10 m odd positive integer
12 return integer v, 0 <= v < m.
15 Let B be the base calc uses for representing integers internally
16 (B = 2^16 for 32-bit machines, 2^32 for 64-bit machines) and N the
17 number of words (base-B digits) in the representation of m. Then
18 rcpow(x,k,m) returns the value of B^-N * (B^N * x)^k % m, w here
19 the inverse implicit in B^-N is modulo m and the modulus operator %
20 gives the least nonnegative residue. Note that rcpow(x,0,m) =
21 rcin(1,m), rcpow(x,1,m) = x % m; rcpow(x,2,m) = rcsq(x,m).
23 The normal use of rcpow() may be said to be that of finding the
24 encoded value of the k-th power of an integer modulo m:
26 rcin(x^k, m) = rcpow(rcin(x,m), k, m),
30 x^k % m = rcout(rcpow(rcin(x,m), k, m), m).
32 If x^k % m is to be evaluated for the same k and m and several
33 values of x, it may be worth while to first evaluate:
35 a = minv(rcpow(1, k, m), m);
39 x^k % m = a * rcpow(x, k, m) % m.
42 If the value of m in rcpow(x,k,m) is being used for the first time
43 in a REDC function, the information required for the REDC
44 algorithms is calculated and stored for future use, possibly
45 replacing an already stored valued, in a table covering up to 5
46 (i.e. MAXREDC) values of m. The runtime required for this is about
47 two times that required for multiplying two N-word integers.
49 Two algorithms are available for evaluating rcpow(x,k,m), the one
50 which is usually faster for small N is used when N <
51 config("redc2"); the other is usually faster for larger N. If
52 config("redc2") is set at about 90 and 0 <= x < m, the runtime
53 required for rcpow(x,k,m) is at most about f times the runtime
54 required for ilog2(k) N-word by N-word multiplications, where f
55 increases from about 1.3 for N = 1 to near 4 for N > 90. More
56 runtime may be required if x has to be reduced modulo m.
59 Using a 64-bit machine with B = 2^32:
63 > print rcout(rcpow((rcin(x,m), m - 1, m), m), pmod(x, m-1, m)
70 void zredcpower(REDC *rp, ZVALUE z1, ZVALUE z2, ZVALUE *res)
73 rcin, rcout, rcmul, rcsq