The workshop was held on November 4th and 5th and there were 70 registrants for the first day and 56 for the second day, consisting of faculty, graduate students and staff. Research Computing was honored to have Bob Crovella, a solutions architect from Nvidia on campus to lead the workshop. Nvidia invented the GPU and is one of the leaders in high performance computing (HPC) for scientific research. The workshop started with introductory material but then soon worked it’s way into advanced topics. Attendance was high and Bob’s presentations were well received as evidenced by the many questions and the lively discussions that took place during the sessions.
The agenda was as follows and the class materials are available via the links below.
Day 1
Morning:
- Intro to GPU computing (no programming)
- Intro to CUDA C/C++ (+ hands-on)
- CUDA basic optimizations (+ hands-on)
Afternoon:
- Working with GPU Libraries (CUBLAS, CUSPARSE) | Windows Example (+hands-on)
- GPU Graphical tools (profiler, nsight EE, debugging) (+ workshop/demo, no hands-on)
Day 2
Morning:
- Intro to OpenACC (+ hands-on)
- OpenACC basic optimizations (+ hands-on)
Afternoon:
- Thrust: GPU programming for the C++ STL programmer (+ hands-on)
- Wrap-up, Q+A, missed topics, next steps
Gzipped Workshop Hands-On Exercises