NO VIDEO AVAILABLE

Presenter: Paul Preney, SHARCNET

Today one can easily write efficient multithreaded C++ code using C++ parallel algorithms (including using multidimensional arrays via mdspan) targetting CPUs as well as NVIDIA GPUs (e.g., using NVIDIA's HPC SDK). This allows one to focus on the algorithms needed in order to perform the desired computations. This presentation will discuss how one can utilize ISO C++ parallel algorithms in C++ code, how multidimensional array views, i.e., "mdspan", can be utilized, and how to run programs using either on CPUs and on NVIDIA GPUs.