CUDA Reading
Class on 2/19 will be a hands-on panel and demo of CUDA. I expect everyone to prepare, have some questions ready, and participate (unless previously discussed an absence). The homework is reading and it will be evaluated with a short quiz in the beginning of class.
What to Read
1. Start by reading this Easy Introduction to CUDA.
Then follow up with at least the following links from the bottom of this introduction:
3. How to Access Global Memory Efficiently.
4. How to Overlap Data Transfers.
CUDA Manual
In addition, you really should strongly consider starting to read the CUDA Programming Guide, which you will almost certainly need to consult as part of Lab 2. The manual includes a lot of useful information about more powerful features of GPUs and CUDA, including synchronization, atomic memory operations, and launching more work from within threads (which NVIDIA calls dynamic parallelism).
