{"id":113552,"date":"2024-02-12T22:02:43","date_gmt":"2024-02-12T20:02:45","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/opublikovan-instrumentarij-zluda-pozvolyayushhij-zapuskat-cuda-prilozheniya-na-gpu-amd"},"modified":"2024-02-12T22:02:43","modified_gmt":"2024-02-12T20:02:45","slug":"opublikovan-instrumentarij-zluda-pozvolyayushhij-zapuskat-cuda-prilozheniya-na-gpu-amd","status":"publish","type":"post","link":"https:\/\/prohoster.info\/ro\/blog\/news\/opublikovan-instrumentarij-zluda-pozvolyayushhij-zapuskat-cuda-prilozheniya-na-gpu-amd","title":{"rendered":"A fost publicat instrumentul ZLUDA, care permite rularea aplica\u021biilor CUDA pe GPU-uri AMD","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

Proiectul ZLUDA a preg\u0103tit o implementare deschis\u0103 a tehnologiei CUDA pentru GPU AMD, care permite executarea aplica\u021biilor CUDA nemodificate cu performan\u021be apropiate de cele ale aplica\u021biilor rulate f\u0103r\u0103 straturi intermediare. Toolkitul publicat ofer\u0103 compatibilitate binar\u0103 cu aplica\u021biile CUDA existente, compilate folosind compilatorul CUDA pentru GPU NVIDIA. Implementarea func\u021bioneaz\u0103 deasupra stivei ROCm dezvoltat\u0103 de AMD \u0219i a runtime-ului HIP (Heterogeneous-computing Interface for Portability). Codul proiectului este scris \u00een limbajul Rust \u0219i este distribuit sub licen\u021bele MIT \u0219i Apache 2.0. Se sus\u021bine func\u021bionarea pe Linux \u0219i Windows. <\/p>\n

Stratul pentru organizarea func\u021bion\u0103rii CUDA pe sistemele cu GPU AMD a fost dezvoltat \u00een ultimii doi ani, dar proiectul are o istorie mai \u00eendelungat\u0103 \u0219i a fost ini\u021bial creat pentru a asigura func\u021bionarea CUDA pe GPU Intel. Schimbarea politicii de suport pentru GPU este explicat\u0103 prin faptul c\u0103, la \u00eenceput, dezvoltatorul ZLUDA era angajat al Intel, dar \u00een 2021, aceast\u0103 companie a considerat c\u0103 oferirea posibilit\u0103\u021bii de a rula aplica\u021bii CUDA pe GPU Intel nu mai este de interes pentru afacere \u0219i nu a mai continuat dezvoltarea ini\u021biativei. <\/p>\n

La \u00eenceputul anului 2022, dezvoltatorul a renun\u021bat la Intel \u0219i a semnat un contract cu compania AMD pentru dezvoltarea unui strat de compatibilitate cu CUDA. Compania AMD a solicitat, \u00een timpul dezvolt\u0103rii, s\u0103 nu fie f\u0103cut public interesul AMD pentru proiectul ZLUDA \u0219i s\u0103 nu se fac\u0103 commit-uri \u00een depozitul public ZLUDA. Dup\u0103 doi ani, compania AMD a decis c\u0103 rularea aplica\u021biilor CUDA pe GPU AMD nu mai reprezint\u0103 un interes pentru afacere, ceea ce, conform contractului, a permis dezvoltatorului s\u0103 \u00ee\u0219i deschid\u0103 lucr\u0103rile. Deoarece produc\u0103torii de GPU au \u00eencetat s\u0103 financeze proiectul, soarta acestuia depinde acum de interesul comunit\u0103\u021bii \u0219i de propunerile de colaborare din partea altor companii. F\u0103r\u0103 suport extern, proiectul ar putea evolua doar \u00een direc\u021biile care intereseaz\u0103 personal autorul, cum ar fi DLSS (Deep Learning Super Sampling). <\/p>\n

\u00cen forma actual\u0103, nivelul de calitate al implement\u0103rii este evaluat ca o versiune alpha. Cu toate acestea, ZLUDA poate fi deja folosit pentru a rula multe aplica\u021bii CUDA, inclusiv Geekbench, 3DF Zephyr, Blender, Reality Capture, LAMMPS, NAMD, waifu2x, OpenFOAM \u0219i Arnold. Este asigurat\u0103 o suport minim pentru primitivele \u0219i bibliotecile cuDNN, cuBLAS, cuSPARSE, cuFFT, NCCL \u0219i NVML. <\/p>\n

Primul start al aplica\u021biilor CUDA sub controlul ZLUDA se realizeaz\u0103 cu \u00eent\u00e2rzieri semnificative, deoarece ZLUDA compileaz\u0103 codul GPU. \u00cen execu\u021biile ulterioare, aceast\u0103 \u00eent\u00e2rziere este absent\u0103, deoarece codul compilat este p\u0103strat \u00een cache. \u00cen timpul execu\u021biei codului compilat, performan\u021ba este aproape de nativ. La rularea pachetului Geekbench pe GPU AMD Radeon 6800 XT, varia\u021bia setului de teste pentru CUDA, efectuat\u0103 cu ajutorul ZLUDA, a ar\u0103tat o performan\u021b\u0103 semnificativ mai bun\u0103 dec\u00e2t varia\u021bia bazat\u0103 pe OpenCL.

\"A<\/center> <\/p>\n

Suportul oficial pentru API-ul CUDA Driver \u0219i p\u0103r\u021bile nedocumentate ale API-ului CUDA studiate prin inginerie invers\u0103 sunt implementate \u00een ZLUDA prin \u00eenlocuirea apelurilor de func\u021bii cu func\u021bii echivalente oferite \u00een runtime-ul HIP, care este foarte similar cu CUDA. De exemplu, func\u021bia cuDeviceGetAttribute() este \u00eenlocuit\u0103 cu hipDeviceGetAttribute(). \u00cen mod similar, compatibilitatea cu bibliotecile NVIDIA, cum ar fi NVML, cuBLAS \u0219i cuSPARSE, este asigurat\u0103; pentru aceste biblioteci, ZLUDA ofer\u0103 biblioteci de transla\u021bie cu acela\u0219i nume \u0219i acela\u0219i set de func\u021bii, construite ca suprapuse peste biblioteci similare AMD. <\/p>\n

Codul GPU al aplica\u021biilor, compilat \u00een reprezentarea PTX (Parallel Thread Execution), este translatat de un compilator special, mai \u00eent\u00e2i \u00een reprezentarea intermediar\u0103 LLVM IR, pe baza c\u0103reia se genereaz\u0103 cod binar pentru GPU AMD.

\"A<\/center>
\n
Sursa: opennet.ro<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"

\u041f\u0440\u043e\u0435\u043a\u0442 ZLUDA \u043f\u043e\u0434\u0433\u043e\u0442\u043e\u0432\u0438\u043b \u043e\u0442\u043a\u0440\u044b\u0442\u0443\u044e \u0440\u0435\u0430\u043b\u0438\u0437\u0430\u0446\u0438\u044e \u0442\u0435\u0445\u043d\u043e\u043b\u043e\u0433\u0438\u0438 CUDA \u0434\u043b\u044f GPU AMD, \u043f\u043e\u0437\u0432\u043e\u043b\u044f\u044e\u0449\u0443\u044e \u0432\u044b\u043f\u043e\u043b\u043d\u044f\u0442\u044c \u043d\u0435\u043c\u043e\u0434\u0438\u0444\u0438\u0446\u0438\u0440\u043e\u0432\u0430\u043d\u043d\u044b\u0435 CUDA-\u043f\u0440\u0438\u043b\u043e\u0436\u0435\u043d\u0438\u044f \u0441 \u043f\u0440\u043e\u0438\u0437\u0432\u043e\u0434\u0438\u0442\u0435\u043b\u044c\u043d\u043e\u0441\u0442\u044c\u044e, \u0431\u043b\u0438\u0437\u043a\u043e\u0439 \u043a \u043f\u0440\u043e\u0438\u0437\u0432\u043e\u0434\u0438\u0442\u0435\u043b\u044c\u043d\u043e\u0441\u0442\u0438 \u043f\u0440\u0438\u043b\u043e\u0436\u0435\u043d\u0438\u0439, \u0432\u044b\u043f\u043e\u043b\u043d\u044f\u0435\u043c\u044b\u0445 \u0431\u0435\u0437 \u043f\u0440\u043e\u0441\u043b\u043e\u0435\u043a. \u041e\u043f\u0443\u0431\u043b\u0438\u043a\u043e\u0432\u0430\u043d\u043d\u044b\u0439 \u0438\u043d\u0441\u0442\u0440\u0443\u043c\u0435\u043d\u0442\u0430\u0440\u0438\u0439 \u043e\u0431\u0435\u0441\u043f\u0435\u0447\u0438\u0432\u0430\u0435\u0442 \u0431\u0438\u043d\u0430\u0440\u043d\u0443\u044e \u0441\u043e\u0432\u043c\u0435\u0441\u0442\u0438\u043c\u043e\u0441\u0442\u044c \u0441 \u0441\u0443\u0449\u0435\u0441\u0442\u0432\u0443\u044e\u0449\u0438\u043c\u0438 CUDA-\u043f\u0440\u0438\u043b\u043e\u0436\u0435\u043d\u0438\u044f\u043c\u0438, \u0441\u043e\u0431\u0440\u0430\u043d\u043d\u044b\u043c\u0438 \u043f\u0440\u0438 \u043f\u043e\u043c\u043e\u0449\u0438 \u043a\u043e\u043c\u043f\u0438\u043b\u044f\u0442\u043e\u0440\u0430 CUDA \u0434\u043b\u044f GPU NVIDIA. \u0420\u0435\u0430\u043b\u0438\u0437\u0430\u0446\u0438\u044f \u0440\u0430\u0431\u043e\u0442\u0430\u0435\u0442 \u043f\u043e\u0432\u0435\u0440\u0445 \u0440\u0430\u0437\u0432\u0438\u0432\u0430\u0435\u043c\u043e\u0433\u043e \u043a\u043e\u043c\u043f\u0430\u043d\u0438\u0435\u0439 AMD \u0441\u0442\u0435\u043a\u0430 ROCm \u0438 runtime HIP (Heterogeneous-computing Interface for Portability). \u041a\u043e\u0434 […]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":1,"featured_media":113553,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[702],"tags":[],"class_list":["post-113552","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\n","aioseo_head_json":{"title":"\ud83e\udd47A fost publicat instrumentul ZLUDA, care permite rularea aplica\u021biilor CUDA pe GPU AMD | ProHoster","description":"Proiectul ZLUDA a preg\u0103tit o implementare deschis\u0103 a tehnologiei CUDA pentru GPU AMD, permi\u021b\u00e2nd rularea aplica\u021biilor CUDA nemodificate cu performan\u021b\u0103 apropiat\u0103 de cea a aplica\u021biilor care ruleaz\u0103 f\u0103r\u0103.","canonical_url":"https:\/\/prohoster.info\/ro\/blog\/news\/opublikovan-instrumentarij-zluda-pozvolyayushhij-zapuskat-cuda-prilozheniya-na-gpu-amd","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"ro_RO","og:site_name":"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b","og:type":"article","og:title":"\ud83e\udd47\u041e\u043f\u0443\u0431\u043b\u0438\u043a\u043e\u0432\u0430\u043d \u0438\u043d\u0441\u0442\u0440\u0443\u043c\u0435\u043d\u0442\u0430\u0440\u0438\u0439 ZLUDA, \u043f\u043e\u0437\u0432\u043e\u043b\u044f\u044e\u0449\u0438\u0439 \u0437\u0430\u043f\u0443\u0441\u043a\u0430\u0442\u044c CUDA-\u043f\u0440\u0438\u043b\u043e\u0436\u0435\u043d\u0438\u044f \u043d\u0430 GPU AMD | ProHoster","og:description":"\u041f\u0440\u043e\u0435\u043a\u0442 ZLUDA \u043f\u043e\u0434\u0433\u043e\u0442\u043e\u0432\u0438\u043b \u043e\u0442\u043a\u0440\u044b\u0442\u0443\u044e \u0440\u0435\u0430\u043b\u0438\u0437\u0430\u0446\u0438\u044e \u0442\u0435\u0445\u043d\u043e\u043b\u043e\u0433\u0438\u0438 CUDA \u0434\u043b\u044f GPU AMD, \u043f\u043e\u0437\u0432\u043e\u043b\u044f\u044e\u0449\u0443\u044e \u0432\u044b\u043f\u043e\u043b\u043d\u044f\u0442\u044c \u043d\u0435\u043c\u043e\u0434\u0438\u0444\u0438\u0446\u0438\u0440\u043e\u0432\u0430\u043d\u043d\u044b\u0435 CUDA-\u043f\u0440\u0438\u043b\u043e\u0436\u0435\u043d\u0438\u044f \u0441 \u043f\u0440\u043e\u0438\u0437\u0432\u043e\u0434\u0438\u0442\u0435\u043b\u044c\u043d\u043e\u0441\u0442\u044c\u044e, \u0431\u043b\u0438\u0437\u043a\u043e\u0439 \u043a \u043f\u0440\u043e\u0438\u0437\u0432\u043e\u0434\u0438\u0442\u0435\u043b\u044c\u043d\u043e\u0441\u0442\u0438 \u043f\u0440\u0438\u043b\u043e\u0436\u0435\u043d\u0438\u0439, \u0432\u044b\u043f\u043e\u043b\u043d\u044f\u0435\u043c\u044b\u0445 \u0431\u0435\u0437.","og:url":"https:\/\/prohoster.info\/ro\/blog\/news\/opublikovan-instrumentarij-zluda-pozvolyayushhij-zapuskat-cuda-prilozheniya-na-gpu-amd","og:image":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:secure_url":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:width":350,"og:image:height":350,"article:published_time":"2024-02-12T20:02:45+00:00","article:modified_time":"2024-02-12T20:02:45+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster","article:author":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":[],"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/posts\/113552","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/comments?post=113552"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/posts\/113552\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/media\/113553"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/media?parent=113552"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/categories?post=113552"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/tags?post=113552"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}