![\"\"](\"https:\/\/www.yazilimperver.com\/wp-content\/uploads\/2023\/09\/img_65132e6d7274d.png\")
<\/p>\nEvet sevgili yaz\u0131l\u0131mperver dostlar\u0131m, yaz\u0131m\u0131n ba\u015fl\u0131\u011f\u0131ndan da anla\u015f\u0131laca\u011f\u0131 \u00fczere, bu yaz\u0131mda \u00e7ok spesifik bir konu \u00fczerine e\u011filece\u011fiz. Daha \u00f6nceki yaz\u0131lar\u0131mda heap ve stack belleklerinden bahsetmi\u015ftim (umar\u0131m \ud83d\ude42 )<\/p>\n
Kabaca, uygulama i\u00e7erisinde, heap, new\/malloc ile al\u0131nan bellek alanlar\u0131 i\u00e7in kullan\u0131l\u0131rken, stack ise tan\u0131mlad\u0131\u011f\u0131n\u0131z yerel de\u011fi\u015fkenler, i\u00e7 i\u00e7e \u00e7a\u011f\u0131rd\u0131\u011f\u0131m\u0131z fonksiyona ili\u015fkin bir tak\u0131m veriler i\u00e7in kullan\u0131lan bellektir (\u00f6rne\u011fin recursive fonksiyonlarda kontrol hatas\u0131 yap\u0131nca “stack overflow” hatas\u0131 da asl\u0131nda buna i\u015faret eder), elbette bundan \u00e7ok daha fazlas\u0131 var ama \u015fimdilik bu kadar\u0131 yeterli.<\/p>\n
A\u015fa\u011f\u0131ya internetten buldu\u011fum bir yans\u0131y\u0131 da b\u0131rakay\u0131m. Burada nelerin stack, nelerin ise heap bellek al\u0131nd\u0131\u011f\u0131n\u0131 g\u00f6stermekte.<\/p>\n
<\/p>\n
\u015eimdi gelelim problemimize; normalde fonksiyon i\u00e7erisinde tan\u0131mlad\u0131\u011f\u0131m\u0131z de\u011fi\u015fkenler pek b\u00fcy\u00fck olmaz (asl\u0131nda olmamal\u0131 da, statik kod analiz ara\u00e7lar\u0131 da zaten bunu sizlere s\u00f6yler) ama bir sebeple buna ihtiyac\u0131m\u0131z olursa ne yapaca\u011f\u0131z? \u0130\u015fte bu yaz\u0131da, buna cevap ar\u0131yor olaca\u011f\u0131z. Hemen bir \u00f6rnek ile bakal\u0131m:<\/p>\n
#include <iostream>\r\n#include <cstdint>\r\n\r\nconstexpr uint32_t cArrSize = 10000000;\r\n\r\nint32_t main(){\r\n double dataFromStack[cArrSize];\r\n int32_t i = 0;\r\n\r\n while(i < cArrSize) {\r\n dataFromStack[i] = i;\r\n i++;\r\n }\r\n return 0;\r\n}<\/pre>\nBu kodu \u00e7al\u0131\u015ft\u0131rd\u0131\u011f\u0131n\u0131zda “[1] 1979 segmentation fault .\/a.out” hatas\u0131 alabilirsiniz (e\u011fer \u00f6zellikle stack boyutunu artt\u0131rmad\u0131ysan\u0131z).<\/p>\n
Peki ne yapaca\u011f\u0131z? \u00d6ncelikle mevcut stack boyutunu \u00f6\u011frenmeliyiz.<\/p>\n
Bunu \u00f6\u011frendikten sonra art\u0131k artt\u0131rabiliriz, bunu da:<\/p>\n
ulimit -s unlimited<\/pre>\nkomutu ile yapabilirsiniz. Bu komutu \u00e7a\u011f\u0131rd\u0131ktan sonra yukar\u0131daki uygulamay\u0131 \u00e7al\u0131\u015ft\u0131rman\u0131z durumunda art\u0131k hata almad\u0131\u011f\u0131n\u0131z\u0131 g\u00f6receksiniz. Elbette bu durum, bu shell oturumu \u00fczerinde \u00e7al\u0131\u015ft\u0131r\u0131lacak b\u00fct\u00fcn uygulamalar i\u00e7in ge\u00e7erli olacakt\u0131r. Bunu s\u00fcrekli hale getirmek i\u00e7in bash veya kulland\u0131\u011f\u0131n\u0131z shell ayarlar\u0131na ekleyebilirsiniz.<\/p>\n
Bu arada, “core dump” konusundan haberdar takip\u00e7ilerim de bu komutu kullanm\u0131\u015f olmalar\u0131 olas\u0131d\u0131r.<\/p>\n
Yukar\u0131daki y\u00f6ntemin yan\u0131nda, a\u015fa\u011f\u0131daki gibi bir fonksiyon arac\u0131l\u0131\u011f\u0131 ile de stack boyutunu g\u00fcncelleyebilirsiniz. Bunu da sistem kaynaklar\u0131ndan ilgili bellek alan\u0131n\u0131 de\u011fi\u015ftirerek yap\u0131yoruz:<\/p>\n
#include <sys\/resource.h>\r\n\r\nvoid increaseStackSize(uint32_t newstackSize){\r\n struct rlimit rl;\r\n int result;\r\n\r\n result = getrlimit(RLIMIT_STACK, &rl);\r\n if (result == 0)\r\n {\r\n if (rl.rlim_cur < newstackSize)\r\n {\r\n rl.rlim_cur = newstackSize;\r\n result = setrlimit(RLIMIT_STACK, &rl);\r\n if (result != 0) {\r\n fprintf(stderr, \"Stack set error! Error code: %d\\n\", result);\r\n }\r\n }\r\n }\r\n}<\/pre>\nSon ekleme ile nihai kodumuz a\u015fa\u011f\u0131daki gibi olacak. Bu kod ile art\u0131k “ulimit -s unlimited” yapman\u0131za gerek kalmayacak.<\/p>\n
#include <iostream>\r\n#include <cstdint>\r\n\r\n#include <sys\/resource.h>\r\n\r\nvoid increaseStackSize(uint32_t newstackSize){\r\n\tstruct rlimit rl;\r\n int result;\r\n\r\n result = getrlimit(RLIMIT_STACK, &rl);\r\n if (result == 0)\r\n {\r\n if (rl.rlim_cur < newstackSize)\r\n {\r\n rl.rlim_cur = newstackSize;\r\n result = setrlimit(RLIMIT_STACK, &rl);\r\n if (result != 0) {\r\n fprintf(stderr, \"Stack set error! Error code: %d\\n\", result);\r\n }\r\n }\r\n }\r\n}\r\n\r\nconstexpr uint32_t cArrSize = 10'000'000;\r\n\r\nvoid testFunc(){\r\n double dataFromStack[cArrSize];\r\n int32_t i = 0;\r\n\r\n while(i < cArrSize) {\r\n dataFromStack[i] = i;\r\n i++;\r\n }\r\n}\r\n\r\nint32_t main(){\r\n\t\tincreaseStackSize(cArrSize*9);\r\n\t\ttestFunc();\r\n return 0;\r\n}<\/pre>\n\u015eimdi burada \u00e7ok ince bir nokta var. \u0130lgili setrlimit fonksiyonunu, main() i\u00e7erisinde \u00e7a\u011f\u0131rd\u0131\u011f\u0131m i\u00e7in e\u011fer\u00a0testFunc()<\/em> i\u00e7eri\u011fini buraya al\u0131rsan\u0131z yine ayn\u0131 hatay\u0131 al\u0131yor olacaks\u0131n\u0131z. Bunun sebebi ilgili stack boyutunun ayarland\u0131ktan sonraki \u00e7a\u011fr\u0131lar i\u00e7in ge\u00e7erli olmas\u0131 sebebi ile. E\u011fer ille main i\u00e7erisinde kullanma ihtiyac\u0131n\u0131z var ise ulimit yine bir \u00e7\u00f6z\u00fcm olabilir.<\/p>\n
\u0130\u015fimiz bitti mi peki? Elbette hay\u0131r, kaynaklar k\u0131sm\u0131na ekledi\u011fim setrlimit sayfas\u0131na bakarsan\u0131z bu \u00e7a\u011fr\u0131n\u0131n sadece main thread i\u00e7in ge\u00e7erli oldu\u011funu g\u00f6receksiniz. Peki birden fazla thread kullan\u0131yorsak ne yapaca\u011f\u0131z? \u0130\u015fte burada da, pthread API’lerini kullan\u0131yor olaca\u011f\u0131z. \u00c7\u00fcnk\u00fc, \u015fu ana kadar (bildi\u011fim\/g\u00f6rd\u00fc\u011f\u00fcm kadar\u0131yla) thread stack boyutunu ayarlayan standart bir y\u00f6ntem yok, var ise (ul\u0131m\u0131t d\u0131\u015f\u0131nda) yorumlar k\u0131sm\u0131ndan katk\u0131da bulunabilirsiniz. A\u015fa\u011f\u0131daki \u00f6rnekte, ana thread yan\u0131nda, olu\u015fturulacak di\u011fer thread’ler i\u00e7in izlemeniz gereken y\u00f6ntemi g\u00f6stermeye \u00e7al\u0131\u015ft\u0131m:<\/p>\n
#include <iostream>\r\n#include <cstdint>\r\n#include <pthread.h>\r\n\r\n#include <sys\/resource.h>\r\n\r\n\/\/ stack boyutunu arttiran fonksiyonumuz\r\nvoid increaseStackSizeForMainThread(uint32_t newstackSize){\r\n\tstruct rlimit rl;\r\n int result;\r\n\r\n result = getrlimit(RLIMIT_STACK, &rl);\r\n if (result == 0)\r\n {\r\n if (rl.rlim_cur < newstackSize)\r\n {\r\n rl.rlim_cur = newstackSize;\r\n result = setrlimit(RLIMIT_STACK, &rl);\r\n if (result != 0) {\r\n fprintf(stderr, \"Stack set error! Error code: %d\\n\", result);\r\n }\r\n }\r\n }\r\n}\r\n\r\nconstexpr uint32_t cArrSize = 10'000'000;\r\n\r\n\/\/ ornek kodumuz\r\nvoid testFunc(){\r\n double dataFromStack[cArrSize];\r\n int32_t i = 0;\r\n\r\n while(i < cArrSize) {\r\n dataFromStack[i] = i;\r\n i++;\r\n }\r\n}\r\n\r\n\/\/ Main thread yaninda da bu ayar ise yariyor mu diye bakalim\r\nvoid* threadTask(void*){\r\n testFunc();\r\n pthread_exit(NULL);\r\n}\r\n\r\nvoid increaseStackSizeForOtherThreads(){\r\n}\r\n\r\nint32_t main(){\r\n\t\tincreaseStackSizeForMainThread(cArrSize*9);\r\n\t\ttestFunc();\r\n \r\n increaseStackSizeForOtherThreads();\r\n\r\n \/\/ pthread icin ayarlayalim \r\n pthread_attr_t attr;\r\n pthread_attr_init( &attr );\r\n\r\n \/\/ Stack boyutunu gecirelim\r\n size_t stacksize = 9 * cArrSize;\r\n pthread_attr_setstacksize( &attr, stacksize );\r\n\r\n pthread_t threadId;\r\n pthread_create( &threadId, &attr, threadTask, nullptr );\r\n void *status;\r\n pthread_join(threadId, &status);\r\n\r\n return 0;\r\n}<\/pre>\n<\/p>\n
pthread’ler i\u00e7in kullan\u0131ma y\u00f6nelik g\u00fczel bir SO sayfas\u0131n\u0131 da kaynaklar k\u0131sm\u0131na ekliyorum arkada\u015flar. Bir sonraki yaz\u0131mda g\u00f6r\u00fc\u015fmek dile\u011fiyle.<\/p>\n
Kaynaklar<\/span><\/strong><\/h2>\n
\n
- https:\/\/www.educative.io\/blog\/stack-vs-heap<\/span><\/strong><\/a><\/li>\n
- https:\/\/learn.microsoft.com\/en-us\/cpp\/build\/reference\/stack-stack-allocations?view=msvc-170<\/a><\/strong><\/span><\/li>\n
- https:\/\/ss64.com\/bash\/ulimit.html<\/a><\/strong><\/span><\/li>\n
- https:\/\/unix.stackexchange.com\/questions\/551491\/ulimit-for-stack-size-per-process-or-per-thread-limit<\/strong><\/a><\/span><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"
Evet sevgili yaz\u0131l\u0131mperver dostlar\u0131m, yaz\u0131m\u0131n ba\u015fl\u0131\u011f\u0131ndan da anla\u015f\u0131laca\u011f\u0131 \u00fczere, bu yaz\u0131mda \u00e7ok spesifik bir konu \u00fczerine e\u011filece\u011fiz. Daha \u00f6nceki yaz\u0131lar\u0131mda heap ve stack belleklerinden bahsetmi\u015ftim (umar\u0131m \ud83d\ude42 ) Kabaca, uygulama i\u00e7erisinde, heap, new\/malloc ile al\u0131nan bellek alanlar\u0131 i\u00e7in kullan\u0131l\u0131rken, stack ise tan\u0131mlad\u0131\u011f\u0131n\u0131z yerel de\u011fi\u015fkenler, i\u00e7 i\u00e7e \u00e7a\u011f\u0131rd\u0131\u011f\u0131m\u0131z fonksiyona ili\u015fkin bir tak\u0131m veriler i\u00e7in kullan\u0131lan bellektir… Continue reading