{"id":115974,"date":"2024-05-21T14:59:30","date_gmt":"2024-05-21T12:59:30","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/proekt-bend-razvivaet-vysokourovnevyj-yazyk-dlya-parallelnyh-vychislenij-na-gpu"},"modified":"2024-05-21T14:59:30","modified_gmt":"2024-05-21T12:59:30","slug":"proekt-bend-razvivaet-vysokourovnevyj-yazyk-dlya-parallelnyh-vychislenij-na-gpu","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/news\/proekt-bend-razvivaet-vysokourovnevyj-yazyk-dlya-parallelnyh-vychislenij-na-gpu","title":{"rendered":"Projekti Bend zhvillon nj\u00eb gjuh\u00eb t\u00eb nivelit t\u00eb lart\u00eb p\u00ebr llogaritjet paralel n\u00eb GPU","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
Jan\u00eb publikuar numrat e par\u00eb t\u00eb projektit Bend, i cili zhvillon nj\u00eb gjuh\u00eb programimi me nivel t\u00eb lart\u00eb p\u00ebr llogaritjet paralele, e cila pozicionohet si nj\u00eb alternativ\u00eb ndaj gjuh\u00ebve me nivel t\u00eb ul\u00ebt, si CUDA dhe Metal, dhe ofron nj\u00eb sintaks\u00eb t\u00eb shprehshme dhe mund\u00ebsi p\u00ebr t\u00eb rritur leht\u00ebsin\u00eb e zhvillimit, karakteristike p\u00ebr gjuh\u00eb si Python dhe Haskell. Kodi i projektit \u00ebsht\u00eb shkruar n\u00eb gjuh\u00ebn Rust dhe shp\u00ebrndahet n\u00ebn licenc\u00ebn Apache 2.0. <\/p>\n
Nga mund\u00ebsit\u00eb e gjuh\u00ebs Bend, theksohet shp\u00ebrndarja e shpejt\u00eb e objekteve, mund\u00ebsia e p\u00ebrdorimit t\u00eb funksioneve t\u00eb larta, mbylljeve, vazhdimeve, rekursivitetit t\u00eb pakufizuar, p\u00ebrputhjes me modelin, p\u00ebrputhjeve rekursive (fold) dhe cikleve (bend), tipeve t\u00eb numrave t\u00eb plot\u00eb, vargjeve dhe listave. Nj\u00eb dy variantet e sintaks\u00ebs mb\u00ebshteten - n\u00eb stilin Python dhe n\u00eb stilin Haskell. N\u00eb programe nuk k\u00ebrkohet t\u00eb specifikohen anotime kontrolli p\u00ebr paralelizimin, krijimin eksplicit t\u00eb rrjedhave dhe vendosjen e bllokimeve. Paralelizimi kryhet automatikisht, p\u00ebr shembull, kur llogaritet shprehja '((1 + 2) + (3 + 4))', operacionet '1 + 2' dhe '3 + 4' do t\u00eb ekzekutohen paralelisht. <\/p>\n
Programet n\u00eb gjuh\u00ebn Bend mund t\u00eb ekzekutohen n\u00eb pajisje q\u00eb ofrojn\u00eb masiv paraleliz\u00ebm, p\u00ebr shembull, n\u00eb GPU, duke demonstruar nj\u00eb rritje praktike linjare n\u00eb performanc\u00eb n\u00eb var\u00ebsi t\u00eb numrit t\u00eb nyjeve llogarit\u00ebse. Kodi n\u00eb gjuh\u00ebn Bend kompozohet n\u00eb nj\u00eb p\u00ebrfaq\u00ebsi nd\u00ebrmjet\u00ebse me nivel t\u00eb ul\u00ebt HVM2 (Higher-order Virtual Machine 2), e cila pastaj kompozohet n\u00eb nj\u00eb p\u00ebrfaq\u00ebsi n\u00eb C dhe CUDA. Aktualisht, projekti mb\u00ebshtet vet\u00ebm ekzekutimin n\u00eb GPU t\u00eb NVIDIA. <\/p>\n
Sa i p\u00ebrket performanc\u00ebs, aplikacioni testues me implementimin e renditjes bitonale, u ekzekutua n\u00eb nj\u00eb rrjedh\u00eb CPU Apple M3 Max n\u00eb 12.15 sekonda, duke p\u00ebrdorur 16 rrjedha - p\u00ebr 0.96 sekonda, dhe duke angazhuar GPU NVIDIA RTX 4090 me 16k rrjedha - p\u00ebr 0.21 sekonda. def sort(d, s, tree): switch d: case 0: return tree case _: (x,y) = tree lft = sort(d-1, 0, x) rgt = sort(d-1, 1, y) return rots(d, s, lft, rgt) def rots(d, s, tree): switch d: case 0: return tree case _: (x,y) = tree return down(d, s, warp(d-1, s, x, y)) \u2026 \n Burimi: opennet.ru<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"