Implementation of floating point operations on fixed point processors — An optimization algorithm and comparative analysis
Subhajit Banerjee Purnapatra, Siddharth Kumar, Subrata Bhattacharya · 2010
A method for representation of floating-point numbers with the help of integers is proposed. Simple mathematical operations involving the former are also defined in the new representation. High level codes developed for fixed-point targets may contain floating point expressions and simple modifications in the code according to the proposed conversions can produce optimized codes consisting of only integers. These codes are seen to execute approximately 10%-40% faster.