This paper will introduce a set of nonproprietary extensions to the American National Standards Institute (ANSI) C programming language that help facilitate the use of C as a high level language (HLL) for DSP and numeric applications. We will define the problems of using C as it exists today for DSP applications and propose solutions for these problems. This paper is an interpretation of the work being done by the Numeric C Extensions Group (NCEG), a working group of the ANSI X3Jll committee. We will use the Analog Devices' ADSP-21000 family of IEEE floating-point digital signal processor as a target to illustrate how these extensions solve the problems mentioned. This paper is intended for anyone developing or researching DSP applications or anyone involved in the development of numeric processing software. I. THE ARGUMENT FOR CThe C programming language has found wide acceptance as the primary programming language for embedded microprocessor software and firmware development. The demonstrated utility of C in embedded applications development can be applied to DSP applications development as well. The arguments for this are compelling: 1) Many engineers and most newly graduated engineers know C already, so the development of applications can start quickly, without the burden of learning the eccentricities of a new assembly language and instruction set for the particular DSP chip the engineer may be using.2) C allows for applications to be developed, prototyped, analyzed and tested in a workstation or PC environment using commercially available compilers and tools. This prototyped code can be used, sometimes unaltered, as the basis for the application software running on the DSP target.3) C is well defined. Therefore, C code should be portable to any DSP chip for which there exists C language support. The stability of C in the future is insured with the adoption of the ANSI standard for 4) The majority of the software developed for DSP applications is really control code requiring sophisticated data structures and complicated control flows [l]. This control code typically represents a small fraction of application execution time, but a significant portion of the software engineering effort. Control code is ideally suited for development in C. Also, C allows easy access to the underlying hardware. Code that needs to access hardware control and status registers and 1/0 ports can be written in C [9]. Using C is easier than assembly language for control code and will decrease the amount of time is takes to develop an application. 5) Software developed in C will typically be more reliable, more maintainable and cheaper to develop than software written in assembly language. The C programmer may also benefit from the existence of third party CASE and systems engineering tools for C based software development. THE ARGUMENT AGAINST CDespite the arguments put forth above, the track record of C as a programming language for DSP applications development has not been that good.In the early eighties, soon after single chip DSP...

No abstract

This paper describes the use of Analog Devices' ADSP-2100 digital signal processor in discrete utterance recognition applications. The practicality of implementing conventional 100, 300, 500, 1000, 2000, and 5000-word speaker-dependent recognizers is explored. The processor's functionality is demonstrated via the coordination and execution of tasks which include d a t a acquisition, silence detection, feature extraction, amplitude normalization, and dynamic programming. A two-pass algorithm is presented which limits the computational burden of large vocabularies in the dynamicprogramming phase. On the first pass decimated prototypes and candidates are compared to obtain a set of best matches. A complete dynamic programming search is then performed on this set during the second pass. An adequate hardware implementation is shown to consist of a small general-purpose ADSP-2100 processor board and separate memory board.

No abstract

No abstract