Evidence
The checkable part: peer-reviewed publications, granted patents, standards context, and the domains this has been used in.
In brief
Different artifacts support different conclusions. Patents establish disclosure and legal status; publications support the methods and evaluations described in them; product records support named implementations and measurements; standards define external objectives; customer cases support operational use in their stated context. No category should be used as a substitute for another. Published items link to their public source, listed for reference and verification; items whose record is not yet public say so in a dated status note on the page that carries them, and are not presented as public product proof.
Comparisons use traceable evidence.
Quantitative figures for third-party technologies are attributed to their published sources. Where no reliable figure is available, we leave it open rather than substitute an estimate.
Start with the evidence you need
| Publications | Papers with named co-authors from Razorcat, AbsInt and the University of Lübeck - each classified as peer-reviewed, conference paper or preprint. |
| Patents | The granted families, with numbers you can look up - and a note on what a patent supports and what it does not establish. |
| Standards | Where trace-based evidence fits in DO-178C, ISO 26262, IEC 61508, EN 50716 and AMC 20-193 - and where it does not. |
| Industries | Where this has been used - and the case for the domains beyond. |
| Measurement records | Documented product measurements - version, configuration, method and result - that public performance figures link to. The first published record: the SQLite -O3 comparison. Until the full index is published, other figures carry a “record pending” label. |
Selected claims and their evidence
Public marks a statement carried by a published artifact anyone can retrieve: a specification, a publication, a granted patent. Configuration-specific marks a measured result that holds for the named target, build and workload and is not a general figure. Where a published measurement is bounded that way, the narrower label applies.
| Claim | Evidence | Status |
|---|---|---|
| For an observation interval whose trace-validity checks pass, control-flow trace supplies the machine-level program-counter sequence the analysis needs. Detected overflow, discontinuity or unsupported events qualify or invalidate the affected interval. | Observability and its limits · Measurement validity | Public |
| Machine-level trace completeness and unambiguous source attribution are separate properties. Optimization can merge, predicate, clone or remove the source structures that attribution relies on while the machine events stay observable. | Source attribution · Observability and its limits | Public |
In an optimized SQLite workload at -O3 on an Arm® Cortex®-M7 target, native trace supported unambiguous source attribution for 77% (5,667 of 7,356) of the decision obligations linked into the image, and for 48.2% (1,470 of 3,047) of the conditions in multi-condition decisions. | SQLite at -O3 | Configuration-specific |
In that same configuration, CEDARtools.Anchor increased .text by 8.4% for the statement and decision class and by 11.5% for the condition/MC/DC class, while static RAM remained effectively unchanged. | SQLite at -O3 | Configuration-specific |
| In that same configuration, the complete GCC coverage images were projected to exceed the available target memory, so no on-target runtime result was produced for them. | SQLite at -O3 | Configuration-specific |
| CEDARtools.TraceEncoder implements selectable N-Trace/Nexus and E-Trace back ends behind one protocol-agnostic front end, integrated with six RISC-V® cores. RTL, testbenches, formal properties and build scripts are published under CERN-OHL-S-2.0. | CTTE preprint · The standards which CTTE implements | Public |
| With hardware and encoder configuration held constant, workload choice alone moved the measured trace cost from 0.21 to 4.90 bits per retired instruction. | CTTE preprint | Configuration-specific |
| Announcement of trace loss is required and defined by all three trace specifications; the time recovery takes is not. Measured across 3,305 announced-loss episodes on six simulation streams and four compression configurations, the distance from the error message to re-anchoring synchronization was one message, four bytes and zero instructions. Not claimed for board streams or the E-Trace back end. | The CTTE extras | Configuration-specific |
| Tool qualification does not transfer automatically because one tool imports another tool’s output. CEDARtools® is not currently supplied with a DO-330 or ISO 26262 tool-qualification package. | Qualification and assurance boundary | Public |
| Granted European and US patents cover trace-data processing and profiling, event processing, and control-flow integrity monitoring. The Hybrid Runtime Analysis architecture behind CEDARtools.Anchor is a published pending application, not a grant. Legal status is stated per publication or grant. | Patents | Public |
| Multi-core WCET evidence usually combines architectural analysis, controlled scenarios and measurement, because interference depends on shared resources and concurrent activity. | Multi-core WCET · Publications | Public |