Ikibazo gihuza ibibazo · Biragoye Urwego

Ikibazo cya CUDA na GPU

Reba ibyibanze bya CUDA, harimo intangiriro, SIMT ikora, kwinjira mububiko, ububiko busangiwe, hamwe nibikoresho byo murwego rwohejuru bya GPU.

Inzira zijyanye no kuyoboraGahunda ya Cuda na GPU
Ikibazo 1 Bya 8

Mu magambo ya CUDA, 'intangiriro' ni iki?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Mu magambo ya CUDA, 'intangiriro' ni iki?
    • Imikorere ikora murwego rumwe na GPU nyinshi
    • Intangiriro ya sisitemu y'imikorere
    • Ubwoko bwa chip chip
    • Sisitemu yo gukonjesha GPU
  2. Icyitegererezo cyo gukora SIMT gisobanura iki?
    • Buhoro Kwinjiza, Kwimura Byinshi
    • Inyigisho imwe, Imitwe myinshi ikoresha amabwiriza amwe kumakuru atandukanye
    • Ishusho imwe, Imiterere myinshi
    • Amabwiriza akurikiranye, Igihe cyintoki
  3. Kuki PyTorch na TensorFlow bidakunze gusaba abakoresha kwandika CUDA mbisi?
    • GPU ntishobora gukoresha kode y'abakoresha
    • CUDA biremewe gukoresha mu buryo butaziguye
    • Bita amasomero meza ya CUDA nka cuDNN na cuBLAS imbere
    • Bakorera gusa kuri CPU
  4. 'Warp divergence' ni iki kandi ni ukubera iki bibabaza imikorere?
    • Ikosa rya neti ita paki
    • GPU yunamye kumubiri munsi yubushyuhe
    • Ikosa risiba kwibuka
    • Imitwe iri murugamba ifata amashami atandukanye kandi igomba gukora muburyo bukurikirana
  5. Ni ukubera iki 'guhuriza hamwe' kwibuka kwibuka ari ngombwa muri CUDA?
    • Utudodo twegeranye dusoma aderesi yegeranye cyane
    • Irabika amakuru mu buryo bwikora
    • Irekura umwanya wa disiki
    • Irahagarika abakunzi ba GPU
  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