C++ Standard Parallelism
summary
Claims Standard C++ now gives a practical, portable on‑ramp to run parallel code on CPUs GPUs and clusters; concrete pieces include parallel algorithms with execution policies, composable async senders and receivers, and multidimensional views via MD span. Examples show parallel transform_reduce word count, a tiled inclusive scan built from bulk/then, a matrix transpose, and mini‑apps that scale to hundreds of GPUs using prototype implementations that compile run and match other back ends' performance.
AI-written summary · about
chapters (12)
- 0:00C++ standard parallelism intro
- 1:46Introduction
- 2:50Common parallel algorithms
- 7:24Ranges and parallel iteration
- 9:29Fork-join limits
- 13:21Senders and receivers
- 23:02Sender factories
- 29:22Sender examples takeaways
- 32:21MD span for multi-dim data
- 43:01Future of numerical algorithms
- 46:38C++ parallelism summary
- 49:54Q&A on schedulers
related talks
01h 07mGeneric Parallel Programming
53mIntroduction to C++ AMP (GPGPU Computing)
54mParallel Algorithms Ranges and oneDPL
48mBridging the Gap: Writing Portable C++ Programs for CPU and GPU
19mSTRADS-AP: Simplifying Distributed Machine Learning Programming without Introducing a New Programming Model
22mConcurrent and Parallel programming in Julia
10mGraph Based Pipeline Parallelism
01h 05mstd::simd: How to Express Inherent Parallelism Efficiently Via Data-parallel Types