NCI AustraliaPawsey Supercomputing Centre and the New Zealand eScience Infrastructure  are the leaders in high-performance compute and data across the Australasian region.

    
Objective: Learn intermediate to advance CUDA optimization techniques for code acceleration.
Prerequisites: Basic experience with C/C++ as well as knowledge of parallel computing fundamentals or basics of CUDA. 
How to Prepare: You can find resources that will help learn basic CUDA and GPU Programming concepts here
 

Agenda

Day 1: 7th Oct 2020  ( 11:30 am – 2:30 pm Aust Eastern Standard Time )

  • Welcome (Moderator): (11:30 am – 11:45 am)
  • Connecting to Hackathon cluster (11:45 am -12:15 pm)
  • CUDA 101 (12:15 – 1:15 pm)
  • GPU Memory Hierarchy ( 1:15 – 2:30 )
    • GPU Memory Hierarchy evolution
    • Global memory

Day 2:  8th Oct 2020  ( 11:30 am – 2:30 pm Aust Eastern Standard Time )

  • GPU Memory Hierarchy ( 11:30 – 12:30 )
    • Shared Memory
  • CUDA Thread Programming (12:30 – 2:30)
    • GPU Occupancy
    •  Atomic operation
    •  Mixed precision

Day 3:  9th Oct 2020  ( 11:30 am – 2:30 pm Aust Eastern Standard Time )

  • CUDA Thread Programming (11:30 – 12:30)
    • Warp Divergence and Warp Synchronous programming
  • Debugging and Profiling(12:30 – 2:30)
    • CUDA Sanitizer
    • Nsight profiling Compute: Roofline Analysis

Please note the agenda may change slightly.

2020-09-21T08:00:00-07:00 296 Bootcamps HPC
Event Focus
HPC

Important Event Dates

  • Application Deadline
  • Day 1
  • Day 2
  • Day 3

Questions?