them.

[Tom 13]

Re: Apple are working on a new chip that doesn't violate that patent

Doubtful. If Chipzilla couldn't work around it, I doubt anybody else can either.

[Frank Zuiderduin]

Branch prediction has been in ARM chips since forever (late 1980s or at least early 1990s). Prior art. Patent should never have been granted.

[Richard 12]

Maybe.

Depends on the specifics.

What type of branch prediction, how does it work etc.

Lots of ways to skin that particular cat - for example, most (GP)GPUs simply execute both branches and throw away the "wrong" answer.

It does seem odd to be going after Apple though, as I'm pretty sure they didn't design the microprocessor in the A7.

Unless they mean some custom bolt-on module, like a GPU or whatever.

Given the vagueness it does seem rather likely they are simply trolling, in the original sense.

[Anonymous Coward]

Re: Maybe.

It does seem odd to be going after Apple though, as I'm pretty sure they didn't design the microprocessor in the A7.

Makes no difference who designed the core, they designed and sell the final product for use in their own devices.

[Anonymous Coward]

Re: Maybe.

It does make a difference.

If you buy a part from a supplier you expect to be able to use that without licencing any patent.

If you build a part and use it then you expect to need to licence a patent if your design infringes a patent or uses a FRAND technology.

[Anonymous Coward]

Re: Maybe.

It does make a difference.

If you buy a part from a supplier you expect to be able to use that without licencing any patent.

However this is not the case here ... the difference is Apple have an "ARM architecture license" so they have rights to design their own implementation of the ARM architecture.

[Anonymous Coward]

Re: Maybe.

>Given the vagueness it does seem rather likely they are simply trolling, in the original sense.

You mean hiding under a bridge?

[Anonymous Coward]

Re: Maybe.

Apple is a licensee of ARM but that doesn't mean Apple cannot change the designs. Qualcomm does that, they have an ARM compatible processor. Also, ARM is a British company and if the university didn't file a patent say in Britain or Europe, then it is fair game. Apple is an American company operating in America and subject to patents in the US. Does ARM have offices in the US? Yes, but ARM didn't reference that patent, Apple did. If Apple never mentioned that patent, then the university wouldn't have known.

[Radbruch1929]

That may apply to (independent) claim 1 and (dependant) claim 2. From (dependedant) claim 3 onwards, the patent is more specific. And the patent addresses prior art in branch prediction under "Operation of the Data Speculation Circuit" and then differentiates its approach from this prior art. A cursory reading of the description and claim 3 (onwards) points to the operation of the prediction table as main differentiatior.

So this can be interesting because the prior art would have to use a prediction table to enable an early collision detection in order to decide what branch to execute.

[pacman7de]

History of Pipelining ..

"Seminal uses of pipelining were in the ILLIAC II project and the IBM Stretch project, though a simple version was used earlier in the Z1 in 1939" ref

[JP19]

Re: History of Pipelining ..

My memory is hazy but I worked on IBM mainframes in the late 70's and they pre-fetched, decoded, and pre-fetched operands for three possible execution paths and guessed at the outcome of branches to select the most likely paths. There was crap load of hardware to detect if stores invalidated any of the pre-fetched information.

I can't help but think that by 1998 any wrinkle on pipe-lining and branch prediction should have been considered obvious to those skilled in the art.

Looks like another example of the patent system making the world a worse rather than better place.

[Anonymous Coward]

Apple licences ARM core, I very much doubt they modify the design in any way.

So this is an attack on ARM users. It's no different to those patent trolls that expect users of GSM chips to licence a patent even though they've just bought an off the shelf chip and used it.

[JeeBee]

No, Apple license the ARM Instruction Set in the form of an architecture license - in the case of the A7 chip, that's the ARMv8 architecture.

The ARM CPU cores in the Apple A6 and A7 are full custom in-house designs, the first was called Swift and the current one is called Cyclone. http://en.wikipedia.org/wiki/Apple_A7

We don't know yet if other architecture licenses (and ARM themselves) also incorporate what this patent describes, and have the appropriate license for the patent if they do. The patent has to be far more specific than "branch prediction with speculative execution" ...

[John Smith 19]

A note on instruction set Vs hardware.

AFAIK ARM licenses it's products at various levels.

At the instruction set level they specify a binary pattern and state what that instruction does in some way, like VHDL. You could run that through a design system which would do an implementation. It would be based on what assumptions the design system made (register blocks with single read/multiple read, decoding strategy etc).

You could get a version licensed to one of the big fabs and add your own hardware blocks (or ones from their catalogue)

Or you could go for completely custom implementation of their instruction set, with your own team.

If Apple have gone that route then they are solely to blame for an infringement since it's their designers decision to do this, whereas if they'd imported a component from a design tool it would be the tool vendors problem.

[petronivs]

Agree on the previous correction. You'll be wanting to refer to the "University of Wisconsin". The 'Madison' isn't necessary, although it's sometimes used locally to distinguish the main campus from all the other University of Wisconsin universities around the state.

For abbreviation, "UW" is often used.

[MrDamage]

Expect a counter-suit

As Apple send in its spies to find out if any of the buildings at the University have rounded corners.

[RaidersOfTheLostArchitect]

Hey-- computer architect here (who also studied at UW Madison).

As many of the comments have erroneously stated, this has nothing to do with branch prediction. It has to do with a circuit that makes better guesses on whether or not to perform data speculation down the instruction stream based on prior successful/unsuccessful speculations. Modern pipelines in processors allow the execution of multiple otherwise sequential instructions to execute at the same time. However, you have to account for the fact that many times, instruction C relies on the output of instruction A-- so if you execute them in parallel you have to guess, or "speculate", on the data used. If done correctly this can get you huge gains, but when you are wrong, it really shoots you in the foot. What is novel about the '752 patent is the mechanism that is described and used to prevent being wrong as much as possible by using prior speculation history to accurately guess whether or not the current instruction is a good candidate for speculation. Prior approaches either used speculation on everything (meaning you were right a lot, but wrong quite a bit too-- which a lot of times actually hurt overall performance) or not at all. This approach really improved the overall performance of speculation in general by attempting to remove the 'wrong' guesses and thus the massive performance hits. (Note the difference here between data speculation versus branch prediction.)