Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit c5f5877c authored by Stephen Hemminger's avatar Stephen Hemminger Committed by David S. Miller
Browse files

[TCP] tcp_cubic: faster cube root 



The Newton-Raphson method is quadratically convergent so
only a small fixed number of steps are necessary.
Therefore it is faster to unroll the loop. Since div64_64 is no longer
inline it won't cause code explosion.

Also fixes a bug that can occur if x^2 was bigger than 32 bits.

Signed-off-by: default avatarStephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent 8570419f

net/ipv4/tcp_cubic.c

+5 −11
Original line number Original line Diff line number Diff line
@@ -96,23 +96,17 @@ static void bictcp_init(struct sock *sk) 
 */

 */

static u32 cubic_root(u64 a)

static u32 cubic_root(u64 a)

{

{

	u32 x, x1;

	u32 x;





	/* Initial estimate is based on:

	/* Initial estimate is based on:

	 * cbrt(x) = exp(log(x) / 3)

	 * cbrt(x) = exp(log(x) / 3)

	 */

	 */

	x = 1u << (fls64(a)/3);

	x = 1u << (fls64(a)/3);





	/*

	/* converges to 32 bits in 3 iterations */

	 * Iteration based on:

	x = (2 * x + (u32)div64_64(a, (u64)x*(u64)x)) / 3;

	 *                         2

	x = (2 * x + (u32)div64_64(a, (u64)x*(u64)x)) / 3;

	 * x    = ( 2 * x  +  a / x  ) / 3

	x = (2 * x + (u32)div64_64(a, (u64)x*(u64)x)) / 3;

	 *  k+1          k         k

	 */

	do {

		x1 = x;

		x = (2 * x + (uint32_t) div64_64(a, x*x)) / 3;

	} while (abs(x1 - x) > 1);





	return x;

	return x;

}

}