QBoard » Artificial Intelligence & ML » AI and ML - PyTorch » What is the best way to learn CUDA?
User Dashboard

What is the best way to learn CUDA?

Back To Topics
Tags : data python datascience
  • 126
    What is the best way to learn CUDA?
      August 24, 2021 4:43 PM IST
    0
  • 461

    f you want to quickly accelerate your application code, then try the Accelerated Libraries like CUBLAS, CuFFT, CuDNN, CULA, ArrayFire, CuSPARSE, OPENCV, etc. This is highly recommended.
    If you want to learn from the basics, try coursera course "Heterogeneous Parallel Programming". There are assignments, quiz etc.

    For a beginner, the book " CUDA by Example" is good to start. As you get comfortable of syntax and pointers, try "Programming Massively Parallel Processors: A Hands-on Approach" book. NVIDIA CUDA Resources will help in getting started. (All the books in http://nvidia.com are removed. You can buy the books in amazon/flipkart easily.)
    C/C++ pointers,& structures handling is needed. Here is one of my CUDA Workshop slides.

    For advanced programming, use CUDA Guide:Online Documentation as reference. Also, GTC Presentations will be helpful for advanced topics.
    After a certain level of expertise in CUDA, you need to start focusing on parallel algorithms. The above books, GPU Computing GEMs - Jade Edition, Emerald Edition, will be helpful.

      August 26, 2021 2:04 PM IST
    0
  • 317

    If you want to quickly accelerate your application code, then try the Accelerated Libraries like CUBLAS, CuFFT, CuDNN, CULA, ArrayFire, CuSPARSE, OPENCV, etc. This is highly recommended.
    If you want to learn from the basics, try coursera course "Heterogeneous Parallel Programming". There are assignments, quiz etc.

    For a beginner, the book " CUDA by Example" is good to start. As you get comfortable of syntax and pointers, try "Programming Massively Parallel Processors: A Hands-on Approach" book. NVIDIA CUDA Resources will help in getting started. (All the books in http://nvidia.com are removed. You can buy the books in amazon/flipkart easily.)
    C/C++ pointers,& structures handling is needed. Here is one of my CUDA Workshop slides.

    For advanced programming, use CUDA Guide:Online Documentation as reference. Also, GTC Presentations will be helpful for advanced topics.
    After a certain level of expertise in CUDA, you need to start focusing on parallel algorithms. The above books, GPU Computing GEMs - Jade Edition, Emerald Edition, will be helpful.

      January 8, 2022 2:33 PM IST
    0
  • 346 3
    Think up a numerical problem and try to implement it. Make sure that you have an NVIDIA card first. :) Download the SDK from NVIDIA web site. Read the "CUDA programming guide", it's less than 200 pages long and sufficiently well written that you should be able to do it in one pass. Pick a sufficiently simple sample and start modifying/rewriting it.
      August 30, 2021 1:18 PM IST
    0