Function: zetamultinit
Section: transcendental
C-Name: zetamultinit
Prototype: Lp
Help: zetamultinit(maxw): initialize data to compute multiple zeta
 values at integral s = [s1,...,sk] for s1 + ... + sk <= maxw.
Doc: initialize data (depending on the precision) used to compute
 multiple zeta values at
 integral points $s = [s_1,\dots,s_k]$ for any $s_1 + \dots + s_k \leq
 \var{maxw}$. The corresponding data is inexpensive to compute or store
 and provides a small speedup (usually about 10\%) when multiple values are to
 be computed at a given accuracy.
 \bprog
 ? for(i = 1, 2^12-1, zetamult(i))
 time = 1,413 ms
 ? T = zetamultinit(13); \\ instantaneous
 ? for(i = 1, 2^12-1, zetamult(i, T)) \\ used cached data
 time = 1,315 ms
 ? zetamultall(12); \\ much faster !
 time = 27 ms

 ? T=zetamultinit(102); sizebyte(T) \\ small even for huge weights
 time = 5 ms.
 %5 = 1440504
 ? for(i = 1, 2^5, zetamult(2^100+i))
 time = 633 ms.
 ? for(i = 1, 2^5, zetamult(2^100+i, T))
 time = 550 ms.
 @eprog\noindent
 For small weights, \kbd{zetamultall} will be much more efficient; but it
 is not an option when the weight gets large.
