Glossary

Key Points

Introduction	A GPU (Graphics Processing Unit) is best at data-parallel, arithmetic-intense calculations CUDA is one of several programming interfaces for general-purpose computing on GPUs Alliance clusters have special GPU-equipped nodes, which must be requested from the scheduler
Hello World	Use nvcc to compile CUDA source files are suffixed .cu Use salloc to get an interactive session on a GPU node for testing
Adding Two Integers	The CPU (the ‘host’) and the GPU (the ‘device’) have separate memory banks This requires explicit copying of data to and from the GPU
Adding vectors with GPU threads	Threads are the lowest level of parallelization on a GPU A kernel function replaces the code inside the loop to be parallelized The CUDA `<<<M,N>>>` notation replaces the loop itself
Using blocks instead of threads	Use `nvprof` to profile CUDA functions Blocks are the batches in which a GPU handles data Blocks are handled by streaming multiprocessors (SMs) Each block can have up to 1024 threads (on our current GPU cards)
Putting It All Together	A typical kernel indexes data using both blocks and threads

FIXME