Um, it's a division by a constant, which GCC optimizes to a multiply.
GCC also knows that the division must have no remainder, so it can do even
better than the usual case.)
For example,
struct x {
unsigned c[13];
};
unsigned foo (struct x *p1, struct x *p2)
{
return p2-p1;
}
On an x86 with egcc -O2 -fomit-frame-pointer produces:
foo:
movl 8(%esp),%eax
subl 4(%esp),%eax
imull $-991146299,%eax,%eax
sarl $2,%eax
ret
A multiply is not a trivial cost, but it's cheaper than a divide.
(And depending on the multiplier, GCC can optimize it further into
a series of shifts and adds.)
Any odd number x has a multiplicative inverse y such that
(x*y) == 1 (mod 2^32). Thus, (k*x)*y == k*(x*y) == k (mod 2^32).hC
and multiplying by y is the same as dividing by x.
And may number at all can be broken down into an odd number and
a power of two, where division is a simple shift.
-- -Colin- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/