Creating new algorithms and modifying old algorithms to use the variable precision floating point simulator

The paper describes a variable precision floating point simulator, its features and how it is used. The simulator was created at the University of Texas by Darioush Samani, Joshua Ellinger, Edward Powers, and Earl Swartzlander, Jr. (1993). It provides a mechanism for the design and testing of algorithms which use floating point precisions other than the standard single and double precision. The simulator allows the various precisions for floating point operations to be modeled in a virtually transparent mechanism to the user. The existing programming skills of the average user are sufficient to successfully use the simulator. To use the simulator the user simply programs his algorithm using standard mathematical operations and then compiles the simulator code into the program. Techniques and examples that show how both new and existing programs can use the simulator are presented. The various features of the simulator are also described.<>