NVIDIA DLI Releases New Accelerated Computing Teaching Kit

The NVIDIA Deep Learning Institute (DLI) recently released the latest version of the Accelerated Computing Teaching Kit. NVIDIA Teaching Kits are complete course solutions for use by educators in a variety of academic disciplines that benefit from GPU-accelerated computing. 

The new modules were developed in collaboration with University of Delaware Professor Sunita Chandrasekaran and University of Illinois at Urbana–Champaign (UIUC) Professor Wen-mei Hwu.

DLI Teaching Kits are available to qualified university educators interested in course solutions across deep learning, accelerated computing, and robotics. Educators can integrate lecture slides, hands-on labs with sample solutions, GPU cloud resources, and more into their curriculum.

“It is critical for the next-generation workforce to be up-to-date with the newer concepts and techniques of CUDA as they learn about scientific applications and massively parallel computing architectures,” Chadrasekaran said.

Here’s a snapshot of the what’s covered in the new modules: 

  • Unified Memory:  Learn how to simplify GPU memory management and teach students additional CUDA paradigms such as memory oversubscribing.
  • Dynamic Parallelism: Show how the application enables GPU threads to launch kernels resulting in nested parallelism (versus traditional flat, bulk parallel programming models). 
  • Multi-GPU:  Explore paradigms like OpenMP for multi-GPU coordination. This is critical for students as more systems become equipped with “fat” nodes, i.e. a single node consisting of multiple GPUs.The module also offers other techniques and strategies students can leverage for multi-GPU programming.  
  • CUDA libraries: Covers libraries such as cuBLAS, cuSOLVER, cuFFT and Thrust, which form an integral part of modern GPU programming and often interoperate with CUDA and other programming models including OpenMP and OpenACC.

Other additions to the Accelerated Computing Teaching Kit include new lab exercises and slide decks for teaching Pinned Memory, Breadth-First Search (BFS) algorithms, and profiling with NSIGHT (which replaced NVPROF in CUDA 10).

“The new content in the updated Teaching Kit reflects several new advancements in GPU Computing,” Hwu said. “ It has been a pleasure working with the NVIDIA teams that have provided excellent insight and content to the educator community.”

To learn more about the DLI Teaching Kits visit the Teaching Kits homepage on our developer portal. Additionally, read how university educators are pulling NVIDIA Teaching Kits into their classrooms