Applications and Uses of FPGAs

The evolution from “traditional” FPGA architectures, mainly consisting of basic standard reconfigurable building blocks (LBs and IL), to more feature- rich, heterogeneous devices is widening the fields of applicability of FPGAs, taking advantage of their current ability to implement entire complex sys- tems in a single chip. FPGAs are not used anymore just for glue logic or emulation purposes, but have also fairly gained their own position as suit- able platforms to deal with increasingly complex control tasks and are also getting, at a very fast pace, into the world of HPC. 

This technological trend has also extended the applicability of FPGAs in their original application domains. For instance, emulation techniques are evolving into mixed solutions, where the behavior of (parts of) a system can be evaluated by combining simulation models with hardware emulation, in what is nowadays referred to as hardware-in-the-loop (HIL). Tools exist, including some of general use in engineering, such as MATLAB ® , which allow this combined simulation/emulation approach to be used to accelerate system validation. 

FPGAs are also increasingly penetrating the area of embedded control systems, because in many cases, they are the most suitable solution to deal with the growing complexity problems to be addressed in that area. Some important fields of application (not only in terms of technological challenges but also in terms of digital systems’ market share) are in automated man- ufacturing, robotics, control of power converters, motion and machinery control, and embedded units in automotive (and all transportation areas in general)—it is worth noting that a modern car has some 70–100 embedded control units onboard. As the complexity of the systems to be controlled grows, microcontroller and DSPs are becoming less and less suitable, and FPGAs are taking the floor. 

A clear proof of the excellent capabilities of current FPGAs is their recent penetration in the area of HPC, where a few years ago, no one would have thought they could compete with software approaches implemented in large processor clusters. However, computing-intensive areas such as big data applications, astronomical computations, weather forecast, financial risk management, complex 3D imaging (e.g., in architecture, movies, virtual real- ity, or video games), traffic prediction, earthquake detection, and automated manufacturing may currently benefit from the acceleration and energy- efficient characteristics of FPGAs. 

One may argue these are not typical applications of industrial embed- ded systems. There is, however, an increasing need for embedded high- performance systems, for example, systems that must combine intensive computation capabilities with the requirements of embedded devices, such as portability, small size, and low-energy consumption. Examples of such applications are complex wearable systems in the range of augmented or virtual reality, automated driving vehicles, and complex vision systems for robots or in industrial plants. The Internet of Things is one of the main forces behind the trend to integrate increasing computing power into smaller and energy-efficient devices, and FPGAs can play an important role in this scenario. 

Given the complexity of current devices, FPGA designers have to deal with many different issues related to hardware (digital and analog circuits), soft- ware (OSs and programming for single- and multicore platforms), tools and languages (such as HDLs, C, C++, SystemC, as well as some with explicit parallelism, such as CUDA or OpenCL), specific design techniques, and knowledge in very diverse areas such as control theory, communications, and signal processing. All these together seem to point to the need for super- engineers (or even super-engineering teams), but do not panic. While it is not possible to address all these issues in detail in a single book, this one intends at least to point industrial electronics professionals who are not specialists in FPGAs to the specific issues related to their working area so that they can first identify them and then tailor and optimize the learning effort to fulfill their actual needs. 

References

Jones, D.H., Powell, A., Bouganis, Ch.-S., and Cheung, P.Y.K. 2010. GPU versus

           FPGA for high productivity computing. In Proceedings of the 20th International

           Conference on Field Programmable Logic and Applications, August 31 to September 2,

           Milano, Italy.

Kaeli, D. and Akodes, D. 2011. The convergence of HPC and embedded systems in our

           heterogeneous computing future. In Proceedings of the IEEE 29th International

           Conference on Computer Design (ICCD), October 9–12, Amherst, MA.

Laprie, J.C. 1985. Dependable computing and fault tolerance: Concepts and terminol-

            ogy. In Proceedings of the 15th Annual International Symposium on Fault-Tolerant

            Computing (FTCS-15), June 19–21, Ann Arbor, MI.

Shuai, C., Jie, L., Sheaffer, J.W., Skadron, K., and Lach, J. 2008. Accelerating compute-

            intensive applications with GPUs and FPGAs. In Proceedings of the Symposium on

            Application Specific Processors (SASP 2008), June 8–9, Anaheim, CA.