乘方(幂)算法比较

#include <stdio.h> #include <math.h> //m>=0, n>=0 int power1(int m, int n) { if(n == 0) return 1; if(n == 1) return m; if(n&1) return m*power1(m, n-1); else { int v = power1(m, n>>1); return v*v; } } int power2(int m, int n) { int ret = 1; int tmp = m; do{ if(n&1) ret *= tmp; n >>= 1; tmp *= tmp; }while (n); return ret; } int hash[10000]; int power3(int m, int n) { return pow((float)m,n); } #include <stdlib.h> #include <time.h> void test_driver() { srand(unsigned(time(NULL))); int times; while(scanf("%d", ×) != EOF) { int arg_num = 3; for(int i = 0; i < arg_num; ++i) { int beg = clock(); for(int j = 0; j < times; ++j) if(i == 0) power1(23, 100000000); else if(i == 1) power2(23, 100000000); else power3(23, 100000000); printf("Time %f/n", (clock()-beg)/1000.0); } } } int main() { test_driver(); return 0; }

 

算法一是一种标准的分治法.

算法二是利用数的比特位特性.

即,

power(m, n) = m^n

n = ai*2^i + ... + a0+2^0(ai...a0为n的二进制表示)

那么,

m^n = m^(ai*2^i + ... + a0*^0) = m^(ai^2) *...*(m^a0*2^2)

 

算法三实际上是和算法二是一致的, 但是由于用到了浮点数的乘法运算和模板, 效率会低很多.

前两种, 算法的效率PK:
10
Time 0.000000
Time 0.000000

1000
Time 0.000000
Time 0.000000

10000
Time 0.000000
Time 0.016000

100000
Time 0.032000
Time 0.203000

1000000
Time 0.219000
Time 2.141000

10000000
Time 2.250000
Time 21.422000