Computational complexity of mathematical operations

This article needs additional citations for verification. Please help improve this article by adding citations to reliable sources. Unsourced material may be challenged and removed. Find sources: "Computational complexity of mathematical operations" – news · newspapers · books · scholar · JSTOR (April 2015) (Learn how and when to remove this message)

The following tables list the

.

Here, complexity refers to the time complexity of performing computations on a multitape Turing machine.^[1] See big O notation for an explanation of the notation used.

Note: Due to the variety of multiplication algorithms, ${\displaystyle M(n)}$ below stands in for the complexity of the chosen multiplication algorithm.

Arithmetic functions

This table lists the complexity of mathematical operations on integers.

Operation	Input	Output	Algorithm	Complexity
Addition	Two ${\displaystyle n}$-digit numbers	One ${\displaystyle n+1}$-digit number	Schoolbook addition with carry	${\displaystyle \Theta (n)}$
Subtraction	Two ${\displaystyle n}$-digit numbers	One ${\displaystyle n+1}$-digit number	Schoolbook subtraction with borrow	${\displaystyle \Theta (n)}$
Multiplication	Two ${\displaystyle n}$-digit numbers	One ${\displaystyle 2n}$-digit number	Schoolbook long multiplication	${\displaystyle O{\mathord {\left(n^{2}\right)}}}$
			Karatsuba algorithm	${\displaystyle O{\mathord {\left(n^{1.585}\right)}}}$
			3-way Toom–Cook multiplication	${\displaystyle O{\mathord {\left(n^{1.465}\right)}}}$
			${\displaystyle k}$-way Toom–Cook multiplication	${\displaystyle O{\mathord {\left(n^{\frac {\log(2k-1)}{\log k}}\right)}}}$
			Mixed-level Toom–Cook (Knuth 4.3.3-T)[2]	${\displaystyle O{\mathord {\left(n\,2^{\sqrt {2\log n}}\,\log n\right)}}}$
			Schönhage–Strassen algorithm	${\displaystyle O{\mathord {\left(n\log n\log \log n\right)}}}$
			Harvey-Hoeven algorithm[3][4]	${\displaystyle O(n\log n)}$
Division	Two ${\displaystyle n}$-digit numbers	One ${\displaystyle n}$-digit number	Schoolbook long division	${\displaystyle O{\mathord {\left(n^{2}\right)}}}$
			Burnikel–Ziegler Divide-and-Conquer Division[5]	${\displaystyle O(M(n)\log n)}$
			Newton–Raphson division	${\displaystyle O(M(n))}$
Square root	One ${\displaystyle n}$-digit number	One ${\displaystyle n}$-digit number	Newton's method	${\displaystyle O(M(n))}$
Modular exponentiation	Two ${\displaystyle n}$-digit integers and a ${\displaystyle k}$-bit exponent	One ${\displaystyle n}$-digit integer	Repeated multiplication and reduction	${\displaystyle O{\mathord {\left(M(n)\,2^{k}\right)}}}$
			Exponentiation by squaring	${\displaystyle O(M(n)\,k)}$
			Exponentiation with Montgomery reduction	${\displaystyle O(M(n)\,k)}$

On stronger computational models, specifically a pointer machine and consequently also a unit-cost random-access machine it is possible to multiply two n-bit numbers in time O(n).^[6]

Algebraic functions

Here we consider operations over polynomials and n denotes their degree; for the coefficients we use a unit-cost model, ignoring the number of bits in a number. In practice this means that we assume them to be machine integers.

Operation	Input	Output	Algorithm	Complexity
Polynomial evaluation	One polynomial of degree ${\displaystyle n}$ with integer coefficients	One number	Direct evaluation	${\displaystyle \Theta (n)}$
			Horner's method	${\displaystyle \Theta (n)}$
Polynomial gcd (over ${\displaystyle \mathbb {Z} [x]}$ or ${\displaystyle F[x]}$)	Two polynomials of degree ${\displaystyle n}$ with integer coefficients	One polynomial of degree at most ${\displaystyle n}$	Euclidean algorithm	${\displaystyle O{\mathord {\left(n^{2}\right)}}}$
			Fast Euclidean algorithm (Lehmer)[citation needed]	${\displaystyle O(M(n)\log n)}$

Special functions

Many of the methods in this section are given in Borwein & Borwein.^[7]

Elementary functions

The elementary functions are constructed by composing arithmetic operations, the exponential function (${\displaystyle \exp }$), the natural logarithm (${\displaystyle \log }$),

(${\displaystyle \sin ,\cos }$), and their inverses. The complexity of an elementary function is equivalent to that of its inverse, since all elementary functions are analytic and hence invertible by means of Newton's method. In particular, if either ${\displaystyle \exp }$ or ${\displaystyle \log }$ in the complex domain can be computed with some complexity, then that complexity is attainable for all other elementary functions.

Below, the size ${\displaystyle n}$ refers to the number of digits of precision at which the function is to be evaluated.

Algorithm	Applicability	Complexity
Taylor series; repeated argument reduction (e.g. ${\displaystyle \exp(2x)=[\exp(x)]^{2}}$) and direct summation	${\displaystyle \exp ,\log ,\sin ,\cos ,\arctan }$	${\displaystyle O{\mathord {\left(M(n)n^{1/2}\right)}}}$
Taylor series; FFT-based acceleration	${\displaystyle \exp ,\log ,\sin ,\cos ,\arctan }$	${\displaystyle O{\mathord {\left(M(n)n^{1/3}(\log n)^{2}\right)}}}$
Taylor series; binary splitting + bit-burst algorithm[8]	${\displaystyle \exp ,\log ,\sin ,\cos ,\arctan }$	${\displaystyle O{\mathord {\left(M(n)(\log n)^{2}\right)}}}$
Arithmetic–geometric mean iteration[9]	${\displaystyle \exp ,\log ,\sin ,\cos ,\arctan }$	${\displaystyle O(M(n)\log n)}$

It is not known whether ${\displaystyle O(M(n)\log n)}$ is the optimal complexity for elementary functions. The best known lower bound is the trivial bound ${\displaystyle \Omega }$${\displaystyle (M(n))}$.

Non-elementary functions

Function	Input	Algorithm	Complexity
Gamma function	${\displaystyle n}$-digit number	Series approximation of the incomplete gamma function	${\displaystyle O{\mathord {\left(M(n)n^{1/2}(\log n)^{2}\right)}}}$
	Fixed rational number	Hypergeometric series	${\displaystyle O{\mathord {\left(M(n)(\log n)^{2}\right)}}}$
	${\displaystyle m/24}$, for ${\displaystyle m}$ integer.	Arithmetic-geometric mean iteration	${\displaystyle O(M(n)\log n)}$
Hypergeometric function ${\displaystyle {}_{p}\!F_{q}}$	${\displaystyle n}$-digit number	(As described in Borwein & Borwein)	${\displaystyle O{\mathord {\left(M(n)n^{1/2}(\log n)^{2}\right)}}}$
	Fixed rational number	Hypergeometric series	${\displaystyle O{\mathord {\left(M(n)(\log n)^{2}\right)}}}$

Mathematical constants

This table gives the complexity of computing approximations to the given constants to ${\displaystyle n}$ correct digits.

Constant	Algorithm	Complexity
Golden ratio, ${\displaystyle \phi }$	Newton's method	${\displaystyle O(M(n))}$
Square root of 2, ${\displaystyle {\sqrt {2}}}$	Newton's method	${\displaystyle O(M(n))}$
Euler's number, ${\displaystyle e}$	Binary splitting of the Taylor series for the exponential function	${\displaystyle O(M(n)\log n)}$
	Newton inversion of the natural logarithm	${\displaystyle O(M(n)\log n)}$
Pi, ${\displaystyle \pi }$	Binary splitting of the arctan series in Machin's formula	${\displaystyle O{\mathord {\left(M(n)(\log n)^{2}\right)}}}$[10]
	Gauss–Legendre algorithm	${\displaystyle O(M(n)\log n)}$[10]
Euler's constant , ${\displaystyle \gamma }$	Sweeney's method (approximation in terms of the exponential integral)	${\displaystyle O{\mathord {\left(M(n)(\log n)^{2}\right)}}}$

Number theory

Algorithms for number theoretical calculations are studied in computational number theory.

Operation	Input	Output	Algorithm	Complexity
Greatest common divisor	Two ${\displaystyle n}$-digit integers	One integer with at most ${\displaystyle n}$ digits	Euclidean algorithm	${\displaystyle O{\mathord {\left(n^{2}\right)}}}$
			Binary GCD algorithm	${\displaystyle O{\mathord {\left(n^{2}\right)}}}$
			Left/right k-ary binary GCD algorithm[11]	${\displaystyle O{\mathord {\left({\frac {n^{2}}{\log n}}\right)}}}$
			Stehlé–Zimmermann algorithm[12]	${\displaystyle O(M(n)\log n)}$
			Schönhage controlled Euclidean descent algorithm[13]	${\displaystyle O(M(n)\log n)}$
Jacobi symbol	Two ${\displaystyle n}$-digit integers	${\displaystyle 0}$, ${\displaystyle -1}$ or ${\displaystyle 1}$	Schönhage controlled Euclidean descent algorithm[14]	${\displaystyle O(M(n)\log n)}$
			Stehlé–Zimmermann algorithm[15]	${\displaystyle O(M(n)\log n)}$
Factorial	A positive integer less than ${\displaystyle m}$	One ${\displaystyle O(m\log m)}$-digit integer	Bottom-up multiplication	${\displaystyle O{\mathord {\left(M\left(m^{2}\right)\log m\right)}}}$
			Binary splitting	${\displaystyle O(M(m\log m)\log m)}$
			Exponentiation of the prime factors of ${\displaystyle m}$	${\displaystyle O(M(m\log m)\log \log m)}$,[16] ${\displaystyle O(M(m\log m))}$[1]
Primality test	A ${\displaystyle n}$-digit integer	True or false	AKS primality test	${\displaystyle O{\mathord {\left(n^{6+o(1)}\right)}}}$[17][18] ${\displaystyle O(n^{3})}$, assuming Agrawal's conjecture
			Elliptic curve primality proving	${\displaystyle O{\mathord {\left(n^{4+\varepsilon }\right)}}}$ heuristically[19]
			Baillie–PSW primality test	${\displaystyle O{\mathord {\left(n^{2+\varepsilon }\right)}}}$[20][21]
			Miller–Rabin primality test	${\displaystyle O{\mathord {\left(kn^{2+\varepsilon }\right)}}}$[22]
			Solovay–Strassen primality test	${\displaystyle O{\mathord {\left(kn^{2+\varepsilon }\right)}}}$[22]
Integer factorization	A ${\displaystyle b}$-bit input integer	A set of factors	General number field sieve	${\displaystyle O{\mathord {\left((1+\varepsilon )^{b}\right)}}}$[nb 1]
			Shor's algorithm	${\displaystyle O(M(b)b)}$, on a quantum computer

Matrix algebra

The following complexity figures assume that arithmetic with individual elements has complexity O(1), as is the case with fixed-precision floating-point arithmetic or operations on a finite field.

Operation	Input	Output	Algorithm	Complexity
Matrix multiplication	Two ${\displaystyle n\times n}$ matrices	One ${\displaystyle n\times n}$ matrix	Schoolbook matrix multiplication	${\displaystyle O(n^{3})}$
			Strassen algorithm	${\displaystyle O{\mathord {\left(n^{2.807}\right)}}}$
			Coppersmith–Winograd algorithm (galactic algorithm )	${\displaystyle O{\mathord {\left(n^{2.376}\right)}}}$
			Optimized CW-like algorithms[23][24][25][26] (galactic algorithms)	${\displaystyle O{\mathord {\left(n^{2.3728596}\right)}}}$
Matrix multiplication	One ${\displaystyle n\times m}$ matrix, and one ${\displaystyle m\times p}$ matrix	One ${\displaystyle n\times p}$ matrix	Schoolbook matrix multiplication	${\displaystyle O(nmp)}$
Matrix multiplication	One ${\displaystyle n\times \left\lceil n^{k}\right\rceil }$ matrix, and one ${\displaystyle \left\lceil n^{k}\right\rceil \times n}$ matrix, for some ${\displaystyle k\geq 0}$	One ${\displaystyle n\times n}$ matrix	Algorithms given in [27]	${\displaystyle O(n^{\omega (k)+\epsilon })}$, where upper bounds on ${\displaystyle \omega (k)}$ are given in [27]
Matrix inversion	One ${\displaystyle n\times n}$ matrix	One ${\displaystyle n\times n}$ matrix	Gauss–Jordan elimination	${\displaystyle O{\mathord {\left(n^{3}\right)}}}$
			Strassen algorithm	${\displaystyle O{\mathord {\left(n^{2.807}\right)}}}$
			Coppersmith–Winograd algorithm	${\displaystyle O{\mathord {\left(n^{2.376}\right)}}}$
			Optimized CW-like algorithms	${\displaystyle O{\mathord {\left(n^{2.373}\right)}}}$
Singular value decomposition	One ${\displaystyle m\times n}$ matrix	One ${\displaystyle m\times m}$ matrix, one ${\displaystyle m\times n}$ matrix, & one ${\displaystyle n\times n}$ matrix	Bidiagonalization and QR algorithm	${\displaystyle O{\mathord {\left(m^{2}n\right)}}}$ (${\displaystyle m\geq n}$)
		One ${\displaystyle m\times n}$ matrix, one ${\displaystyle n\times n}$ matrix, & one ${\displaystyle n\times n}$ matrix	Bidiagonalization and QR algorithm	${\displaystyle O{\mathord {\left(mn^{2}\right)}}}$ (${\displaystyle m\leq n}$)
QR decomposition	One ${\displaystyle m\times n}$ matrix	One ${\displaystyle m\times n}$ matrix, & one ${\displaystyle n\times n}$ matrix	Algorithms in [28]	${\displaystyle O{\mathord {\left(mn^{1+{\frac {1}{4-\omega }}}\right)}}}$ (${\displaystyle m\geq n}$)
Determinant	One ${\displaystyle n\times n}$ matrix	One number	Laplace expansion	${\displaystyle O(n!)}$
			Division-free algorithm[29]	${\displaystyle O{\mathord {\left(n^{4}\right)}}}$
			LU decomposition	${\displaystyle O(n^{3})}$
			Bareiss algorithm	${\displaystyle O{\mathord {\left(n^{3}\right)}}}$
			Fast matrix multiplication[30]	${\displaystyle O{\mathord {\left(n^{2.373}\right)}}}$
Back substitution	Triangular matrix	${\displaystyle n}$ solutions	Back substitution[31]	${\displaystyle O{\mathord {\left(n^{2}\right)}}}$

In 2005, Henry Cohn, Robert Kleinberg, Balázs Szegedy, and Chris Umans showed that either of two different conjectures would imply that the exponent of matrix multiplication is 2.^[32]

Transforms

Algorithms for computing transforms of functions (particularly integral transforms) are widely used in all areas of mathematics, particularly analysis and signal processing.

Operation	Input	Output	Algorithm	Complexity
Discrete Fourier transform	Finite data sequence of size ${\displaystyle n}$	Set of complex numbers	Schoolbook	${\displaystyle O(n^{2})}$
			Fast Fourier transform	${\displaystyle O(n\log n)}$

Notes

References

Further reading

• ISBN 978-0-521-19469-3
  .
• ISBN 978-0-201-89684-8
  .