HKHLR - HPC Hessen

HiPerCH 12 - Module 3

HiPerCH 12 - Module 3

Introduction to CUDA Programming

 

Content

This course introduces students to the principles and practice of parallel programming for NVIDIA GPUs using CUDA.Through a combination of lectures and practicals, students learn about Nvidia GPU architectures, CUDA C++, fundamental CUDA algorithms and optimization techniques.   

 

Agenda

Day 1 (Wednesday, September 23)

  • 09:00 Welcome
    Overview of the course
    Connecting to the Host Cluster required for the hands-on sessions
  • 09:30 Lecture 1: Introduction to GPU Architecture and Cuda C++
    GPU Architecture and fundamental differences with CPU
    Basic concepts, syntax and API needed or heterogenous programming with CUDA C++
    How to write GPU kernels and manage GPU thread groups
    CUDA API and error Handling
  • 10:45 Break
  • 11:00 Hands-on
    Participants can complete example exercises meant to reinforce the presented concepts
  • 13:00 Lunch Break
  • 14:00 Lecture 2: Fundamental CUDA Optimization
    Optimization strategies related to kernel launch configurations & hiding latency
    GPU Memory Hierarchy
    Global memory throughput and use of shared memory
  • 15:15 Break
  • 15:30 Hands-on
  • 17:30 Closing Day One

Day 2 (Thursday, September 24)

  • 09:00 Lecture 3: CUDA Programming & Parallel Patterns
    Thread execution, memory access and atomic operations
    Fundamental Parallel algorithms: reduction, scan and histogram
    GPU Managed memory: API and performance optimization
  • 10:15 Break
  • 10:30 Hands-on
  • 12:30 Lunch Break
  • 13:30 Lecture 4: CUDA Concurrency
    CUDA concurrency: Motivation and possible scenarios
    Pinned memory
    CUDA Streams concept and API
    Overlapping Computation and data transfer
  • 14:45 Break
  • 15:00 Hands-on
  • 17:00 Final Remarks

 

Trainer(s)

  • Fouzhan Hosseini (NAG)
  • Jacob Senior (NAG)
  • Eleni Vlachopoulou (NAG)

 

Participation

  • The target audience is people who are reasonably competent programmers, and in particular we assume they are familiar with basic C, but no other knowledge is assumed.
  • Sessions will be held online via Zoom. Contact data will be made available after successful registration.
  • The hands-on session will be carried out through an SSH connection on the Lichtenberg cluster at Technical University Darmstadt. Participants shall use their own computer with either Linux/MacOS or Windows with MobaXterm installed. Participants without account to Lichtenberg will be provided with a guest account for the course.
  • Participants will be provided access to Lichtenberg prior to the course to be able to familiarize themselves with the environment and the material.
  • NAG will provide PDFs of the lecture material. Example code and solutions, as source files, will be supplied as appropriate. Printed course materials are not offered.
  • Practical example classes will be conducted via Slack. Attendees will have the opportunity to post questions, compiler error messages, code snippets etc. for discussion.
  • The maximum number of participants is 30.

Participating Universities