Speaker
Description
Description:
The goal of this workshop is to introduce students to CUDA and provide them with an understanding of parallel programming. CUDA is more than a few new keywords. One must understand SIMD and the pitfalls of serialization.
Students will leave with basic CUDA skills and some OpenACC knowledge plus useful machine learning and big data tools as well. My “from Hello World to exascale machine learning in one slide” will also be covered as data parallel training fits on GPUs nicely.
Target Audience:
Anyone with C/C++ programming skills in the Unix environment who wishes to learn about parallel programming and CUDA. The material will be 60% beginner, 30% intermediate, and 10% advanced.
Prerequisites:
C/C++ along with an ability to edit and compile programs in a Unix environment
Special requirements:
Users will have access to a CHPC system with GPUs.
Attendees should bring their own laptops. The ability to view pdf or PowerPoint files is required.
Outline of full syllabus:
08:00 Registration
09:00 Introduction and morning talk (30 minutes)
Login details and extracting the workshop material (15 minutes)
Section 01: Parallel intro and a first CUDA program)
Section 02: Profiling on a GPU
10:30 Morning Refreshment Break
11:00 Section 03: More CUDA and the Thrust Interface
Section 04: “From Hello World to TF/s machine learning”
12:30 Lunch
13:30 Afternoon talk
Section 05: Controlling parallel resources
Section 06: C++ objects and transparent host/GPU data movement
15:00 Afternoon Refreshment Break
15:30 Section 07: Task level parallelism on a GPU
Section 08: Managing big data, CUDA as a scripting language via dynamic load/link
17:00 End of Day
Additional Comments:
Students can work at their own pace.
Introductory students will learn the basics of CUDA and the profiler as well as how to think in parallel and understand the impact of parallel hardware on performance.
Intermediate/advanced students will hone their thinking about parallel programming and the limitations and advantages of GPU hardware. Extra credit exercises will challenge them.
All students will learn how to use machine learning and the ability to explore this hot field and leave with a tool that allows them to train and predict using their own data sets and neural network architectures. Further, they learn how to work with and collaborate using big data.
Presenter Biography
Rob Farber was a pioneer in the field of neural networks while on staff as a scientist in the Theoretical Division at Los Alamos National Laboratory.
He is active in the field and works with companies and national laboratories as a consultant plus teaches about HPC and AI technology worldwide.
Rob can be reached at info@techenablement.com
http://www.techenablement.com/rob-farber/
