Extended precision

About: Extended precision is a research topic. Over the lifetime, 515 publications have been published within this topic receiving 9582 citations. The topic is also known as: Extended precision floating point number.

A floating-point technique for extending the available precision

Abstract: A technique is described for expressing multilength floating-point arithmetic in terms of singlelength floating point arithmetic, i.e. the arithmetic for an available (say: single or double precision) floating-point number system. The basic algorithms are exact addition and multiplication of two singlelength floating-point numbers, delivering the result as a doublelength floating-point number. A straight-forward application of the technique yields a set of algorithms for doublelength arithmetic which are given as ALGOL 60 procedures.

A Software Package for the Numerical Integration of ODEs by Means of High-Order Taylor Methods

Abstract: This paper revisits the Taylor method for the numerical integration of initial value problems of Ordinary Differential Equations (ODEs). The main goal is to present a computer program that outputs a specific numerical integrator for a given set of ODEs. The generated code includes a function to compute the jet of derivatives of the solution up to a given order plus adaptive selection of order and step size at run time. The package provides support for several extended precision arithmetics, including user-defined types. The paper discusses the performance of the resulting integrator in some examples, showing that it is very competitive in many situations. This is especially true for integrations that require extended precision arithmetic. The main drawback is that the Taylor method is an explicit method, so it has all the limitations of these kind of schemes. For instance, it is not suitable for stiff systems.

Algorithms for quad-double precision floating point arithmetic

Abstract: A quad-double number is an unevaluated sum of four IEEE double precision numbers, capable of representing at least 212 bits of significand. We present the algorithms for various arithmetic operations (including the four basic operations and various algebraic and transcendental operations) on quad-double numbers. The performance of the algorithms, implemented in C++, is also presented.

Precimonious: tuning assistant for floating-point precision

Abstract: Given the variety of numerical errors that can occur, floating-point programs are difficult to write, test and debug. One common practice employed by developers without an advanced background in numerical analysis is using the highest available precision. While more robust, this can degrade program performance significantly. In this paper we present Precimonious, a dynamic program analysis tool to assist developers in tuning the precision of floating-point programs. Precimonious performs a search on the types of the floating-point program variables trying to lower their precision subject to accuracy constraints and performance goals. Our tool recommends a type instantiation that uses lower precision while producing an accurate enough answer without causing exceptions. We evaluate Precimonious on several widely used functions from the GNU Scientific Library, two NAS Parallel Benchmarks, and three other numerical programs. For most of the programs analyzed, Precimonious reduces precision, which results in performance improvements as high as 41%.