Java’da \u00e7oklu i\u015f par\u00e7ac\u0131kl\u0131 (multithreaded) uygulamalar geli\u015ftirirken, payla\u015f\u0131lan de\u011fi\u015fkenlerin g\u00fcvenli bir \u015fekilde g\u00fcncellenmesi en kritik konulardan biridir. Java’da Peki neden b\u00f6yle bir s\u0131n\u0131fa ihtiyac\u0131m\u0131z var? Bir \u00f6rnekle a\u00e7\u0131klayal\u0131m: Bir sunucunun gelen istek say\u0131s\u0131n\u0131 tuttu\u011fu bir saya\u00e7 d\u00fc\u015f\u00fcn\u00fcn. Bu saya\u00e7, E\u011fer iki i\u015f par\u00e7ac\u0131\u011f\u0131 (Thread A ve Thread B) ayn\u0131 anda bu Beklentimiz sayac\u0131n 2 olmas\u0131yken, sonu\u00e7 1 oldu. Bu t\u00fcr hatalar, \u00e7oklu i\u015f par\u00e7ac\u0131kl\u0131 uygulamalarda tespit edilmesi ve d\u00fczeltilmesi en zor hatalardand\u0131r.<\/p>\n Geleneksel olarak bu sorunu \u00e7\u00f6zmek i\u00e7in Yukar\u0131da bahsedilen yar\u0131\u015f durumu, payla\u015f\u0131lan ve de\u011fi\u015ftirilebilir (mutable) durumun oldu\u011fu her yerde ortaya \u00e7\u0131kabilir. CAS, i\u015flemci seviyesinde atomik olarak ger\u00e7ekle\u015ftirilen bir komuttur. \u00dc\u00e7 ana arg\u00fcman al\u0131r:<\/p>\n CAS i\u015flemi \u015fu \u015fekilde \u00e7al\u0131\u015f\u0131r: “E\u011fer bellek konumundaki V de\u011feri, A de\u011ferine e\u015fitse, o zaman V’yi B de\u011feriyle g\u00fcncelle. Aksi takdirde, g\u00fcncelleme yapma.” Bu i\u015flem tek bir atomik ad\u0131mda ger\u00e7ekle\u015fti\u011fi i\u00e7in, bu kontrol ve g\u00fcncelleme aras\u0131nda ba\u015fka hi\u00e7bir i\u015f par\u00e7ac\u0131\u011f\u0131 araya giremez.<\/p>\n E\u011fer CAS i\u015flemi ba\u015far\u0131l\u0131 olursa (V == A), Bu mekanizma, 1. Kurucu Metotlar:<\/strong><\/p>\n 2. De\u011fer Okuma ve Yazma:<\/strong><\/p>\n 3. Atomik Art\u0131rma ve Azaltma:<\/strong><\/p>\n 4. Atomik Ekleme ve Kar\u015f\u0131la\u015ft\u0131rma\/Takas Etme (CAS):<\/strong><\/p>\n Bu metotlar\u0131n her biri, kendi i\u00e7inde atomik olarak y\u00fcr\u00fct\u00fcl\u00fcr ve birden fazla i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n ayn\u0131 anda eri\u015fmesi durumunda dahi do\u011fru sonucu garantiler.<\/p>\n 1. G\u00fcvenli Saya\u00e7lar:<\/strong> En temel ve yayg\u0131n kullan\u0131m senaryosudur. Bir uygulaman\u0131n ka\u00e7 i\u015flem yapt\u0131\u011f\u0131n\u0131, ka\u00e7 kullan\u0131c\u0131n\u0131n giri\u015f yapt\u0131\u011f\u0131n\u0131 veya ka\u00e7 iste\u011fin al\u0131nd\u0131\u011f\u0131n\u0131 takip etmek i\u00e7in kullan\u0131l\u0131r.<\/p>\n 2. ID \u00dcretimi:<\/strong> Benzersiz kimlikler (ID) \u00fcretmek i\u00e7in kullan\u0131labilir. Her yeni nesne veya i\u015flem i\u00e7in artan bir say\u0131sal ID sa\u011flamak.<\/p>\n 3. Basit Bayraklar\/Durum G\u00f6stergeleri:<\/strong> Bir kayna\u011f\u0131n kullan\u0131mda olup olmad\u0131\u011f\u0131n\u0131 veya bir i\u015flemin tamamlan\u0131p tamamlanmad\u0131\u011f\u0131n\u0131 belirtmek i\u00e7in bir bayrak olarak kullan\u0131labilir.<\/p>\n 1. Her ikisi de haf\u0131za tutarl\u0131l\u0131\u011f\u0131 sa\u011flar, yani bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n yapt\u0131\u011f\u0131 de\u011fi\u015fikli\u011fi di\u011ferleri g\u00f6r\u00fcr. Ancak 2. 3. \u00c7o\u011fu zaman, \u00f6zellikle d\u00fc\u015f\u00fck ve orta \u00e7eki\u015fmeli senaryolarda 4. Sadece Hay\u0131r, 5. <\/body><\/p>\n","protected":false},"excerpt":{"rendered":"Java’da \u00e7oklu i\u015f par\u00e7ac\u0131kl\u0131 (multithreaded) uygulamalar geli\u015ftirirken, payla\u015f\u0131lan de\u011fi\u015fkenlerin g\u00fcvenli bir \u015fekilde g\u00fcncellenmesi en kr","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"csco_page_header_type":"","csco_page_load_nextpost":"","csco_page_subscribe_form":"","csco_page_contact_form":"","footnotes":""},"categories":[1],"tags":[],"class_list":{"0":"post-35891","1":"post","2":"type-post","3":"status-publish","4":"format-standard","6":"category-genel","7":"cs-entry","8":"cs-video-wrap"},"yoast_head":"\nint<\/code> gibi ilkel veri tipleri \u00fczerinde basit g\u00f6r\u00fcnen art\u0131rma\/azaltma i\u015flemleri bile, birden fazla i\u015f par\u00e7ac\u0131\u011f\u0131 ayn\u0131 anda eri\u015fti\u011finde beklenmedik ve hatal\u0131 sonu\u00e7lara yol a\u00e7abilir. \u0130\u015fte bu noktada Java’n\u0131n java.util.concurrent.atomic<\/code> paketi ve bu paketin \u00f6nemli bir \u00fcyesi olan AtomicInteger<\/code> devreye girer. Bu rehberde, AtomicInteger<\/code>‘\u0131n ne oldu\u011funu, neden bu kadar \u00f6nemli oldu\u011funu, nas\u0131l \u00e7al\u0131\u015ft\u0131\u011f\u0131n\u0131 ve ger\u00e7ek d\u00fcnya senaryolar\u0131nda nas\u0131l kullan\u0131laca\u011f\u0131n\u0131 detayl\u0131 bir \u015fekilde inceleyece\u011fiz.<\/p>\n
\nGiri\u015f: AtomicInteger Nedir ve Neden \u00d6nemlidir?<\/h2>\n
AtomicInteger<\/code>, bir int<\/code> de\u011ferini atomik (b\u00f6l\u00fcnemez) operasyonlarla manip\u00fcle etmek i\u00e7in tasarlanm\u0131\u015f bir s\u0131n\u0131ft\u0131r. Atomik operasyonlar, bir i\u015flem dizisinin tamam\u0131n\u0131n tek bir ad\u0131mda, kesintiye u\u011framadan ger\u00e7ekle\u015fti\u011fi anlam\u0131na gelir. Ba\u015fka bir deyi\u015fle, bir i\u015f par\u00e7ac\u0131\u011f\u0131 atomik bir i\u015flem yaparken, di\u011fer i\u015f par\u00e7ac\u0131klar\u0131 bu i\u015flemi tamamlanana kadar g\u00f6remez veya m\u00fcdahale edemez.<\/p>\nint requestCount = 0;<\/code> olarak tan\u0131mlanm\u0131\u015f olsun. Her istek geldi\u011finde requestCount++<\/code> i\u015flemiyle art\u0131r\u0131l\u0131yor. G\u00f6r\u00fcn\u00fc\u015fte basit olan bu i\u015flem, asl\u0131nda \u00fc\u00e7 ayr\u0131 CPU talimat\u0131ndan olu\u015fur:<\/p>\n\n
requestCount<\/code> de\u011ferini bellekten oku.<\/li>\nrequestCount<\/code> de\u011fi\u015fkenine geri yaz.<\/li>\n<\/ol>\nrequestCount++<\/code> i\u015flemini yapmaya \u00e7al\u0131\u015f\u0131rsa, bir “yar\u0131\u015f durumu” (race condition) ortaya \u00e7\u0131kar. \u00d6rne\u011fin:<\/p>\n\n
requestCount<\/code> ba\u015flang\u0131\u00e7ta 0 olsun.<\/li>\nrequestCount<\/code> de\u011ferini (0) okur.<\/li>\nrequestCount<\/code> de\u011ferini (hala 0) okur.<\/li>\nrequestCount<\/code>‘a yazar. (requestCount<\/code> \u015fimdi 1).<\/li>\nrequestCount<\/code>‘a yazar. (requestCount<\/code> hala 1).<\/li>\n<\/ul>\nsynchronized<\/code> anahtar kelimesi kullan\u0131l\u0131rd\u0131. \u00d6rne\u011fin:<\/p>\n\npublic class Counter {\n private int count = 0;\n\n public synchronized void increment() {\n count++;\n }\n\n public synchronized int getCount() {\n return count;\n }\n}\n<\/code><\/pre>\nsynchronized<\/code> blo\u011fu veya metodu, ayn\u0131 anda yaln\u0131zca bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n belirli bir kod b\u00f6l\u00fcm\u00fcne eri\u015fmesine izin vererek yar\u0131\u015f durumlar\u0131n\u0131 engeller. Ancak synchronized<\/code>‘\u0131n da kendi dezavantajlar\u0131 vard\u0131r: daha y\u00fcksek overhead (ek y\u00fck), potansiyel kilitlenmeler (deadlock) ve daha az \u00f6l\u00e7eklenebilirlik. AtomicInteger<\/code> ise, daha d\u00fc\u015f\u00fck seviyeli ve genellikle daha performansl\u0131 olan “kilitsiz” (lock-free) veya “non-blocking” algoritmalar kullanarak bu sorunlar\u0131 \u00e7\u00f6zer.<\/p>\n
\nConcurrency Sorunlar\u0131 ve AtomicInteger’\u0131n Rol\u00fc<\/h2>\n
AtomicInteger<\/code>, bu t\u00fcr durumlar i\u00e7in \u00f6zel olarak tasarlanm\u0131\u015ft\u0131r ve temel olarak tek bir int<\/code> de\u011ferinin thread-safe bir \u015fekilde g\u00fcncellenmesini sa\u011flar. Bunu yaparken, CPU’nun donan\u0131msal deste\u011fiyle \u00e7al\u0131\u015fan “Kar\u015f\u0131la\u015ft\u0131r ve Takas Et” (Compare-And-Swap – CAS) ad\u0131 verilen bir mekanizmadan yararlan\u0131r.<\/p>\nCAS (Compare-And-Swap) Prensibi<\/h3>\n
\n
true<\/code> d\u00f6ner ve de\u011fer g\u00fcncellenir. E\u011fer ba\u015far\u0131s\u0131z olursa (V != A), false<\/code> d\u00f6ner ve de\u011fer g\u00fcncellenmez, \u00e7\u00fcnk\u00fc ba\u015fka bir i\u015f par\u00e7ac\u0131\u011f\u0131 de\u011feri biz kontrol ettikten sonra de\u011fi\u015ftirmi\u015ftir. Bu durumda, genellikle i\u015flemi yeniden denemek i\u00e7in bir d\u00f6ng\u00fc kullan\u0131l\u0131r (spin-lock benzeri bir yakla\u015f\u0131m).<\/p>\nAtomicInteger<\/code> gibi s\u0131n\u0131flar\u0131n dahili olarak kilitsiz bir \u015fekilde \u00e7al\u0131\u015fmas\u0131n\u0131 sa\u011flar. \u0130\u015f par\u00e7ac\u0131klar\u0131 birbirini kilitlemek yerine, de\u011fer beklendi\u011fi gibi de\u011filse i\u015flemi yeniden denerler. Bu, \u00f6zellikle d\u00fc\u015f\u00fck ve orta seviyede \u00e7eki\u015fme (contention) olan durumlarda synchronized<\/code> bloklar\u0131ndan \u00e7ok daha verimli olabilir.<\/p>\n
\nAtomicInteger Nas\u0131l \u00c7al\u0131\u015f\u0131r? (\u0130\u00e7 Yap\u0131s\u0131 ve Metotlar\u0131)<\/h2>\n
AtomicInteger<\/code> s\u0131n\u0131f\u0131, volatile int value;<\/code> \u015feklinde bir i\u00e7 de\u011fi\u015fken tutar ve t\u00fcm operasyonlar\u0131n\u0131 bu value<\/code> \u00fczerinde CAS mekanizmas\u0131 ile ger\u00e7ekle\u015ftirir. volatile<\/code> anahtar kelimesi, de\u011fi\u015fkenin de\u011ferinin her zaman ana bellekten okunmas\u0131n\u0131 ve yaz\u0131lmas\u0131n\u0131 garanti ederek \u00f6nbellek tutars\u0131zl\u0131klar\u0131n\u0131 \u00f6nler.<\/p>\nTemel Metotlar ve Kullan\u0131mlar\u0131<\/h3>\n
AtomicInteger<\/code>‘\u0131n sa\u011flad\u0131\u011f\u0131 temel metotlar \u015funlard\u0131r:<\/p>\n\nimport java.util.concurrent.atomic.AtomicInteger;\n\npublic class AtomicIntegerExample {\n public static void main(String[] args) {\n \/\/ Ba\u015flang\u0131\u00e7 de\u011feri 0 olan bir AtomicInteger olu\u015fturur\n AtomicInteger counter = new AtomicInteger(); \n System.out.println(\"Initial counter: \" + counter.get()); \/\/ \u00c7\u0131kt\u0131: 0\n\n \/\/ Ba\u015flang\u0131\u00e7 de\u011feri 10 olan bir AtomicInteger olu\u015fturur\n AtomicInteger anotherCounter = new AtomicInteger(10);\n System.out.println(\"Another counter: \" + anotherCounter.get()); \/\/ \u00c7\u0131kt\u0131: 10\n }\n}\n<\/code><\/pre>\n\n
get()<\/code>: Mevcut de\u011feri d\u00f6nd\u00fcr\u00fcr.<\/li>\nset(newValue)<\/code>: De\u011feri belirtilen yeni de\u011fere ayarlar.<\/li>\ngetAndSet(newValue)<\/code>: Mevcut de\u011feri d\u00f6nd\u00fcr\u00fcr ve ard\u0131ndan de\u011feri yeni de\u011fere ayarlar.<\/li>\n<\/ul>\n\nAtomicInteger value = new AtomicInteger(5);\nSystem.out.println(\"Current value: \" + value.get()); \/\/ \u00c7\u0131kt\u0131: 5\n\nvalue.set(7);\nSystem.out.println(\"New value: \" + value.get()); \/\/ \u00c7\u0131kt\u0131: 7\n\nint oldValue = value.getAndSet(10);\nSystem.out.println(\"Old value: \" + oldValue); \/\/ \u00c7\u0131kt\u0131: 7\nSystem.out.println(\"Updated value: \" + value.get()); \/\/ \u00c7\u0131kt\u0131: 10\n<\/code><\/pre>\n\n
incrementAndGet()<\/code>: De\u011feri 1 art\u0131r\u0131r ve art\u0131r\u0131lm\u0131\u015f yeni de\u011feri d\u00f6nd\u00fcr\u00fcr.<\/li>\ngetAndIncrement()<\/code>: De\u011feri 1 art\u0131r\u0131r, ancak art\u0131rma i\u015fleminden \u00f6nceki eski de\u011feri d\u00f6nd\u00fcr\u00fcr.<\/li>\ndecrementAndGet()<\/code>: De\u011feri 1 azalt\u0131r ve azalt\u0131lm\u0131\u015f yeni de\u011feri d\u00f6nd\u00fcr\u00fcr.<\/li>\ngetAndDecrement()<\/code>: De\u011feri 1 azalt\u0131r, ancak azaltma i\u015fleminden \u00f6nceki eski de\u011feri d\u00f6nd\u00fcr\u00fcr.<\/li>\n<\/ul>\n\nAtomicInteger count = new AtomicInteger(0);\n\nint newValue1 = count.incrementAndGet(); \/\/ count = 1\nSystem.out.println(\"Incremented (new value): \" + newValue1); \/\/ \u00c7\u0131kt\u0131: 1\n\nint oldValue1 = count.getAndIncrement(); \/\/ count = 2\nSystem.out.println(\"Incremented (old value): \" + oldValue1); \/\/ \u00c7\u0131kt\u0131: 1\n\nint newValue2 = count.decrementAndGet(); \/\/ count = 1\nSystem.out.println(\"Decremented (new value): \" + newValue2); \/\/ \u00c7\u0131kt\u0131: 1\n\nint oldValue2 = count.getAndDecrement(); \/\/ count = 0\nSystem.out.println(\"Decremented (old value): \" + oldValue2); \/\/ \u00c7\u0131kt\u0131: 1\n<\/code><\/pre>\n\n
addAndGet(delta)<\/code>: De\u011fere belirtilen delta<\/code> de\u011ferini ekler ve yeni de\u011feri d\u00f6nd\u00fcr\u00fcr.<\/li>\ngetAndAdd(delta)<\/code>: De\u011fere belirtilen delta<\/code> de\u011ferini ekler, ancak ekleme i\u015fleminden \u00f6nceki eski de\u011feri d\u00f6nd\u00fcr\u00fcr.<\/li>\ncompareAndSet(expectedValue, updateValue)<\/code>: CAS i\u015flemini ger\u00e7ekle\u015ftirir. E\u011fer mevcut de\u011fer expectedValue<\/code>‘a e\u015fitse, de\u011feri updateValue<\/code> ile de\u011fi\u015ftirir ve true<\/code> d\u00f6nd\u00fcr\u00fcr. Aksi takdirde, hi\u00e7bir de\u011fi\u015fiklik yapmaz ve false<\/code> d\u00f6nd\u00fcr\u00fcr.<\/li>\nweakCompareAndSet(expectedValue, updateValue)<\/code>: compareAndSet<\/code>‘e benzer ancak daha zay\u0131f haf\u0131za tutarl\u0131l\u0131\u011f\u0131 garantileri sunar. Baz\u0131 durumlarda daha h\u0131zl\u0131 olabilir, ancak genellikle compareAndSet<\/code> tercih edilir.<\/li>\n<\/ul>\n\nAtomicInteger balance = new AtomicInteger(100);\n\nint newBalance = balance.addAndGet(50); \/\/ balance = 150\nSystem.out.println(\"Added (new balance): \" + newBalance); \/\/ \u00c7\u0131kt\u0131: 150\n\nint oldBalance = balance.getAndAdd(-20); \/\/ balance = 130\nSystem.out.println(\"Added (old balance): \" + oldBalance); \/\/ \u00c7\u0131kt\u0131: 150\n\n\/\/ CAS \u00d6rne\u011fi\nboolean casResult1 = balance.compareAndSet(130, 200); \/\/ Mevcut de\u011fer 130 oldu\u011fu i\u00e7in ba\u015far\u0131l\u0131\nSystem.out.println(\"CAS successful (130 -> 200)? \" + casResult1); \/\/ \u00c7\u0131kt\u0131: true\nSystem.out.println(\"Current balance: \" + balance.get()); \/\/ \u00c7\u0131kt\u0131: 200\n\nboolean casResult2 = balance.compareAndSet(150, 250); \/\/ Mevcut de\u011fer 150 olmad\u0131\u011f\u0131 i\u00e7in ba\u015far\u0131s\u0131z\nSystem.out.println(\"CAS successful (150 -> 250)? \" + casResult2); \/\/ \u00c7\u0131kt\u0131: false\nSystem.out.println(\"Current balance: \" + balance.get()); \/\/ \u00c7\u0131kt\u0131: 200 (de\u011fi\u015fmedi)\n<\/code><\/pre>\n
\nAtomicInteger Kullan\u0131m Senaryolar\u0131 ve Pratik \u0130pu\u00e7lar\u0131<\/h2>\n
AtomicInteger<\/code>, basit saya\u00e7lar ve durum g\u00f6stergeleri gibi bir\u00e7ok yayg\u0131n \u00e7oklu i\u015f par\u00e7ac\u0131kl\u0131 senaryoda m\u00fckemmel bir \u00e7\u00f6z\u00fcmd\u00fcr.<\/p>\nYayg\u0131n Kullan\u0131m Senaryolar\u0131<\/h3>\n
\nimport java.util.concurrent.atomic.AtomicInteger;\nimport java.util.concurrent.ExecutorService;\nimport java.util.concurrent.Executors;\nimport java.util.concurrent.TimeUnit;\n\npublic class RequestCounter {\n private static AtomicInteger requestCount = new AtomicInteger(0);\n private static final int NUM_THREADS = 10;\n private static final int NUM_REQUESTS_PER_THREAD = 1000;\n\n public static void main(String[] args) throws InterruptedException {\n ExecutorService executor = Executors.newFixedThreadPool(NUM_THREADS);\n\n for (int i = 0; i < NUM_THREADS; i++) {\n executor.submit(() -> {\n for (int j = 0; j < NUM_REQUESTS_PER_THREAD; j++) {\n requestCount.incrementAndGet(); \/\/ Atomik art\u0131rma\n }\n });\n }\n\n executor.shutdown();\n executor.awaitTermination(1, TimeUnit.MINUTES);\n\n \/\/ Toplam beklenen istek say\u0131s\u0131: NUM_THREADS * NUM_REQUESTS_PER_THREAD\n System.out.println(\"Final request count: \" + requestCount.get()); \/\/ Beklenen: 10000\n }\n}\n<\/code><\/pre>\n\nimport java.util.concurrent.atomic.AtomicInteger;\n\npublic class IdGenerator {\n private static AtomicInteger nextId = new AtomicInteger(0);\n\n public static int getNextId() {\n return nextId.incrementAndGet();\n }\n\n public static void main(String[] args) {\n System.out.println(\"Generated ID: \" + getNextId()); \/\/ 1\n System.out.println(\"Generated ID: \" + getNextId()); \/\/ 2\n \/\/ Farkl\u0131 thread'lerden \u00e7a\u011fr\u0131ld\u0131\u011f\u0131nda dahi benzersiz ve s\u0131ral\u0131 ID'ler \u00fcretir.\n }\n}\n<\/code><\/pre>\nPerformans Kar\u015f\u0131la\u015ft\u0131rmas\u0131 ve Ne Zaman Kullan\u0131lmal\u0131\/Kullan\u0131lmamal\u0131<\/h3>\n
AtomicInteger<\/code> genellikle synchronized<\/code> bloklar\u0131na g\u00f6re daha iyi performans sunar, \u00f6zellikle d\u00fc\u015f\u00fck ve orta derecede i\u015f par\u00e7ac\u0131\u011f\u0131 \u00e7eki\u015fmesi olan durumlarda. Bunun nedeni, synchronized<\/code>'\u0131n i\u015fletim sistemi d\u00fczeyinde kilitlenme (mutex) mekanizmalar\u0131n\u0131 kullan\u0131rken, AtomicInteger<\/code>'\u0131n donan\u0131m destekli CAS operasyonlar\u0131na dayanmas\u0131d\u0131r. CAS, bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n kilitlenmesini gerektirmez, ba\u015far\u0131s\u0131z olursa sadece yeniden dener.<\/p>\nNe Zaman Kullan\u0131lmal\u0131:<\/h4>\n
\n
int<\/code> de\u011ferinin atomik olarak g\u00fcncellenmesi, okunmas\u0131 veya de\u011fi\u015ftirilmesi gerekti\u011finde.<\/li>\nsynchronized<\/code>'\u0131n getirdi\u011fi ek y\u00fckten ka\u00e7\u0131nmak ve daha iyi \u00f6l\u00e7eklenebilirlik elde etmek istendi\u011finde.<\/li>\nNe Zaman Kullan\u0131lmamal\u0131:<\/h4>\n
\n
List<\/code>, Map<\/code> veya \u00f6zel nesneler gibi karma\u015f\u0131k veri yap\u0131lar\u0131 \u00fczerinde atomik i\u015flemler yapmak i\u00e7in AtomicInteger<\/code> uygun de\u011fildir. Bu t\u00fcr durumlar i\u00e7in ConcurrentHashMap<\/code>, CopyOnWriteArrayList<\/code> veya ReentrantLock<\/code> gibi daha \u00fcst d\u00fczey konkurent koleksiyonlar veya kilit mekanizmalar\u0131 kullan\u0131lmal\u0131d\u0131r.<\/li>\nAtomicInteger<\/code> veya ba\u015fka de\u011fi\u015fkenin birbiriyle ili\u015fkili olarak atomik bir \u015fekilde g\u00fcncellenmesi gerekiyorsa, her bir AtomicInteger<\/code>'\u0131n kendi operasyonu atomik olsa bile, t\u00fcm bu operasyonlar dizisi atomik olmaz. Bu durumda hala synchronized<\/code> veya ReentrantLock<\/code> gibi daha geni\u015f kapsaml\u0131 bir kilit mekanizmas\u0131na ihtiyac\u0131n\u0131z olabilir.<\/li>\n<\/ul>\nAtomicInteger<\/code>'\u0131n sa\u011flad\u0131\u011f\u0131 \"happens-before\" garantileri sayesinde, bir i\u015f par\u00e7ac\u0131\u011f\u0131 taraf\u0131ndan yap\u0131lan bir AtomicInteger<\/code> g\u00fcncellemesinin di\u011fer i\u015f par\u00e7ac\u0131klar\u0131 taraf\u0131ndan g\u00f6r\u00fcn\u00fcr olmas\u0131 sa\u011flan\u0131r. Bu, volatile<\/code> anahtar kelimesinin sa\u011flad\u0131\u011f\u0131 haf\u0131za tutarl\u0131l\u0131\u011f\u0131na benzer bir davran\u0131\u015ft\u0131r ve \u00f6nbellek tutars\u0131zl\u0131klar\u0131n\u0131 engeller.<\/p>\n
\nSonu\u00e7 ve S\u0131k\u00e7a Sorulan Sorular (SSS)<\/h2>\n
AtomicInteger<\/code>, Java'n\u0131n konkurent programlama ara\u00e7 kutusunun vazge\u00e7ilmez bir par\u00e7as\u0131d\u0131r. Basit int<\/code> de\u011ferleri \u00fczerinde g\u00fcvenli, verimli ve kilitsiz atomik operasyonlar sa\u011flamas\u0131 sayesinde, \u00f6zellikle saya\u00e7lar, ID \u00fcrete\u00e7leri ve basit durum y\u00f6netimi gibi senaryolarda synchronized<\/code> anahtar kelimesine \u00fcst\u00fcn bir alternatif sunar. Do\u011fru kullan\u0131ld\u0131\u011f\u0131nda, \u00e7oklu i\u015f par\u00e7ac\u0131kl\u0131 uygulamalar\u0131n\u0131z\u0131n hem do\u011frulu\u011funu hem de performans\u0131n\u0131 \u00f6nemli \u00f6l\u00e7\u00fcde art\u0131rabilir.<\/p>\nS\u0131k\u00e7a Sorulan Sorular (SSS)<\/h3>\n
AtomicInteger<\/code> ve volatile int<\/code> aras\u0131ndaki fark nedir?<\/strong><\/p>\nvolatile int<\/code> sadece okuma ve yazma i\u015flemlerinin g\u00f6r\u00fcn\u00fcrl\u00fc\u011f\u00fcn\u00fc garanti eder, atomikli\u011fi de\u011fil. Yani volatile int i = 0; i++;<\/code> i\u015flemi hala bir yar\u0131\u015f durumuna neden olabilir \u00e7\u00fcnk\u00fc i++<\/code> atomik de\u011fildir. AtomicInteger<\/code> ise hem g\u00f6r\u00fcn\u00fcrl\u00fc\u011f\u00fc hem de incrementAndGet()<\/code> gibi karma\u015f\u0131k operasyonlar\u0131n atomikli\u011fini garanti eder.<\/p>\nAtomicInteger<\/code> ve synchronized<\/code> aras\u0131ndaki fark nedir?<\/strong><\/p>\nsynchronized<\/code>, bir kod blo\u011funa veya metoda ayn\u0131 anda yaln\u0131zca bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n eri\u015fmesine izin veren bir kilitleme mekanizmas\u0131d\u0131r. Bu, birden fazla de\u011fi\u015fken veya karma\u015f\u0131k i\u015flemler \u00fczerinde atomikli\u011fi sa\u011flamak i\u00e7in genel bir \u00e7\u00f6z\u00fcmd\u00fcr. AtomicInteger<\/code> ise tek bir int<\/code> de\u011feri \u00fczerinde kilitsiz (non-blocking) atomik operasyonlar sa\u011flamak i\u00e7in \u00f6zel olarak tasarlanm\u0131\u015ft\u0131r. Genellikle synchronized<\/code>'dan daha hafiftir ve daha iyi performans sunar, ancak kapsam\u0131 daha s\u0131n\u0131rl\u0131d\u0131r.<\/p>\nAtomicInteger<\/code> her zaman synchronized<\/code>'dan daha m\u0131 h\u0131zl\u0131d\u0131r?<\/strong><\/p>\nAtomicInteger<\/code> daha h\u0131zl\u0131d\u0131r \u00e7\u00fcnk\u00fc kilitlenme maliyetinden ka\u00e7\u0131n\u0131r. Ancak \u00e7ok y\u00fcksek \u00e7eki\u015fmeli senaryolarda, i\u015f par\u00e7ac\u0131klar\u0131n\u0131n CAS ba\u015far\u0131s\u0131zl\u0131klar\u0131 nedeniyle s\u00fcrekli yeniden denemesi (spinning) CPU zaman\u0131n\u0131 bo\u015fa harcayabilir ve bu durumda synchronized<\/code> veya ReentrantLock<\/code> gibi bloklama kilitleri daha iyi performans g\u00f6sterebilir. Performans her zaman \u00f6l\u00e7\u00fclerek do\u011frulanmal\u0131d\u0131r.<\/p>\nint<\/code> i\u00e7in mi atomik s\u0131n\u0131flar var?<\/strong><\/p>\njava.util.concurrent.atomic<\/code> paketi di\u011fer ilkel tipler ve nesneler i\u00e7in de atomik s\u0131n\u0131flar sunar:<\/p>\n\n
AtomicLong<\/code>: long<\/code> de\u011ferleri i\u00e7in.<\/li>\nAtomicBoolean<\/code>: boolean<\/code> de\u011ferleri i\u00e7in.<\/li>\nAtomicReferenceAtomicStampedReferenceAtomicMarkableReferenceDoubleAdder<\/code>, LongAdder<\/code>, DoubleAccumulator<\/code>, LongAccumulator<\/code>: \u00c7ok y\u00fcksek \u00e7eki\u015fmeli senaryolarda toplama i\u015flemleri i\u00e7in daha iyi performans sunan s\u0131n\u0131flar.<\/li>\n<\/ul>\nAtomicInteger<\/code>'\u0131n thread g\u00fcvenli\u011fi nas\u0131l sa\u011flan\u0131r?<\/strong><\/p>\nAtomicInteger<\/code>, dahili olarak CAS (Compare-And-Swap) ad\u0131 verilen \u00f6zel i\u015flemci komutlar\u0131n\u0131 kullanarak thread g\u00fcvenli\u011fini sa\u011flar. Bu komutlar, bir de\u011ferin okunmas\u0131, kar\u015f\u0131la\u015ft\u0131r\u0131lmas\u0131 ve g\u00fcncellenmesinin tek bir atomik ad\u0131mda ger\u00e7ekle\u015fmesini garanti eder. E\u011fer de\u011fer beklenen gibi de\u011filse, i\u015flem ba\u015far\u0131s\u0131z olur ve \u00e7a\u011f\u0131ran i\u015f par\u00e7ac\u0131\u011f\u0131 i\u015flemi yeniden dener. Bu \"kilitsiz\" yakla\u015f\u0131m, kilit mekanizmalar\u0131na ihtiya\u00e7 duymadan veri tutarl\u0131l\u0131\u011f\u0131n\u0131 sa\u011flar.<\/p>\n