mikolalysenko / two-sum

Computes the sum of two floats

two-sum

Computes the sum of two floating point numbers as a non-overlapping sequence using Knuth's method.

• D.E. Knuth. "The Art of Computer Programming: Seminumerical Algorithms". Volume 2. 1981

Install

Using npm:

	npm install two-sum

Example

var twoSum = require("two-sum")

//Add two wildly different sized floats
var result = twoSum(1e64, 1e-64)
console.log(result)

//Prints:
//  [1e-64, 1e64]

API

require("two-sum")(a, b[, result])

Computes a non-overlapping sequence representing the sum of a and b.

• a is a number
• b is a number
• result is an optional length 2 array encoding the result of the sum of a and b

Returns A length 2 array representing the non-overlapping sequence encoding the sum of a and b. The first term has smaller magnitude than the second.

Credits

Based on an idea from JRS robust geometric predicates paper.

Implementation (c) 2013 Mikola Lysenko. MIT License