1: \begin{abstract}

2: We consider some of Jonathan and Peter Borweins' contributions to the

3: high-precision computation of $\pi$ and the elementary functions, with

4: particular reference to their book

5: {\PAGM} (Wiley, 1987).

6: Here ``AGM'' is the

7: \emph{arithmetic-geometric mean} of Gauss and Legendre.

8: Because the AGM converges quadratically, it can be combined with

9: fast multiplication algorithms 

10: to give fast algorithms for the {$n$-bit} computation of

11: $\pi$, and more generally

12: the elementary functions. % $\exp$, $\log$, $\sin$, $\tan$ etc.

13: These algorithms run in ``almost linear'' time

14: $O(M(n)\log n)$, where

15: $M(n)$ % $M(n) = O(n \log n (\log\log  n)^{O(1)})$

16: is the time for $n$-bit multiplication.

17: We outline some of the results and algorithms given

18: in {\PAGM}, and present some related (but new) results.

19: In particular, we improve the published error bounds for some

20: quadratically and quartically convergent algorithms for $\pi$,

21: such as the Gauss-Legendre algorithm.

22: We show that an iteration of 

23: the Borwein-Borwein quartic algorithm for $\pi$ is

24: equivalent to two iterations of the Gauss-Legendre quadratic algorithm

25: for $\pi$, in the sense that they produce exactly 

26: the same sequence of approximations to $\pi$ if performed using

27: exact arithmetic.

28: \keywords{arithmetic-geometric mean,

29: Borwein-Borwein algorithm,

30: Bor\-wein-Borwein quartic algorithm,

31: Brent-Salamin algorithm,

32: Chudnovsky algorithm,

33: computation of $\pi$,

34: computational complexity,

35: elliptic integrals,

36: \hbox{equivalence} of algorithms for $\pi$,

37: evaluation of elementary functions,

38: Gauss-Legendre \hbox{algorithm},

39: linear convergence,

40: quadratic convergence,

41: quartic convergence,

42: Ramanujan-Sato algorithms,

43: Sasaki-Kanada algorithm,

44: theta functions.

45: }

46: \end{abstract}

47: