Kedis ku xidhan hagaha · Adag Heerka

CUDA iyo Kediska Barnaamijyada GPU

Hubi fikradaha asaasiga ah ee CUDA, oo ay ku jiraan kernels, fulinta SIMT, helitaanka xusuusta, xusuusta la wadaago, iyo qalabka barnaamijka GPU-ga heerka sare ah.

Wadooyinka haga ee la xidhiidhaBarnaamijyada Cuda iyo GPU
Su'aal 1 ee 8

Eray bixinta CUDA, waa maxay 'kernel'?

Horudhac su'aal kasta8 su'aalo

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

  1. Eray bixinta CUDA, waa maxay 'kernel'?
    • Shaqaale ay barbar socdaan silsilado badan oo GPU ah
    • Xuddunta nidaamka hawlgalka
    • Nooca chip xusuusta
    • Habka qaboojinta GPU
  2. Waa maxay macnaha qaabka fulinta SIMT?
    • Soo gelis qunyar socod ah, Bedel badan
    • Tilmaan-bixin keli ah, taxane badan ayaa ku socodsiiya tilmaamo isku mid ah xogo kala duwan
    • Sawir Keliya, Qoraallo Badan
    • Tilmaamaha isku xigxiga, Wakhtiga Buugga
  3. Waa maxay sababta PyTorch iyo TensorFlow ay dhif ugu baahan yihiin isticmaalayaasha inay qoraan CUDA cayriin?
    • GPU-yadu ma socodsiin karaan koodka isticmaalaha
    • CUDA waa sharci darro in si toos ah loo isticmaalo
    • Waxay ugu yeeraan maktabadaha CUDA la hagaajiyay sida cuDNN iyo cuBLAS gudaha
    • Kaliya waxay ku shaqeeyaan CPU-yada
  4. Waa maxay kala duwanaanshuhu waa maxay sababta ay u dhaawacdo waxqabadka?
    • Cilad xidhiidhineed oo soo riddo baakado
    • GPU jir ahaan kulaylka hoostiisa ku foorarsada
    • Cayayaanka tirtira xusuusta
    • Xadhkaha dagaalku waxa ay qaataan laamo kala duwan waana in ay si taxane ah u socdaan
  5. Waa maxay sababta gelitaanka xusuusta 'isku-duuban' ay muhiim ugu tahay CUDA?
    • Mawduucyada ku dheggan akhrinta ciwaannada ku dheggan ayaa aad waxtar u leh
    • Si toos ah ayay xogta u siraysaa
    • Waxay xoraynaysaa booska diskooga
    • Waxay curyaamisaa taageerayaasha GPU-ga
  6. What is OpenAI Triton in relation to CUDA programming?
    • A CUDA driver that replaces NVIDIA GPUs
    • A Python-like language and compiler for writing efficient GPU kernels
    • A model-serving server that only loads TensorFlow checkpoints
    • A framework that converts GPUs into CPUs
  7. What does a CUDA kernel launch configuration specify?
    • A model tokenizer vocabulary
    • The number of training labels
    • The grid and block dimensions used to organize thread execution
    • The host operating system
  8. Why might a CUDA kernel use shared memory?
    • To make data available across every cluster without communication
    • To replace all registers
    • To store the model permanently on the GPU
    • To let threads in a block reuse data from a faster on-chip memory space