This workshop presents an introduction to using CUDA to harness the power of an Nvidia graphics card. We will cover the essential basics required to write, compile, run and monitor a program.
Prerequisites
- An Alliance account
- Basic C knowledge
- Basic Linux command line competence
| Setup | Download files required for the lesson | ||
| Day 1 | 14:00 | 1. Introduction |
What is a GPU, and why do we use them?
What is CUDA? |
| 14:20 | 2. Hello World | How to compile and run Hello World | |
| 14:40 | 3. Adding Two Integers | How does GPU memory work? | |
| 15:05 | 4. Adding vectors with GPU threads | How can we parallelize code on a GPU? | |
| 15:30 | 5. Using blocks instead of threads |
What is the difference between blocks and threads?
Is one faster than the other? |
|
| 16:00 | Finish | ||
| Day 2 | 14:00 | 6. Putting It All Together |
How do blocks and threads work together?
What other GPU programming resources are there? |
| 14:30 | 7. GPU Hardware Architecture | What are the capabilities of the GPU that I’m using? | |
| 14:45 | 8. Thread Addressing | How can I address multidimensional data structures more intuitively? | |
| 14:55 | 9. Memory Performance | How does memory performance impact overall performance when using GPUs? | |
| 15:15 | 10. Exercise: Julia Set | Exercise to apply what we have learned so far. | |
| 15:50 | 11. Where To Go Next? |
What software libraries can I use so that I don’t need to write my own kernels?
Where can I find additional resources? |
|
| 15:55 | Finish |
The actual schedule may vary slightly depending on the topics and exercises chosen by the instructor.