bench: 21'f — explicit-mode pair, full alloc+dec vs malloc+free

Closes the apples-to-apples gap from 21'e. Adds:
- examples/bench_list_sum_explicit.ailx — same algorithm and sizes
  as bench_list_sum, fully (borrow)/(own)/(drop-iterative)
  annotated so codegen emits proper inc/dec instrumentation.
- bench/reference/list_sum_explicit_free.c — same algorithm
  with explicit free() walking the chain after sum.

The full alloc+dec vs malloc+free comparison reveals two non-
trivial conclusions:

1. AILang's full RC pipeline is only 26% slower than glibc
   malloc+free on this workload (rc/c = 1.26x). The implicit-
   mode comparison's 1.42x was misleading — it counted neither
   pipeline's free path. The fair ratio is 1.26x, materially
   better than the previous read.

2. RC's dec is cheaper per cell than glibc free(). AILang
   dec-tax: ~3 ns/cell. C free-tax: ~5.5 ns/cell. Plausible
   cause: ailang_rc_dec operates on a known-shape cell with a
   fixed-offset refcount and a static per-type drop fn — no
   free-list bucketing, no header introspection, no global lock.

bump's advantage expresses fully: bench_list_sum_explicit.bump/c
= 0.42x means AILang at bump is 2.4x faster than C malloc+free.
Sets a useful upper bound on a slab/pool RC allocator's potential.

The 21'-family arc — bench-regression infrastructure — is now
substantively complete: 21'a (bench/check.py), 21'b (corpus
widening), 21'c (compile_check.py), 21'd (pure-compute fixtures
+ harness hardening), 21'e (cross-language hand-C), 21'f (explicit
apples-to-apples). 63 runtime metrics + 18 compile metrics + 25
cross-lang metrics under regression coverage. Any future iter
that regresses any axis beyond tolerance gets caught at the next
family close.

Remaining queue is back to substantive language work — Family 21
(typeclasses / polymorphic ADTs at runtime / pattern-binding
generalisation) is now an orchestrator-level fork that needs
direct user input.

bench/baseline.json

@@ -54,6 +54,16 @@
       "gc_rss_kb":     { "baseline": 13624,  "tolerance_pct": 15 },
       "bump_rss_kb":   { "baseline": 13860,  "tolerance_pct": 15 },
       "rc_rss_kb":     { "baseline": 13880,  "tolerance_pct": 15 }
+    },
+    "bench_list_sum_explicit": {
+      "gc_s":          { "baseline": 0.139,  "tolerance_pct": 10 },
+      "bump_s":        { "baseline": 0.048,  "tolerance_pct": 10 },
+      "rc_s":          { "baseline": 0.152,  "tolerance_pct": 10 },
+      "gc_over_bump":  { "baseline": 2.89,   "tolerance_pct": 8  },
+      "rc_over_bump":  { "baseline": 3.14,   "tolerance_pct": 8  },
+      "gc_rss_kb":     { "baseline": 103772, "tolerance_pct": 5  },
+      "bump_rss_kb":   { "baseline": 97636,  "tolerance_pct": 5  },
+      "rc_rss_kb":     { "baseline": 142424, "tolerance_pct": 8  }
     }
   },
 

bench/baseline_cross_lang.json

@@ -6,89 +6,111 @@
   "fixtures": {
     "bench_list_sum": {
       "ail_rc_s": {
-        "baseline": 0.141597,
+        "baseline": 0.138906,
         "tolerance_pct": 15
       },
       "ail_bump_s": {
-        "baseline": 0.048038,
+        "baseline": 0.050245,
         "tolerance_pct": 15
       },
       "c_s": {
-        "baseline": 0.095312,
+        "baseline": 0.09749,
         "tolerance_pct": 15
       },
       "rc_over_c": {
-        "baseline": 1.485614,
+        "baseline": 1.42483,
         "tolerance_pct": 12
       },
       "bump_over_c": {
-        "baseline": 0.504003,
+        "baseline": 0.515393,
         "tolerance_pct": 12
       }
     },
     "bench_tree_walk": {
       "ail_rc_s": {
-        "baseline": 0.097028,
+        "baseline": 0.100173,
         "tolerance_pct": 15
       },
       "ail_bump_s": {
-        "baseline": 0.038905,
+        "baseline": 0.038673,
         "tolerance_pct": 15
       },
       "c_s": {
-        "baseline": 0.037159,
+        "baseline": 0.036567,
         "tolerance_pct": 15
       },
       "rc_over_c": {
-        "baseline": 2.611185,
+        "baseline": 2.739449,
         "tolerance_pct": 12
       },
       "bump_over_c": {
-        "baseline": 1.046998,
+        "baseline": 1.057606,
         "tolerance_pct": 12
       }
     },
     "bench_compute_intsum": {
       "ail_rc_s": {
-        "baseline": 0.00044,
+        "baseline": 0.000362,
         "tolerance_pct": 15
       },
       "ail_bump_s": {
-        "baseline": 0.000393,
+        "baseline": 0.000371,
         "tolerance_pct": 15
       },
       "c_s": {
-        "baseline": 0.000374,
+        "baseline": 0.000361,
         "tolerance_pct": 15
       },
       "rc_over_c": {
-        "baseline": 1.177715,
+        "baseline": 1.002676,
         "tolerance_pct": 12
       },
       "bump_over_c": {
-        "baseline": 1.050856,
+        "baseline": 1.028255,
         "tolerance_pct": 12
       }
     },
     "bench_compute_collatz": {
       "ail_rc_s": {
-        "baseline": 0.056878,
+        "baseline": 0.057994,
         "tolerance_pct": 15
       },
       "ail_bump_s": {
-        "baseline": 0.056547,
+        "baseline": 0.057159,
         "tolerance_pct": 15
       },
       "c_s": {
-        "baseline": 0.05748,
+        "baseline": 0.05776,
         "tolerance_pct": 15
       },
       "rc_over_c": {
-        "baseline": 0.989523,
+        "baseline": 1.00405,
         "tolerance_pct": 12
       },
       "bump_over_c": {
-        "baseline": 0.983756,
+        "baseline": 0.989593,
+        "tolerance_pct": 12
+      }
+    },
+    "bench_list_sum_explicit": {
+      "ail_rc_s": {
+        "baseline": 0.150754,
+        "tolerance_pct": 15
+      },
+      "ail_bump_s": {
+        "baseline": 0.0495,
+        "tolerance_pct": 15
+      },
+      "c_s": {
+        "baseline": 0.119204,
+        "tolerance_pct": 15
+      },
+      "rc_over_c": {
+        "baseline": 1.264667,
+        "tolerance_pct": 12
+      },
+      "bump_over_c": {
+        "baseline": 0.415253,
         "tolerance_pct": 12
       }
     }

bench/cross_lang.py

@@ -47,10 +47,11 @@ OUTDIR = ROOT / "target" / "cross_lang"
 # (ailang fixture stem, C reference stem) — both must exist as
 # `examples/<ail>.ail.json` and `bench/reference/<c>.c`.
 CORPUS = [
-    ("bench_list_sum",        "list_sum"),
-    ("bench_tree_walk",       "tree_walk"),
-    ("bench_compute_intsum",  "compute_intsum"),
-    ("bench_compute_collatz", "compute_collatz"),
+    ("bench_list_sum",          "list_sum"),
+    ("bench_tree_walk",         "tree_walk"),
+    ("bench_compute_intsum",    "compute_intsum"),
+    ("bench_compute_collatz",   "compute_collatz"),
+    ("bench_list_sum_explicit", "list_sum_explicit_free"),
 ]
 
 

bench/reference/list_sum_explicit_free.c

@@ -0,0 +1,76 @@
+// Hand-C reference for bench_list_sum_explicit — full malloc + free.
+//
+// Companion to list_sum.c (which leaks); this version adds explicit
+// `free()` calls to mirror AILang's explicit-mode RC dec-tax. Pairs
+// with examples/bench_list_sum_explicit.ailx via bench/cross_lang.py.
+//
+// The free pass walks the list after sum, freeing each cell. This is
+// genuinely O(N) work that bench_list_sum.c never paid; the rc/c
+// ratio for this fixture is therefore the honest answer to "how
+// much does RC's dec-and-free pipeline cost vs hand-C's malloc-free
+// pipeline on the same workload?".
+//
+// Note: we sum first, then free, rather than fold-with-free. This
+// matches the AILang fixture's structure: sum_list owns the chain
+// and consumes it; the dec walk happens implicitly as sum_acc
+// pattern-matches LCons and re-binds the tail to the recursive call.
+// AILang's pattern there is "consume during sum" — codegen emits
+// dec on the consumed cells as the match arm closes. The C version
+// could in principle interleave free into the sum loop too; doing it
+// in two passes is structurally cleaner and the cost is the same
+// (one O(N) walk + one O(N) walk-and-free).
+//
+// Build: clang -O2 -o list_sum_explicit_free list_sum_explicit_free.c
+// Expected stdout (one int per line):
+//   4999950000
+//   499999500000
+//   4499998500000
+
+#include <stdio.h>
+#include <stdlib.h>
+
+typedef struct cell {
+    long head;
+    struct cell *tail;
+} cell_t;
+
+static cell_t *cons_n(long n) {
+    cell_t *acc = NULL;
+    for (long i = n - 1; i >= 0; i--) {
+        cell_t *c = (cell_t *) malloc(sizeof(cell_t));
+        c->head = i;
+        c->tail = acc;
+        acc = c;
+    }
+    return acc;
+}
+
+static long sum_list(const cell_t *xs) {
+    long acc = 0;
+    while (xs) {
+        acc += xs->head;
+        xs = xs->tail;
+    }
+    return acc;
+}
+
+static void free_list(cell_t *xs) {
+    while (xs) {
+        cell_t *next = xs->tail;
+        free(xs);
+        xs = next;
+    }
+}
+
+static void run_one(long n) {
+    cell_t *xs = cons_n(n);
+    printf("%ld\n", sum_list(xs));
+    free_list(xs);
+}
+
+int main(void) {
+    run_one(100000);
+    run_one(1000000);
+    run_one(3000000);
+    return 0;
+}

bench/run.sh

@@ -61,7 +61,7 @@ mkdir -p "$OUTDIR"
 
 # Compile both modes for both fixtures up front so the bench loop only
 # measures runtime, not build time.
-fixtures=(bench_list_sum bench_tree_walk bench_closure_chain bench_hof_pipeline bench_compute_collatz)
+fixtures=(bench_list_sum bench_tree_walk bench_closure_chain bench_hof_pipeline bench_compute_collatz bench_list_sum_explicit)
 modes=(gc bump rc)
 echo ">>> compiling fixtures (-O2)"
 for f in "${fixtures[@]}"; do