## Detecting Perfect Powers In Essentially Linear Time (1998)

### Cached

### Download Links

- [cr.yp.to]
- [www.ams.org]
- [www.ams.org]
- [cr.yp.to]
- DBLP

### Other Repositories/Bibliography

Venue: | Math. Comp |

Citations: | 41 - 12 self |

### BibTeX

@ARTICLE{Bernstein98detectingperfect,

author = {Daniel J. Bernstein},

title = {Detecting Perfect Powers In Essentially Linear Time},

journal = {Math. Comp},

year = {1998},

volume = {67},

pages = {1253--1283}

}

### Years of Citing Articles

### OpenURL

### Abstract

This paper (1) gives complete details of an algorithm to compute approximate kth roots; (2) uses this in an algorithm that, given an integer n>1, either writes n as a perfect power or proves that n is not a perfect power; (3) proves, using Loxton's theorem on multiple linear forms in logarithms, that this perfect-power decomposition algorithm runs in time (log n) . 1.

### Citations

4675 | Matrix Analysis - HORN, JOHNSON - 1990 |

2438 |
The Design and Analysis of Computer Algorithms
- Aho, Hopcroft, et al.
- 1974
(Show Context)
Citation Context ...s is almost exclusively M-time. The same is true in practice. See [15, section 4.3.3] for a discussion of fast multiplication algorithms. For the best known bounds on multiplication speed see [29] or =-=[1]-=-; note that it is possible to build a different algorithm, achieving the same bounds, out of the method of [24]. 4. Floating-point arithmetic A positive floating-point number is a positive integer div... |

1869 | Numerical Recipes in C: The Art of Scientific Computing - Press, Teukolsky, et al. - 1992 |

913 | A Course in Computational Algebraic Number Theory - Cohen - 1993 |

675 | The Art of Computer Programming, volume 2: Seminumerical Algorithms - Knuth - 1988 |

450 | Elements of algebraic topology - Munkres - 1993 |

382 | Commutative Ring Theory - Matsumura - 1986 |

332 | The Art of Computer Programming, volume 1: Fundamental Algorithms - Knuth - 1973 |

296 |
Approximate formulas for some functions of prime numbers
- Rosser, Schoenfeld
- 1962
(Show Context)
Citation Context ... for example, when n ≈ 32768—then F (n) will be noticeably larger than expected. I will get a lower bound on F (n) by changing u to 1, but I cannot get an upper bound in any analogous way. Notes. =-=See [28] for bo-=-unds on ϑ. See [3, section 2.7] for a general approach to obtaining bounds on functions such as ϑ, ϑ2, andℓ. 15. Analysis of F (n) Lemma 15.1 gives a lower bound for F (n), roughly lg lg n−lg l... |

250 | A Course in Arithmetic - Serre - 1973 |

134 | Numerical recipes in C: The art of scienti c computing - Press, Teukolsky, et al. - 1992 |

133 | Algebraic Number Theory - Fröhlich, Taylor - 1991 |

126 |
The Development of the Number Field Sieve
- Lenstra, Lenstra
- 1993
(Show Context)
Citation Context ...heorem 1 by showing that F (n) is essentially linear in log n. Part VI surveys several practical improvements. Motivation. Before attempting to factor n with algorithms such as the number field sieve =-=[16]-=-, one should make sure that n is not a perfect power, or at least not a prime power. This is a practical reason to implement some power-testing method, though not necessarily a quick one. Speed is mor... |

100 | Algorithmic algebraic number theory - Pohst, Zassenhaus - 1989 |

85 | Fast multiple-precision evaluation of elementary functions
- Brent
- 1976
(Show Context)
Citation Context ...(k) multiplications; this is useful in practice. See [15, section 4.6.3] and [11, section 1.2] for further discussion. For large k it is probably better to compute r k as exp(k log r) bythemethods of =-=[9], which take es-=-sentially linear time. PART II. ROOTS 7. Some overly specific inequalities Lemma 7.1. If κ>0and 0 <ɛ<1then (1 + ɛ/4κ) 2κ < 1+ɛ. Lemma 7.2. If κ ≥ 1 then 7/8 ≤ (1 − 1/8κ) κ . Lemma 7.3. ... |

76 | A Course of Modern Analysis, 4th edition - Whittaker, Watson - 1927 |

74 |
Algorithmic Number Theory
- Bach, Shallit
- 1996
(Show Context)
Citation Context ...omplete details of an algorithm to compute approximate kth roots; (2) uses this in an algorithm that, given an integer n>1, either writes n as a perfect power or proves that n is not a perfect power; =-=(3) p-=-roves, using Loxton’s theorem on multiple linear forms in logarithms, that this perfect-power decomposition algorithm runs in time (log n) 1+o(1) . 1. Introduction An integer n>1isaperfect power if ... |

74 |
On a problem of Oppenheim concerning Factorisatio Numerorum
- Canfield, os, et al.
- 1983
(Show Context)
Citation Context ...at most the number of products ≤ lg U of those primes, which is at most the number of products ≤ lg U of the first m prime numbers, which in turn can be estimated by analytic techniques. See [28] =-=and [10]. -=-20. Final F (n) analysis In this section, I use Lemma 19.4 to bound the function F (n) introduced in section 12. This upper bound is in (lg n) 1+o(1) . Lemma 20.1. Fix n ≥ exp exp 1000. Setu=loglog2... |

69 | The Number Field Sieve
- Lenstra, Jr, et al.
- 1990
(Show Context)
Citation Context ...t new. It is stated in, e.g., [17, section 2.4]. But God is in the details: without good methods for computing n 1/k and for checking whether x k = n, Algorithm X is not so attractive. The authors of =-=[17] go on t-=-o say that one can “save time” by adding a battery of tests to Algorithm X. Variants of Algorithm X are also dismissed in [11, page 38] (“This is clearly quite inefficient”) and [4]. Observe t... |

41 |
Schnelle Multiplikation großer Zahlen, Computing 7
- Schönhage, Strassen
- 1971
(Show Context)
Citation Context ...lgorithms is almost exclusively M-time. The same is true in practice. See [15, section 4.3.3] for a discussion of fast multiplication algorithms. For the best known bounds on multiplication speed see =-=[29]-=- or [1]; note that it is possible to build a different algorithm, achieving the same bounds, out of the method of [24]. 4. Floating-point arithmetic A positive floating-point number is a positive inte... |

18 |
A hyperelliptic smoothness test
- Lenstra, Pila, et al.
- 1993
(Show Context)
Citation Context ...essarily a quick one. Speed is more important in other applications. According to [18] there is a theoretically interesting method of finding all small factors of n (to be presented in a successor to =-=[19]-=-) for which perfect-power classification can be a bottleneck. Received by the editor October 11, 1995 and, in revised form, April 10, 1997. 1991 Mathematics Subject Classification. Primary 11Y16; Seco... |

17 |
private communication
- Lenstra
- 1999
(Show Context)
Citation Context ...ower, or at least not a prime power. This is a practical reason to implement some power-testing method, though not necessarily a quick one. Speed is more important in other applications. According to =-=[18]-=- there is a theoretically interesting method of finding all small factors of n (to be presented in a successor to [19]) for which perfect-power classification can be a bottleneck. Received by the edit... |

17 |
Fast polynomial transform algorithms for digital convolution
- Nussbaumer
- 1980
(Show Context)
Citation Context ... multiplication algorithms. For the best known bounds on multiplication speed see [29] or [1]; note that it is possible to build a different algorithm, achieving the same bounds, out of the method of =-=[24]-=-. 4. Floating-point arithmetic A positive floating-point number is a positive integer divided by a power of 2. The computer can store a pair (a, n), with a an integer and n a positive integer, to repr... |

15 |
Schnelle Multiplikation gro?er Zahlen, Computing 7
- Schonhage, Strassen
- 1971
(Show Context)
Citation Context ...lgorithms is almost exclusively M-time. The same is true in practice. See [15, section 4.3.3] for a discussion of fast multiplication algorithms. For the best known bounds on multiplication speed see =-=[29-=-] or [1]; note that it is possible to build a dierent algorithm, achieving the same bounds, out of the method of [24]. 4. Floating-point arithmetic A positivesoating-point number is a positive integer... |

13 | The complexity of multiple-precision arithmetic
- Brent
- 1976
(Show Context)
Citation Context ...rithms themselves are short and straightforward. The basic outline of my method is well known, as is its approximate run time. For Newton’s method with increasing precision see [9] (which popularize=-=d [8]) or-=- [7, section 6.4]. For the specific case of inversion see also [15, Algorithm 4.3.3–R] or [1, page 282]. For a controlled number of steps of binary search as preparation for Newton’s method see [4... |

9 |
Fast compact prime number sieves (among others
- Pritchard
- 1983
(Show Context)
Citation Context ...g enough work into the subroutines, I have made Algorithm X quite fast—so fast, in fact, that typical modifications will slow it down. I use the Sieve of Eratosthenes to enumerate the primes p<f. Se=-=e [27]-=- for faster methods. Note that the best order of operations in Algorithm X depends on the distribution of inputs; for example, if the input source is very likely to produce 37th powers, then p = 37 sh... |

8 | The theory of linear forms in logarithms - Baker - 1977 |

6 |
Sieve algorithms for perfect power testing, Algorithmica 9
- Bach, Sorenson
- 1993
(Show Context)
Citation Context ...⌈lg 8k⌉)µ(2g + ⌈lg 8k⌉) <P(k)(4g + ⌈lg 2g⌉⌈lg 8k⌉)µ(2g + ⌈lg 8k⌉)sDETECTING PERFECT POWERS IN ESSENTIALLY LINEAR TIME 1263 Notes. My use of increasing precision is at the heart o=-=f my improvement over [4]-=-. A 50-digit number that starts with 9876 is not an 11th power; the remaining 46 digits are irrelevant. In general, Algorithm C does not inspect many more bits of n than are necessary to distinguish n... |

5 |
editors), The complexity of computational problem solving
- Anderssen, Brent
- 1976
(Show Context)
Citation Context ...3–1283 S 0025-5718(98)00952-1 DETECTING PERFECT POWERS IN ESSENTIALLY LINEAR TIME DANIEL J. BERNSTEIN Abstract. This paper (1) gives complete details of an algorithm to compute approximate kth roots=-=; (2) u-=-ses this in an algorithm that, given an integer n>1, either writes n as a perfect power or proves that n is not a perfect power; (3) proves, using Loxton’s theorem on multiple linear forms in logari... |

4 | The Theory of Linear Forms - Baker - 1977 |

3 |
The number sieve
- Lenstra, Lenstra, et al.
- 1993
(Show Context)
Citation Context ...t new. It is stated in, e.g., [17, section 2.4]. But God is in the details: without good methods for computing n 1=k and for checking whether x k = n, Algorithm X is not so attractive. The authors of =-=[17] go on to -=-say that one can \save time" by adding a battery of tests to Algorithm X. Variants of Algorithm X are also dismissed in [11, page 38] (\This is clearly quite inecient") and [4]. Observe that... |

1 | Masser (editors), Transcendence theory: advances and applications - Baker, William - 1977 |

1 |
Some problems involving powers of integers, Acta Arithmetica 46
- Loxton
- 1986
(Show Context)
Citation Context ...ohn Loxton, Michael Parks, Igor Shparlinski, Jon Sorenson, and the referee for their helpful comments. Lenstra pointed out that the method of Part III had a p-adic analogue; Shparlinski suggested that=-=[20]-=-wouldhelpintheF(n) analysis; Loxton supplied the corrected proof of [20, Theorem 1] shown in section 19. PART I. ARITHMETIC 3. Integer arithmetic I represent a positive integer n inside a computer as ... |

1 | der Poorten, Multiplicative dependence in number fields, Acta Arithmetica 42 - Loxton, van - 1983 |

1 | der Poorten, Multiplicative dependence in number Acta Arithmetica 42 - Loxton, van - 1983 |