C++ Standard Parallelism
summary
Standard C++ already provides portable parallelism: transform_reduce word count shows a single parallel algorithm mapping to CPUs or GPUs; execution policies (seq, unseq, par, par_unseq) control allowed parallelism; mdspan in C++23 gives non-owning multi-dimensional views and slicing for tiled kernels; an asynchronous model senders and receivers composes task graphs and lets one line swap between single-thread, multi-GPU, or distributed runs — prototypes scale real apps (LULESH, Palabos) up to 512 GPUs.
AI-written summary · about
chapters (5)
- 0:00Start
- 0:54Conventions
- 36:35Maxwell's Equations
- 38:53Standard Algorithms
- 55:46Pillars of Standard Parallelism
related talks
58mC++ Standard Parallelism
57mC++17 Parallel Algorithms
57m“The C++17 Parallel Algorithms Library and Beyond"
01h 10mC++17 ParallelSTL: A Standardization Experience Report for CPU and GPU on SYCL
01h 33mGPU programming with modern C++
50mC++ Parallel Programming Models
01h 02m“Parallelism in Modern C++"
01h 02mInside Visual C++' Parallel Algorithms