Spinlocks and Read-Write Locks
Spinlocks and Read-Write Locks Most parallel programming in some way will involve the use of locking at the lowest levels. Locks are primitives that provide mutual exclusion that allow data structures to remain in consistent states. Without locking, multiple threads of execution may simultaneously modify a data structure. Without a carefully thought out (and usually complex) lock-free algorithm, t
次期C++ã«å°Žå…¥ã•ã‚Œã‚‹ãƒ¡ãƒ¢ãƒªãƒãƒªã‚¢ã«ã¤ã„ã¦è§£èª¬ã—ã¦ã¿ã‚‹ - yamasaã®ãƒã‚¿å¸³
å‰ã®ã‚¨ãƒ³ãƒˆãƒªã§æ¬¡æœŸC++標準(通称C++0x)ã«atomicåž‹ã¨ãƒ¡ãƒ¢ãƒªãƒãƒªã‚¢ãŒå°Žå…¥ã•ã‚Œã‚‹ã¨ã„ã†è©±ã‚’ã—ã¾ã—ãŸã€‚今回ã¯ãã®C++ã§ã®å®Ÿè£…ã«ã¤ã„ã¦ã€ã‚‚ã†å°‘ã—æ·±ã追ã„ã‹ã‘ã¦ã¿ã¾ã™ã€‚ スライド資料ã§ã¯ã€Œatomicæ“作 ＋ acquire/releaseãƒãƒªã‚¢ã€ãŒåŸºæœ¬ã§ã‚ã‚‹ã¨æ›¸ãã¾ã—ãŸãŒã€å®Ÿéš›ã«æ¬¡æœŸC++ã«å°Žå…¥ã•ã‚Œã‚‹äºˆå®šã®atomic APIã¯ã€ã‚‚ã†å°‘ã—複雑ãªä»•æ§˜ã«ãªã£ã¦ã„ã¾ã™ã€‚一番ã®é•ã„ã¯ã€ãƒ¡ãƒ¢ãƒªãƒãƒªã‚¢ã®ç¨®é¡žãŒå¢—ãˆã¦ã„ã‚‹ã“ã¨ã§ã™ã€‚ 次期C++標準ã®ç¾åœ¨ã®ãƒ‰ãƒ©ãƒ•ãƒˆã§ã¯ã€ãƒ¡ãƒ¢ãƒªãƒãƒªã‚¢ã®ç¨®é¡žã‚’表ã™enumåž‹ã®å®šç¾©ã¯ä»¥ä¸‹ã®ã‚ˆã†ã«ãªã£ã¦ã„ã¾ã™ã€‚ namespace std { typedef enum memory_order { memory_order_relaxed, memory_order_consume, memory_order_acquire, memory_order_release, m
Lock-freeãƒ‡ãƒ¼ã‚¿æ§‹é€ ã§ä½¿ç”¨ã•ã‚Œã‚‹ABAé¿ã‘カウンタã«ã¤ã„ã¦
ãã¾ãŽ @kumagi boost.atomicã¯64bitå¹…ã®æ™‚ã«ä¸Šä½16bitã«ã‚«ã‚¦ãƒ³ã‚¿ã‚’埋ã‚込むã£ã¦ã„ã†ã®ãŒè³›å¦ã‚ã‚‹ã¿ãŸã„ã ã‘ã‚Œã©ã€å対æ„見ã®äººã¯ãã®bitæ•°ã¾ã§å¿…è¦ãªãƒ¡ãƒ¢ãƒªã‚’使ã†ã‚“ã§ã—ょã†ã‹ã€‚48ビットã‚ã‚Œã°262テラãƒã‚¤ãƒˆã¾ã§è¡Œã‘ã‚‹ã¯ãšâ€¦ã€‚ 2011-11-04 22:48:06 ã‚„ã¾ã• @yamasa @kumagi ç§ã¯ã‚«ã‚¦ãƒ³ã‚¿ãŒ16bitã§ã‚る点ãŒæ°—ã«ãªã‚Šã¾ã™ã。ãŸã¨ãˆã°3000クãƒãƒƒã‚¯ã«1ã®å‰²åˆã§ã‚«ã‚¦ãƒ³ãƒˆã‚¢ãƒƒãƒ—ã•ã‚Œã‚‹ã¨ä»®å®šã™ã‚‹ã¨ã€16bitカウンタãŒä¸€å‘¨ã™ã‚‹ã®ã«196Mサイクルã€2GHzã®CPUã ã¨100mså¼±ã§ã™ã€‚ 2011-11-05 00:13:25
HTMã¯ãƒ¡ãƒ¢ãƒªç®¡ç†ã®ç‚ºã«ç”Ÿã¾ã‚Œã¦ããŸã‚“ã よï¼ã€€Î©Î©ï¼œãªã€ãªã‚“ã ã£ã¦ãƒ¼ - Software Transactional Memo
HTMã®ç¬¬ä¸€äººè€…ã«ã—ã¦ã€obstruction-freeã‚„wait-freeãªã©ã®åŽ³å¯†ãªå®šç¾©ã‚„ã€CAS命令ã®æ•°å¦çš„ãªæ„義を証明ã—ã€The Art of Multiprocessor Programmingã®è‘—者ã§ã€ãƒãƒƒã‚¯ãƒ•ãƒªãƒ¼ã§ã‚‚STMã§ã‚‚å…«é¢å…臂ã®æ´»èºã‚’ã—ã¦ã„ã‚‹Maurice Herlihy先生ãŒé€£åã—ã¦ã„る最近ã®è«–文をæµã—èªã¿ã€‚ Aleksandar Dragojevic, Maurice Herlihy, Yossi Lev, Mark Moir: On the power of hardware transactional memory to simplify memory management. PODC 2011: 99-108 「トランザクショナルメモリã®ãƒ‘ワーã§ãƒ¡ãƒ¢ãƒªç®¡ç†ã‚’簡略化ã™ã‚‹ã€ã¨ã„ã†æ„Ÿã˜ã€‚ GCを使ã‚ãªã„環境下ã«ãŠã„ã¦ãƒ¡ãƒ¢ãƒªç®¡ç†ã¯é¬¼é–€ã«ãªã£ã¦ã¦ã€ãã®ã›ã„ã§C++ã®ãƒ
Quiescent State Based Reclamationã¨RCUã®è©±ã¨HTMã®è©±
ãã¾ãŽ @kumagi ブãƒã‚°æ›¸ã„㟠「userspace RCU(QSBR)ã®ä½¿ã„æ–¹ã¨è§£èª¬ã€ http://t.co/mvjPoU6M51 安全ã«ãƒãƒƒã‚¯ãƒ•ãƒªãƒ¼ã™ã‚‹ãŸã‚ã®å±€æ‰€çš„ãªã‚¬ãƒ™ãƒ¼ã‚¸ã‚³ãƒ¬ã‚¯ã‚¿ã¨ã‚‚言ãˆã‚‹ã€‚ Sadayuki Furuhashi @frsyuki @kumagi 全スレッドãŒrcu_quiescent_stateを定期的ã«å‘¼ã°ãªã„ã¨æˆç«‹ã—ãªã„æ°—ãŒã—ã¾ã™ãŒã€ãƒ‡ãƒ¼ã‚¿ã®èªã¿å‡ºã—ã‚’è¡Œã‚ãªã„スレッドã¨ã‹ã€ãŸã¾ã«è¡Œã‚ãªããªã‚‹ã‚¹ãƒ¬ãƒƒãƒ‰ã¯ã€ã©ã†ã„ã†ã‚¿ã‚¤ãƒŸãƒ³ã‚°ã§rcu_quiescent_stateを呼んã§ã„ã‚‹ã‚“ã§ã—ょ？ ãã¾ãŽ @kumagi @frsyuki ã‰ã€ãã®æ¬ 点ã«æ°—ã¥ã„ã¦ã—ã¾ã†ã¨ã¯ã•ã™ãŒã§ã™ã。他ã®UserspaceRCUãŒã¿ã‚“ãªRW-Lockã®ä»£ã‚ã‚Šã«ã‚‚使ãˆã‚‹ï¼ˆã—ã‹ã‚‚ä½ç½®ã‚„ã‚·ã‚°ãƒãƒãƒ£ã‚‚一致ã™ã‚‹ã®ã§ãƒã‚¤ãƒŠãƒªã®ãƒªãƒ³ã‚¯å…ˆã‚’変ãˆã‚‹ã ã‘ã§è‰¯ã„）一方ã§QSBRã¯ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ä¸ã«ä¾µå…¥çš„ã«å¤‰
userspace RCU(QSBR)ã®ä½¿ã„æ–¹ã¨è§£èª¬ - ãã¾ãƒ¡ãƒ¢
http://lttng.org/urcu|Userspace RCU ã¨ã„ã†å¤§å¤‰ã‚¯ãƒ¼ãƒ«ãªãƒ—ãƒã‚¸ã‚§ã‚¯ãƒˆãŒã‚ã‚Šã¾ã™ã€‚ 「RCU（リードコピーアップデート）をユーザースペースã§è¡Œã†ã‚‚ã®ã€ã¨ã„ã†äº‹ã§ã€ãã“ã ã‘èžãã¨ãªã‚“ã®ã“ã£ã¡ã‚ƒã¨ã„ã†æ„Ÿã˜ã€‚ リードコピーアップデートã£ã¦ä½•ã‚ˆ リードコピーアップデートãã®ã‚‚ã®ã®æ£ã—ã„説明ã¯Wikipedia*1 ã§ã‚‚èªã‚“ã§ã‚‚らã†ã¨ã—ã¦ã€Wikipediaã‚’èªã‚€ã®ã™ã‚‰é¢å€’ãªäººã®ç‚ºã«èª¬æ˜Žã™ã‚‹ã¨ã€Œã¿ã‚“ãªã§å…±æœ‰ã—ã¦ã‚‹ãƒ‡ãƒ¼ã‚¿ã‚’free()ã—ã¦ã‚‚良ã„タイミングを見極ã‚る技術ã€ã§ã™ã€‚ 特定ã®æ§‹é€ 体をã¿ã‚“ãªã§å…±æœ‰ã—ã¦æ›¸ãæ›ãˆãŸã‚Šã—ãŸã„ãªã€€â†’　Lockã¨ã‚Œã°ã‚ˆãã？ ã™ã”ãé »ç¹ã«ã‚¢ã‚¯ã‚»ã‚¹ã™ã‚‹ã—èªã¿å‡ºã—ã®æ–¹ãŒå¤šã„ã‹ã‚‰Lockã¯ã‚„ã ãªã€€â†’　Read-Write Lockã§ã‚ˆãã？ Read-Write Lockã ã¨Atomic命令多ã™ãŽã¦ãƒ‘フォーマンスã§ãªã„ãªã€€â†’　Lockç„¡ãã—ã¦ãƒ‡ãƒ¼
1
リリースã€éšœå®³æƒ…å ±ãªã©ã®ã‚µãƒ¼ãƒ“スã®ãŠçŸ¥ã‚‰ã›
最新ã®äººæ°—エントリーã®é…ä¿¡
j次ã®ãƒ–ックマーク
kå‰ã®ãƒ–ックマーク
lã‚ã¨ã§èªã‚€
eコメント一覧を開ã
oページを開ã
Studying Intel TSX Performance
{{#tags}}- {{label}}
{{/tags}}