Multi-Core ARM Processors

Multi-Core ARM Processors: Parallel Programming for High-Performance Embedded Systems Michael SOUVALL
In an era where single-core performance has plateaued and embedded applications demand more intelligence, lower power, and stricter timing guarantees, multi-core ARM architectures have become the foundation of modern embedded design. Yet most resources treat parallelism as a desktop/server topic ignoring the brutal constraints of real embedded systems: microsecond deadlines, microamp budgets, thermal walls, safety certification, security boundaries, and the need for reliability that lasts years in the field.
This book bridges that gap.
Written by an engineer who has shipped multi-core designs across bare-metal Cortex-M, RTOS-based MCUs, and Linux Cortex-A SoCs, it delivers a complete, battle-tested progression from architecture fundamentals to production-grade parallel systems.
You will learn how to:

• Understand ARM multi-core realities: Cortex-M symmetry, big.LITTLE heterogeneity, cache coherency costs, weak memory ordering
• Choose the right concurrency model: shared-memory vs message-passing, task vs data vs pipeline parallelism
• Master synchronization: atomics, barriers, spinlocks vs blocking, avoiding races/deadlocks
• Build bare-metal and RTOS multi-core code: boot, inter-core communication, affinity, deterministic scheduling
• Use POSIX threads and OpenMP effectively in embedded Linux with affinity, tuning, and pitfalls
• Optimize memory and cache performance: eliminate false sharing, tile loops, prefetch, profile misses
• Design power-aware systems: DVFS, core gating, energy-efficient scheduling, balancing throughput vs battery life
• Debug and profile rigorously: jitter measurement, tracing, perf, powertop, interpreting multi-core metrics
• Apply real-world patterns through case studies: sensor pipelines, image processing, signal processing, networking acceleration
• Navigate advanced topics: heterogeneous systems, accelerators/NPUs, virtualization, safety-critical/secure design
• Know exactly when NOT to use parallelism and how to avoid common fatal mistakes