BarraCUDA Open-source CUDA nuƒoƒoƒula si ɖoa taɖodzinu na AMD GPUwo

\u003ch2\u003eBarraCUDA Ʋuʋu CUDA nuƒoƒoƒula si le taɖodzinu ɖom na AMD GPUwo\u003c/h2\u003e \u003cp\u003eGitHub nudzraɖoƒe sia si le ʋuʋu ɖi la tsi tre ɖi na akpa vevi aɖe le developer ecosystem me. Dɔa ɖe egbegbe ŋgɔyiyi ƒe nuwɔnawo kple nuwɔwɔ aduadu ƒe kɔpiwɔwɔ fia.\u003c/p\u003e \u003ch3\u003eMɔ̃ɖaŋununya ƒe Nɔnɔmewo\u003c/h3\u003e \u003cp\u003eAnɔ eme be nudzraɖoƒea ƒe akpa aɖewoe nye:\u003c/p\u003e \u003cul\u003e \u003cli\u003eKɔda, si woŋlɔ ɖi nyuie\u003c/li\u003e \u003cli\u003eREADME si me kɔ kple zazã ƒe kpɔɖeŋuwo\u003c/li\u003e \u003cli\u003eTaɖoɖo kplɔkplɔ kple nudzɔdzɔ ƒe mɔfiamewo\u003c/li\u003e \u003cli\u003eNu yeyewo kple beléle na wo edziedzi\u003c/li\u003e \u003c/ul\u003e \u003ch3\u003eNutoa me ƒe Ŋusẽkpɔɖeamedzi\u003c/h3\u003e \u003cp\u003eDɔwɔwɔ siwo le ʋuʋu ɖi abe esiawo ene doa sidzedze mama ɖe ŋgɔ eye wowɔa mɔ̃ɖaŋununya ƒe nu yeyewo dodo ɖe ŋgɔ kabakaba to kɔda si ŋu woate ŋu aɖo kple nuwɔwɔ aduadu ƒe ŋgɔyiyi me.\u003c/p\u003e

Nyabiase Siwo Wobiana Enuenu

Nukae nye BarraCUDA eye aleke wòwɔa dɔe?

BarraCUDA nye nuƒoƒoƒula si le ʋuʋu ɖi si ɖea NVIDIA CUDA ƒe kɔpi gɔme be wòawɔ dɔ le eɖokui si le AMD GPUwo dzi. Ewɔa dɔ to CUDA dzɔtsoƒe faɛlwo me toto kple GPU mɔfiame siwo sɔ kple AMD la dodo ɖe go me, si wɔnɛ be dɔwɔlawo te ŋu gbugbɔ zãa CUDA codebases siwo li xoxo evɔ womagbugbɔ wo ŋlɔ na ROCm alo HIP o. Esia xea mɔ na dometsotso si nɔ anyi ɣeyiɣi didi aɖe le GPU kɔmpiuta zazã me, si naa mɔ̃ɖaŋudɔwɔlawo te ŋu trɔna ɖe nɔnɔmewo ŋu evɔ wometsɔa agbalẽdzraɖoƒe kple dɔwɔnu siwo wowɔ xoxo siwo wowɔ wònyo wu ƒe lãwo ƒe agbenɔnɔ ƒe ɖoɖo gã la sa vɔe o.

Ðe mateŋu azã BarraCUDA le nuwɔwɔ ƒe mɔ̃ ƒe nusɔsrɔ̃ ƒe dɔwɔwɔ mea?

BarraCUDA nye dzɔtsoƒe si le ʋuʋu ɖi fifia eye wòle tɔtrɔm, si wɔe be wòsɔ nyuie wu na dodokpɔ kple dɔwɔwɔ siwo mele vevie o le afɔɖeɖe sia me. Le wɔwɔme AI kple ML pɔmpiwo gome la, wokafui be woatsɔe akpe ɖe mɔ̃ sesẽ aɖe ŋu. Mewayz naa modules siwo wu 207 siwo wotsɔ wɔ ɖekae — siwo dometɔ aɖewoe nye AI dɔwɔnuwo, automation, kple analytics — ɖe $19/ɣleti nu, si naa gɔmeɖoanyi si li ke si dzi woate ŋu atsɔ dodokpɔ GPU dɔwɔnuwo abe BarraCUDA ene aƒo xlãe dedie.

AMD GPU kawoe BarraCUDA doa alɔe?

Kpekpeɖeŋu nɔ te ɖe dɔa ƒe ŋgɔyiyi ƒe nɔnɔme si li fifia dzi, gake zi geɖe la, BarraCUDA tɔa ŋku AMD GCN kple RDNA xɔtuɖaŋu GPU siwo sɔ kple ROCm kɔmpiuta dɔwɔɖoɖo ƒe ƒuƒoƒoa. Esia lɔ kaɖi xɔŋkɔ siwo tso RX 6000 kple RX 7000 ƒe hatsotsowo me kpakple Instinct datacenter GPUwo ɖe eme. Kpɔ nudzraɖoƒea ƒe README kple ɖekawɔwɔ ƒe matriki ɣesiaɣi, esime xɔtunu siwo wodo alɔe le kekem ɖe enu kple dodo ɖesiaɖe esime ʋuʋu-dzɔtsoƒe habɔbɔa le megbenya yeyewo kpem.

Aleke BarraCUDA sɔ kple AMD ƒe HIP ʋuʋu dɔwɔnu si dziɖuɖua da asi ɖo?

AMD ƒe HIPify dɔwɔnu bia be woatsɔ asi atrɔ CUDA ƒe kɔda ɖe HIP nyagɔmeɖegbalẽ me, si bia tso dɔwɔla ƒe agbagbadzedze kple kɔda ƒe tɔtrɔ me. BarraCUDA ɖoe be yeawɔ nuƒoƒoƒu mɔ si me kɔ wu, si ana dɔwɔlawo naƒo CUDA ƒe kɔpi nu ƒu tẽ tsoƒe-dzidzenu ƒe ʋuʋu manɔmee. Esia na wòdoa dzidzɔ na ƒuƒoƒo siwo kpɔa kɔpiwɔƒe gãwo dzi ŋutɔ. Dɔwɔla siwo le dɔwɔnu geɖe ƒe AI ƒuƒoƒo tum hã ateŋu akpɔ asixɔxɔ le mɔ̃wo abe Mewayz ene, si ƒoa modules 207+ nu ƒu ɖe $19/ɣleti me, si ɖea dɔwɔnu ƒe kɔsɔkɔsɔ bliboa ƒe sesẽ dzi kpɔtɔna kpe ɖe egbɔkpɔnuwo abe BarraCUDA.

ŋu