A couple of points about this program:
Karl F. Larsen wrote:
> double power(double a, int b)
> {
> double pow = 1.0;
> int i;
>
> if ( a == 0 )
> return 0;
Performing `==' on a double is often something to be avoided. In this
case, it shouldn't matter; the program will deal with the case where a
is zero. In general, it's often better to use
if (fabs(a - b) < EPSILON)
than
if (a == b)
where EPSILON is some small value.
> if ( b == 0 )
> return 1;
This shouldn't need to be a special case; `pow' is initialised to 1.0,
so if b is zero, the for loops will never be executed, and 1.0 would
be returned.
> if ( b < 0)
> {
> b = b * -1; /* makes b a positive number */
Using `b = -b' may be more efficient than performing a multiply.
> for( i=1; i <= b; i++)
> pow = pow * a;
This is a less than ideal way to compute powers. The time taken is
proportional to b, whereas you can easily do it in a time proportional
to log[2](b).
The original algorithm is equivalent to computing a * b by adding a to
zero b times. In practice, multiplication is performed by breaking b
into powers of 2 (b0 + 2.b1 + 4.b2 + ..., where each bi is zero or
one), so
a*b = a*(b0 + 2.b1 + 4.b2 + 8.b3 + ...)
= b0.a + 2.b1.a + 4.b2.a + 8.b3.a + ...
= b0.a + 2(b1.a + 2(b2.a + 2(b3.a + ...
Similarly, to compute an integral power, you break the exponent into
powers of two. Then:
a^b = a^(b0 + 2.b1 + 4.b2 + 8.b3 + ...)
= b0.a * b1.a^2 * b2.a^4 * b3.a^8 * ...
= b0.a * a^2.(b1 * a^2.(b2 * a^2.(b3 * ...
Which can be computed using:
double pow(double a, int b)
{
double pow = 1.0;
int i;
for (i = 0; i < 8*sizeof(int); i++)
{
if (b & (1U << i)) /* if the ith bit of b is set */
pow *= a;
a *= a; /* a = a^2 */
}
return pow;
}
However, all of this is general programming theory, rather than being
specific to C.
> So this is where I am. And a question arises: Is this a Sig where
> my level of C programming development should be? Sure looks like You and
> others are way ahead of me.
Well, there are the newsgroups comp.lang.c and comp.lang.c.moderated.
I don't know about mailing lists.
--
Glynn Clements <[EMAIL PROTECTED]>
