Re-baseline bench/baseline_compile.json, last ratified 2026-05-15
(20add51, mut.4-tidy). check_ms shifted +35..47% uniformly across all
12 fixtures (~1.8-3.3ms -> ~2.4-4.7ms); build_O0_ms rose ~10%
(90->100ms) but stayed inside its 20% tolerance and so never tripped.
RATIFY, not a regression. The decisive signal is that the check_ms
drift is *uniform* across every fixture regardless of complexity
(hello +35%, nested_pat +35%). Fixture-independent drift means the
cost lives in shared per-invocation work, not in any fixture-specific
checker path. Two concrete shared-cost growths fully account for it:
- examples/prelude.ail grew 116 -> 153 lines (+32%); every
`ail check` parses and typechecks the full prelude.
- the kernel_stub module is now injected alongside prelude on every
check (landed in raw-buf prep.3, 9339279, #33) — a whole extra
module parsed + typechecked per invocation.
Both are real modules added by wanted features, not a quadratic
blow-up. There is no fixable type-checker regression hiding here; the
+0.8ms is the parse+typecheck cost of more shared source. The throughput
baseline (bench/baseline.json) was already re-ratified 2026-05-28
(de4399d); this brings the compile baseline back in sync.
Corrects the record: the #44 cycle-close audit body (7321826) wrongly
attributed the compile_check exit-1 to transient machine load. main is
sacrosanct so 7321826 cannot be amended; #45 and this commit are the
forward correction.
Verified: compile_check.py -n5 -> 24 metrics, 0 regressed, 24 stable.
Tidy iter addressing the audit-mut-local drift. Plus paired baseline
update on bench/baseline_compile.json (audit-skill discipline).
Architect [high] items closed:
1. CheckError::MutVarCapturedByLambda rejects lambdas whose body
free vars include a mut-var of the enclosing mut_scope_stack.
Uses the existing ailang_core::desugar::free_vars_in_term walker
(which honours Term::Match pattern bindings). The scan runs only
when mut_scope_stack is non-empty.
2. crates/ailang-codegen/src/lambda.rs: the capture-not-in-locals
path that previously raised CodegenError::Internal blaming the
typechecker now uses unreachable!. The companion comment block
was rewritten to state the current reality.
Architect [medium] items closed (stale mut.1-stub history comments):
3. docs/DESIGN.md §'Term (expression)' mut/assign block: the
trailing paragraph describing the iter mut.1 stub state was
replaced with one describing the current mut.3-end-state. The
inline jsonc comment on {'t': 'assign'} was updated to drop the
'deferred to mut.2/mut.3' language.
4. crates/ailang-codegen/src/lib.rs: the stale comment block above
the real Term::Mut arm describing the mut.1 stub was removed
entirely.
Other:
- Short-circuit on empty mut_scope_stack in synth's Term::Var arm:
the iter mut.2 prepend now skips the iter-and-find walk when
the stack is empty, eliminating any per-Var-resolution overhead
for the common case (programs with no mut blocks). The
short-circuit did NOT close the check_ms regression — see ratify
below.
- Spec docs/specs/2026-05-15-mut-local.md §'Out of scope' amended
with the lambda-capture rejection bullet.
- Negative fixture examples/test_mut_var_captured_by_lambda.ail.json
+ driver test extension in crates/ailang-check/tests/
mut_typecheck_pin.rs (6th test) +
crates/ailang-core/tests/carve_out_inventory.rs EXPECTED bumped
12 → 13.
Bench-regression ratify:
bench/compile_check.py check_ms showed a uniform ~30-50% relative
shift across the entire 11-fixture suite (~0.5ms absolute on a
1.4-1.5ms baseline). The uniformity across fixtures of very
different Var counts argues for a fixed-cost-per-invocation tax,
not a Var-proportional hot path. The Term::Var short-circuit
falsified the hot-path hypothesis. The plausible remaining causes
(synth-parameter-passing through ~19 recursive sites, and binary-
size startup tax from mut-* adding ~1400 LOC to typecheck/codegen)
are both feature-cost, not pathological. Ratified by paired
journal entry; bench/baseline_compile.json updated to the post-
mut-local numbers via 'bench/compile_check.py --update-baseline'.
bench/check.py continues to show the established tail-latency-noise
envelope from audit-pd (2026-05-14) — no separate ratify needed.
bench/cross_lang.py clean.
Tests: 594 → 598 green (4 new lib.rs mod tests + 1 driver test +
1 fixture-corpus uptake).
Journal: docs/journals/2026-05-15-iter-mut.4-tidy.md.
mut-local milestone end-to-end status:
- mut.1 (7b92719): AST + Form A surface.
- mut.2 (b24718a): typecheck.
- mut.3 (03fb633): codegen + e2e.
- mut.4-tidy (this commit): audit-drift close + bench ratify.
mut-local milestone CLOSED.
Refs: docs/specs/2026-05-15-mut-local.md,
docs/plans/2026-05-15-iter-mut.4-tidy.md.
Milestone-23 close audit. Architect clean on substance; one
pre-existing DESIGN.md grammar stub (line 1692, dangling 'The
original' from 2026-05-10 design-md-consolidation) caught and
fixed.
Bencher localised the compile_check.py regression to H2 (the 5 new
polymorphic prelude Def::Fns iter 23.5 added — each carries forall
+ class constraint + match/not body, all typechecked on every
workspace check). H1 (linearity workspace-visibility extension +
contains_rigid_var) is negligible at today's 2-module workspace.
The regression is the feature's measured cost. Ratified per audit
skill (--update-baseline + journal entry). 0/24 metrics regressed
post-update.
Milestone 23 (Eq/Ord prelude) is closed.
Three drift items from ailang-architect, plus one false-positive
surfaced during verification:
1. DESIGN.md silent on closure-pair 4.14x finding (21'b).
Decision-10 ratified: "Workload scope of the 1.3x target"
paragraph scopes the retirement gate to linear/tree/poly-ADT
workloads; closure-pair carve-out documented as
representational cost (closure cell + env struct = 2 allocs
per step) until a slab/pool answer ships.
2. bench/compile_check.py corpus drift. Three fixtures added
(bench_compute_intsum, bench_compute_collatz,
bench_list_sum_explicit), re-baselined. Now 12 fixtures x
2 ops = 24 compile-time metrics.
3. baseline.json convention not codified. Note field gains
"max-of-distribution gets wider band than percentile"
convention discovered in 21'd.
4. (verification finding) bench_compute_intsum cross_lang
tolerances at 15%/12% fire on subprocess-spawn jitter for
sub-millisecond fixtures. Widened to 35% across all five
intsum metrics; convention recorded in baseline_cross_lang
note field (sub-ms fixtures need looser bands).
All three bench gates re-run sequentially after edits:
bench/check.py — 63 metrics; 63 stable
bench/compile_check.py — 24 metrics; 24 stable
bench/cross_lang.py — 25 metrics; 25 stable
Total: 112 metrics under regression coverage, all green.
288 tests passing, 3 ignored. No Rust changes.
Closes the second axis the user named: every typechecker / codegen
perf change was previously invisible to the tidy-iter gate. With
Family 21 typeclasses (queued) and 21'b's poly-ADT additions both
pushing on the typechecker, naive substitution loops would have
landed silently and decayed the compile path.
bench/compile_check.py is a separate script from bench/check.py
because the methodology differs: sub-process spawn timing on small
workloads (1ms scale for `ail check`, 65ms for `ail build`) vs.
allocator-stress on large ones (multi-second). Tolerances differ
by an order of magnitude (25% / 20% here vs. 5-15% there).
Empirically: ail check is sub-ms across the corpus, dominated by
subprocess spawn (~5-10ms on Linux); ail build is 63-69ms,
dominated by clang's link step. The bench is a catastrophe
detector (10x slowdowns visible) — finer regressions need a
profiler. CLAUDE.md updated to list both scripts as co-equal
tidy-iter gates alongside the architect drift report; exit 0/1/2
semantics are uniform across both.
JOURNAL queue: 21'd (pure-compute fixtures) and 21'e (cross-
language reference / hand-C ratio) remain to land the LLVM-
linkable performance claim.
Brummel