Six codegen-side optimization passes against step() and friends. No
user-facing API changes; all 296 tests pass under no-default, tokio,
embassy, and tokio+journal. Cumulative on h8_cross_tree (deepest
dispatch in the suite): -82% instructions / -72% cycles vs the
pre_lca_fix baseline.
* __PATH_DATA + __PATH_RANGE: per-(from, to) precomputed exit/enter
sequences. Up/self/lateral semantics fold into one indexed lookup;
removes lca() and is_up_transition() runtime helpers and the
per-call heapless::Vec<u16, 16> ancestry build. Hierarchy depth no
longer bounded by Vec capacity.
* __DISPATCH[(state, event)]: precomputed handler resolution. The
runtime __PARENT bubble walk + per-hop __handler_lookup call in
step() collapses to one indexed lookup. Bubble resolution is now
O(1) regardless of depth -- bubble_top (4-hop) is within 3
instructions of bubble_leaf (handler on current).
__handler_lookup retained for timer dispatch.
* __TERMINATE_DATA: per-state precomputed exit chain (root inclusive
-- chart-root entry/exit actions are valid). do_terminate() indexes
the slice instead of walking __PARENT, and uses Deque::clear()
(O(1) in heapless 0.8) instead of pop_front-loop. terminate_event
bench: -22.5% instr / -23.4% cycles.
* TimerTable empty-skip: pop_expired / min_remaining / decrement /
cancel_state each get inlined is_empty() early-out with the
non-empty body marked #[cold].
* Codegen-static timer-machinery elide: when chart has no Duration
triggers, every TimerTable call is omitted from step() and
exit_state(). step() returns None directly. dispatch_trigger and
the timer-fired arms remain but are unreachable; LLVM DCEs them.
* TimerEntry.remaining_ns: u128 -> u64 (~584-year ceiling). Halves
entry size; replaces u128 saturating-sub with u64. The smaller
machine struct improves cache density on every bench, even
timer-free ones.
Bench coverage: four new benches added to cover paths the original
suite missed -- bubble_leaf/mid/top (handler resolution at varying
ancestor depths), timer_fire_dispatch (only bench exercising
pop_expired's non-empty branch and dispatch_trigger), terminate_event
(only bench exercising do_terminate), bubble_cross_tree (deep bubble
+ 10-state cross-tree, the realistic HSM worst case).
See hsmc/benches/baseline.txt for the per-pass delta blocks and
docs/003. latency-analysis.md sections 1, 9, and 12 for the
cost-surface analysis.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The runtime `lca()` helper used a `heapless::Vec<u16,32>` ancestry-set
with a linear `contains()` per node — O(H²) and capped at H ≤ 32. The
codegened `__DEPTH` table already had everything needed for an O(H)
depth-walk with no allocation and no hierarchy ceiling. Switch to that.
Also adds an iai-callgrind dispatch bench harness (Linux-only dev-dep)
so future optimizations can be measured deterministically by instruction
count. Seven benches cover: cold start, lateral H=2/4/8, up, self, and a
self-emit chain. Baseline numbers checked in at `hsmc/benches/baseline.txt`.
The latency analysis (`docs/003. latency-analysis.md`) is rounded out
with event-entry-point and cooperative-driving audits — three new
sections documenting the four event injection paths, the run-loop's
cooperative shape, and three correctness invariants (sticky overflow
flag, biased-select trap, cancel-safety on durings).
Bench deltas (cargo bench --bench dispatch -- --baseline=pre_lca_fix):
h2_cold_start -1.9% / -0.3% (priming step, no lca() — noise)
h2_warm_lateral -22.8% / -22.1%
h4_cross_tree -25.9% / -10.6%
h4_up +1.0% / +0.4% (up-transition, no lca() — noise)
h4_self -21.5% / -11.7%
h8_cross_tree -46.2% / -41.0% ← headline: deepest LCA stress
chain_self_emit -15.4% / -10.4%
Improvement scales with hierarchy depth, exactly as the O(H²) → O(H)
algorithmic change would predict.
All tests pass under --no-default-features, --features tokio, and
--features embassy.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
swaitsandClaude Opus 4.7