Validation records
These records explain what evidence supports a scientific or release claim and what remains unqualified. They are not tutorials or implementation roadmaps. Results are specific to their recorded revision, workload, and device.
| Question | Record |
|---|---|
| What supports the scientific baseline? | Scientific correctness |
| Which adaptive method and splitting choices are qualified? | Adaptive integration assessment |
| What supports recording and exact continuation? | Simulation workflow |
| What supports independent CPU/CUDA optics? | Reference corpus and CUDA qualification |
| Is the native migration approved for release? | v0.3.0 gates: maintainer/release approval remains pending |
| What supports the mass representation? | Mass precision |
| What supports the fixed-four-substep condensation method? | Stiffness study and direct parity |
| What supports the two-million-particle environmental example? | Environmental sweep |
| Which direct coagulation mechanisms were validated? | Coagulation physics |
| What do historical one-thread-per-box timings show? | Dated coagulation benchmarks and legacy viewer |
| Is resident capture performance qualified? | Qualification gates: UNSHIPPED/BLOCKED |
| How must resident timing and memory be measured? | Benchmark/memory record |
| How must capture profiling be collected? | Profiling record |
| What qualifies custom processes and registered fields? | Extension qualification |
Missing reviewed artifacts are unavailable and unmeasured. A passing source test or closed implementation plan cannot populate a missing measurement row.
Explore this topic: Verification and comparison