{"id":183759,"date":"2026-10-12T16:53:43","date_gmt":"2026-10-12T14:53:43","guid":{"rendered":"https:\/\/prohoster.info\/blog\/news\/linus-torvalds-eksperimentiruet-s-generacziej-izmenenij-k-yadru-cherez-ai"},"modified":"2026-10-12T16:53:48","modified_gmt":"2026-10-12T14:53:48","slug":"linus-torvalds-experiments-with-ai-generated-kernel-changes","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/news\/linus-torvalds-experiments-with-ai-generated-kernel-changes","title":{"rendered":"Linus Torvalds eksperimenton me gjenerimin e ndryshimeve n\u00eb b\u00ebrtham\u00eb p\u00ebrmes AI","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
Linus Torvalds ka p\u00ebrgatitur nj\u00eb grup patch-e p\u00ebr n\u00ebn-sistemin e menaxhimit t\u00eb memories n\u00eb kernelin Linux, duke realizuar optimizimin p\u00ebr t\u00eb lexuar blloqe t\u00eb vogla t\u00eb dh\u00ebnash nga cache-i i faqeve pa ndryshimin atomik t\u00eb numrit t\u00eb referencave (refcount) n\u00eb struktur\u00ebn folio. N\u00eb tekstin e komitit \u00ebsht\u00eb e sh\u00ebnuar se ndryshimet dhe provat e lidhura me t\u00eb jan\u00eb krijuar nga asistenti AI Claude Code n\u00ebn mbik\u00ebqyrjen e Torvaldsit. Ndryshimi \u00ebsht\u00eb shtuar n\u00eb nj\u00eb deg\u00eb t\u00eb ve\u00e7ant\u00eb t\u00eb kernelit noref-filemap-read-with-llm dhe p\u00ebr momentin nuk \u00ebsht\u00eb e qart\u00eb n\u00ebse do t\u00eb trasferohet n\u00eb deg\u00ebn kryesore apo \u00ebsht\u00eb thjesht nj\u00eb eksperiment t\u00eb brendsh\u00ebm.<\/p>\n
Gjat\u00eb testimit, fal\u00eb patches, arriti t\u00eb rritej kapaciteti i leximit t\u00eb skedar\u00ebve t\u00eb vegj\u00ebl n\u00eb konfigurimin me 8 rrjedha nga 9.9-10.4 n\u00eb 30-33 milion operacione n\u00eb sekond\u00eb (rritje 3-fish), me 4 rrjedha \u2014 nga 8.9-11.6 n\u00eb 25-26 milion\/sec (rritje 2-fish), nd\u00ebrsa n\u00eb ngarkes\u00ebn me nj\u00eb rrjedh\u00eb nga 7.9-8.1 n\u00eb 8.8-8.9 milion\/sec. Rritja e performanc\u00ebs shpjegohet me faktin se kur ka qasje t\u00eb nj\u00ebkohshme n\u00eb menj\u00ebhersh\u00ebm, ndodh nj\u00eb konkurrenc\u00eb e fort\u00eb nd\u00ebrmjet rrjedhave p\u00ebr qasje n\u00eb cache. P\u00ebr shembull, kur 8 rrjedha p\u00ebrpiqen t\u00eb lexojn\u00eb t\u00eb dh\u00ebna 16 byte nj\u00ebher\u00ebsh, 99.92% e koh\u00ebs shpenzohet jo p\u00ebr leximin e vet\u00eb, por p\u00ebr pritjen e qasjes n\u00eb num\u00ebruesin e referenc\u00ebs.<\/p>\n
N\u00eb patch-in p\u00ebr t\u00eb garantuar sakt\u00ebsin\u00eb e t\u00eb dh\u00ebnave, n\u00eb vend t\u00eb vendosjes s\u00eb bllokimeve t\u00eb ve\u00e7anta, ndryshimeve t\u00eb numrave t\u00eb referencave ose bufrim t\u00eb nd\u00ebrmjet\u00ebm, aplikohet nj\u00eb variant i anashkalimit i cili p\u00ebrdor \u00e7lirimin e vonuar t\u00eb memories me sinkronizim p\u00ebrmes mekanizmit RCU (Read-Copy-Update). Metoda e propozuar n\u00eb patch garanton q\u00eb faqja e memories nuk do t\u00eb lirohet fizikisht derisa tufa q\u00eb lexon t\u00eb p\u00ebrfundoj\u00eb kopjimin e t\u00eb dh\u00ebnave. \n Burimi: opennet.ru<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"