Floating-point Number LISP

There exist many applications where both symbolic and floating-point computations are required at the same time. Image recognition, for example, requires huge amounts of numeric calculation to extract features from raw image data before the actual recognition is performed according to those features. Systems for symbolic computation should perform type-checking on numeric calculations. Although the overhead can be eliminated by declarations to the compiler, our observation shows that it is not always easy to make the appropriate declarations. The overhead can be eliminated by special hardware, but use of special machines sometimes makes system development difficult. This paper describes a LISP system, suitable for both symbolic and numeric computation, which has been developed under VAX/VMS. It runs simple differential-equation problems nearly as fast as VAX Fortran with no declarations. This system, called Real Number Lisp, benefits people who need both the flexibility of a symbolic computation system and the capability of a numeric computation system at the same time. The same techniques can be used with other symbolic computing languages, where dynamic type-checking is required.