Comments and docs across this codebase reference several parallel ID systems —
invariant labels, phase codes, perf-arc milestones, optimisation candidates,
and task numbers. This table decodes them so a reader on docs.rs (who does not
have CHANGELOG.md or the docs/checkpoints/ snapshots, both excluded from the
crates.io tarball) can follow a comment like Э9 (P7.1, task #160) without a
dead reference.
Each row is: ID family → one-line meaning → where to read more.
| ID family | Meaning | Read more |
|---|---|---|
| I1–I7 | Region/Handle-face invariants (I1 resolution, I2 tombstone, I3 bounded stale-handle detection, I4 accounting, I5 drop-once ownership, I6 slot reuse and bounded growth, I7 instance isolation). | INVARIANTS.md |
| M1–M11 | Malloc-face (allocator) invariants: M1 validity, M2 no double-free/UAF, M3 no-overlap, M4 alignment/size fidelity, M5 reentrancy-freedom, M6 OS-return (decommit), M7 owner routing, M8 generational coherence, M9–M11 cross-thread reclamation/epoch obligations. | INVARIANTS.md, PHASE35_DECOMMIT_DESIGN.md |
| Phase 8–78 | The allocator build-out phases (Phase 8 segment substrate, 10 cross-thread free, 11 GlobalAlloc face, 35 decommit, 78 NUMA). Numbers are chronological milestones, not a version. |
ALLOC_PLAN.md, PLAN.md |
| 3b-I / 3b-II | The two legacy concurrent-tier sub-phases: 3b-I (arc-swap RCU, zero-unsafe) and 3b-II (crossbeam-epoch per-slot atomics, the confined hand organ). |
PLAN.md, src/concurrent/ |
| P0–P8 | The 0.3.x FASTBIN (per-thread tcache magazine) design sweep sub-phases (P0 baseline … P6 CAP sweep … P7/P8 batched carve). Dotted forms like P6.1 / P7.1 are refinements within a P-phase. | FASTBIN_DESIGN.md |
| Ф0–Ф6 | PERF-3 run-encoded-freelist arc phases (Cyrillic "Ф" = phase): Ф1 storage/layout (RunStack, feature alloc-runfreelist — now removed), Ф2/Ф3 encode/drain, Ф5 go/no-go cost-benefit gate (NO-GO). The experiment's negative result is preserved as institutional memory. |
design/RUN_ENCODED_FREELIST_PLAN.md, perf/PERF3_RUN_FREELIST_EXPERIMENT.md |
| PERF-3 / PERF-4 | Named performance arcs: PERF-3 = run-encoded freelist (the Ф-series); PERF-4 = decommit-churn / hysteresis before OS release (tasks #216/#217). | design/RUN_ENCODED_FREELIST_PLAN.md, CHANGELOG.md |
| Э1–Э11 | ALLOC_BENCH perf-arc milestones (Cyrillic "Э"): e.g. Э1 bump-direct/batched carve, Э6 magazine oracle, Э9 the P7.1 refinement. Track individual measured wins in the single-thread bench arc. | ALLOC_BENCH.md |
| OPT-A…H | The eight flamegraph-derived optimisation candidates (OPT-A skip re-stamp, OPT-B O(1) contains_base hash, OPT-C lazy stamp, OPT-E large cache, …), each with estimated/actual impact. |
PROFILE_FLAMEGRAPHS.md, ALLOC_BENCH.md |
| X2 / X7 | Cross-thread-hardening items; X7 is the per-granule generational ring (the documented accepted counter-wrap exception on the free face). | DURABILITY.md, FASTBIN_DESIGN.md |
| W2–W7 | 0.3.0 work items (the "W" release-hardening series): e.g. W3 = the alloc-stats per-hit-counter gate, W4 = batched carve_batch. |
CHANGELOG.md |
| A1 / A2 / C2 / R1 / R2 / H1 / MUST-N | 0.3.0 correctness/robustness work items in HeapCore realloc & cross-thread reclaim: A1/A2 reclaim-on-growth, C2 in-place regression fix, MUST-1 realloc-growth reclaim obligation, H1 (task #167) the opt-in hardened interior-pointer guard. |
src/registry/heap_core.rs, CHANGELOG.md |
| SEC-N | Security-review findings (see docs/security/). |
docs/security/, CHANGELOG.md |
| task #NNN | A session task id (#103–#217 and up). The authoritative log is CHANGELOG.md; task ids also appear in commit messages. Not resolvable from the published tarball alone — this is expected (the tracker is dev-only). |
CHANGELOG.md, git history |
Note: CHANGELOG.md, scripts/, .github/, and docs/checkpoints/ are
excluded from the published crates.io tarball (see Cargo.toml exclude), so a
docs.rs reader will not have them. This glossary is the in-tarball fallback for
decoding the ID references that remain in source comments.