Industry Tutorial IT3
Title: GPU Code Generation from MATLAB – Virtual Lab
Requirements (Who Should Attend): Engineers experienced in MATLAB with a need to deploy algorithms to a GPU for acceleration in MATLAB or a stand-alone system. Knowledge of CUDA and/ or GPU architectures is not required but may be useful.
Each engineer should have a computer, access to a network, and a sufficiently large monitor. It’s recommended to have a resolution of greater than 2048×1536. A simple laptop screen will not be enough. Having a separate screen for the WebEx is ideal, though not necessary. No GPU is required. The virtual lab will be run virtually through WebEx and the tools will be provided with GPU-enabled instances of MATLAB Online. MATLAB Online recommends using the Chrome internet browser.
See this link for recommended browsers: https://www.mathworks.com/support/requirements/browser-requirements.html
Description: Learn how to generate CUDA code automatically from MATLAB to run on NVIDIA desktop and embedded GPUs. MATLAB is the ideal environment for exploring, developing, and prototyping algorithms. GPUs are the hardware of choice for many applications, such as signal, image processing, and deep learning, that benefit from the parallel processing they offer. GPU Coder offers a direct route to transition from MATLAB development to deployment on GPUs via the generation of CUDA code.
This virtual lab will guide you through hands-on exercises design to ramp you up quickly on GPU Coder. Through these exercises, you will experience a typical workflow that can then be applied to your projects.
- Prototype and accelerate implementations with automatic CUDA code generation
- Enhance performance through code refactoring and design pattern pragmas
- Generate CUDA from deep learning networks for acceleration and implementation
- Deploy generated code to desktop and embedded GPUs
Organizers and speakers: Jack Ferrari, The MathWorks, USA
Biographies: Jack Ferrari is a product marketing engineer focused on supporting a group of code generation tools, including GPU Coder, at MathWorks. Jack holds a bachelor’s degree in Mechanical Engineering from Boston University.