Definition of "ulp"

I sent her a text asking as nonchalantly as possible whether cows ever go for you. [...] The answer I got was, 'Occasionally they do, and they can be bloody scary.' Ulp.

2009, Charlie Connelly, And Did Those Feet: Walking through 2000 Years of British and Irish History‎, London: Little, Brown and Company

And where would the Grammys be without Motown maverick Stevie Wonder, continuing his slide into musical irrelevance with a mind-boggling collaboration with, ulp, teen idols the Jonas Brothers, who massacred their own "Burnin' Up" and Stevie's "Superstition" with boyish aplomb.

2009 February 10, Joel Rubinoff, “Train wreck keeps rollin’ along”, in Toronto Star‎, Toronto, Ont.: Toronto Star Newspapers, archived from the original on 12 March 2016

Little—ulp—little red horns curving out of his brow. My mouth was totally dry.

2012, Jennifer Stevenson, It’s Raining Angels and Demons (Slacker Demons; book 2), Lancaster, Oh.: Calliope

The basic concept of two vectors "agreeing to k ulps" (units in the last place) of each entry or of the largest entry allows us to express relationships among computed values and solutions at perturbed data.

1980, Webb Miller, Celia Wrathall, “Software for Roundoff Analysis”, in Software for Roundoff Analysis of Matrix Algorithms, New York, N.Y., London: Academic Press, section 4.2 (Rounding and Perturbations of the Computational Problem), page 86

Floating point operations are correct to within half an ulp, and the calculation of uvw by two floating point multiplications will be correct within about one ulp (ignoring second-order terms).

1981, Donald E[rvin] Knuth, “Arithmetic”, in The Art of Computer Programming (Addison-Wesley Series in Computer Science and Information Processing), 2nd edition, volume 2 (Seminumerical Algorithms), Reading, Mass.: Addison-Wesley, page 217; 3rd edition, Reading, Mass.: Addison-Wesley, 1997, section 4.2.2 (Accuracy of Floating Point Arithmetic)

The difference between two consecutive machine numbers is called an ulp (unit in the last place, i.e., one digit in the least significant place). The size of an ulp varies depending on where you are in the set of machine numbers. [...] Ideally no function should ever return a result with error exceeding half of an ulp since the distance from the true result to the nearest machine number is always less than half of an ulp, the worst case being when it is exactly halfway between two machine numbers.

1996, “Numerical Mathematics”, in Victor Adamchik et al., edited by Emily Martin, Mathematica 3.0 Standard Add-on Packages, Champaign, Ill.: Wolfram Media; Cambridge, Cambridgeshire: Cambridge University Press, page 369

If, instead of considering ulps of the "exact" value x {\displaystyle x} , we consider ulps of the floating-point value X {\displaystyle X} , we have a property that is very similar to Property 2, with the interesting difference that it now holds for any value of the radix.

2010, Jean-Michel Muller et al., “Definitions and Basic Notions”, in Handbook of Floating-Point Arithmetic, Boston, Mass., Basel: Birkhäuser, section 2.6.1 (The Ulp Function), page 34