​

While Moore's law continues to remain valid, processor speed has reached a plateau for quite some time owing to the limitations of CMOS technology. High performance computing increasingly relies on massive parallelism in order to make computationally tractable bigger and bigger problems. Parallel computing has a long history, but until recently parallel machines have been the domain of a small number of wellfunded supercomputing centres. Recently, however, inexpensive commodity GPUs containing thousands of processors have become available. NVIDIA has developed a generalpurpose programming language for GPUs known as CUDA, allowing a large range of problems to be run on GPUs. This Mini-school will provide an introduction to parallel programming concepts emphasising hands-on CUDA programming experience.

More information: https://bit.ly/3utwPxG​