2

Here is a patent (US10255656B2) that claims an apparatus with specific functionality. I would like to focus on the first claim for now:

A multiprocessor comprising:

a register file to store operands; and

a plurality of processing cores, each core having execution logic to perform mixed precision multi-dimensional matrix fused multiply-accumulate (FMAC) operations, the execution logic to execute one or more instructions to:

multiply a first 16-bit floating point (FP16) operand with a second FP16 operand to obtain a first 32-bit floating point (FP32) intermediate product;

multiply a third FP16 operand with a fourth FP16 operand to obtain a second FP32 intermediate product; and

add the second FP32 intermediate product with the first FP32 intermediate product to generate a FP32 sum result.

It is specified that each core of the apparatus (microprocessor) has logic to perform the specific operations. Do I infringe the first claim if I perform the same operations on some other device from other manufacturer by implementing the operations in a programming language? What if my PC contains also those hardware elements? Steps and results of the operations would be the same.

Next, there is the second independent claim:

A method to facilitate execution of mixed precision multi-dimensional matrix fused multiply-accumulate (FMAC) operations comprising:

receiving a first 16-bit floating point (FP16) operand and a second FP16 operand at one or more processing cores;

multiplying the first FP16 operand with the second FP16 operand to obtain a first 32-bit floating point (FP32) intermediate product;

multiplying a third FP16 operand with a fourth FP16 operand to obtain a second FP32 intermediate product; and

adding the second FP32 intermediate product with the first FP32 intermediate product to generate a FP32 sum result.

If I want to perform such operations by implementing the operations in a programming language, will I infringe the second claim? I am asking this because of the "receiving" part of the method. My software implementation surely wouldn't have that part included but I am concerned with the "receiving" part might be included on a hardware level of my implementation. Would I be liable for the infringement then?

Note that when I say operations I mean exactly to multiplication and addition parts of the claims.

2
  • 1
    If you are going to ask a question about a claim, it would be helpful to quote the whole claim verbatim in the question.
    – Eric S
    Apr 30 at 13:06
  • Thank your for the information! I updated it.
    – Marko
    Apr 30 at 13:40

2 Answers 2

1

The claims to devices (a multiprocessor) can be infringed by making selling offering for sale importing or using such a piece of hardware. The closest you could get to infringing would be using one. If you were using an infringing device made by AMD it is AMD who would be trouble for making, selling etc. even though use is technically infringement.

On the face of it a program might do the steps specified in the method claims but multiplying four floating point numbers has been done for decades so assertions made during the prosecution of the patent must have implicitly narrowed the meaning of the steps to not apply to the prior art.

Separate from your question I will point out that the “each core” wording is horribly limiting. In a claim, case law (ResQNet, Inc. v. Lansa, In ) says each means every. Adding a fixed point core to a multi core chip will sidestep those claims. The Intel lawyers know that so it must be needed for novelty/non-obviousness.

I am a retired patent agent and a co-inventor of a multiprocessor patent.

5
  • Thank you for the answer. Do you think is it even possible to infringe the method without having the specified apparatus when looking at the whole method's construction even if I could implement the computation steps (but not the receiving step)?
    – Marko
    Apr 30 at 20:28
  • If I get a chance I’ll study the patent and its history.
    – George White
    Apr 30 at 20:32
  • There's no need to do that, I just wanted to see your opinion based on the claims only.
    – Marko
    Apr 30 at 20:34
  • The claims are interpreted in light of the specification and the prosecution history.
    – George White
    Apr 30 at 23:07
  • Receiving says the core accepts a value it does not say who or what performs an operation transmitting or providing it or otherwise how the value arrives there. So your program putting it in a register or the core retrieving it from memory would all, ultimately, involve the core “accepting” it. Without reading the specification.
    – George White
    Apr 30 at 23:14
0

This is really not my field. That said, I think I'm safe saying you won't be sued by Intel if you run any program you want on your PC even if you aren't using an Intel processor. The patent seems to me to be very specific to microprocessor hardware design not the software that might be run on the microprocessor. If Intel is going to sue anyone for infringement, it would be AMD, IBM or Apple or some other manufacturer of competitive microprocessors, not end users. That said, I am not a lawyer and this is not legal advice.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .