1da8754d45
**問題:**
- Larson 4T で 100% SEGV (1T は 2.09M ops/s で完走)
- System/mimalloc は 4T で 33.52M ops/s 正常動作
- SS OFF + Remote OFF でも 4T で SEGV
**根本原因: (Task agent ultrathink 調査結果)**
```
CRASH: mov (%r15),%r13
R15 = 0x6261 ← ASCII "ba" (ゴミ値、未初期化TLS)
```
Worker スレッドの TLS 変数が未初期化:
- `__thread void* g_tls_sll_head[TINY_NUM_CLASSES];` ← 初期化なし
- pthread_create() で生成されたスレッドでゼロ初期化されない
- NULL チェックが通過 (0x6261 != NULL) → dereference → SEGV
**修正内容:**
全 TLS 配列に明示的初期化子 `= {0}` を追加:
1. **core/hakmem_tiny.c:**
- `g_tls_sll_head[TINY_NUM_CLASSES] = {0}`
- `g_tls_sll_count[TINY_NUM_CLASSES] = {0}`
- `g_tls_live_ss[TINY_NUM_CLASSES] = {0}`
- `g_tls_bcur[TINY_NUM_CLASSES] = {0}`
- `g_tls_bend[TINY_NUM_CLASSES] = {0}`
2. **core/tiny_fastcache.c:**
- `g_tiny_fast_cache[TINY_FAST_CLASS_COUNT] = {0}`
- `g_tiny_fast_count[TINY_FAST_CLASS_COUNT] = {0}`
- `g_tiny_fast_free_head[TINY_FAST_CLASS_COUNT] = {0}`
- `g_tiny_fast_free_count[TINY_FAST_CLASS_COUNT] = {0}`
3. **core/hakmem_tiny_magazine.c:**
- `g_tls_mags[TINY_NUM_CLASSES] = {0}`
4. **core/tiny_sticky.c:**
- `g_tls_sticky_ss[TINY_NUM_CLASSES][TINY_STICKY_RING] = {0}`
- `g_tls_sticky_idx[TINY_NUM_CLASSES][TINY_STICKY_RING] = {0}`
- `g_tls_sticky_pos[TINY_NUM_CLASSES] = {0}`
**効果:**
```
Before: 1T: 2.09M ✅ | 4T: SEGV 💀
After: 1T: 2.41M ✅ | 4T: 4.19M ✅ (+15% 1T, SEGV解消)
```
**テスト:**
```bash
# 1 thread: 完走
./larson_hakmem 2 8 128 1024 1 12345 1
→ Throughput = 2,407,597 ops/s ✅
# 4 threads: 完走（以前は SEGV）
./larson_hakmem 2 8 128 1024 1 12345 4
→ Throughput = 4,192,155 ops/s ✅
```
**調査協力:** Task agent (ultrathink mode) による完璧な根本原因特定
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
79 lines
3.9 KiB
Bash
Executable File
79 lines
3.9 KiB
Bash
Executable File
#!/usr/bin/env bash
|
||
set -euo pipefail
|
||
|
||
# Larson defaults wrapper for HAKMEM Tiny.
|
||
# Usage:
|
||
# scripts/run_larson_defaults.sh [tput|pf] [DUR] [THREADS]
|
||
# Examples:
|
||
# scripts/run_larson_defaults.sh # 2s, threads=1,4, tput profile
|
||
# scripts/run_larson_defaults.sh pf 10 4 # 10s, threads=4, low PF profile
|
||
|
||
MODE=${1:-tput} # tput (throughput) or pf (lower page-fault/sys)
|
||
DUR=${2:-2}
|
||
THREADS_CSV=${3:-1,4}
|
||
|
||
ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")"/.. && pwd)"
|
||
cd "$ROOT_DIR"
|
||
|
||
# Ensure binaries
|
||
[[ -x ./larson_system ]] || make -s larson_system >/dev/null
|
||
[[ -x ./larson_hakmem ]] || make -s larson_hakmem >/dev/null
|
||
|
||
# Common env for HAKMEM Tiny + Larson
|
||
export HAKMEM_LARSON_TINY_ONLY=${HAKMEM_LARSON_TINY_ONLY:-1}
|
||
export HAKMEM_LARSON=${HAKMEM_LARSON:-1}
|
||
export HAKMEM_WRAP_TINY=${HAKMEM_WRAP_TINY:-1}
|
||
export HAKMEM_DISABLE_BATCH=${HAKMEM_DISABLE_BATCH:-1}
|
||
export HAKMEM_TINY_META_ALLOC=${HAKMEM_TINY_META_ALLOC:-1}
|
||
export HAKMEM_TINY_META_FREE=${HAKMEM_TINY_META_FREE:-1}
|
||
export HAKMEM_TINY_USE_SUPERSLAB=${HAKMEM_TINY_USE_SUPERSLAB:-1}
|
||
export HAKMEM_TINY_MUST_ADOPT=${HAKMEM_TINY_MUST_ADOPT:-1}
|
||
export HAKMEM_TINY_SS_ADOPT=${HAKMEM_TINY_SS_ADOPT:-1}
|
||
export HAKMEM_TINY_TRIM_SS=${HAKMEM_TINY_TRIM_SS:-1}
|
||
|
||
# Publish→mail→adopt pipeline (fast-tier + mailbox) — defaults safe for bench
|
||
# These are critical; keep here so we never forget them.
|
||
export HAKMEM_TINY_FAST_CAP=${HAKMEM_TINY_FAST_CAP:-64}
|
||
export HAKMEM_TINY_FAST_SPARE_PERIOD=${HAKMEM_TINY_FAST_SPARE_PERIOD:-8}
|
||
export HAKMEM_TINY_TLS_LIST=${HAKMEM_TINY_TLS_LIST:-1}
|
||
# Mailbox slow discovery to avoid missed registrations in short runs
|
||
export HAKMEM_TINY_MAILBOX_SLOWDISC=${HAKMEM_TINY_MAILBOX_SLOWDISC:-1}
|
||
export HAKMEM_TINY_MAILBOX_SLOWDISC_PERIOD=${HAKMEM_TINY_MAILBOX_SLOWDISC_PERIOD:-256}
|
||
|
||
# Sukesuke (signal dump) defaults off; counters at exit only to avoid early noise
|
||
export HAKMEM_TINY_SUKESUKE=${HAKMEM_TINY_SUKESUKE:-0}
|
||
export HAKMEM_TINY_DUMP_ATEXIT_ONLY=${HAKMEM_TINY_DUMP_ATEXIT_ONLY:-1}
|
||
export HAKMEM_TINY_COUNTERS_DUMP=${HAKMEM_TINY_COUNTERS_DUMP:-1}
|
||
|
||
# Profile-specific knobs
|
||
if [[ "$MODE" == "tput" ]]; then
|
||
# Throughput-leaning defaults
|
||
export HAKMEM_TINY_SS_FORCE_LG=${HAKMEM_TINY_SS_FORCE_LG:-21} # 2MB
|
||
export HAKMEM_TINY_PUBLISH_THROTTLE=${HAKMEM_TINY_PUBLISH_THROTTLE:-8}
|
||
export HAKMEM_TINY_DRAIN_THRESHOLD=${HAKMEM_TINY_DRAIN_THRESHOLD:-4}
|
||
# Prefer mmap over adopt for raw tput until publish pipeline is proven
|
||
export HAKMEM_TINY_MUST_ADOPT=${HAKMEM_TINY_MUST_ADOPT:-0}
|
||
# SS cache/precharge ON also for tput（syscall抑制で張り付き解消を狙う）
|
||
export HAKMEM_TINY_SS_CACHE=${HAKMEM_TINY_SS_CACHE:-8}
|
||
export HAKMEM_TINY_SS_PRECHARGE=${HAKMEM_TINY_SS_PRECHARGE:-1}
|
||
export HAKMEM_TINY_TRIM_SS=${HAKMEM_TINY_TRIM_SS:-0}
|
||
else
|
||
# Lower page-fault/sys defaults
|
||
export HAKMEM_TINY_SS_FORCE_LG=${HAKMEM_TINY_SS_FORCE_LG:-20} # 1MB
|
||
export HAKMEM_TINY_PUBLISH_THROTTLE=${HAKMEM_TINY_PUBLISH_THROTTLE:-8}
|
||
export HAKMEM_TINY_DRAIN_THRESHOLD=${HAKMEM_TINY_DRAIN_THRESHOLD:-4}
|
||
# Enable adopt on PF/sys profile to reduce mmap/sys
|
||
export HAKMEM_TINY_MUST_ADOPT=${HAKMEM_TINY_MUST_ADOPT:-1}
|
||
export HAKMEM_TINY_SS_CACHE=${HAKMEM_TINY_SS_CACHE:-4}
|
||
export HAKMEM_TINY_SS_PRECHARGE=${HAKMEM_TINY_SS_PRECHARGE:-1}
|
||
fi
|
||
|
||
echo "[Larson Defaults] mode=$MODE dur=$DUR threads=$THREADS_CSV"
|
||
echo "[Larson Defaults] HAKMEM_LARSON_TINY_ONLY=$HAKMEM_LARSON_TINY_ONLY HAKMEM_WRAP_TINY=$HAKMEM_WRAP_TINY"
|
||
echo "[Larson Defaults] SS_FORCE_LG=$HAKMEM_TINY_SS_FORCE_LG THROTTLE=$HAKMEM_TINY_PUBLISH_THROTTLE DRAIN=$HAKMEM_TINY_DRAIN_THRESHOLD CACHE=$HAKMEM_TINY_SS_CACHE PRECHARGE=$HAKMEM_TINY_SS_PRECHARGE"
|
||
echo "[Larson Defaults] FAST_CAP=$HAKMEM_TINY_FAST_CAP FAST_SPARE_PERIOD=$HAKMEM_TINY_FAST_SPARE_PERIOD TLS_LIST=$HAKMEM_TINY_TLS_LIST"
|
||
echo "[Larson Defaults] MAILBOX_SLOWDISC=$HAKMEM_TINY_MAILBOX_SLOWDISC PERIOD=$HAKMEM_TINY_MAILBOX_SLOWDISC_PERIOD"
|
||
|
||
# Delegate to perf runner (system/mimalloc/hakmem)
|
||
DUR=$DUR THREADS=$THREADS_CSV scripts/run_larson_perf.sh
|