Eureka on the web
When you want the answer in the next five minutes.
The agent works the prompt against patents and technical literature, citing every source.
Run your analysis now →Filing growth compares 2021 (683 records) with 2024 (465) — a three-year span. 2024 is the most recent year we treat as complete: publication lags filing by roughly 18 months, so 2025 onwards are still filling in and any growth rate that ends there would understate the field. Top-5 share is the combined record count of the five largest assignees divided by all 6,021 records in scope (CR5), not by the ranked leaders only.
Intermediate representation (IR) design sits at the core of how compilers translate source code into something a machine can execute efficiently. This landscape tracks patent activity around static single assignment form, compiler optimization passes, code lowering, pass ordering, and the supporting infrastructure of invariant verification and debug information that keeps an IR trustworthy as it moves through transformation stages. The scope spans 6,021 published records filed between 2015 and mid-2026.
The dataset captures both classic CPU-targeted compiler work and the newer wave of IR design built for graph transformation on AI accelerators — a split visible in the IPC composition, where general digital data processing (G06F) dominates but AI-model computing (G06N) and image/graph processing (G06T) carry a meaningful share.
Two views of the same 6,021-record dataset: how filing volume moved year over year, and which IPC subclasses the underlying technology falls into.
Filings rose from 275 in 2017 to a peak of 871 in 2022, then declined to 465 by 2024 — a 32% drop across 2021 to 2024. Years from 2025 onward are still filling in under an 18-month publication lag, so the apparent tail-off should not be read as the field slowing; the 2026 figure of 48 is a partial year.
G06F (electric digital data processing) appears in 77.1% of the 6,021 records, confirming this is fundamentally compiler-and-runtime territory. G06T (21.2%) and G06N (16.2%) point to IR work increasingly built around image/graph pipelines and AI-model execution, while H05K, H04L, G06V, G09G and H04N each register single-digit shares, marking narrower application niches rather than core claim territory.
Shares are the percentage of the 6,021 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
This page is one run against one query. Ask Eureka your own question about intermediate representation design and every answer comes back with the patent numbers behind it.
Try EurekaThe invention describes creating a rank-n static single assignment intermediate language from a rank-(n-1) SSA intermediate language, where n is a positive integer, by searching the lower-rank language for a rank-1 definition that reaches at least one use, then renaming that definition and its use to a compiler temporary.Filed by Intel Corporation, granted 1999-12-07 — one of the earliest filings in this dataset to formalise multi-rank SSA construction for symbolic memory references.


| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20090018996A1 | Cross-category view of a dataset using an analytic platform | 1,559 |
| 2 | US7089272B1 | Specializing write-barriers for objects in a garbage collected heap | 565 |
| 3 | US20080288889A1 | Data visualization application | 528 |
| 4 | US20080294996A1 | Customized retailer portal within an analytic platform | 517 |
| 5 | US6721941B1 | Collection of timing and coverage data through a debugging interface | 499 |
| 6 | US6760903B1 | Coordinated application monitoring in a distributed computing environment | 471 |
| 7 | US20050273772A1 | Method and apparatus of streaming data transformation using code generator and translator | 425 |
| 8 | US6785673B1 | Method for converting relational data into XML | 415 |
| 9 | US6507947B1 | Programmatic synthesis of processor element arrays | 406 |
| 10 | US6427234B1 | System and method for performing selective dynamic compilation using run-time information | 402 |
Citation counts reflect influence within the searched corpus and favour older filings; they are not a measure of current commercial relevance.
Each row carries its publication number; clicking a row searches Eureka by that number.
When you want the answer in the next five minutes.
The agent works the prompt against patents and technical literature, citing every source.
Run your analysis now →When it has to run inside your own pipeline.
Patent search, landscape analysis and assignee resolution as MCP tools. Drop them into any agent framework, or call REST directly.
Browse MCP servers →Three patterns stand out once concentration, timing and classification are read together.
Five assignees hold 3,988 of the 6,021 records in scope. That density means core SSA-construction and pass-ordering claims are largely spoken for; new entrants need to look at adjacent implementation detail rather than the foundational mechanics.
Filings peaked at 871 in 2022 before easing to 465 by 2024, a 32% decline over that span. Recent-year assignee momentum shows steep pullbacks across the largest filers, though this reads more like normalization after a filing surge than an exit from the field.
Nearly a fifth of records (21.2%) fall under G06T image data processing and 16.2% under G06N AI-model computing — evidence that IR design has moved well beyond classic scalar-code compilers into graph-based accelerator pipelines.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to intermediate representation design, with the prior art for and against each one.
| Assignee | Co-assignee | Shared families |
|---|---|---|
| International Business Machines Corporation (IBM) | IBM DEUTSCHLAND GMBH | 16 |
| Intel Corp | YAO ANBANG | 10 |
| Intel Corp | XU LIN | 9 |
| Intel Corp | WANG SHANDONG | 9 |
| Intel Corp | WANG LIBIN | 9 |
| Intel Corp | HU PING | 9 |
| Intel Corp | GUO YIWEN | 9 |
| Intel Corp | CHENG WENHUA | 9 |
Only 10 co-assignee pairs appear in the dataset, and the strongest pairings link a single corporate parent to its own subsidiary or named inventors rather than to independent third parties — collaboration across separate companies is not a meaningful feature of this landscape.
The ranked leaders span classic CPU and systems vendors alongside newer AI-accelerator entrants, but recent-year momentum has cooled sharply even among the largest filers.
The leading assignee's 1,825 records dwarf the fifth-place total of 187, a gap wide enough that most competitive dynamics in this landscape are really about the distance between the top filer and everyone else.
Recent-year momentum shows steep drops across the largest names, with several major assignees recording zero filings in the latest tracked year. Given publication lag, this likely overstates the slowdown, but it is a real shift from the 2021-2022 filing pace.
The United States receives the large majority of filings tracked here, with EPO, WIPO/PCT, UK and Germany forming a secondary tier. That distribution suggests the commercially decisive jurisdiction for freedom-to-operate work is the US, with EPO filings worth checking for parallel European coverage.
| Assignee | Recent year | YoY |
|---|---|---|
| NVIDIA Corp | 35 | -73% |
| SambaNova Systems Inc | 4 | -83% |
| Intel Corp | 2 | -98% |
| International Business Machines Corporation (IBM) | 0 | -100% |
| Microsoft Corp | 0 | — |
| Microsoft Technology Licensing LLC | 0 | — |
| Oracle International Corp | 0 | -100% |
| Digital Equipment Corp | 0 | — |
The concentration and composition patterns here raise questions that benefit from claim-level and citation-level follow-up.
With 66.2% of records held by five assignees, the practical question is which specific claim elements they actually hold versus what remains open around SSA construction, pass ordering and debug-info preservation.
Explore assignee claim mapsThe 16.2% G06N and 21.2% G06T shares suggest IR design work is migrating toward graph-based accelerator pipelines; a focused search on that intersection would sharpen where new filings are landing.
Run a focused technology searchOne assignee leads the dataset with 1,825 of the 6,021 records in scope, well ahead of the fifth-place holder at 187. The top five assignees together account for 66.2% of all records, so the field is heavily concentrated rather than evenly spread across many filers. Anyone assessing freedom-to-operate in this space should expect most foundational claim territory to trace back to a small handful of companies, with a long tail of much smaller filers below them.
Filings rose from 275 in 2017 to a peak of 871 in 2022, then fell to 465 by 2024 — a 32% decline over that three-year span. That said, publication typically lags actual filing by around 18 months, so the lower counts seen in 2025 and 2026 in this dataset understate real activity rather than proving the field has cooled. The fairest read is that 2024 marks the most recent year with a reasonably complete picture, and it still shows a level well above the 2017 starting point.
General digital data processing (G06F) appears in 77.1% of the 6,021 records, which is expected given IR sits inside compiler and runtime infrastructure. Beyond that core, image data processing (G06T, 21.2%) and AI-model computing (G06N, 16.2%) are the two largest secondary classes, reflecting how much IR design work now targets graph-based and accelerator-oriented execution rather than purely scalar CPU code. Smaller shares in printed circuits, digital transmission, video recognition, display control and pictorial communication mark narrower application-specific filings.
The lower-share IPC classes — display control circuits at 1.5% and pictorial communication at 1.4% of the 6,021 records — point to areas where IR-specific claim density is thin relative to the core G06F territory. Debug-information preservation across lowering passes and invariant verification tailored to graph-form IR are two functionally important problems that show up far less often than core SSA-construction claims. These are reasonable areas to search in more depth before assuming they are unclaimed, but the raw class shares suggest less crowding than the top of the field.
US5999735A, filed by Intel and granted in 1999, claims a specific method for constructing a rank-n static single assignment intermediate language from a rank-(n-1) SSA form by finding a rank-1 definition that reaches a use and renaming both to a compiler temporary. Any freedom-to-operate assessment needs to check the patent's actual expiry and claim scope rather than assume it still controls all SSA construction — its citation count of 565 shows it has been influential as prior art, not that its claims remain broadly enforceable. Design-arounds typically avoid the exact rank-based renaming mechanism it describes while still achieving SSA form through alternative construction algorithms.
Go past this page: query the whole intermediate representation design corpus yourself, in your own scope.
Every answer comes back with patent numbers you can open.
Disclaimer. This page is generated from Patsnap Eureka data drawn from a limited snapshot of global patent and scientific-literature records, and is provided for general information and reference only.
Patent data carries inherent limitations: recent filings (typically the most recent 18–24 months) are under-counted due to standard publication lag; counts may be reported at either a patent-family or a patent-record basis and are not always directly comparable; classification, applicant-name, and citation data may contain errors, duplicates, or omissions; and the underlying search query defines and constrains the scope shown. As a result, the analysis may be incomplete or inaccurate and may not reflect the full technology landscape.
Nothing on this page constitutes an exhaustive prior-art, novelty, freedom-to-operate, or validity search, nor does it constitute legal, financial, investment, or professional advice, and it should not be relied upon as such. Any patent, commercial, or strategic decision should be verified independently and reviewed with qualified patent, legal, and domain professionals. Patsnap makes no warranties, express or implied, as to the accuracy, completeness, or fitness for any particular purpose of the information presented.
Machine translation. Assignee and organisation names originally recorded in Chinese, Japanese or Korean have been rendered into English by an AI translation step so that the tables stay readable. These renderings are best-effort and may not match a company's registered English name; the original name is what the underlying patent record carries, and it is what any Eureka query launched from this page uses.
Help us improve
Something wrong, out of date, or missing on this page? Tell us and it goes straight to the team that maintains it.
This page