N\u00eb frontendin e kompilatorit Rust, q\u00eb kryen detyra si analiz\u00eb sintaksore, verifikim t\u00eb tipeve dhe analiz\u00eb t\u00eb huazimeve, \u00ebsht\u00eb implementuar mb\u00ebshtetje p\u00ebr ekzekutimin paralel, duke lejuar nj\u00eb reduktim t\u00eb konsideruesh\u00ebm t\u00eb koh\u00ebs s\u00eb kompilimit. Paralelizimi \u00ebsht\u00eb tashm\u00eb i disponuesh\u00ebm n\u00eb nd\u00ebrtimet e nat\u00ebs t\u00eb Rust dhe aktivizohet p\u00ebrmes opsionit «-Z threads=8». N\u00eb deg\u00ebn stabile, k\u00ebt\u00eb mund\u00ebsi pritet t\u00eb p\u00ebrfshihen n\u00eb vitin 2024. <\/p>\n
Puna p\u00ebr t\u00eb reduktuar koh\u00ebn e kompilimit n\u00eb Rust ka vazhduar p\u00ebr disa vite. Gjat\u00eb 10 muajve t\u00eb par\u00eb t\u00eb vitit 2023, koha e kompilimit u zvog\u00eblua n\u00eb mesatare me 13%, konsumimi maksimal i memories u ul me 15%, dhe madh\u00ebsia e skedareve t\u00eb gjeneruara u zvog\u00eblua me 7%. N\u00eb k\u00ebt\u00eb faz\u00eb, p\u00ebrshpejtimi \u00ebsht\u00eb arritur p\u00ebrmes optimizimeve t\u00eb vet\u00eb kompilatorit. Pas k\u00ebsaj, zhvilluesit kaluan n\u00eb pun\u00ebn p\u00ebr p\u00ebrshpejtimin p\u00ebrmes paralelizimit t\u00eb operacioneve gjat\u00eb kompilimit. <\/p>\n
Deri tani, paralelizimi n\u00eb Rust \u00ebsht\u00eb siguruar kryesisht n\u00eb nivelin e proceseve, p\u00ebr shembull, menaxheri i paketave Cargo mund t\u00eb nis\u00eb disa procese rustc p\u00ebr p\u00ebrpunimin e nj\u00ebkohsh\u00ebm t\u00eb disa paketave. Mb\u00ebshtetje p\u00ebr paralelizimin \u00ebsht\u00eb gjithashtu e pranishme n\u00eb an\u00ebn e backend-it, i cili kryen operacione t\u00eb lidhura me gjenerimin e kodit \u2014 backend-i Rust mund t\u00eb gjeneroj\u00eb kod n\u00eb pjes\u00eb, t\u00eb cilat m\u00eb pas LLVM mund t'i p\u00ebrpunoj\u00eb paralelisht. Frontend-i, nga ana tjet\u00ebr, ende mund t\u00eb p\u00ebrpunoj\u00eb kodin burimor vet\u00ebm n\u00eb modin nj\u00eb-fijor. <\/p>\n
P\u00ebr t\u00eb mb\u00ebshtetur paralelizmin, frontend-i \u00ebsht\u00eb ridizajnuar p\u00ebr t\u00eb p\u00ebrdorur bibliotek\u00ebn Rayon dhe \u00ebsht\u00eb rimodeluar ndjesh\u00ebm; p\u00ebr shembull, shum\u00eb pjes\u00eb tani sinkronizohen p\u00ebrmes mutex-eve dhe bllokimeve t\u00eb leximit\/shkrimit, nd\u00ebrsa n\u00eb kod p\u00ebrdoren llojet atomike. Gjat\u00eb testimit t\u00eb performanc\u00ebs, implementimi i ri paralel mund t\u00eb realizonte kompaktimin deri n\u00eb 2% m\u00eb ngadal\u00eb kur operohej n\u00eb nj\u00eb nj\u00ebsi (-Z threads=1), por kur numri i thread-eve tejkalonte nj\u00eb, shpejt\u00ebsia rritej ndjesh\u00ebm. P\u00ebr shembull, me 8 thread (-Z threads=8), n\u00eb disa raste, koha e kompaktimit mund t\u00eb reduktohej me 50%. <\/p>\n
Megjithat\u00eb, rezultati varet shum\u00eb nga konfigurimet e mjedisit dhe kodi q\u00eb kompilohet \u2014 p\u00ebr programet shum\u00eb t\u00eb vogla, t\u00eb cilat tashm\u00eb kompilohet shpejt, kompaktimi n\u00eb modin shum\u00ebth\u00ebn\u00eb mund t\u00eb jet\u00eb m\u00eb i ngadalsh\u00ebm. P\u00ebr m\u00eb tep\u00ebr, konsumimi i memories n\u00eb modin shum\u00ebth\u00ebn\u00eb mund t\u00eb rritet ndjesh\u00ebm; p\u00ebr shembull, n\u00eb testime \u00ebsht\u00eb v\u00ebn\u00eb re nj\u00eb rritje e konsumimit t\u00eb memories deri n\u00eb 35%. \n Burimi: opennet.ru<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"