Software Quickly Builds Fixed-Point DSP Algorithms

The ART Library product extends the C and C++ languages to provide classes that encapsulate the characteristics of fixed-point data types and operators. With it, designers quickly and easily can recast floating-point C algorithms in a bit-accurate, fixed-point representation. DSP uses a great deal of complex math, such as MAC operations. The floating-point results of these operations tend to be very large numbers with a virtually infinite number of precision bits. To effect a silicon-efficient implementation of these floating-point algorithms, they must be converted to a fixed-point representation with a fixed word width and fixed number of precision bits. Because there is no support for fixed-point algorithms in the C language, designers produce ad-hoc code by trial and error until they achieve the desired result.