# Open IP integration evidence

This original compilation compares 20 public IP records across 20 integration attributes. Sources were checked on 24 October 2026 (America/Denver); individual UTC retrievals fall on 25 October. Every record has an exact repository commit. The table is a documentation audit, not a simulation, formal proof, implementation, security assessment or tapeout-readiness score.

## Records and independence

The 20 records cover 19 repositories. OpenTitan AES and HMAC are distinct IP blocks at the same repository revision. Related PULP, LiteX and bus-library records are not asserted to be independent projects or evidence votes. Collection rows identify their selected module where clocks, resets or widths cannot sensibly describe every module.

The PULP APB I2C adapter replaces the earlier unpinned OpenCores project-page entry. Its inherited controller attribution is retained. The controller files have permissive notices; the APB wrapper’s licence remains unverified. This unresolved scope is a filterable finding, not clearance to copy the wrapper.

## Field evidence

Read-only Git ref lookups established the pinned commits. Recursive repository-tree metadata identified candidate files; the files themselves were then opened at those revisions. Positive facts are tied to those opened files through `field_evidence`, with exact repository paths and URLs. The source ledger retains dates, HTTP outcomes and hashes without distributing source bodies or core code.

The schema covers function, language, buses, clock/reset semantics, widths, configuration, register descriptions, interrupts, software collateral, memory interfaces, dependencies, manifests, testbenches, simulators, formal collateral, timing constraints, examples, licence scope, maintenance and limits. Register descriptions are labelled by their actual form: prose/CSR tables, HJSON, SystemRDL, CMSIS-SVD or RTL definitions. None is relabelled IP-XACT without observed evidence.

A null means **not verified in this audit**, not absent. An opened testbench, assertion or proof configuration is collateral, not a passing result. Simulator names in a repository describe upstream flows; those flows were not run here. Driver-path references are distinguished from inspected driver headers and generic software examples. Board XDC/SDC examples are not complete constraints for another integration.

The four finder filters use the row’s `filter_values` arrays: interface, licence, collateral type and language. They indicate observed categories. A generated language is labelled as generated, and configurable interfaces are not assumed to coexist in one elaborated design. Empty or unknown fields cannot be interpreted as negative capability evidence.

## Existing references and added value

The inspected [OpenTitan AES interface documentation](https://github.com/lowRISC/opentitan/tree/master/hw/ip/aes), [PULP AXI module inventory](https://github.com/pulp-platform/axi), and [SERV integration documentation](https://github.com/olofk/serv) each describe one ecosystem or project. This comparison adds a shared, revision-pinned integration schema across CPU, bus, networking, storage and cryptographic IP, with explicit unknowns and field-level provenance. The fixed-commit links in JSON are the authoritative snapshot locators; the general links here are orientation only.

## Licence and maintenance boundaries

The original observations and compilation are **CC BY 4.0**. Attribute them as “Open IP integration evidence — original compilation, 2026-10-24” and link to this method. Upstream HDL, software, licence text, patents and ISA rights remain separate. No upstream core or licence-file body is included.

Licence evidence is scoped: Caliptra’s SHA wrapper retains a BSD-2-Clause notice while the controller/register files use Apache-2.0; Rocket Chip has Berkeley and SiFive notices; Microwatt’s softcore licence separately discusses physical-implementation rights. PULP APB I2C has an unresolved wrapper licence. These are recorded limitations, not legal conclusions.

PicoRV32 explicitly reports inactive development. The two Alex Forencich collections explicitly report deprecation in favor of another repository. Their pins remain useful integration evidence, but the table does not promote them as maintained starting points. No broader maintenance score was inferred from recent commits alone.

## Reproduction and integrity

Inspect the exact source paths at the recorded commits and compare returned bytes to ledger hashes. A new upstream commit requires a new observation rather than silently replacing this snapshot. Review changes in configuration, dependency versions and module selection before reuse. `hashes.txt` excludes itself and is refreshed after CSV export; JSON is authoritative for nulls and nested evidence. No former magazine articles, circuits or figures were consulted or reconstructed.