`
` UNITED STATES PATENT AND TRADEMARK OFFICE
`______________________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`______________________
`MICROSOFT CORPORATION,
`Petitioner,
`v.
`DIRECTSTREAM, LLC,
`Patent
`Owner.
`_______________________
`Cases IPR2018-01605, IPR2018-01606, IPR2018-01607
`Patent 7,620,800 B2
`__________________________
`
`PATENT OWNER DIRECTSTREAM, LLC’S
`SUR-REPLY IN SUPPORT OF PATENT OWNER’S RESPONSE
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`TABLE OF CONTENTS
`
`
`
`I. PETITIONER PURPOSEFULLY AVOIDS THE TECHNICAL
`BACKGROUND OF THE ’800 PATENT ..................................................... 1
`A. The ’800 Patent: Methods for Enhancing Parallelism and
`Performance in Reconfigurable Computing Systems, Using FPGAs.
` .................................................................................................................... 1
`B. Patent Owner’s Experts Provide the Board the True State of the Art
`and Opinions Consistent With An Actual Methodology. .......................... 2
`C. Petitioner and its Expert Provide No Evidence or Support for Their
`Understanding of the State of the Art or What a POSITA Would
`Have Been Motivated to Do In Order to Meet Their Burden of Proof.
` .................................................................................................................... 4
`II. PETITIONER’s EXPERT’S REPLY TESTIMONY AND EVIDENCE IS
`CONCLUSORY, UNTIMELY, AND NOT RELIABLE ............................... 6
`A. Dr. Stone Fails to Provide Facts and Data to Support His Opinions
`and Purposefully Misconstrues Patent Owner’s Expert Testimony. .......... 6
`B. Petitioner’s Reply and New Testimony is Outside the Scope of
`Patent Owner’s Response or Should Have Been Provided in His
`Original Declaration. .................................................................................. 8
`III. PATENT OWNER’S CLAIM CONSTRUCTIONS .......................................... 9
`A. “stream communication” .......................................................................... 10
`B. “computational loop” [wherein only functional units needed to solve
`the calculation are instantiated] ................................................................ 15
`1. Computational Loops Are Not Infinite Loops ................................ 15
`2. Looping controlled by the Sun workstation .................................... 17
`C. “pass computed data seamlessly” ............................................................. 20
`D. “seamless” vs. “systolic” and “data driven” ............................................. 21
`
`
`
`i
`
`
`
`IV. OBJECTIONS TO THESE PROCEEDINGS ................................................... 21
`APPENDIX 1 ........................................................................................................... 25
`LIST OF EXHIBITS ................................................................................................ 28
`
`
`
`
`
`
`
`ii
`
`
`
`TABLE OF AUTHORITIES
`
`
`CASES
`Align Tech., Inc. v. ClearCorrect Operating, LLC,
`745 Fed. App’x 361 (Fed. Cir. 2018) ......................................................................... 4
`
`
`Arthrex, Inc. v. Smith & Nephew, Inc.,
`941 F.3d 1320 (Fed. Cir. 2019) ........................................................................... 22, 23
`
`
`Beacon Adhesives, Inc. v. U.S.,
`134 Fed. Cl. 26 (Fed Cl. 2017) .................................................................................. 13
`
`
`In re IPR Licensing, Inc.,
`942 F.3d 1363 (Fed. Cir. 2019) ................................................................................... 9
`
`
`In re Nuvasive, Inc.,
`841 F.3d 966 (Fed. Cir. 2016) ..................................................................................... 9
`
`
`Intel Corp. v. VIA Techs., Inc.,
`319 F.3d 1357 (Fed. Cir. 2003) ................................................................................. 11
`
`
`IPR Licensing, Inc. v. ZTE Corp.,
`685 Fed. App’x 933 (Fed. Cir. 2017) ..................................................................... 5, 6
`
`
`Knauf Insulation, Inc. v. Rockwool Int’l A/S,
`No. 2018-1810, et al., 2019 WL 5152356 (Fed. Cir. Oct. 15, 2019) ........................ 4
`
`
`Personal Web Techs., LLC v. Apple, Inc.,
`917 F.3d 1376 (Fed. Cir. 2019) ................................................................................... 4
`STATUTES
`35 U.S.C. §312(a)(3) ........................................................................................................ 9
`
`U.S. Constitution, Article II, Section 2, Clause 2 ......................................................... 23
`RULES
`FED. R. EVID. 401, 402, 702 and 703 .............................................................................. 3
`
`
`
`iii
`
`
`
`REGULATIONS
`37 C.F.R. §42.23 .......................................................................................................... 8, 9
`
`37 C.F.R. §42.65 .......................................................................................................... 3, 6
`
`37 C.F.R. §§42.22 ............................................................................................................ 9
`
`Office Patent Trial Practice Guide,
`77 Fed. Reg. 48,756 (Aug. 14, 2012) ......................................................................... 8
`
`
`
`iv
`
`
`
`
`
`I. PETITIONER PURPOSEFULLY AVOIDS THE TECHNICAL
`BACKGROUND OF THE ’800 PATENT
`Petitioner’s Reply arguments are premised on purposeful mischaracterizations
`
`of the inventions disclosed in the ’800 Patent, the state of the art, and claim construction
`
`legal standards in order to misinterpreting the scope of the patent. See Reply, passim.
`
`When examined in context of the specifications and actual state of art, Petitioner fails
`
`to meet its burden of proof challenging the patentability of the disputed claims.
`
`A. The ’800 Patent: Methods for Enhancing Parallelism and Performance in
`Reconfigurable Computing Systems, Using FPGAs.
`Using hindsight bias, Petitioner misquotes the patent specifications to claim
`
`that the disclosed inventions relate to forms of “general” FPGA-based computing.
`
`Id., 2. This is simply untrue—the invention disclosed in the patent is how to achieve
`
`enhanced parallelism and performance using reconfigurable systems, such as with
`
`FPGAs.
`
`The patent specification makes clear the ’800 Patent claims techniques for
`
`enhancing parallelism and performance in reconfigurable computing systems.
`
`EX1005 at 1:39-43. And, contrary to Petitioner’s characterizations, parallel
`
`computing is in the field of high-performance computing. EX2112¶¶40-41, 57-59,
`
`78, 123-131; EX2166¶¶17-30; EX2101¶¶11-79. At the time of the invention and as
`
`the specifications teach, “most large software applications achieve[d] high
`
`performance operation through the use of parallel processing” that required
`
`
`
`1
`
`
`
`“multiple processors to work simultaneously on the same problem.” EX1005 at
`
`1:44-52. The problem was that “as more and more performance is required, so is
`
`more parallelism, resulting in ever larger systems” to the point that “[c]lusters
`
`exist…that have tens of thousands of processors and can occupy football fields of
`
`space.” Id. at 1:55-58. “Systems of such a large physical size present many obvious
`
`downsides, including, among other factors, facility requirements, power, heat
`
`generation and reliability.” Id. at 1:58-61.
`
`The inventors of the ’800 Patent realized that this problem could be solved by
`
`“a processor technology…that offers orders of magnitude more parallelism per
`
`processor.” Id. at 1:65-66. And that this type of processor technology is “possible
`
`through the use of a reconfigurable processor” because reconfigurable processors
`
`can “instantiate as many functional units as may be required to solve the problem up
`
`to the total capacity of the integrated circuit chips they employ.” Id. at 2:1-7
`
`(emphasis added).
`
`B. Patent Owner’s Experts Provide the Board the True State of the Art and
`Opinions Consistent With An Actual Methodology.
`To further avoid the field of endeavor the patent teaches and problems in the art
`
`at the time, Petitioner in its Reply attempts to side-step the actual state of the art.
`
`Instead, Petitioner criticizes Patent Owner and its experts and fact witnesses for
`
`including corroborating evidence to support their testimony or evidence of secondary
`
`indicia of non-obviousness, such as relevant facts and data demonstrating the true state
`2
`
`
`
`
`
`of art of high-performance computing, parallelism, and FPGA technology at the time.
`
`And, Petitioner further criticizes Patent Owner for citing such evidence in the Response
`
`or discounts its significance. See Reply, 1, 3.
`
`As shown in Appendix 1, attached hereto, the evidence relied upon by Patent
`
`Owner supports each of its experts’ or witnesses’ underlying methodology, opinions,
`
`or facts and data as required under 37 C.F.R. §42.65 and FED. R. EVID. 401, 402, 702
`
`and 703. More importantly, the state of the art is critical to show what a POSITA at
`
`the time would understand, be motivated to do with the prior art, or demonstrate the
`
`secondary indicia of non-obviousness—which Petitioner wholly failed to provide. See
`
`Response, 3-30.
`
`Petitioner’s argument effectively requests the Board ignore this relevant
`
`evidence and technological background in order to bolster its unsupported expert
`
`testimony, attorney argument, POSITA “knowledge,” and inconsistent legal standards
`
`advanced in its Petition. Realizing that its Petition and expert “testimony,” which the
`
`Board relied upon to institute these proceedings, are wholly deficient of any
`
`corroborating evidence as well as facts and data to meet its burden of proof, see
`
`Response, 18-30, Petitioner urges the Board to ignore Patent Owner’s evidence.
`
`Petitioner’s Reply is a de facto Motion to Exclude Patent Owner’s evidence because it
`
`is too much information to read. See Reply at 2. Petitioner’s criticism is misplaced.
`
`Patent Owner provided the Board with thorough declaration testimony and supporting
`
`
`
`3
`
`
`
`documentation to show the clear state of the art at the time of the invention along with
`
`establishing the proper standard for the knowledge of a POSITA at the relevant point
`
`in time. To the contrary, Petitioner’s attempt to avoid Patent Owner’s evidence
`
`demonstrates a lack of support for its positions other than cherry-picked expert
`
`testimony. See Response, 18-30.
`
`As shown in Appendix 1, Petitioner’s characterizations are without merit and
`
`reversible error if the Board does choose to ignore or strike. See, e.g., Knauf Insulation,
`
`Inc. v. Rockwool Int’l A/S, No. 2018-1810, et al., 2019 WL 5152356, at *4 (Fed. Cir.
`
`Oct. 15, 2019) (vacating Board invalidity ruling based on improper interpretation of
`
`prior art reference at the time of invention); Personal Web Techs., LLC v. Apple, Inc.,
`
`917 F.3d 1376, 1382-83 (Fed. Cir. 2019) (same); Align Tech., Inc. v. ClearCorrect
`
`Operating, LLC, 745 Fed. App’x 361, 364-65 (Fed. Cir. 2018) (same).
`
`C. Petitioner and its Expert Provide No Evidence or Support for Their
`Understanding of the State of the Art or What a POSITA Would Have Been
`Motivated to Do In Order to Meet Their Burden of Proof.
`Similar to the flawed argument advanced by the patent challenger in IPR
`
`Licensing, Petitioner in its Reply at 49-52, essentially attempts to shift its burden of
`
`proof onto Patent Owner as to why a POSITA would have combined the asserted
`
`combinations. In summary, Petitioner’s argument is that Patent Owner allegedly “has
`
`advanced no evidence to suggest that” use of FPGAs in the development of parallel
`
`computing “would have been beyond the level of ordinary skill at the relevant time”
`
`
`
`4
`
`
`
`or that the combinations of disparate art would have been obvious. See IPR Licensing,
`
`Inc. v. ZTE Corp., 685 Fed. App’x 933, 939-40 (Fed. Cir. 2017).
`
`Petitioner, in conclusory fashion, alleges its expert, Dr. Stone, “includes
`
`numerous reasons to combine each cited combination…and an explanation of how
`
`each combination would be made.” And then, Petitioner mischaracterizes Patent
`
`Owner’s argument in its Response by asserting Patent Owner somehow conceded to
`
`these “explanations” by Dr. Stone. Again, Petitioner’s argument is untrue.
`
`Patent Owner made clear it was hindsight reasoning, resulting in the creation
`
`of a flawed POSITA, that allows Petitioner and its expert to improperly assume the
`
`proposed combinations would “just work.” Response, 18-30, 114-120. What
`
`Petitioner fails to show in its Petition, or address in its Reply, is why a POSITA would
`
`have been motivated to make the combinations it relies upon, without the benefit of
`
`hindsight. Instead, Petitioner’s Reply tries to shift the burden onto Patent Owner by
`
`asserting Patent Owner “presents no evidence or affirmative argument to contest” the
`
`conclusory opinions of its expert. Reply at 51.
`
`As the Federal Circuit admonished, such a burden shift is misplaced by
`
`Petitioner. The Federal Circuit recently made clear the test is not claiming evidence is
`
`needed what “skilled artisans could have done if motivated;” instead, the correct
`
`standard Petitioner and its expert must demonstrate is “what [skilled artisans] would
`
`have been motivated to do.” ZTE, 685 Fed. App’x 939-40. Here, Petitioner has no
`
`
`
`5
`
`
`
`evidence that any of the proposed combinations are obvious, such that a POSITA
`
`would have been motived to do the combinations, and therefore has not met its burden.1
`
`II. PETITIONER’S EXPERT’S REPLY TESTIMONY AND EVIDENCE IS
`CONCLUSORY, UNTIMELY, AND NOT RELIABLE
`
`
`A. Dr. Stone Fails to Provide Facts and Data to Support His Opinions and
`Purposefully Misconstrues Patent Owner’s Expert Testimony.
`
`
`
`Dr. Stone is not a reliable expert. His opinions are grounded in cherry-picked
`
`and attorney-provided facts and assumptions. As shown in its Response, Patent Owner
`
`provides the Board with numerous examples of his flawed reasoning, lack of
`
`understanding of proper patent scope, lack of any identifiable methodology to support
`
`his opinions, and complete failure to disclose any facts or data to support his opinions.
`
`See Response, 18-30. In an attempt to rehabilitate Dr. Stone, Petitioner provides a
`
`reply “declaration.” But, like his original testimony, this new testimony is equally
`
`unreliable and should be given no weight. See 37 C.F.R §42.65.
`
`
`1 Instead, the evidence provided by Patent Owner about the state of the art
`
`demonstrates the opposite, which Petitioner wants this Board to ignore. A POSITA
`
`trying to enhance parallelism and performance in high-performance computing
`
`systems would not even consider the use of FPGAs at the time of the invention. See
`
`Patent Owner Response, passim.
`
`
`
`6
`
`
`
`For example, similar to Petitioner’s initial testimony, Dr. Stone selectively
`
`chose which facts to rebut while simultaneously showing neither an understanding of
`
`Patent Owner’s theories, evidence, or data nor asking for any additional information
`
`or making any attempt to address the breadth of Patent Owner’s evidence on the state
`
`of the art. See EX2178, 8:1-10:2.
`
`Dr. Stone also had no idea that Patent Owner did not rely upon or use certain
`
`“evidence” in its Response that was in fact first introduced by Petitioner in its Reply.
`
`See EX1074, and citations thereto in Reply. Instead, Petitioner launders this new
`
`exhibit and theories in its Reply via Dr. Stone by mischaracterizing Patent Owner’s
`
`evidence while falsely claiming Patent Owner relied on this exhibit to make opinions
`
`and arguments outside the scope of the Response. See EX2178, 10:4-13:9 (discussing
`
`EX1074 and contents, first introduced by Petitioner in Reply, and outside the scope of
`
`the Patent Owner Response or evidence).
`
`Finally, Dr. Stone did not attempt (nor was allowed to by Petitioner) to review
`
`the totality of Patent Owner’s evidence for his reply testimony, see EX2178 16:25-
`
`17:18, which is evidence contradicting his other opinions. Compare EX1076¶¶15-17,
`
`with EX1078 63:13-66:22 (Dr. El-Ghazawi explaining the irrelevance of the Gokhale
`
`article cited in the Halverson reference and how Halverson teaches how the Sun
`
`workstation in Splash2 shows no anticipation).
`
`
`
`7
`
`
`
`In contrast, Patent Owner’s experts and fact witnesses not only provided detailed
`
`analysis of their opinions and testimony, but also provided the Board with underlying
`
`facts and data that support their testimony. Ironically, it is this detailed analysis and
`
`evidence Patent Owner provided that Petitioner now complains of as “excessive” as
`
`discussed above.
`
`B. Petitioner’s Reply and New Testimony is Outside the Scope of Patent
`Owner’s Response or Should Have Been Provided in His Original
`Declaration.
`Petitioner’s Reply and Dr. Stone’s Reply testimony, EX1076, along with at
`
`least EX1074, EX1077, and EX1079, raises new issues and introduces new evidence
`
`and should therefore be excluded pursuant to 37 C.F.R. §42.23(b). (“A reply may
`
`only respond to arguments raised in the corresponding opposition or patent owner
`
`response.”)
`
`The Trial Practice Guide explains:
`
`[A] reply that raises a new issue or belatedly presents evidence will not be
`considered and may be returned. The Board will not attempt to sort proper
`from improper portions of the reply.
`
`Office Patent Trial Practice Guide, 77 Fed. Reg. 48,756, 48,767 (Aug. 14, 2012); see
`
`also Blackberry Corp. v. Mobilemedia Ideas, LLC, Case No. IPR2013-00036 (JYC)
`
`(Jul. 26, 2013) (Paper 40).
`
`Moreover, Petitioner improperly relies in its Reply at 43-44 and fn. 13 on
`
`reply testimony that could have been solicited during re-direct examination during
`
`
`
`8
`
`
`
`Dr. Stone’s initial depositions. Compare EX1076¶27 with EX2178, 6:22-7:22.
`
`Petitioner cannot remedy a fatal flaw in its Petition or testimony through new
`
`evidence and arguments in a reply brief or declaration, which would violate the
`
`Administrative Procedure Act. See In re Nuvasive, Inc., 841 F.3d 966, 973 (Fed.
`
`Cir. 2016) (consideration of a new argument in reply violated the APA); see also 35
`
`U.S.C. §312(a)(3); 37 C.F.R. §§42.22(a), 42.23. As discussed with more specificity
`
`below, Petitioner improperly uses reply testimony preventing Patent Owner from
`
`providing any contrary evidence. See also In re IPR Licensing, Inc., 942 F.3d 1363,
`
`1369 (Fed. Cir. 2019) (“To hold otherwise would allow the Board’s final written
`
`decision to rest on arguments that a patent owner has no ability to rebut or
`
`anticipate…”); Hulu, LLC v. Sound View Innovations, Case No. IPR2018-01039
`
`(Dec. 20, 2019) (Paper 29 at 14) (citing 37 C.F.R. §42.23, “A reply may only respond
`
`to arguments raised in the corresponding opposition, patent owner preliminary
`
`response, or patent owner response.”).
`
`III. PATENT OWNER’S CLAIM CONSTRUCTIONS
`As noted above, Petitioner’s objective in its Reply, and to support its flawed
`
`theories in its Petition, is to divorce patent scope from both the intrinsic and extrinsic
`
`evidence Patent Owner provided, which help define key terms a POSITA would
`
`understand. But, as will be discussed in greater detail below, once Petitioner’s
`
`arguments are dissected, it is clear that Patent Owner’s constructions properly define
`
`
`
`9
`
`
`
`claim scope, and the references advanced by Petitioner are neither anticipating nor
`
`show obvious combinations. See Response, passim.
`
`A. “stream communication”
`Petitioner’s Reply builds upon the flaws in its Petition. First, contrary to
`
`Petitioner’s Reply, the Board declined to construe the term “stream communication”
`
`in its institution decision. Reply, 3. In failing to recognize this, Petitioner failed to
`
`articulate an analysis of why its proposed construction is correct, instead focusing
`
`almost exclusively on attacking Patent Owner’s construction.
`
`Second, Petitioner’s Reply relies on semantic games about adjectives—
`
`claiming that the phrase “stream communication connection” precludes “stream
`
`communication” from being a noun. But, Petitioner points to no such requirement in
`
`law or in English grammar. Instead, “stream communication” defines what type of
`
`connection is being instantiated in dependent Claim 15, and this use of a noun to further
`
`specify the type of another noun is a common usage in English.2
`
`Third, Petitioner’s Reply ignores the plain claim language that states the “stream
`
`communication” type of connection in dependent Claim 15 further specifies the
`
`structure that is instantiated in independent Claim 1. Petitioner’s construction does
`
`
`2 For example, the noun “door” in “door handle” defines a specific type of handle as
`
`opposed to a “drawer handle” (also a noun) or “bicycle handle” (also a noun).
`
`
`
`10
`
`
`
`not specify any structure or configuration thereof, and in Reply, Petitioner is silent on
`
`this flaw. Response, 51-52; Reply, 30-34.
`
`Fourth, and in stark contrast to the plain language of the claims, Petitioner’s
`
`Reply continues to ignore the canon of claim construction that a dependent claim scope
`
`should be differentiated from its independent claim. Petitioner continues to construe
`
`the phrase to be a connection for “communication of a data sequence,” which is no
`
`different than the systolic or data driven requirements of independent claim 1. See
`
`Response, 51; Reply, 30-34.
`
`Fifth, Petitioner’s Reply asks the Board to improperly ignore numerous citations
`
`demonstrating consistency between the intrinsic evidence and a POSITA’s
`
`understanding of the phrase “stream communication” based upon extrinsic evidence.
`
`This is improper when the intrinsic record is silent on the specific meaning of a
`
`technical term. See Intel Corp. v. VIA Techs., Inc., 319 F.3d 1357, 1367 (Fed. Cir.
`
`2003) (“When an analysis of intrinsic evidence resolves any ambiguity in a disputed
`
`claim term, it is improper to rely on extrinsic evidence to contradict the meaning so
`
`ascertained.”). Petitioner attempts to argue “extrinsic prior art cannot be used to
`
`contradict the disclosure of the ’800 Patent,” (Reply, 32) which is only true if: (i) the
`
`intrinsic record is clear and unambiguous and (ii) the extrinsic evidence actually
`
`contradicts the intrinsic record. Petitioner presents no coherent arguments, authority,
`
`or evidence to support either.
`
`
`
`11
`
`
`
`The intrinsic record does not unambiguously define “stream communication.”
`
`In fact, Petitioner’s own Petition conceded this: “The term ‘stream communication’ is
`
`not used in the ’800 Patent except in its claims, nor is it used in the incorporated
`
`references.” Petition, 20. The only allegedly supporting phrase referenced by
`
`Petitioner is different (stream of operands), which at best clarifies what is in the stream
`
`(operands), not the communication connection that must be instantiated (structure). It
`
`is also from a different patent referring to a different invention pertaining to internet
`
`communications. Because of the intrinsic record’s silence in defining a stream
`
`communication connection, Patent Owner provided numerous citations to references
`
`around the time of the invention that all consistently provide support for a POSITA’s
`
`understanding of the associated structure, consistent with Patent Owner’s proposed
`
`construction. Response, 53-62 citing:
`
` U.S. Patent No. 8,589,666 (assigned to Patent Owner and describing the prior
`
`art understanding of stream as a data path as shown in fig.1)
`
` Patent Owner’s own product documentation3 and supporting declaration of the
`
`inventor of the ’800 Patent
`
` Argonne National Laboratory article, defining stream as a data structure
`
`
`3 Petitioner disputes the public availability date, however the reference illustrates
`
`that the concept of a stream was already known in the prior art to a POSITA.
`
`
`
`12
`
`
`
` U.S. Patent No. 5,748,613
`
` European Patent No. 1820309
`
` U.S. Patent No. 8,543,746
`
` U.S. Patent No. 8,352,4564
`
` U.S. Appl. No. 2010/0070730
`
`The intrinsic record also is consistent with this extrinsic evidence, which
`
`Petitioner fails to dispute. At best, Petitioner attempts to argue a statement by the
`
`examiner regarding a “stream between processors” does not mention a queue. But
`
`again, this refers to the contents (“data is transferred systolically in at least one
`
`stream”), not to the structure of the stream communication connection that must be
`
`instantiated. Reply, 34. Even Petitioner’s own brief concedes “the examiner
`
`noted…the claimed ‘stream communication connection’ [is] established as the
`
`interconnections are made,” but the examiner provides no statements on what specific
`
`
`4 Petitioner engages in hyperbole to argue it is “illegal” to refer to prior art that post-
`
`date a patent’s filing date. Notwithstanding the obvious lack of an applicable
`
`criminal statute, courts specifically contemplate the use of such prior art to
`
`understand the state of the art and how a POSITA would interpret the patent
`
`disclosures. See, e.g., Beacon Adhesives, Inc. v. U.S., 134 Fed. Cl. 26, 34 (Fed Cl.
`
`2017).
`
`
`
`13
`
`
`
`structure is established. Reply, 34; see also Response, 66-67. Contrary to Petitioner’s
`
`assertion, the examiner’s silence cannot “contradict” the extrinsic evidence; it is by
`
`definition silent one way or the other. If anything, the intrinsic record demonstrates
`
`consistency with the extrinsic evidence. Response, passim.
`
`Finally, Petitioner’s Reply fails to address the inconsistency between the
`
`unreasonable broadness of its construction and its own expert’s testimony restricting
`
`the scope of the claimed invention. Specifically, Petitioner’s own expert agreed that
`
`the independent claims (whose limitations necessarily apply to the dependent claims
`
`on stream communication connection) exclude storing values in memory by sending
`
`them off chip. Response, 69. In Reply, Petitioner curiously disputes this point by
`
`arguing its expert was not referring to the dependent claim limitation, but provides no
`
`analysis or legal authority in support of why a limitation in an independent claim
`
`should not apply to the dependent claims. This highlights the unreasonable broadness
`
`in Petitioner’s proposed construction that it seeks to preserve such an interpretation of
`
`the claims.
`
`Neither Splash2 nor any of the other prior art disclose stream communication as
`
`properly construed, and Microsoft does not dispute this. See Reply, 37-38.
`
`Accordingly, Microsoft failed to meet its burden of proof to invalidate at least
`
`dependent claims 15, 22, and 28.
`
`
`
`14
`
`
`
`B. “computational loop” [wherein only functional units needed to solve the
`calculation are instantiated]
`1. Computational Loops Are Not Infinite Loops
`Petitioner’s Reply highlights the flaws in its Petition and the importance of
`
`clarifying the Board’s construction of computational loop in light of those flaws.
`
`First, Petitioner falsely claims “DirectStream’s expert, Dr. Homayoun, never
`
`offers an interpretation of ‘computational loop’… [and] DirectStream could not
`
`convince him to support its position.” Reply, 36-37. This is false. EX2112¶207 (Dr.
`
`Homayoun opining that “A computational loop evaluates each piece of data multiple
`
`times, ‘a fixed number of times or until some condition is true or false,’” and more
`
`importantly, opining throughout this section how Petitioner’s flawed view of the claim
`
`term results in a flawed invalidity analysis).
`
`Second, Petitioner incorrectly claims Patent Owner does not identify any
`
`supporting intrinsic evidence. Reply, 35. Patent Owner cites to numerous figures and
`
`passages in the patent in support, none of which Petitioner refutes or addresses. See
`
`Response, 71 and 84. Specifically, Fig.4B of the patent (which patentee used to
`
`distinguish from the prior art depicted in Fig.4A) depicts two sets of loops: the red
`
`loops that form Loop A and Loop B, and the blue program loop that repeats the
`
`execution of both in each “phase.” Specifically, the patent states that in “Phase 1,”
`
`both loops are active with Loop A working on dimension 1 of the data, and Loop B
`
`working on dimension 0 of the data. Then in the next phase (“Phase 2”), both loops
`
`
`
`15
`
`
`
`are again active with Loop A working on dimension 2 of the data, and Loop B working
`
`on dimension 1 of the data.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Thus, the ’800 Patent unambiguously describes the two loops A and B operating a
`
`number of times (looping) on each dimension of data.
`
`In contrast, Petitioner’s Reply reiterates its flawed interpretation that the red
`
`computational loops need only “execute[] instructions on one piece of data, and then
`
`execute[] those very same instructions on a next piece of data.” Reply, 35. This
`
`argument effectively deletes the red loops as follows:
`
`
`
`
`
`16
`
`
`
`From 1 to N
`
`CODE A
`
`CODE B
`
`CODE A
`CODE B
`
`CODE A
`CODE B
`
`
`
`Only the blue program loop would be needed to cycle through all of the datum and
`
`execute the code once per datum. In essence, Petitioner attempts to improperly rewrite
`
`the figures and specification using the Board’s claim construction. Additionally,
`
`Petitioner’s arguments contradict and would exclude from the claims the ’800 Patent’s
`
`embodiments and figures. Petitioner’s Reply highlights the need to clarify the Board’s
`
`construction to avoid Petitioner’s unreasonable construction of the claims under either
`
`Phillips or BRI.
`
`2. Looping controlled by the Sun workstation
`Petitioner and Dr. Stone concede in reply that the only basis for claiming
`
`Splash2 discloses looping is relying on the infinite “loop-endloop” in Figs. 8.7 and
`17
`
`
`
`
`
`8.12 to allegedly compare of each of the genetic sequences of datum. But Microsoft’s
`
`Petition does not explain how these sequences arrive, in Splash2, even though
`
`Microsoft and its expert acknowledge the sequences must be “streamed through the
`
`array.” Petition, 33; see also EX1003¶134 (“More specifically, for
`
`this
`
`implementation… two genetic sequences are shifted in opposite directions through
`
`multiple processing elements of the Splash2 system…. The source and target
`
`sequences enter the array on opposite ends. Microsoft does not dispute that Splash2
`
`requires a host Sparc computer. Petition, 30. This workstation controls the sequences
`
`of data sent into Splash2 is also consistent with the disclosures in Halverson. See
`
`EX2169 at 14-15; EX2166¶¶42-43.
`
`Even Petitioner and its expert’s attempt to distinguish this argument by
`
`reinterpreting Gokhale is unavailing. Despite Dr. Stone professing without any
`
`support or analysis that Splash2 is not a SIMD structure, Splash2 clearly states it
`
`operates as a SIMD structure, which stands for single instruction, multiple data.
`
`EX1007, 125 (“The Splash2 system supports several models of computation, including
`
`PEs executing a single instruction on multiple data (SIMD mode) and PEs executing
`
`multiple instructions on multiple data (MIMD mode).”)
`
`Tellingly, neither Dr. Stone nor Petitioner claim that there are either multiple
`
`instructions or only a single data being processed in Splash2. In fact, Dr. Stone and
`
`Petitioner concede that there is only one “code” being run by any of the processing
`
`
`
`18
`
`
`
`elements—Figs.8.7 and 8.12, depending on if it is a unidirectional or bidirectional
`
`example. EX1003¶135 (discussing the bidirectional example, Dr. Stone states
`
`“Splash2 also discloses the code executed by each processing element in the
`
`system…” and then pasting an image of the code in Fig.8.7); EX1003¶139 (discussing
`
`the unidirectional example, Dr. Stone states “Splash2… also discloses the code
`
`executed by each processing element...” and then pasting an image of the code in
`
`Fig.8.12); Petition, 31 (“…the code description of Figs. 8.7 and 8.12 into an FPGA
`
`object module describing the hardware configuration of each PE in the array…”)
`
`Thus, there can only be a single instruction set. And, Petitioner and Microsoft need
`
`there to be multiple data, otherwise their entire basis for claiming there is “looping”
`
`would fall apart. See Reply, 35 (“A Skilled Artisan would therefore understand that
`
`each processing element executes its loop instructions until it ‘computes the distances
`
`along a particular diagonal matrix.’”); see also EX1076¶12.
`
`As demonstrated above, this would also be a nonsensical interpretation of
`
`Fig.4B, having the blue program loop cycle from 1 to n and each computational loop
`
`also cycl