GPU with Cuda

Dates: 21-22 November 2017

Location: STFC, Daresbury

Visitors to Daresbury please report to reception in the first instance.

Timetable

Please note that the exact content of the slides may be updated before the course.

A link from this page to the web-based hands-on exercise system will be made live at the time of the course.

Day 1

Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.

• 10:00 Introduction (Slides)
• 10:20 GPU Concepts/Architectures (Slides)
• 11:00 Break
• 11:20 CUDA Programming (Slides)
• 12:00 A first CUDA Exercise
• 13:00 Lunch
• 14:00 Some Optimisations (Slides
• 14:20 CUDA Exercise
• 15:00 Break
• 15:20 CUDA contant and shared memory
• 16:00 CUDA Exercise
• 17:00 Close

Day 2

Unless otherwise indicated all material is Copyright © EPCC, The University of Edinburgh, and is only made available for private study.

• 09:00 Recap/Questions
• 09:10 OpenCL / Directives
• 10:00 OpenCL or Directives exercise
• 11:00 Break
• 11:30 Guest presentation from Alan Gray of NVIDIA: NVIDIA Volta
• 12:30 Questions/Continue Exercises
• 13:00 Lunch
• 14:00 Case Study in Scientific Applications I: GPU acceleration strategies for algebraic multi-grid CFD codes (Samuel Antao, IBM)
• 15:00 Break
• 15:20 Case Study II: Porting DL_MESO (DPD) on NVIDIA GPUs (Jony Castagna, STFC)
• 16:00 Close

Course Chat

The chat page is a live collaborative online document using Dropbox Paper which we will use to share links, information and comments. All course participants are encouraged to contribute.