Fixpt: a C++ method for development of fixed point digital signal processing algorithms

This paper describes a C++ class, fixpt /spl minus/ a new data type used for development of fixed point algorithms. Fixpt is designed to reduce the development time incurred during the conversion of floating point algorithms into equivalent fixed point implementations on fixed point processors or custom designed ASIC hardware. Fixpt allows each variable in a C++ program to be declared with a user defined whole and fractional size. For example, the C++ statement "fixpt var1 (6,10);" creates a fixpt variable with 6 bits for the whole part and 10 bits for the fractional part. During program execution, the fixpt class handles detection and management of overflow and underflow conditions, thus allowing the programmer to quickly identify which variables are incorrectly sized. The fixpt class also provides extensive debugging aids and user definable overflow function. Basic functionality of the class, along with comparisons between floating point and fixpt algorithms, such as audio decompression, are presented.<>