Wisconsin Alumni Research Foundation v. Apple Inc.
Updated: Nov 10, 2020
Wisconsin Alumni Research Foundation v. Apple Inc.
2017-2265, 2017-2380
Decided: September 28, 2018
Case Summary
Computer programs include lists of instructions that are written in “program
order”—and the instructions are typically executed in that order. In some
cases, a program’s performance may be improved by running its instructions
out of order. But this may also cause problems, such as when an instruction
later in the program depends on a calculation made by an earlier instruction.
The Wisconsin Alumni Research Foundation owns U.S. Patent No.
5,781,752, which covers specific techniques for running programs out of
order. Although the claimed methods are complex, the key dispute in
WARF’s suit against Apple relates to a claim limitation in which a
prediction is made as to whether—when a “particular” load instruction is
executed out of program order—problems will arise because that
“particular” instruction depends on calculations from another instruction that
has not yet been executed.
The District Court ruled that the claimed “particular” load instruction should
be construed to cover only a single, identified load instruction. Although
Apple showed at trial that its accused method associates predictions with
groups of load instructions (not a “particular” or single load instruction) the
Wisconsin jury found infringement, and awarded WARF more than $200
million in damages. The jury also rejected Apple’s effort to invalidate the
asserted claims; and the trial judge denied Apple’s post-trial motions to
overturn the jury’s findings.
In reversing the infringement verdict, the United States Court of Appeals for
the Federal Circuit explained that the claim language referring to a
prediction associated with a “particular” load instruction means that the
prediction must be associated with a single load instruction. A prediction
that is associated with more than one load instruction—as is the case in
Apple’s accused method—does not meet this limitation; therefore, the jury’s
verdict was not supported by substantial evidence.
The panel then turned to Apple’s invalidity argument; the key issue here
relates to the nature of the predictions made in the asserted method claims:
WARF argued that the predictions were “dynamic” in the sense that after the
claimed prediction (as to whether out of order execution of a particular load
instruction will create data dependency problems), events that take place
later in the program’s execution must be used to “update” that prediction.
The patent owner also said that although Apple’s asserted prior art reference
made data dependency predictions, these predictions were never updated.
The Federal Circuit agreed with WARF, noting that all of the disclosed
embodiments in the asserted patent described “dynamic” predictions that are
capable of receiving updates. And because Apple’s prior art reference only
described predictions that were not capable of receiving updates, the panel
affirmed the lower court’s validity ruling.
Download the PDF.