{"id":184097,"date":"2026-11-02T23:40:40","date_gmt":"2026-11-02T21:40:40","guid":{"rendered":"https:\/\/prohoster.info\/blog\/news\/vypusk-svt-av1-4-3-0-kodirovshhika-dlya-formata-video-av1"},"modified":"2026-11-02T23:40:47","modified_gmt":"2026-11-02T21:40:47","slug":"svt-av1-4-3-0-av1-video-encoder-released","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/news\/svt-av1-4-3-0-av1-video-encoder-released","title":{"rendered":"L\u00ebshimi i SVT-AV1 4.3.0, koduesit p\u00ebr formatin e videos AV1","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
\u00cbsht\u00eb publikuar versioni 4.3.0 i bibliotek\u00ebs SVT-AV1 (Teknologjia Video e Shkallueshme AV1) me implementime t\u00eb encoder dhe decoder p\u00ebr formatin e kodimit t\u00eb videos AV1, p\u00ebr shpejtimin e t\u00eb cilave p\u00ebrdoren zgjerimet e pranishme n\u00eb procesor\u00ebt x86_64 dhe ARM p\u00ebr paralelizimin e p\u00ebrpunimeve n\u00eb nivel harduerik. Projekti \u00ebsht\u00eb krijuar nga kompanit\u00eb Intel n\u00eb partneritet me Netflix me q\u00ebllim t\u00eb arritjes s\u00eb nj\u00eb niveli performance, t\u00eb p\u00ebrshtatsh\u00ebm p\u00ebr rikodimin e videos n\u00eb koh\u00eb reale dhe p\u00ebrdorimin n\u00eb sh\u00ebrbime q\u00eb ofrojn\u00eb video sipas k\u00ebrkes\u00ebs (VOD). Aktualisht, zhvillimi po realizohet n\u00ebn udh\u00ebheqjen e aleanc\u00ebs Open Media (AOMedia), q\u00eb mbik\u00ebqyr zhvillimin e formatit t\u00eb kodimit t\u00eb videos AV1. P\u00ebrpara, projekti \u00ebsht\u00eb zhvilluar brenda projektit OpenVisualCloud, i cili gjithashtu zhvillon koduesit SVT-HEVC dhe SVT-VP9. Kodi shp\u00ebrndahet n\u00ebn licenc\u00ebn BSD. <\/p>\n
SVT-AV1 mund t\u00eb nd\u00ebrtohet p\u00ebr sisteme t\u00eb bazuara n\u00eb arkitektur\u00ebn e \u00e7do lloji, p\u00ebr t\u00eb cilat ekziston nj\u00eb kompilator q\u00eb mb\u00ebshtet standardin C99, por performanca m\u00eb e mir\u00eb arrihet n\u00eb sistemet x86_64, p\u00ebr t\u00eb cilat aplikohet optimizimi i assembler-it n\u00eb baz\u00eb t\u00eb udh\u00ebzimeve SIMD (preferohet q\u00eb CPU t\u00eb mb\u00ebshtes\u00eb AVX2, por si minimum mjafton edhe SSE2). Konsumi i memories varet nga numri i b\u00ebrthamave t\u00eb procesor\u00ebve t\u00eb angazhuara gjat\u00eb kodimit, t\u00eb rregulluara nga opcioni «—lp». P\u00ebr shkak t\u00eb kompleksitetit t\u00eb algoritmeve t\u00eb aplikuara n\u00eb AV1, kodimi i k\u00ebtij formati k\u00ebrkon ndjesh\u00ebm m\u00eb shum\u00eb burime se formatet e tjera. <\/p>\n
Nd\u00ebr sh\u00ebmbujt e ndryshimeve n\u00eb versionin e ri t\u00eb SVT-AV1:<\/p>\n
\n
Kjo ka \u00e7uar n\u00eb optimizimin e kompromisave cil\u00ebsi\/shpejt\u00ebsi p\u00ebr t\u00eb gjitha presetet, duke lejuar nj\u00eb ulje t\u00eb ngarkes\u00ebs n\u00eb CPU prej 5-8% duke ruajtur t\u00eb nj\u00ebjtin nivel cil\u00ebsie, si dhe p\u00ebrshpejtimin e koh\u00ebs s\u00eb kodimit me 10-11% n\u00eb m\u00ebnyr\u00ebn e shum\u00ebproceseve. Jan\u00eb p\u00ebrshpejtuar operacionet p\u00ebr p\u00ebrjashtimin e m\u00ebnyrave t\u00eb kodimit joefikase (MD — Vendimi p\u00ebr Mod\u00ebn), llogaritjen e vektor\u00ebve t\u00eb l\u00ebvizjes (MV — Vektori i L\u00ebvizjes) dhe parametrave t\u00eb filtrave gjat\u00eb kodimit, llogaritja e madh\u00ebsis\u00eb s\u00eb koeficient\u00ebve t\u00eb frekuenc\u00ebs, k\u00ebrkimi i zhvendosjeve n\u00ebnpiksel, kopjimi i blloqeve dhe filtrimi i detajeve t\u00eb panevojshme me an\u00eb t\u00eb algoritmit RDOQ (Kvantizimi i Optimizuar n\u00eb lidhje me Norm\u00ebn-Distortion). N\u00eb m\u00ebnyr\u00ebn e shum\u00ebproceseve p\u00ebr kursimin e burimeve, algoritmi i p\u00ebrcaktimit t\u00eb transmetimit t\u00eb p\u00ebrmbajtjes s\u00eb ekranit tani aplikohet vet\u00ebm p\u00ebr kadra I.\n
Shtuar nj\u00eb flags «—fgs-table», q\u00eb aktivizon mb\u00ebshtetje p\u00ebr tabelat e pik\u00ebs s\u00eb filmave (film grain), dhe API FILM_GRAIN_EVENT p\u00ebr t\u00eb vendosur parametrat e pik\u00ebs n\u00eb lidhje me kadra.\n
Shtuar shpejt\u00ebsin\u00eb e kodimit pa humbje.\n
N\u00eb modalitetin e kodimit me vonesa t\u00eb ulta, \u00ebsht\u00eb realizuar paralelizmi i p\u00ebrpunimit t\u00eb blloqeve (tile).\n
P\u00ebr sistemet x86, jan\u00eb realizuar trajtues 8-bit t\u00eb CDEF mbi baz\u00ebn e instrukcioneve AVX2 dhe SSE4.1, si dhe transformime t\u00eb drejta 16-bit mbi baz\u00ebn e instrukcioneve AVX2 dhe AVX-512. Shtuar kod i optimizuar me instrukcione SSE4.1 p\u00ebr parashikimin nd\u00ebrkadror, pun\u00ebn me paletat dhe transformimet. Optimizuar trajtuesit p\u00ebr kquantizim, kopjimin e blloqeve, llogaritjen e deformimeve dhe vler\u00ebsimin e ndryshimeve subpiksel.\n
P\u00ebr sistemet ARM jan\u00eb shtuar variante trajtuesish t\u00eb p\u00ebrshpejtuara me instrukcione procesori Neon.\n<\/ul>\n