tomoaki/hakmem
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
552 Commits 2 Branches 0 Tags
b2724e6f5d848dc96dba4b4b75b332e891e95ca1
Commit Graph

12 Commits

Author SHA1 Message Date
Moe Charm (CI)
bdfa32d869 Phase v4-mid-SEGV: C6 v4 を SmallSegment 専用に切り替え、TinyHeap SEGV を解決 
問題: C6 v4 が TinyHeap のページを共有することで iters >= 800k で freelist 破壊 → SEGV 発生

修正内容:
- c6_segment_alloc_page_direct(): C6 専用ページ割当 (SmallSegment v4 経由, TinyHeap 非共有)
- c6_segment_release_page_direct(): C6 専用ページ返却
- cold_refill_page_v4() で C6 を分岐: SmallSegment 直接使用
- cold_retire_page_v4() で C6 を分岐: SmallSegment に直接返却
- fastlist state reset 処理追加 (L392-399)

結果:
✅ iters=1M, ws <= 390 で SEGV 消失
✅ C6-only: v4 OFF ~47M → v4 ON ~43M ops/s (−8.5%, 安定)
✅ Mixed: v4 ON で SEGV なし (小幅回帰許容)

方針: C6 v4 は研究箱として安定化完了。本線には載せない (既存 mid/pool v1 使用)。

🤖 Generated with Claude Code

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2025-12-11 02:39:32 +09:00
Moe Charm (CI)
e486dd2c55 Phase v4-mid-6: Implement C6 v4 TLS Fastlist (Gated) 
- Implemented TLS fastlist logic for C6 in smallobject_hotbox_v4.c (alloc/free).
- Added SmallC6FastState struct and g_small_c6_fast TLS variable.
- Gated the fastlist logic with HAKMEM_SMALL_HEAP_V4_FASTLIST (default OFF) due to observed instability in mixed workloads.
- Fixed a memory leak in small_heap_free_fast_v4 fallback path by calling hak_pool_free.
- Updated CURRENT_TASK.md with phase report.
 2025-12-11 01:44:08 +09:00
Moe Charm (CI)
dd974b49c5 Phase v4-mid-2, v4-mid-3, v4-mid-5: SmallObject HotBox v4 implementation and docs update 
Implementation:
- SmallObject HotBox v4 (core/smallobject_hotbox_v4.c) now fully implements C6-only allocations and frees, including current/partial management and freelist operations.
- Cold Iface (tiny_heap based) for page refill/retire is integrated.
- Stats instrumentation (v4-mid-5) added to small_heap_alloc_fast_v4 and small_heap_free_fast_v4, with a new header file core/box/smallobject_hotbox_v4_stats_box.h and atexit dump function.

Updates:
- CURRENT_TASK.md has been condensed and updated with summaries of Phase v4-mid-2 (C6-only v4), Phase v4-mid-3 (C5-only v4 pilot), and the stats implementation (v4-mid-5).
- docs/analysis/SMALLOBJECT_V4_BOX_DESIGN.md updated with A/B results and conclusions for C6-only and C5-only v4 implementations.
- The previous CURRENT_TASK.md content has been archived to CURRENT_TASK_ARCHIVE_20251210.md.
 2025-12-11 01:01:15 +09:00
Moe Charm (CI)
3b4449d773 Phase v4-mid-1: C6-only v4 route + page_meta_of() Fail-Fast validation 
Implementation:
- SMALL_SEGMENT_V4_* constants (SIZE=2MiB, PAGE_SIZE=64KiB, MAGIC=0xDEADBEEF)
- smallsegment_v4_page_meta_of(): O(1) mask+shift lookup with magic validation
  - Computes segment base: addr & ~(2MiB - 1)
  - Verifies SmallSegment magic number
  - Calculates page_idx: (addr - seg_base) >> PAGE_SHIFT (16)
  - Returns non-NULL sentinel for now (full page_meta[] in Phase v4-mid-2)

Stubs for C6-only phase:
- small_heap_alloc_fast_v4(): C6 returns NULL → pool v1 fallback
- small_heap_free_fast_v4(): C6 calls page_meta_of() for Fail-Fast, then pool v1 fallback

Documentation:
- ENV_PROFILE_PRESETS.md: Add "C6_ONLY_SMALLOBJECT_V4" research profile
  - HAKMEM_SMALL_HEAP_V4_ENABLED=1, HAKMEM_SMALL_HEAP_V4_CLASSES=0x40
  - Expected: Throughput ≈ 28–29M ops/s (same as v1)

Build:
- ビルド成功（警告のみ）
- Backward compatible, alloc/free stubs fall back to pool v1

Sanity:
- C6-heavy with v4 opt-in: segv/assert なし
- page_meta_of() lookup working correctly
- Performance unchanged (expected for stub phase)

Status:
- C6-only v4 route now available via ENV opt-in
- Phase v4-mid-2: SmallHeapCtx v4 full implementation with A/B

🤖 Generated with Claude Code

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2025-12-10 23:37:45 +09:00
Moe Charm (CI)
e3e4cab833 Cleanup: Unify type naming and Cold Iface architecture 
Refactoring:
- Type naming: Rename small_page_v4 → SmallPageMeta, small_class_heap_v4 → SmallClassHeap, small_heap_ctx_v4 → SmallHeapCtx
- Keep backward compatibility aliases for existing code
- SmallSegment struct unified, clean forward declarations
- Cold Iface: Remove vtable (SmallColdIfaceV4 struct) in favor of direct function calls
- Simplify refill_page/retire_page to direct calls, not callbacks
- smallobject_hotbox_v4.c: Update to call small_cold_v4_* functions directly

Documentation:
- Add docs/analysis/ENV_CLEANUP_CANDIDATES.md
- Categorize ENVs: KEEP (production), RESEARCH (opt-in), DELETE (obsolete)
- v2 code: Keep as research infrastructure (complete, safe, gated)
- v4 code: Research scaffold for future mid-level allocator

Build:
- ビルド成功（警告のみ）
- Backward compatible, all existing code still works

🤖 Generated with Claude Code

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2025-12-10 23:30:32 +09:00
Moe Charm (CI)
52c65da783 Phase v4-mid-0: Small-object v4 型・IF 足場（箱化モジュール化） 
- SmallHeapCtx/SmallPageMeta/SmallClassHeap typedef alias 追加
- SmallSegment struct (base/num_pages/owner_tid/magic) を smallsegment_v4_box.h に定義
- SmallColdIface_v4 direct function prototypes (refill/retire/remote_push/drain)
- smallobject_hotbox_v4.c の internal/public API 分離（small_segment_v4_internal）
- direct function stubs 実装（SmallColdIfaceV4 delegate 形式）
- ENV OFF デフォルト（ENABLED=0/CLASSES=0）で既存挙動 100% 不変
- ビルド成功・sanity 確認（mixed/C6-heavy、segv/assert なし）
- CURRENT_TASK.md に Phase v4-mid-0 記録

🤖 Generated with Claude Code

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2025-12-10 23:23:07 +09:00
Moe Charm (CI)
2a13478dc7 Optimize C6 heavy and C7 ultra performance analysis with refined design refinements 
- Update environment profile presets and visibility analysis
- Enhance small object and tiny segment v4 box implementations
- Refine C7 ultra and C6 heavy allocation strategies
- Add comprehensive performance metrics and design documentation

🤖 Generated with Claude Code

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2025-12-10 22:57:26 +09:00
Moe Charm (CI)
f2ce7256cd Add v4 C7/C6 fast classify and small-segment v4 scaffolding 2025-12-10 19:14:38 +09:00
Moe Charm (CI)
3261025995 Phase v4-4: pilot C6 v4 route with opt-in gate 2025-12-10 18:18:05 +09:00
Moe Charm (CI)
7be30c0b5a Avoid full-list scans for C7 v4 and tighten partial reuse 2025-12-10 18:04:32 +09:00
Moe Charm (CI)
860d934d71 Tune C7 v4 partial reuse for mixed perf 2025-12-10 18:03:28 +09:00
Moe Charm (CI)
cbd33511eb Phase v4-3.1: reuse C7 v4 pages and record prep calls 2025-12-10 17:58:42 +09:00