Topic 10. OpenMP: parallel regions and loops, reductions, tasks; thread affinity and NUMA

Goal: become familiar with the fork–join model and OpenMP directives; learn to parallelize loops with different kinds of iteration scheduling, use reductions (including user-defined ones), synchronization, and tasks for recursive algorithms, and vectorize loops with the simd directive; master first-touch data initialization, thread binding with the OMP_PLACES and OMP_PROC_BIND variables, and measuring the speedup of C++ programs.

Lecture contents

1. The fork–join model and parallel regions — The fork–join model and building OpenMP programs · The parallel region
2. Variables, loops, and reductions — Variable scope · Parallel loops · Reductions
3. Synchronization, tasks, and simd — Synchronization · Sections and tasks · simd vectorization
4. Performance, NUMA, and thread affinity — Performance and scalability · The NUMA architecture · Thread affinity
5. Examples and common mistakes — Program examples · Common mistakes

Practice and review

• Practice — 3 worked examples with complete code and output
• Tasks — 30 variants, each with tasks at three difficulty levels
• Summary — conclusions, self-check questions, and useful links