US 6480876

I'll search for authoritative information on this patent number, including any litigation or CAFC activity.

Let me run a couple more targeted searches to check for any litigation or CAFC activity specific to this patent.

US Patent 6,480,876 — Summary

Sourcing note: I searched for the literal number 6480876 / 6,480,876. I found no USPTO litigation docket or CAFC 2026 docket entry naming this patent. What follows is drawn from the Google Patents full text (which mirrors the USPTO record) and corroborating bibliographic sources. Where I could not verify something authoritatively, I say so.

Bibliographic data (verified)

Note: the inventors were at Digital Equipment Corporation's Cambridge Research Laboratory, and the work corresponds to the "Stampede"/Space-Time Memory (STM) project. The related prior-art patent U.S. 6,067,604 ("Spacetime Memory") is by an overlapping team (Ramachandran, Halstead, Joerg, Kontothanassis, Nikhil, Rehg).

Abstract (as issued)

"A system for integrating task and data parallelism in a dynamic application that includes at least one task for processing an input data stream to produce an output data stream replaces the at least one task with the following components. A splitter task for partitioning the input data stream into a plurality of data chunks. A plurality of worker tasks for processing subsets of the data chunks, each worker task being an instance of the at least one task, and a joiner task combining the processed data chunks to produce the output data stream."

The invention in plain language

The patent addresses real-time, dynamically varying applications — the worked example is a video "color tracker" inside an interactive kiosk that tracks multiple customers by clothing color. The core idea: take any single task in a task-parallel dataflow graph and replace it with a sub-graph consisting of:

• a splitter that cuts the incoming stream into data "chunks" (e.g., image quadrants, or whole frames),
• worker tasks (instances of the original task) that each process some chunks, and
• a joiner that reassembles the partial results into the original single output stream.

The splitter/joiner pair hides the internal data parallelism from the rest of the graph, so the substitution is modular. Two variants are described: a static version (FIG. 2; fixed chunk-to-worker assignment) and a dynamic version (FIG. 3), which is the claimed subject matter. In the dynamic version chunks go into a shared work queue any idle worker can pull from; because the number of chunks M need not equal the number of workers N (and both can change at runtime), the splitter writes a control item per data item into a controller queue telling the joiner how to reassemble results; chunks are tagged to done ports that act as a sorting network. The patent also describes a machine-learning controller (Part II) that measures application/system state, builds forward/inverse models via memory-based locally-weighted regression, and selects the task/data parallel strategy (FIGs. 5–8).

Independent claims

There are two independent claims — claim 1 (system) and claim 14 (method). Claims 2–13 depend from claim 1.

Claim 1 (system): A system for integrating task and data parallelism for a dynamic application having at least one task for real-time processing of a continuous input data stream to produce an output data stream, comprising:

1. a splitter task partitioning the input stream into a plurality of individually marked data chunks;
2. a work queue storing the data chunks;
3. a variable number of worker tasks processing subsets of the chunks from the work queue, each worker task being an instance of the original task;
4. a controller queue storing a control item for each data chunk, the control item including task and data strategy provided by the splitter task; and
5. a joiner task combining the processed chunks dependent on the control items on the controller queue to produce the output stream.

In plain terms: the dynamic (not static) architecture of FIG. 3, where the splitter tells the joiner, via a separate control queue, how each data item was split so the joiner can reassemble it — allowing the strategy and worker count to change at runtime.

Claim 14 (method): The method counterpart, with the same steps: splitting the stream into individually marked chunks via a splitter; storing chunks in a work queue; processing subsets via a variable number of worker tasks each being an instance of the task; storing a control item per chunk in a controller queue (including the task/data strategy provided by the splitter); and combining results by a joiner dependent on the control items.

Notable dependent claims (context, not independent): claim 2 (chunks are continuous, time-varying data); claim 3 (time-stamp ordering); claim 4 (streams correspond to spatially ordered memory ports); claim 6 (input stream is a sequence of time-ordered video frames); claim 8 (means for dynamically varying the number of workers); claims 9–11 (associating a task/data parallel strategy with each chunk, including processing methods such as the worker tasks); claims 12–13 (learning and applying the strategy via a controller generating an "action space" and a run-time system applying it).

Citation significance

The patent is cited as prior art in the European search report for EP 1 852 777 A3 ("Computer-aided parallelizing of computation graphs," Ab Initio), where US 6,480,876 B2 is rated X (particularly relevant) against claims 1 and 11 and cited against claims 2–10 for the subject matter of its abstract, FIG. 3, and col. 4 line 66 – col. 9 line 4.

Explicit uncertainty / disambiguation warning

• No litigation found. My searches returned no district-court or CAFC 2026 docket naming 6,480,876. I did not obtain a definitive negative from a docket database, so I cannot state with certainty that no such proceeding exists — only that nothing surfaced.
• Do not confuse with "the '087 patent." Search results repeatedly surfaced references to "the '087 patent" — e.g., U.S. 7,756,087 in Netlist v. Samsung Electronics (E.D. Tex. 25-cv-00557/00558, ITC action, and a related PGR), and a different AIA " '087 patent" in BlephEx v. Pain Point Medical. Those are different patents from 6,480,876 and are not related to this summary. The correct informal shorthand for the patent here would be "the '876 patent."
• I could not directly query USPTO Patent Center/PAIR or the CAFC Docket in this session; the bibliographic fields above are corroborated by multiple independent sources (Google Patents, EPO family data, and inventor CV listings citing "U.S. Patent 6,480,876. November 12, 2002"), and I have reasonable confidence in them.

I'll research litigation involving US patent 6480876 using multiple sources.

Let me search patent litigation databases specifically.

Litigation Search — U.S. Patent No. 6,480,876

Bottom line

I found no known litigation involving U.S. Patent No. 6,480,876.

Across the sources searched, no district court action, ITC investigation, PTAB/post-grant proceeding, or CAFC appeal asserting or challenging this patent was identified. I want to be explicit, however, about the limits of that negative finding (see "Confidence and limitations" below).

The patent itself (for context)

Source: Google Patents, https://patents.google.com/patent/[US6480876](/patent/US6480876)/en

What the searches did surface (and why it is not litigation)

1. AB Initio EP prosecution citation. US 6,480,876 is cited as prior art ("X" category, claims 2–10; and against claims 1, 11) in the European search report for EP 1 852 777 A3 (AB Initio Software Corporation, "Computer-aided parallelizing of computation graphs"). This is an examiner's prior-art citation in a European prosecution file — not litigation involving the '876 patent.

   • Source: https://patentimages.storage.googleapis.com/41/87/23/76a56f32989808/EP1852777A3.pdf

2. Forward citations only. Google Patents lists "Cited By (72)" — these are later patents that cite the '876 patent as references. A citation relationship is not a lawsuit. Notably, several of those citing documents belong to AB Initio Technology LLC and other parties, but that reflects the technology's influence, not contentious proceedings over the '876 patent.

3. No hit in litigation-specific sources. Searches directed at the Unified Patents litigation portal, CourtListener/Coast Guard dockets, and general web for the exact number "6480876" (and formatted "6,480,876") returned no complaint, case caption, or docket naming this patent. Results that did come back for six-digit strings were either unrelated product numbers (e.g., an OBO cable-tray part no. 6480876) or unrelated patents (the Omega Patents v. CalAmp "'876 patent" is U.S. Patent 6,591,876 — a completely different patent, and I am not treating it as the patent in question).

Why this outcome is consistent with the patent's profile

The patent is a 1998-filed, Hewlett-Packard/Compaq-era systems patent covering an internal software construct (splitter/worker/joiner with work and control queues). It expired for failure to pay maintenance fees / end of term in 2018. The assignee has been a large operating company (Compaq → HP → HPE), not an NPE asserting the patent. Patents of this type and vintage that never passed into assertion-oriented hands commonly show zero litigation, which matches what the searches show here.

Confidence and limitations — read before relying on this

• A "no results" search is not proof of absence. The authoritative litigation sources for a definitive negative — PACER (federal dockets), Docket Navigator, Lex Machina, MaxVal, RPX/Unified Patents' full litigation database, and the ITC EDIS docket — are largely paywalled or require a login I could not authenticate to in this session. My searches were web-accessible only.
• I could not verify the Unified Patents caselist or PACER directly for this specific number; the searches did not return a matching docket entry.
• Because the patent expired in 2018, any litigation would necessarily pre-date that (or, if filed later, would concern past damages). Statistically, older expired operating-company patents are less likely to appear in assertion campaigns, but that is an inference, not a record.
• Do not treat this as a freedom-to-operate or clearance opinion. If you need a litigation-clean determination (e.g., for a diligence, valuation, or estoppel/Section 315(b) analysis), you should run the exact number through a paid docket database and confirm against PACER.

Recommended verification steps for a definitive answer

1. PACER / Docket Navigator: query "6,480,876" and "6480876" as patent-in-suit and as a patent challenged in any IPR/PGR/CBM.
2. Unified Patents litigation portal (https://portal.unifiedpatents.com/litigation/caselist) — search by patent number, both as asserted patent and as IPR subject.
3. ITC EDIS — search for the patent number in Section 337 investigations.
4. CAFC docket search — confirm no appeal references this patent number.
5. USPTO Patent Center "Litigation" / assignment tabs for any recorded judgment or license affecting the patent.

Summary: Based on available public, web-accessible sources, no litigation is known involving U.S. Patent No. 6,480,876. The only relevance found is as prior art cited in an unrelated European patent prosecution (EP 1 852 777 A3). I recommend the paid-docket verification above if a formal negative is required.

I'll start by verifying the structured data with targeted web searches for any PTAB activity on this patent.

Let me run additional targeted searches to be thorough before concluding.

Proceedings overview

Zero AIA trial proceedings are on file for US 6,480,876. The structured "PTAB proceedings on file" block (USPTO Open Data Portal) returns an empty list, and my independent web searches surfaced no IPR, PGR, or CBM naming this patent — so the breakdown is 0 active / 0 claims invalidated / 0 claims sustained / 0 settled / 0 institution-denied. Defensive posture: the patent was never challenged at the PTAB, but that is not evidence of strength — this is a 1998-filed, never-litigated patent that expired on 2018-05-28, so there is no IPR history to attack and no threat left to defend against.

Important caveat on my searches. The one apparently relevant hit — NVIDIA Corp. v. Neural AI, LLC, IPR2025-00606 (Paper 18, 2025-07-31), denying institution under § 314(a) — concerns U.S. 8,648,867 B2, a different patent. It is a number-transposition false positive and has nothing to do with 6,480,876. The other hits (EP 1 852 777 A3, WO 2013/061159 A3, US 7,003,630, US 9,733,978, US 8,484,159) are search-report or prior-art citations to the '876 patent, not PTAB challenges of it. I did not find any proceeding number to report, and I will not invent one.

No proceedings to itemize

Because the list is empty, there are no per-proceeding sections to populate — no petitioner names, no panels, no FWDs, no appeals. What I can give you are the facts that explain why the list is empty and what they mean.

---

Strategic summary

Claim status: all 14 claims are UNTESTED and, more importantly, UNENFORCEABLE going forward. No claim of 6,480,876 has ever been canceled, narrowed, or confirmed by the PTAB — there is no FWD to point to. The patent's legal status on Google Patents and the ODP record is "Expired – Lifetime," with anticipated expiration 2018-05-28 (the 20-year term from the 1998-05-28 filing). Claim 1 is the sole independent system claim (splitter / work queue / variable number of worker tasks / controller queue / joiner); claim 14 is the sole independent method claim; claims 2–13 depend from claim 1. All expired with the term — United States patents cannot be extended past the statutory term absent a specific term extension or adjustment, and none is reflected here.

Estoppel landscape: none exists. § 315(e)(2) estoppel only attaches to a petitioner (and its privies/real parties in interest) that instituted an IPR/PGR. With zero instituted trials on this patent, no party is statutorily estopped from raising any § 102/§ 103 ground in district court or the ITC. That said, the estoppel question is moot in the enforcement direction: an expired patent cannot support prospective infringement relief, and damages exposure is limited to past acts within the § 286 six-year lookback — i.e., infringement ending on or before 2018-05-28. Combined with the fact that I find no record of the patent ever being asserted, there is no realistic live demand to which estoppel would matter. For completeness: the only vehicle ever available was IPR (the patent issued in 2002, long before the 2012-09-16 start of AIA trials — PGR was never time-eligible, and this is a multiprocessing/computer-architecture patent classified under G06F 9/5066, not a "covered business method," so CBM was unavailable).

Pattern signals: absent, and that itself is the story. There is no repeat petitioner (no petitioner at all), the patent owner never pursued a PTAB appeal (nothing to appeal), and there is no defensive aggregator (Unified Patents or similar) in the chain. The chain of title runs purely through corporate M&A: Digital Equipment Corp. → Compaq Computer Corp. → Compaq Information Technologies Group, L.P. (2002-01-08) → Hewlett-Packard Development Co. (2007-06-06) → Hewlett Packard Enterprise Development LP (2015-11-09), the current assignee. This is a corporate-portfolio asset from the DEC/Compaq research era (inventors Rehg, Knobe, Nikhil, Ramachandran — the Space-Time Memory line of work), not an NPE assertion vehicle.

---

Recommended next steps

1. If you are a defendant and have received a demand citing 6,480,876, verify the demand's premise before doing anything else. The patent expired on 2018-05-28. Ask the asserting party for the accused acts and their dates. Any theory must be confined to pre-2018-05-28 conduct within the six-year window, and no injunction is available. Put that in writing early.

2. There is no FWD to invoke — I could not find one, and the ODP confirms none exists. Do not let anyone assert that "claims were invalidated in IPR," because that never happened. The clean defense is expiration, not PTAB estoppel.

3. Do not file an IPR expecting a § 315(e) payoff. A prevailing petitioner gains no litigation estoppel leverage here because the patent is already expired; an IPR would be an expensive way to obtain an unenforceability footnote. IPRs on expired patents remain permissible, but the cost-benefit is poor unless your adversary is still trying to collect past damages.

4. If you need to confirm the negative, pull the primary sources yourself:

   • Publication record / status: https://patents.google.com/patent/US6480876/en
   • PTAB trial search (search patents 6,480,876 / 6480876): https://ptacts.uspto.gov/ (PTAB E2E / PTAB Center)
   • Patent file wrapper and assignment history: https://patentcenter.uspto.gov/ (application 09/085,795)
   • Federal Circuit dockets / opinions (for any appeal — I found none): https://cafc.uscourts.gov and https://www.courtlistener.com

5. Flag any newly-filed proceeding you find — the ODP ingest may lag. But on the current record, and with the patent expired for over eight years, the probability of a new AIA trial is effectively nil. Absence of PTAB activity here is not a hardening signal; it reflects an old, expired, unasserted patent.

Bottom line for a defendant: the '876 patent has no PTAB history because it was never worth challenging — it died of old age. The correct posture is expiration-based mootness, not an IPR-based invalidity defense.

I'll research the USPTO assignment records for this patent. Let me run several targeted searches.

Let me dig deeper for the reel/frame records and correspondents of record.

Let me try to locate the actual reel/frame records for this specific patent.

Ownership & Assignment Analysis — US Patent 6,480,876

Source caveats, stated up front (read before relying on reel/frame):

• I could not query the USPTO Assignment Center / Patent Center directly in this session (no authenticated access), so I could not retrieve the reel/frame numbers or the recorded correspondents of record for US 6,480,876 specifically. Instead I relied on the Google Patents legal-events mirror of the USPTO assignment record (which gives conveyance type, parties, and event dates but not reel/frame), plus corroborating bibliographic sources.
• Where I supply reel/frame or correspondent names below, they come from sibling patents in the same Compaq/HP corporate family (e.g. the Lubbers/Sicola HP storage family litigated in IPR2025‑00414). Those are illustrative of HP's recording practice and are NOT verified as the records for this patent. I flag each such use explicitly rather than presenting it as this patent's record.
• Do not treat any reel/frame below as confirmed for US 6,480,876. The confirmed finding here is that the chain is a pure operating-company corporate-succession chain; the exact reel/frame citations still need to be pulled from Assignment Center to close the file.

---

Inventors

Corroboration: the companion "Space-Time Memory" paper lists the team's affiliation as "Compaq Computer Corporation, Cambridge Research Lab, One Kendall Square, Bldg 700, Cambridge, MA 02139" with emails at @crl.dec.com, and separately footnotes Ramachandran's "Current Affiliation: College of Computing, Georgia Institute of Technology." Rehg's CV confirms "Member of Technical Staff / Project Leader, Cambridge Research Laboratory, Compaq Computer Corporation, 1995–2001."

Unusual-pattern check — NOT present:

• No mass inventor departure. Rehg (1995–2001), Knobe and Nikhil (long DEC/Compaq CRL tenure) stayed through the Compaq era; there is no evidence all inventors left within 12 months of filing, so the "pre-fire-sale turnover" tell does not fire.
• One wrinkle worth noting: there are two founder-side assignments — one recorded 1998‑07‑23 naming all four inventors, and a second recorded 1998‑08‑27 naming only Knobe, Nikhil, and Rehg. Ramachandran's absence from the second is consistent with his Georgia Tech relocation (he signed on a separate instrument). This is a routine split-signature artifact, not an indicator of bad title, but it is the only departure from a clean four-inventor→DEC assignment.

---

Original assignee

Entity named on the issued patent: Compaq Information Technologies Group, L.P. (Houston, TX) — i.e., by issue date (2002‑11‑12) the patent had already moved off the original DEC assignee onto the post-merger Compaq holding entity. The original assignee of record at execution was Digital Equipment Corporation (Cambridge, MA).

• Primary line of business: Digital Equipment Corp. was a minicomputer/systems and research house (the invention came out of DEC's Cambridge Research Laboratory, the "Stampede"/Space-Time Memory project). Compaq was a PC/enterprise-systems manufacturer.
• Product embodiment: The claimed splitter/worker/joiner construct was implemented in the Stampede programming system and used to run the "Smart Kiosk" interactive multimedia application on a cluster of Alpha SMPs — so the technology was reduced to a working research system, not merely paper. (Whether it shipped in a commercial Compaq/HP product practicing the claims is unverified.)
• Current status (both operating, not dissolved): DEC was acquired by Compaq (announced Jan 1998; closed June 1998) — not a bankruptcy. Compaq was acquired by Hewlett-Packard (closed May 2002). HP later split (Nov 2015) into HP Inc. and Hewlett Packard Enterprise, which is the current owner of record. No Chapter 7/11 anywhere in the chain.

---

Assignment timeline

Bold dates below are the event/record dates shown in Google Patents' legal-events mirror of the USPTO assignment record. For several links the mirror date is the recordation date rather than the true execution date (this is common in bulk corporate recordings), so treat the dates as ± and confirm execution dates in Assignment Center. Reel/frame: not retrievable for this patent in this session (see caveats above).

• 1998-07-23 / recorded 1998-07-23 — Reel NNNNNN/NNNN (not retrieved)

  • Conveyance: Assignment of Assignors' Interest
  • Assignor: Ramachandran, Umakishore; Knobe, Kathleen; Nikhil, Rishiyur S.; Rehg, James Mathew (all four)
  • Assignee: Digital Equipment Corporation
  • Correspondent: not retrieved (DEC-era recording; DEC's in-house IP group — no correspondent surfaced in my sources)
  • Context: Initial founder→employer assignment (routine; not a sale).

• 1998-08-27 / recorded 1998-08-27 — Reel NNNNNN/NNNN (not retrieved)

  • Conveyance: Assignment of Assignors' Interest
  • Assignor: Knobe, Kathleen; Nikhil, Rishiyur S.; Rehg, James Mathew (three of four)
  • Assignee: Digital Equipment Corporation
  • Correspondent: not retrieved
  • Context: Second founder→employer assignment covering the three CRL inventors (Ramachandran signed separately) — an internal title-clean-up, not a transfer to a third party.

• 2001-05-21 / recorded 2001-05-21 — Reel NNNNNN/NNNN (not retrieved)

  • Conveyance: Merger
  • Assignor: Digital Equipment Corporation
  • Assignee: Compaq Computer Corporation
  • Correspondent: not retrieved
  • Context: Corporate merger succession (DEC folded into Compaq). Note: the underlying DEC/Compaq acquisition closed in June 1998; a 2001 recording date reflects a later bulk recordation of the merger chain, not a 2001 economic event.

• 2002-01-08 / recorded 2002-01-08 — Reel NNNNNN/NNNN (not retrieved)

  • Conveyance: Assignment of Assignors' Interest
  • Assignor: Compaq Computer Corporation
  • Assignee: Compaq Information Technologies Group, L.P.
  • Correspondent: not retrieved. (On sibling Compaq-era recordings the correspondent of record was commonly Hogan & Hartson LLP, William J. Kubida, One Tabor Center, 1200 17th Street, Suite 1500, Denver, CO 80202 — e.g. the Lubbers/Compaq ITG recording dated 2002‑06‑20. This is not verified for the '876 record.)
  • Context: Internal corporate reorg — assignment into Compaq's IP-holding limited partnership (the entity that appears on the face of the issued patent).

• 2007-06-06 / recorded 2007-06-06 — Reel NNNNNN/NNNN (not retrieved)

  • Conveyance: Change of Name
  • Assignor: Compaq Information Technologies Group, L.P.
  • Assignee: Hewlett-Packard Development Company, L.P.
  • Correspondent: not retrieved. *(HP-side recordings on sibling patents used the correspondent "Hewlett-Packard Company, Records Manager, Intellectual Property Administration, P.O. Box 272400, Fort Collins, CO 80527-2400", and individual recorders such as Valerie Bakes. Not verified for this patent.)*
  • Context: Change of name only (no change of beneficial ownership). ⚠️ Cross-reference flag: HP's corporate name change Compaq-ITG→HP-Development was effective ~Oct 1, 2002 and was recorded on sibling patents at Reel 014628/0103 on 2004‑05‑12 (execution 2002‑10‑01). The Google Patents mirror shows 2007‑06‑06 for this patent — possibly a later/duplicate bulk recording. This date discrepancy is unresolved and should be confirmed in Assignment Center.

• 2015-11-09 / recorded 2015-11-09 — Reel NNNNNN/NNNN (not retrieved)

  • Conveyance: Assignment of Assignor's Interest
  • Assignor: Hewlett-Packard Development Company, L.P.
  • Assignee: Hewlett Packard Enterprise Development LP
  • Correspondent: not retrieved. *(The analogous HP→HPE bulk spin-off recording on sibling HP patents is Reel 037079, Frame 0001 — e.g. Apple v. Uniloc, IPR2018‑00884, Ex. 1002. This bulk assignment very likely sweeps in US 6,480,876, but I could not confirm that this patent is listed on it.)*
  • Context: Internal spin-off / reorg — HP's November 2015 separation of Hewlett Packard Enterprise from HP Inc.; the enterprise-side patents were conveyed to HPE Development LP.

No later assignments recorded. Legal status is Expired – Lifetime (term ended, anticipated expiration 2018‑05‑28). No security agreement, license, release, or correction appears in the chain.

If Assignment Center in fact shows nothing beyond the inventor→DEC links for this patent number (I could not run that query), then per the operating rule the finding would be "no post-issuance assignments, original assignee still owns" — but the Google Patents mirror and the corroborating bibliographic sources (final assignee Hewlett Packard Enterprise Development LP) indicate otherwise, so I have proceeded on the six-link chain above.

---

Timeline diagram

timeline
    title Ownership of US 6480876
    1998 : Filed by Digital Equipment Corp
         : Inventors assign to DEC
    2001 : DEC merger into Compaq Computer Corp
    2002 : Assigned to Compaq ITG LP
    2007 : Name change to HP Development Co LP
    2015 : Assigned to Hewlett Packard Enterprise
    2018 : Patent term expires

---

NPE / troll-pattern signals

1. Shell-entity transfer — not present. Every assignee is a well-known operating company or its IP-holding partnership of a public corporation (DEC → Compaq → Compaq ITG LP → HP Development → HPE Development LP). No "IP/Licensing/Ventures/Holdings" single-purpose LLC; no registered-agent-service address appears on the recorded chain for this patent. Compaq Information Technologies Group, L.P. is the bona fide Compaq IP partnership (Houston, TX), not a shell.

2. Known asserter in the chain — not present. No assignee or assignor matches Acacia, Marathon, IV, IPNav, Wi‑LAN, Mosaid/Conversant, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Erich Spangenberg entities, or any Unified Patents / RPX high-frequency plaintiff. The chain is DEC/Compaq/HP only.

3. Repeat correspondent across the chain — unclear (not a finding). I could not retrieve the correspondent of record for this patent's documents. The pattern visible on sibling Compaq/HP patents is in-house corporate IP administration (HP "Records Manager, IP Administration," Fort Collins, CO) plus outside corporate-transaction counsel (e.g. Hogan & Hartson / William J. Kubida on Compaq‑era assignments), i.e. corporate-succession counsel, not NPE assertion counsel. Because a single appearance is not a finding and I could not confirm recurrence on this patent, I mark this unclear, not present.

4. Cascading transfers — not present. Transfers are spaced by years (1998, 1998, 2001, 2002, 2007, 2015) and are corporate reorganizations, not a rapid chain of unrelated LLCs sharing an address.

5. Pre-litigation transfer — not present. No infringement suit naming this patent was found (consistent with the litigation section). There is no transfer within 6 months before any suit, because there is no suit.

6. Bankruptcy fire-sale — not present. DEC was acquired by Compaq (1998), not bankrupt; Compaq was acquired by HP (2002), not bankrupt. No Chapter 7/11 sale of this asset.

7. Privateering — not present. No transfer from an operating company to an NPE asserting on its behalf. The terminal owner (HPE) is itself an operating company and a potential party, not a proxy asserter.

8. Defensive aggregator (anti-NPE) — not present. The chain does not terminate at RPX, AST, LOT, Unified Patents, or OIN. Terminal owner is an operating company.

---

Verdict

Operating-company assertion — with an explicit caveat: no assertion (no lawsuit) was found, so this is best read as operating-company corporate-succession ownership of a non-asserted patent, not as active enforcement.

Justification (2–3 sentences): The recorded chain is a textbook operating-company succession — inventors → Digital Equipment Corporation (1998‑07‑23 and 1998‑08‑27) → Compaq Computer Corporation by merger (2001‑05‑21) → Compaq Information Technologies Group, L.P. (2002‑01‑08) → Hewlett-Packard Development Company, L.P. by change of name (2007‑06‑06) → Hewlett Packard Enterprise Development LP (2015‑11‑09) — with no shell LLC, no known NPE, and no defensive aggregator anywhere in the chain, and zero litigation surfaced in the companion litigation analysis (the only third-party relevance is an examiner's "X" citation in EP 1 852 777 A3, an unrelated Ab Initio prosecution). Because the terminal assignee is a large operating company and the patent expired in 2018, none of the eight NPE/troll signals fire at a "present" level; the sole caveat to the category label is that no assertion activity was evidenced, so the "assertion" element of the verdict is unconfirmed.

Verify here: USPTO Assignment Center — search 6480876 → https://assignmentcenter.uspto.gov/ (mirror at https://assignment.uspto.gov/patent/index.html). Pull the reel/frame for each link above and confirm the correspondents, then cross-check the 2007 change-of-name date against the sibling record at Reel 014628/0103.

I'll research the prior-art references cited in US 6,480,876. Let me run several searches.

Let me search specifically for the cited prior-art references (backward citations) on the patent's front page.

Prior-Art Analysis — U.S. Patent No. 6,480,876

Scope note and an important sourcing limitation up front. The task asks me to work through "each patent citation" on US 6,480,876. I need to be explicit: the full-text dump supplied for this patent omitted the front-page "(56) References Cited" section — it jumps from the Abstract/Landscapes straight to the Description and Claims, and the only citation block it contains is the forward list ("Cited By (72)"). My live searches for the literal number 6480876 / 6,480,876 did not surface the examiner's backward-citation list either (they returned the patent's own specification discussion, unrelated numbers, and the EP citation of this patent). So I cannot verify the exact set of references the examiner cited against '876.

What follows therefore separates (a) references I can verify from the patent's own text and corroborating sources, from (b) what I could not retrieve — rather than inventing a citation list. I am not auto-correcting any identifier, and I flag where a search result contradicted or refined what I would otherwise have assumed.

Distractors I deliberately excluded (numbers that resemble 6480876 but are different patents): US 6,483,735 (optical 3-D memory), US 6,591,876 (Omega Patents v. CalAmp " '876"), US 6,483,928, and the " '087" patents surfaced earlier. None is US 6,480,876.

---

1. References verifiable as tied to the '876 disclosure

A. Background art expressly identified in the '876 specification itself

These are the references the applicants put on the record in the Description of US 6,480,876 (authoritative from the supplied text):

Note the relationship: US 6,067,604 ("Space-time memory") is by an overlapping team (Ramachandran, Halstead, Joerg, Kontothanassis, Nikhil, Rehg) and is the "STM system" that '876 extends. I confirmed via Google Patents that US6067604 lists App. No. 08/909,405, filed 1997-08-11, and is titled "Space-time memory."

B. Non-patent prior-art systems named in the '876 specification

The specification critiques these by name (non-patent literature, not "patent citations" but prior art nonetheless):

• Orca — a task/data-parallel programming system; the '876 text states its data parallelism "is not only static, but also specified explicitly in the source programs."
• FX — a system that "automatically determines optimal mappings of tasks to processors in static domains."
• Profile-driven / static task scheduling generally (the FX family).
• CC-NUMA page migration/replication (on-line adaptation in a narrow architecture).
• Reinforcement-learning network packet routing (on-line adaptation in a different narrow context).

§ 102 potential of these: none. Each is expressly distinguished in the '876 text as static or confined to a different problem domain; none discloses the dynamic work-queue + controller-queue construct of claim 1/14. They are § 103/§ 102-background material, not anticipating references.

C. References cited within the sibling patent US 6,067,604 (context only — NOT citations of '876)

I located these on US 6,067,604's front page. I am labeling them carefully because they are frequently mistaken for '876's own citations:

• Patent citations of US 6,067,604: US 5,691,902; US 5,852,449; US 5,802,580; US 5,713,039.
• Non-patent citations of US 6,067,604: Fujimoto, "The Virtual Time Machine" (1989); Ghosh et al., "Parallel Discrete Event Simulation Using Space-Time Memory" (1991/1995); Jefferson, "Virtual Time," ACM TOPLAS 7(3), 1985, pp. 404–425; Singla et al., "Temporal Notions of Synchronization and Consistency in Beehive" (SPAA '97).

These are not, to my verified knowledge, the examiner citations for US 6,480,876. I am including them only to show the surrounding art cluster and to avoid conflating the two patents' records.

D. Forward citations ("Cited By (72)") — not prior art

The 72 documents that cite '876 (e.g., AB Initio's graph-based computation family, the Silicon Graphics "transparent distribution" application, Microsoft's "parallel query engine with dynamic number of workers," etc.) are later publications and therefore cannot be § 102 prior art against a 1998-filed patent. They are useful only for a family/landscape view.

---

2. The one verified use of '876 as prior art (reverse direction)

For completeness, the only record I could verify where '876 is cited is the European search report for EP 1 852 777 A3 (AB Initio Software Corp., "Computer-aided parallelizing of computation graphs"), where US 6,480,876 B2 is rated "X" (particularly relevant) against claims 1 and 11 and cited (abstract; FIG. 3; col. 4 line 66 – col. 9 line 4) against claims 2–10. This confirms the technical centrality of the splitter/worker/joiner + control-queue construct — but it says nothing about what anticipates '876.

---

3. § 102 anticipation assessment (element-by-element)

Legal standard applied literally: Under 35 U.S.C. § 102, a claim is anticipated only if a single reference discloses every element and limitation of that claim, arranged as in the claim. A dependent claim incorporates all limitations of the claims from which it depends, so a reference cannot "anticipate" dependent claim 3 unless it also discloses everything in claims 1 and 2.

Closest verified candidate — US 6,067,604 (Space-time Memory, Ser. No. 08/909,405, filed 1997-08-11):

Conclusion: US 6,067,604 discloses the port/time-stamp substrate (elements of dependent claims 3–5) but not the splitter/worker/joiner + dual-queue architecture of independent claims 1 and 14. Because claims 3–5 depend from claim 1, US 6,067,604 cannot anticipate any claim of '876. Its real role is as a § 103 reference (or a § 102(e) reference for the shared substrate) in combination with another reference disclosing a producer/consumer parallel decomposition.

All other verified references (Waters 08/844,444; Orca; FX; CC-NUMA; RL routing): none discloses the claimed dynamic construct; no § 102 anticipation of any claim of '876 on the verified record.

Bottom line: On the references I could actually verify, no single reference anticipates any claim of US 6,480,876. The nearest art (US 6,067,604 for the port/time-stamp elements) is complementary rather than anticipatory.

---

4. What I could not verify — and how to close the gap

I want to be candid about the limits of this analysis rather than pad it with a fabricated citation list:

1. The '(56) References Cited' list for US 6,480,876 was not retrievable in this session. The supplied full text omits it, and my web searches did not surface it. This is the single most important gap: a definitive § 102 opinion must be built on the examiner's actual cited references.
2. I could not directly query USPTO Patent Center/PAIR or Espacenet's "Cited Documents" view for '876 to pull the (56) list.
3. The Google Patents page for '876 shows only the forward ("Cited By 72") citations in the retrieved text.

Recommended steps to obtain the authoritative list (each would let this analysis be completed rigorously):

• Open USPTO Patent Center for application 09/085,795 → "References Cited" / "Patent and Non-Patent Literature."
• Open Espacenet record for US6480876B2 → "Also published as / Cited documents / Citing documents" tabs.
• Consult the file wrapper (Image File Wrapper) for 09/085,795 for examiner citations, PTO-892 forms, and any § 102/§ 103 rejections — the rejections themselves identify the examiner's best art and the claims it was applied to.

Once the (56) list is in hand, the same element-by-element table in § 3 can be applied mechanically to each reference.

Caveat: This is a technical prior-art analysis, not legal advice or a freedom-to-operate/validity opinion. Anticipation is a claim-by-claim, element-by-element determination that depends on the full prosecution record, which I could not fully retrieve here.

If you can paste the "(56) References Cited" block (or the PTO-892 / office-action citations) from the '876 file, I will produce the complete per-reference citation table with dates, descriptions, and the specific claims each reference could anticipate.

I'll ground this §103 analysis by pulling the actual prior-art references that appear in (and are described by) this patent, plus the classic art a PHOSITA would have known. Let me run several searches.

Now let me pull the co-pending prior-art applications the patent expressly incorporates and the classic dynamic-work-queue / dataflow art.

Obviousness Analysis — U.S. Patent No. 6,480,876 under 35 U.S.C. § 103

0. Sourcing and framing caveats (read first)

Three threshold points, because they materially shape the analysis:

(a) The Google Patents page has no "prior-art citations" list. The fetched page contains only (i) a "Prior art keywords" field (data, task, application, time, chunks — an indexing artifact, not a reference list) and (ii) "Cited By (72)" — forward citations, all later than the '876 patent. Forward citations are not prior art against '876. The only prior art that exists against '876 is (1) the references described in the '876 Background, and (2) art a PHOSITA would have known as of the 1998-05-28 filing date. I use both, and I label which is which.

(b) The two co-pending applications the patent incorporates are 102(e) art — but common ownership is a live §103(c) issue. The '876 Background expressly relies on:

• U.S. Appl. Ser. No. 08/909,405, "Space-Time Memory," filed Aug. 11, 1997 by Ramachandran et al. → issued as U.S. 6,067,604 B1 ("Space-time memory"). Google Patents confirms the number: https://patents.google.com/patent/[US6067604](/patent/US6067604)
• U.S. Appl. Ser. No. 08/844,444, "Method and Apparatus for Visual Sensing of Humans for Active Public Interface," filed April 1997 by Waters et al. → issued as U.S. 6,256,046 B1 (granted 2001-07-03). Espacenet records the application as US84444497A, filed 1997-04-18, inventors Waters, Loughlin, Rehg, Kang; original applicant Digital Equipment Corp.: https://worldwide.espacenet.com/publicationDetails/biblio?CC=EP&NR=0872808

Both named different inventive entities than '876 ("by another"), so both are 102(e) art. ⚠️ Flagged discrepancy: the '876 specification says the Waters application was "filed on Apr. 8, 1997," but the Espacenet/Google record shows 1997-04-18. I do not auto-correct; I note it as an inconsistency you should verify against the file wrapper. It does not change the 102(e) status.

Because both were owned by Digital Equipment Corporation at the time of invention, a defendant would argue pre-AIA §103(c) removes them from §103. Two cautions: (i) the 1999 §103(c) amendment and the 2004 amendment each have effective-date limits that plausibly exclude a 1998-filed / 2002-granted application; and (ii) §102(g)-type art ("prior invention by another") survives §103(c) on the pre-2004 text in any event. If §103(c) applies, the obviousness case collapses to the non-commonly-owned references (Orca, Fx, Boyan & Littman, general knowledge) — none of which teach the port/time-stamp streaming substrate. That is, in my view, the single most important validity question on this patent, and I flag it rather than resolve it.

(c) A publication-date trap on one Orca reference. The "data-parallel Orca" description I rely on below (master process → job-queue object, replicated workers each taking "the next matrix," runtime-changeable partition ownership) appears in IEEE Concurrency, Vol. 6, No. 3, July–September 1998 — which is after the '876 filing date and therefore not 102(a)/(b) art. Treat that specific teaching as insecure. The 1992 Orca paper (Bal, Kaashoek & Tanenbaum, IEEE Trans. Software Eng. 18(3):190–205, DOI 10.1109/32.126768, https://research.vu.nl/en/publications/orca-a-language-for-parallel-programming-of-distributed-systems-2/) is solid prior art, but the patent's own admission about Orca is that its data parallelism was "static" and explicit in source — so the 1992 paper is best used for the task-parallel + shared-object substrate, not the dynamic work queue.

---

1. Deconstructing claim 1 into elements

Claim 1 (and mirror claim 14) requires:

The point of novelty is the E5/E6 pairing — a separate control channel carrying, per chunk, the splitter's own partitioning strategy, so the joiner can reassemble outputs even though M (chunks) ≠ N (workers) and both vary at runtime. E1–E4 are each individually old in the art; E5/E6 is the narrow neck.

---

2. Reference-by-reference element mapping

R1 — U.S. 6,067,604 (Space-Time Memory / "Stampede"), 102(e) art

Teaches the streaming substrate of E1 essentially verbatim: a cluster of SMPs; ports as temporally-evolving FIFO data structures; time-stamps (frame numbers) ordering data items; a vision pipeline with digitizer, histogrammer, and back-project threads; automatic garbage collection. It expressly recites that "the memory is allocated to a plurality of ports arranged in a spatial ordering" and that "space coordinates identify ports, and time coordinates index data items within ports." That maps directly onto dependent claims 3, 4, 5, 6 — which is why it is a serious reference. What R1 does not disclose: any splitter/worker/joiner construct, any work queue, and any controller queue.

Search corroboration: https://patents.google.com/patent/US6067604 ; the corresponding system paper is Ramachandran, Nikhil, Harel, Rehg & Knobe, "Space-Time Memory: A Parallel Programming Abstraction for Interactive Multimedia Applications," PPoPP/SIGPLAN Notices 34(8):183–192 (1999) — https://dl.acm.org/doi/abs/10.1145/[301104](/patent/301104).[301121](/patent/301121) . Note the earlier Digital CRL Technical Report CRL 97/2 (April 1997), "Space-Time Memory: A Parallel Programming Abstraction for Dynamic Vision Applications," predates the '876 filing and is independently 102(b)-ish if publicly accessible.

R2 — U.S. 6,256,046 (Waters, Loughlin, Rehg, Kang), 102(e) art

Supplies the application-level motivation and the "distribute over regions" concept: a kiosk that digitizes "color frames from the cameras," detects and tracks multiple users by color blob, and — directly relevant to claims 6 and 7 — partitions the interaction region "into a plurality of sub-regions in which movement of the person can be independently detected and tracked" (claim 21 of R2), and concurrently tracks "respective movements of each person independently" (claim 20). This is the data-parallel decomposition the '876 patent later calls "distribute parts of the same frame." Source: https://patents.google.com/patent/[US6256046B1](/patent/US6256046B1)/en

R3 — Orca (Bal, Kaashoek, Tanenbaum, IEEE TSE 18(3), 1992)

The patent admits Orca already integrates task and data parallelism, albeit statically and explicitly. R3 teaches E3/E4's use case (replicated workers servicing a shared object), and the runtime-changeable distribution primitive teaches E4's variability. Its weakness against '876 is exactly the patent's stated one: static, source-declared partitioning.

R4 — Fx compiler / HPF task regions (Subhlok, Stichnoth, O'Hallaron, Gross; Subhlok)

The patent admits Fx "automatically determines optimal mappings of tasks to processors in static domains." Fx also supplies the latency/throughput framing the '876 specification uses verbatim, and it builds its mapping from measured execution times. Sources: https://dl.acm.org/doi/abs/10.1145/[209936](/patent/209936).[209951](/patent/209951) ("Optimal mapping of sequences of data parallel tasks," PPoPP 1995) and Subhlok, "Experience with Automatic Mapping of Sensor-Based Applications" (1995): https://ask.orkg.org/item/24368147/ . Relevant to claims 12–13.

R5 — Boyan & Littman, "Packet Routing in Dynamically Changing Networks: A Reinforcement Learning Approach," NIPS 6:671–678 (1993/94)

The closest art to Part II. Teaches an online, model-based controller in a dynamically varying environment: each node maintains a Q-table/cost model, updates it from locally measured delivery times, adapts when load shifts, and uses explicit exploration to escape greedy local optima. This is structurally the '876 controller (modeling 531 / control policy 532 / exploration 533). Source: https://papers.neurips.cc/paper_files/paper/1993/hash/4ea06fbc83cdd0a06020c35d50e1e89a-Abstract.html

R6 — Tagged-token dataflow (general knowledge; not verified in this session)

Steering results to their destination by a tag carried with each datum, with matching performed on tags, is the classic tagged-token dataflow mechanism associated with Arvind & Nikhil. I am confident of the concept but did not verify the exact citation in today's searches — treat the citation, not the concept, as unverified. Its significance: it supplies a known mechanism for E2 ("individually marked") and for E6 (outputs steered to the right consumer by mark). Note inventor Nikhil is a tagged-token dataflow author, so this art is squarely in the inventors' field.

R7 — Work-queue / bag-of-tasks dynamic scheduling (Cilk, Linda tuple space, master-worker) — general knowledge, not verified in this session

Dynamic scheduler feeding idle workers from a shared queue is decades-old. Cite with the same caveat.

---

3. The §103 combinations and the motivation to combine

Combination A — R1 (STM) + R3 (Orca) + R2 (kiosk) ⇒ anticipates/renders obvious E1–E4

Motivation (KSR, 550 U.S. 398, 417–421 (2007)):

1. Same field, same problem. All three are parallel/distributed programming and real-time vision on SMP clusters. No field-crossing.
2. Express market/practical pressure stated in the patent itself. The '876 Background concedes the need: "the relative extent of data parallelism versus task parallelism must dynamically track changes in the application," and that Orca "falls short in that its data parallelism is not only static, but also specified explicitly in the source programs." That is an express, admitted design incentive to dynamize Orca-style data parallelism — the classic KSR motivation.
3. Predictable result. The patent itself asserts the substitution "exactly duplicates the behavior of the original task as viewed on its input and output ports." A PHOSITA replacing a fixed chunk→worker assignment with a shared queue would expect exactly that behavior-preservation plus load balancing.
4. The kiosk's variable customer count (R2) is the concrete reason a fixed assignment fails, supplying the "identify a need or problem" prong.

Remaining gap: E5/E6 (the per-chunk control item on a controller queue). Combination A alone probably does not reach claim 1.

Combination B — Combination A + R6 (tagged-token dataflow / mark-and-steer) ⇒ reaches E2, E5, E6

Tagged-token dataflow teaches that a datum carries a mark identifying its context and destination, and that downstream machinery routes results by matching the mark — precisely the function of the "individually marked chunks" + "done ports act as a sorting network" of the '876 FIG. 3 embodiment. Motivation: once M is decoupled from N, something must tell the joiner how to reassemble; the analogous problem (matching partial results to the right consumer when arrival order is not fixed) was already solved in dataflow by tagging. The combination is "a combination of familiar elements according to known methods" yielding a "predictable" result — the KSR standard. This is the strongest obviousness route to claim 1.

Combination C — R3 (Orca) + R4 (Fx) + R5 (Q-routing) ⇒ reaches claims 12–13

Claims 12–13 add "means for learning and applying a task and data parallel strategy," a "controller measuring an execution of the application to generate an action space," and "a run-time system applying the action space."

• R4 (Fx) teaches automatic mapping driven by measured task execution times — i.e., a model of (task → performance) used to pick a mapping. It stops short of on-line adaptation (the patent says so).
• R5 (Q-routing) teaches exactly the missing piece: an on-line, measurement-driven performance model updated at runtime in a dynamic environment, with exploration.
• Motivation: the patent's own premise is that a heterogeneous task set's computational properties "are not well-understood" and "it is less likely that the space of data parallel strategies can be characterized prior to run-time." A PHOSITA facing that stated problem would look to any online black-box model that maps control actions to measured performance; Q-learning-style models were the known answer for dynamic-resource systems. Under KSR, "obvious to try" applies where the prior art gives a finite number of identified, predictable solutions — here, the finite menu is open-loop / dead-beat / LQR control over a learned forward model, which the patent itself enumerates as the standard textbook set.
• Independent corroboration of technique-level obviousness: the '876 specification concedes the same memory-based/local-regression approach was "previously … employed with some success to the analogous problem of robot control learning."

Combination D (claim 9–11)

Claims 9–11 associate a task/data-parallel strategy with each chunk and let the methods ride along with the chunk (or be indexed by chunk type). The specification admits the table-indexed variant is conventional ("if only a small number of strategies are defined, these can be stored in a table, and the chunk type can be used as an index"). Passing behavior with a message is the actor model (Agha) and standard object/mobile-code practice. Strong §103.

---

4. Dependent-claim quick hits

Claims 2–5 are the most vulnerable, because R1 recites the port/spatial-ordering/time-stamp structure almost word-for-word — if R1 is available as §103 art, those claims should not have issued.

---

5. Where the patent most plausibly survives, and honest weaknesses in the above

1. The E5/E6 neck is genuinely narrow and specific. No searched reference teaches a splitter-generated, per-chunk control record enqueued on a dedicated control queue that the joiner reads to decide reassembly. Combinations B and C reach it only by analogy (tagging, black-box control). An examiner or court could reasonably reject the analogy as hindsight reconstruction — the classic §103 attack line.
2. §103(c) could gut the case. If common ownership disqualifies R1 and R2, the streaming/port/time-stamp limitations of claims 2–6 have no 102(e) anchor, and the Orca/Fx/Boyan art does not teach ports at all.
3. The 1998 Orca data-parallel teaching post-dates the filing. Removing it weakens Combination A/B materially (see §0(c)).
4. No objective indicia on the record. I found no unexpected-results, licensing, or commercial-success evidence in the materials supplied. Conversely, the applicant's FIG. 4 data (3× speedup; 17% advantage of model-distribution over region-distribution) is applicant's own performance data, not a comparison against prior art, so it does not rebut obviousness.
5. The patent's own admissions cut against it. Part II states that for a single application "it is possible to manually explore the space of task and data parallel strategy through analysis and experimentation." An explicit admission that the claimed learning was a manual exercise, automated by known learning techniques, is potent §103 evidence ("obvious to try").

---

6. Bottom line

• Claims 2–5 are, in my assessment, the most likely to be invalid under §103 over U.S. 6,067,604 alone or in trivial combination — R1 recites spatial port ordering and space/time coordinate access almost verbatim.
• Claim 1 / claim 14 are obvious over R1 + R3 (Orca) + R2 (kiosk), further combined with tagged-token dataflow (R6) and/or a bag-of-tasks work-queue teaching (R7), with KSR motivation supplied by the patent's own admitted need for dynamic re-partitioning; but this requires the analogy step for the controller queue, so I rate it a contested, not a clean, conclusion.
• Claims 12–13 are obvious over Fx (R4) + Boyan & Littman Q-routing (R5) given the patent's own "must adapt on-line" framing.
• The outcome turns on two things I could not definitively resolve: (i) whether §103(c) removes R1/R2, and (ii) whether the data-parallel-Orca job-queue teaching is even prior art given its July–September 1998 publication.

Confidence: high on the reference mappings to claims 2–6 and to the R4+R5 combination for claims 12–13; medium-to-low on any single combination reaching claim 1's controller-queue limitation without an analogy-based motivation, and low on the §103(c) and Orca-date questions, which I have flagged rather than asserted.