If you're going to be working on any aspect of multicore embedded system design—be it systems architecture, SoC development, or software programming—a newly published book titled "Real World Multicore Embedded Systems" will be an excellent guide. In this 600-page encyclopedic resource, 14 authors, including 2 from Cadence, cover just about every aspect of multicore embedded system development.
The book was edited by Bryon Moyer, editor and writer for EE Journal, and is available from Elsevier or Amazon. EDN recently ran a series of articles adapted from a chapter by Frank Schirrmeister, senior director for the System Development Suite at Cadence, that gives a comprehensive overview of multicore architectures.
In his introduction in Chapter 1, Moyer notes that multicore architectures have been around for many years in desktop computers and supercomputers, but adoption of multicore has lagged in the mainstream embedded world. But for embedded systems today, Moyer writes, "multiple cores have become an unavoidable reality." Fortunately, there are new tools, libraries, and language features that make embedded multicore programming "less of a walk in the wild."
This book introduces readers to a wealth of practical, how-to information about embedded multicore system development. It covers system architecture, hardware, firmware, software, and tools. Key topics include concurrency, architecture, infrastructure, operating systems, virtualization, applications software, hardware acceleration, and system-level considerations. Here's a brief overview of the subsequent chapters:
Chapter 2—The Promise and Challenges of Concurrency: Most of the book is dedicated to managing the challenges of concurrency, and in this chapter Bryon Moyer introduces basic concepts of concurrency, data and functional parallelism, and dependencies.
Chapter 3—Multicore Architectures: Frank Schirrmeister shows why multicore architectures are increasingly required and reviews different types of architectures such as heterogeneous, homogenous, symmetric multiprocessing (SMP), and asymmetric multiprocessing (AMP). He discusses processing elements, communications architectures, memory architectures, and programming challenges.
Chapter 4—Memory Models for Embedded Multicore Architecture: Gitu Jain, software engineer at Synopsys, explains the memory types and memory architectures used in embedded multicore systems. He also covers cache coherency and transactional memory.
Chapter 5—Design Considerations for Multicore SoC Interconnections: Sanjay Deshpande, design manager at Freescale Semiconductor, explains communication activity in multicore SoCs, requirements and topologies of SoC traffic, latency and bandwidth, and interconnection networks. He covers building blocks such as links, clocking, and switches, and explains the various types of multicore SoC networks you may encounter.
Chapter 6—Operating Systems in Multicore Platforms: There is no standard way of configuring an OS for an embedded multicore platform, and many choices to be made. Bryon Moyer discusses SMP and AMP, controlling OS behavior, and debugging a multicore system.
Chapter 7—System Virtualization in Multicore Systems: David Kleidermacher, CTO of Green Hills Software, covers applications of system virtualization, hypervisor architectures, hardware assists, and I/O virtualization. A case study examines power architecture virtualization with the Freescale P4080 processor.
Chapter 8—Communication and Synchronization Libraries: Parallelization of embedded applications requires the use of communication and synchronization primitives. Max Domeika, technical project manager at Intel, shows how libraries have been developed to provide them.
Chapter 9—Programming Languages: Gitu Jain of Synopsys overviews programming languages for multicore embedded systems including C, C++, Java, Python, and Ada.
Chapter 10—Tools: Kenn Leucke from Boeing Test and Evaluation surveys real-time operating systems (RTOS), communications tools, parallelizing serial software tools, software development tools, and benchmarking tools from a number of providers.
Chapter 11—Partitioning Programs for Multicore Systems: Bryon Moyer and Paul Stravers (chief architect, Vector Fabrics) tackle the hard challenge of partitioning programs, including legacy software, for concurrent operation on multicore systems. They discuss dependencies, synchronizing data, and using tools to simplify partitioning.
Chapter 12—Software Synchronization: In this chapter Tom Dickens, Boeing associate technical fellow, shows how to synchronize multiple concurrently executing algorithms. The discussion includes the need for synchronization, how synchronization is achieved, language support for implementation, side effects of synchronization, and common problems.
Chapter 13—Hardware Accelerators: Bryon Moyer teams up with Yosinori Watanabe, senior architect at Cadence, to describe hardware accelerators at a system level. They focus on architectural alternatives and show how accelerators can be invoked by software programs.
Chapter 14—Multicore Synchronization Hardware: Jim Holt, systems architect at Freescale, discusses instruction set support for synchronization, hardware support for synchronization, hardware support for lock-free programming, and memory subsystem considerations.
Chapter 15—Bare-Metal Systems: Some embedded multicore systems eliminate the OS altogether. Sanjay Lal, co-founder of Kyma Systems, shows how these "bare metal" systems work and how they can be built.
Chapter 16—Multicore Debug: Neal Stollon, principal engineer with HDL Dynamics, discusses debug instrumentation, trace methods for multicore debug analysis, and debug flows and subsystems. He compares some commercial approaches.