Debug Counters Implementation - Clean History

Major Features:
- Debug counter infrastructure for Refill Stage tracking
- Free Pipeline counters (ss_local, ss_remote, tls_sll)
- Diagnostic counters for early return analysis
- Unified larson.sh benchmark runner with profiles
- Phase 6-3 regression analysis documentation

Bug Fixes:
- Fix SuperSlab disabled by default (HAKMEM_TINY_USE_SUPERSLAB)
- Fix profile variable naming consistency
- Add .gitignore patterns for large files

Performance:
- Phase 6-3: 4.79 M ops/s (has OOM risk)
- With SuperSlab: 3.13 M ops/s (+19% improvement)

This is a clean repository without large log files.

🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>

benchmarks/scripts/utils/run_larson.sh

@@ -0,0 +1,170 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# Reproducible larson runner for hakmem/system/mimalloc.
+#
+# Usage:
+#   scripts/run_larson.sh [runtime_sec] [threads]
+# Examples:
+#   scripts/run_larson.sh                 # default: 10s, threads=1 4
+#   scripts/run_larson.sh 10 1            # 10s, 1 thread
+#
+# Optional env vars:
+#   HAKMEM_WRAP_TINY=0|1
+#   HAKMEM_WRAP_TINY_REFILL=0|1
+#   HAKMEM_TINY_MAG_CAP=INT
+#   HAKMEM_SAFE_FREE=0|1
+#   HAKMEM_EVO_SAMPLE=INT  (0 disables evo recording; default 0)
+#   MIMALLOC_SO=/path/to/libmimalloc.so.2 (optional; if not set, auto-detect)
+
+usage() {
+  cat << USAGE
+Usage: scripts/run_larson.sh [options] [runtime_sec] [threads_csv]
+
+Options:
+  -d SECONDS     Runtime seconds (default: 10)
+  -t CSV         Threads CSV, e.g. 1,4 (default: 1,4)
+  -c NUM         Chunks per thread (default: 10000)
+  -r NUM         Rounds (default: 1)
+  -m BYTES       Min size (default: 8)
+  -M BYTES       Max size (default: 1024)
+  -s SEED        Random seed (default: 12345)
+  -p PRESET      Preset: burst|loop (sets -c/-r)
+  -w             Include WRAP_TINY=1 runs (default: off)
+  -h             Show this help
+
+Env overrides (alternative to flags):
+  MIN, MAX, CHUNK_PER_THREAD, ROUNDS, SEED
+  HAKMEM_* toggles per README
+USAGE
+}
+
+# Defaults
+RUNTIME="10"
+THREADS_ARG="1,4"
+
+# Workload defaults (burst preset)
+MIN="${MIN:-8}"
+MAX="${MAX:-1024}"
+CHUNK_PER_THREAD="${CHUNK_PER_THREAD:-10000}"
+ROUNDS="${ROUNDS:-1}"
+SEED="${SEED:-12345}"
+
+PRESET=""
+
+INCLUDE_WRAP=0
+while getopts ":d:t:c:r:m:M:s:p:wh" opt; do
+  case $opt in
+    d) RUNTIME="$OPTARG" ;;
+    t) THREADS_ARG="$OPTARG" ;;
+    c) CHUNK_PER_THREAD="$OPTARG" ;;
+    r) ROUNDS="$OPTARG" ;;
+    m) MIN="$OPTARG" ;;
+    M) MAX="$OPTARG" ;;
+    s) SEED="$OPTARG" ;;
+    p) PRESET="$OPTARG" ;;
+    w) INCLUDE_WRAP=1 ;;
+    h) usage; exit 0 ;;
+    :) echo "Missing argument for -$OPTARG" >&2; usage; exit 2 ;;
+    *) usage; exit 2 ;;
+  esac
+done
+shift $((OPTIND-1))
+
+# Backward-compatible positional args
+if [[ $# -ge 1 ]]; then RUNTIME="$1"; fi
+if [[ $# -ge 2 ]]; then THREADS_ARG="$2"; fi
+
+case "$PRESET" in
+  burst|BURST)
+    CHUNK_PER_THREAD=10000; ROUNDS=1 ;;
+  loop|LOOP)
+    CHUNK_PER_THREAD=100; ROUNDS=100 ;;
+  "" ) : ;;
+  *) echo "Unknown preset: $PRESET" >&2; exit 2 ;;
+esac
+
+# Params matching our standard runs (larson reads: runtime, min, max, chunks/thread, rounds, seed, threads)
+# Show resolved parameters for reproducibility
+echo "[CFG] runtime=${RUNTIME}s threads={${THREADS_ARG}} min=${MIN} max=${MAX} chunks/thread=${CHUNK_PER_THREAD} rounds=${ROUNDS} seed=${SEED}"
+
+ROOT_DIR="$(cd "$(dirname "$0")/.." && pwd)"
+LIB_HAKMEM="$ROOT_DIR/libhakmem.so"
+LARSON_BIN="$ROOT_DIR/mimalloc-bench/bench/larson/larson"
+
+if [[ ! -x "$LARSON_BIN" ]]; then
+  echo "[ERR] Larson binary not found at: $LARSON_BIN" >&2
+  echo "      Did you sync submodule/build bench?" >&2
+  exit 1
+fi
+
+if [[ ! -f "$LIB_HAKMEM" ]]; then
+  echo "[INFO] libhakmem.so not found; building..."
+  (cd "$ROOT_DIR" && make -j4 shared >/dev/null)
+fi
+
+abs_hakmem="$(readlink -f "$LIB_HAKMEM")"
+
+detect_mimalloc() {
+  if [[ -n "${MIMALLOC_SO:-}" && -f "$MIMALLOC_SO" ]]; then
+    echo "$MIMALLOC_SO"
+    return 0
+  fi
+  # try common paths or ldconfig
+  for p in \
+    /usr/lib/x86_64-linux-gnu/libmimalloc.so.2 \
+    /lib/x86_64-linux-gnu/libmimalloc.so.2; do
+    [[ -f "$p" ]] && { echo "$p"; return 0; }
+  done
+  if command -v ldconfig >/dev/null 2>&1; then
+    so="$(ldconfig -p | awk '/libmimalloc.so/ {print $4; exit}')"
+    [[ -n "$so" && -f "$so" ]] && { echo "$so"; return 0; }
+  fi
+  return 1
+}
+
+mimalloc_so=""
+if mimalloc_so=$(detect_mimalloc); then
+  :
+else
+  mimalloc_so=""
+fi
+
+run_case() {
+  local label="$1"; shift
+  local preload="$1"; shift
+  local threads="$1"; shift
+
+  echo "\n== $label | ${threads}T | ${RUNTIME}s =="
+  if [[ -n "$preload" ]]; then
+    env LD_PRELOAD="$preload" "$LARSON_BIN" "$RUNTIME" "$MIN" "$MAX" "$CHUNK_PER_THREAD" "$ROUNDS" "$SEED" "$threads" 2>&1 | tail -n 3
+  else
+    "$LARSON_BIN" "$RUNTIME" "$MIN" "$MAX" "$CHUNK_PER_THREAD" "$ROUNDS" "$SEED" "$threads" 2>&1 | tail -n 3
+  fi
+}
+
+IFS=',' read -r -a THREADS <<< "$THREADS_ARG"
+
+for t in "${THREADS[@]}"; do
+  # system malloc
+  run_case "system malloc" "" "$t"
+
+  # mimalloc (optional)
+  if [[ -n "$mimalloc_so" ]]; then
+    run_case "mimalloc" "$mimalloc_so" "$t"
+  else
+    echo "\n== mimalloc | ${t}T | ${RUNTIME}s =="
+    echo "[SKIP] libmimalloc not found"
+  fi
+
+  # hakmem default
+  run_case "hakmem (default)" "$abs_hakmem" "$t"
+
+  # hakmem wrap tiny (optional)
+  if [[ "$INCLUDE_WRAP" -eq 1 ]]; then
+    echo "\n== hakmem (HAKMEM_WRAP_TINY=1) | ${t}T | ${RUNTIME}s =="
+    HAKMEM_WRAP_TINY=1 LD_PRELOAD="$abs_hakmem" "$LARSON_BIN" "$RUNTIME" "$MIN" "$MAX" "$CHUNK_PER_THREAD" "$ROUNDS" "$SEED" "$t" 2>&1 | tail -n 3
+  fi
+done
+
+echo "\nDone."