From df3271f3361b61ce02da0026b4a53e63bc2720cb Mon Sep 17 00:00:00 2001 From: Stephen Hemminger Date: Tue, 13 Dec 2005 23:13:28 -0800 Subject: [TCP] BIC: CUBIC window growth (2.0) Replace existing BIC version 1.1 with new version 2.0. The main change is to replace the window growth function with a cubic function as described in: http://www.csc.ncsu.edu/faculty/rhee/export/bitcp/cubic-paper.pdf Signed-off-by: Stephen Hemminger Signed-off-by: David S. Miller --- net/ipv4/tcp_cubic.c | 445 +++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 445 insertions(+) create mode 100644 net/ipv4/tcp_cubic.c (limited to 'net/ipv4/tcp_cubic.c') diff --git a/net/ipv4/tcp_cubic.c b/net/ipv4/tcp_cubic.c new file mode 100644 index 0000000..bb5dc4b --- /dev/null +++ b/net/ipv4/tcp_cubic.c @@ -0,0 +1,445 @@ +/* + * TCP CUBIC: Binary Increase Congestion control for TCP v2.0 + * + * This is from the implementation of CUBIC TCP in + * Injong Rhee, Lisong Xu. + * "CUBIC: A New TCP-Friendly High-Speed TCP Variant + * in PFLDnet 2005 + * Available from: + * http://www.csc.ncsu.edu/faculty/rhee/export/bitcp/cubic-paper.pdf + * + * Unless CUBIC is enabled and congestion window is large + * this behaves the same as the original Reno. + */ + +#include +#include +#include +#include + + +#define BICTCP_BETA_SCALE 1024 /* Scale factor beta calculation + * max_cwnd = snd_cwnd * beta + */ +#define BICTCP_B 4 /* + * In binary search, + * go to point (max+min)/N + */ +#define BICTCP_HZ 10 /* BIC HZ 2^10 = 1024 */ + +static int fast_convergence = 1; +static int max_increment = 16; +static int beta = 819; /* = 819/1024 (BICTCP_BETA_SCALE) */ +static int initial_ssthresh = 100; +static int bic_scale = 41; +static int tcp_friendliness = 1; + +module_param(fast_convergence, int, 0644); +MODULE_PARM_DESC(fast_convergence, "turn on/off fast convergence"); +module_param(max_increment, int, 0644); +MODULE_PARM_DESC(max_increment, "Limit on increment allowed during binary search"); +module_param(beta, int, 0644); +MODULE_PARM_DESC(beta, "beta for multiplicative increase"); +module_param(initial_ssthresh, int, 0644); +MODULE_PARM_DESC(initial_ssthresh, "initial value of slow start threshold"); +module_param(bic_scale, int, 0644); +MODULE_PARM_DESC(bic_scale, "scale (scaled by 1024) value for bic function (bic_scale/1024)"); +module_param(tcp_friendliness, int, 0644); +MODULE_PARM_DESC(tcp_friendliness, "turn on/off tcp friendliness"); + + +/* BIC TCP Parameters */ +struct bictcp { + u32 cnt; /* increase cwnd by 1 after ACKs */ + u32 last_max_cwnd; /* last maximum snd_cwnd */ + u32 loss_cwnd; /* congestion window at last loss */ + u32 last_cwnd; /* the last snd_cwnd */ + u32 last_time; /* time when updated last_cwnd */ + u32 bic_origin_point;/* origin point of bic function */ + u32 bic_K; /* time to origin point from the beginning of the current epoch */ + u32 delay_min; /* min delay */ + u32 epoch_start; /* beginning of an epoch */ + u32 ack_cnt; /* number of acks */ + u32 tcp_cwnd; /* estimated tcp cwnd */ +#define ACK_RATIO_SHIFT 4 + u32 delayed_ack; /* estimate the ratio of Packets/ACKs << 4 */ +}; + +static inline void bictcp_reset(struct bictcp *ca) +{ + ca->cnt = 0; + ca->last_max_cwnd = 0; + ca->loss_cwnd = 0; + ca->last_cwnd = 0; + ca->last_time = 0; + ca->bic_origin_point = 0; + ca->bic_K = 0; + ca->delay_min = 0; + ca->epoch_start = 0; + ca->delayed_ack = 2 << ACK_RATIO_SHIFT; + ca->ack_cnt = 0; + ca->tcp_cwnd = 0; +} + +static void bictcp_init(struct sock *sk) +{ + bictcp_reset(inet_csk_ca(sk)); + if (initial_ssthresh) + tcp_sk(sk)->snd_ssthresh = initial_ssthresh; +} + +/* 65536 times the cubic root */ +static const u64 cubic_table[8] + = {0, 65536, 82570, 94519, 104030, 112063, 119087, 125367}; + +/* + * calculate the cubic root of x + * the basic idea is that x can be expressed as i*8^j + * so cubic_root(x) = cubic_root(i)*2^j + * in the following code, x is i, and y is 2^j + * because of integer calculation, there are errors in calculation + * so finally use binary search to find out the exact solution + */ +static u32 cubic_root(u64 x) +{ + u64 y, app, target, start, end, mid, start_diff, end_diff; + + if (x == 0) + return 0; + + target = x; + + /* first estimate lower and upper bound */ + y = 1; + while (x >= 8){ + x = (x >> 3); + y = (y << 1); + } + start = (y*cubic_table[x])>>16; + if (x==7) + end = (y<<1); + else + end = (y*cubic_table[x+1]+65535)>>16; + + /* binary search for more accurate one */ + while (start < end-1) { + mid = (start+end) >> 1; + app = mid*mid*mid; + if (app < target) + start = mid; + else if (app > target) + end = mid; + else + return mid; + } + + /* find the most accurate one from start and end */ + app = start*start*start; + if (app < target) + start_diff = target - app; + else + start_diff = app - target; + app = end*end*end; + if (app < target) + end_diff = target - app; + else + end_diff = app - target; + + if (start_diff < end_diff) + return (u32)start; + else + return (u32)end; +} + +static inline u32 bictcp_K(u32 dist, u32 srtt) +{ + u64 d64; + u32 d32; + u32 count; + u32 result; + + /* calculate the "K" for (wmax-cwnd) = c/rtt * K^3 + so K = cubic_root( (wmax-cwnd)*rtt/c ) + the unit of K is bictcp_HZ=2^10, not HZ + + c = bic_scale >> 10 + rtt = (tp->srtt >> 3 ) / HZ + + the following code has been designed and tested for + cwnd < 1 million packets + RTT < 100 seconds + HZ < 1,000,00 (corresponding to 10 nano-second) + + */ + + /* 1/c * 2^2*bictcp_HZ */ + d32 = (1 << (10+2*BICTCP_HZ)) / bic_scale; + d64 = (__u64)d32; + + /* srtt * 2^count / HZ + 1) to get a better accuracy of the following d32, + the larger the "count", the better the accuracy + 2) and avoid overflow of the following d64 + the larger the "count", the high possibility of overflow + 3) so find a "count" between bictcp_hz-3 and bictcp_hz + "count" may be less than bictcp_HZ, + then d64 becomes 0. that is OK + */ + d32 = srtt; + count = 0; + while (((d32 & 0x80000000)==0) && (count < BICTCP_HZ)){ + d32 = d32 << 1; + count++; + } + d32 = d32 / HZ; + + /* (wmax-cwnd) * (srtt>>3 / HZ) / c * 2^(3*bictcp_HZ) */ + d64 = (d64 * dist * d32) >> (count+3-BICTCP_HZ); + + /* cubic root */ + d64 = cubic_root(d64); + + result = (u32)d64; + return result; +} + +/* + * Compute congestion window to use. + */ +static inline void bictcp_update(struct bictcp *ca, u32 cwnd) +{ + u64 d64; + u32 d32, t, srtt, bic_target, min_cnt, max_cnt; + + ca->ack_cnt++; /* count the number of ACKs */ + + if (ca->last_cwnd == cwnd && + (s32)(tcp_time_stamp - ca->last_time) <= HZ / 32) + return; + + ca->last_cwnd = cwnd; + ca->last_time = tcp_time_stamp; + + srtt = (HZ << 3)/10; /* use real time-based growth function */ + + if (ca->epoch_start == 0) { + ca->epoch_start = tcp_time_stamp; /* record the beginning of an epoch */ + ca->ack_cnt = 1; /* start counting */ + ca->tcp_cwnd = cwnd; /* syn with cubic */ + + if (ca->last_max_cwnd <= cwnd) { + ca->bic_K = 0; + ca->bic_origin_point = cwnd; + } else { + ca->bic_K = bictcp_K(ca->last_max_cwnd-cwnd, srtt); + ca->bic_origin_point = ca->last_max_cwnd; + } + } + + /* cubic function - calc*/ + /* calculate c * time^3 / rtt, + * while considering overflow in calculation of time^3 + * (so time^3 is done by using d64) + * and without the support of division of 64bit numbers + * (so all divisions are done by using d32) + * also NOTE the unit of those veriables + * time = (t - K) / 2^bictcp_HZ + * c = bic_scale >> 10 + * rtt = (srtt >> 3) / HZ + * !!! The following code does not have overflow problems, + * if the cwnd < 1 million packets !!! + */ + + /* change the unit from HZ to bictcp_HZ */ + t = ((tcp_time_stamp + ca->delay_min - ca->epoch_start) + << BICTCP_HZ) / HZ; + + if (t < ca->bic_K) /* t - K */ + d32 = ca->bic_K - t; + else + d32 = t - ca->bic_K; + + d64 = (u64)d32; + d32 = (bic_scale << 3) * HZ / srtt; /* 1024*c/rtt */ + d64 = (d32 * d64 * d64 * d64) >> (10+3*BICTCP_HZ); /* c/rtt * (t-K)^3 */ + d32 = (u32)d64; + if (t < ca->bic_K) /* below origin*/ + bic_target = ca->bic_origin_point - d32; + else /* above origin*/ + bic_target = ca->bic_origin_point + d32; + + /* cubic function - calc bictcp_cnt*/ + if (bic_target > cwnd) { + ca->cnt = cwnd / (bic_target - cwnd); + } else { + ca->cnt = 100 * cwnd; /* very small increment*/ + } + + if (ca->delay_min > 0) { + /* max increment = Smax * rtt / 0.1 */ + min_cnt = (cwnd * HZ * 8)/(10 * max_increment * ca->delay_min); + if (ca->cnt < min_cnt) + ca->cnt = min_cnt; + } + + /* slow start and low utilization */ + if (ca->loss_cwnd == 0) /* could be aggressive in slow start */ + ca->cnt = 50; + + /* TCP Friendly */ + if (tcp_friendliness) { + u32 scale = 8*(BICTCP_BETA_SCALE+beta)/3/(BICTCP_BETA_SCALE-beta); + d32 = (cwnd * scale) >> 3; + while (ca->ack_cnt > d32) { /* update tcp cwnd */ + ca->ack_cnt -= d32; + ca->tcp_cwnd++; + } + + if (ca->tcp_cwnd > cwnd){ /* if bic is slower than tcp */ + d32 = ca->tcp_cwnd - cwnd; + max_cnt = cwnd / d32; + if (ca->cnt > max_cnt) + ca->cnt = max_cnt; + } + } + + ca->cnt = (ca->cnt << ACK_RATIO_SHIFT) / ca->delayed_ack; + if (ca->cnt == 0) /* cannot be zero */ + ca->cnt = 1; +} + + +/* Keep track of minimum rtt */ +static inline void measure_delay(struct sock *sk) +{ + const struct tcp_sock *tp = tcp_sk(sk); + struct bictcp *ca = inet_csk_ca(sk); + u32 delay; + + /* No time stamp */ + if (!(tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr) || + /* Discard delay samples right after fast recovery */ + (s32)(tcp_time_stamp - ca->epoch_start) < HZ) + return; + + delay = tcp_time_stamp - tp->rx_opt.rcv_tsecr; + if (delay == 0) + delay = 1; + + /* first time call or link delay decreases */ + if (ca->delay_min == 0 || ca->delay_min > delay) + ca->delay_min = delay; +} + +static void bictcp_cong_avoid(struct sock *sk, u32 ack, + u32 seq_rtt, u32 in_flight, int data_acked) +{ + struct tcp_sock *tp = tcp_sk(sk); + struct bictcp *ca = inet_csk_ca(sk); + + if (data_acked) + measure_delay(sk); + + if (!tcp_is_cwnd_limited(sk, in_flight)) + return; + + if (tp->snd_cwnd <= tp->snd_ssthresh) + tcp_slow_start(tp); + else { + bictcp_update(ca, tp->snd_cwnd); + + /* In dangerous area, increase slowly. + * In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd + */ + if (tp->snd_cwnd_cnt >= ca->cnt) { + if (tp->snd_cwnd < tp->snd_cwnd_clamp) + tp->snd_cwnd++; + tp->snd_cwnd_cnt = 0; + } else + tp->snd_cwnd_cnt++; + } + +} + +static u32 bictcp_recalc_ssthresh(struct sock *sk) +{ + const struct tcp_sock *tp = tcp_sk(sk); + struct bictcp *ca = inet_csk_ca(sk); + + ca->epoch_start = 0; /* end of epoch */ + + /* Wmax and fast convergence */ + if (tp->snd_cwnd < ca->last_max_cwnd && fast_convergence) + ca->last_max_cwnd = (tp->snd_cwnd * (BICTCP_BETA_SCALE + beta)) + / (2 * BICTCP_BETA_SCALE); + else + ca->last_max_cwnd = tp->snd_cwnd; + + ca->loss_cwnd = tp->snd_cwnd; + + return max((tp->snd_cwnd * beta) / BICTCP_BETA_SCALE, 2U); +} + +static u32 bictcp_undo_cwnd(struct sock *sk) +{ + struct bictcp *ca = inet_csk_ca(sk); + + return max(tcp_sk(sk)->snd_cwnd, ca->last_max_cwnd); +} + +static u32 bictcp_min_cwnd(struct sock *sk) +{ + return tcp_sk(sk)->snd_ssthresh; +} + +static void bictcp_state(struct sock *sk, u8 new_state) +{ + if (new_state == TCP_CA_Loss) + bictcp_reset(inet_csk_ca(sk)); +} + +/* Track delayed acknowledgment ratio using sliding window + * ratio = (15*ratio + sample) / 16 + */ +static void bictcp_acked(struct sock *sk, u32 cnt) +{ + const struct inet_connection_sock *icsk = inet_csk(sk); + + if (cnt > 0 && icsk->icsk_ca_state == TCP_CA_Open) { + struct bictcp *ca = inet_csk_ca(sk); + cnt -= ca->delayed_ack >> ACK_RATIO_SHIFT; + ca->delayed_ack += cnt; + } +} + + +static struct tcp_congestion_ops cubictcp = { + .init = bictcp_init, + .ssthresh = bictcp_recalc_ssthresh, + .cong_avoid = bictcp_cong_avoid, + .set_state = bictcp_state, + .undo_cwnd = bictcp_undo_cwnd, + .min_cwnd = bictcp_min_cwnd, + .pkts_acked = bictcp_acked, + .owner = THIS_MODULE, + .name = "cubic", +}; + +static int __init cubictcp_register(void) +{ + BUG_ON(sizeof(struct bictcp) > ICSK_CA_PRIV_SIZE); + return tcp_register_congestion_control(&cubictcp); +} + +static void __exit cubictcp_unregister(void) +{ + tcp_unregister_congestion_control(&cubictcp); +} + +module_init(cubictcp_register); +module_exit(cubictcp_unregister); + +MODULE_AUTHOR("Sangtae Ha, Stephen Hemminger"); +MODULE_LICENSE("GPL"); +MODULE_DESCRIPTION("CUBIC TCP"); +MODULE_VERSION("2.0"); -- cgit v1.1 From 89b3d9aaf46791177c5a5fa07a3ed38a035b5ef5 Mon Sep 17 00:00:00 2001 From: Stephen Hemminger Date: Wed, 21 Dec 2005 19:32:08 -0800 Subject: [TCP] cubic: precompute constants Revised version of patch to pre-compute values for TCP cubic. * d32,d64 replaced with descriptive names * cube_factor replaces srtt[scaled by count] / HZ * ((1 << (10+2*BICTCP_HZ)) / bic_scale) * beta_scale replaces 8*(BICTCP_BETA_SCALE+beta)/3/(BICTCP_BETA_SCALE-beta); Signed-off-by: Stephen Hemminger Signed-off-by: David S. Miller --- net/ipv4/tcp_cubic.c | 133 ++++++++++++++++++++++----------------------------- 1 file changed, 57 insertions(+), 76 deletions(-) (limited to 'net/ipv4/tcp_cubic.c') diff --git a/net/ipv4/tcp_cubic.c b/net/ipv4/tcp_cubic.c index bb5dc4b..44fd408 100644 --- a/net/ipv4/tcp_cubic.c +++ b/net/ipv4/tcp_cubic.c @@ -16,7 +16,7 @@ #include #include #include - +#include #define BICTCP_BETA_SCALE 1024 /* Scale factor beta calculation * max_cwnd = snd_cwnd * beta @@ -34,15 +34,20 @@ static int initial_ssthresh = 100; static int bic_scale = 41; static int tcp_friendliness = 1; +static u32 cube_rtt_scale; +static u32 beta_scale; +static u64 cube_factor; + +/* Note parameters that are used for precomputing scale factors are read-only */ module_param(fast_convergence, int, 0644); MODULE_PARM_DESC(fast_convergence, "turn on/off fast convergence"); module_param(max_increment, int, 0644); MODULE_PARM_DESC(max_increment, "Limit on increment allowed during binary search"); -module_param(beta, int, 0644); +module_param(beta, int, 0444); MODULE_PARM_DESC(beta, "beta for multiplicative increase"); module_param(initial_ssthresh, int, 0644); MODULE_PARM_DESC(initial_ssthresh, "initial value of slow start threshold"); -module_param(bic_scale, int, 0644); +module_param(bic_scale, int, 0444); MODULE_PARM_DESC(bic_scale, "scale (scaled by 1024) value for bic function (bic_scale/1024)"); module_param(tcp_friendliness, int, 0644); MODULE_PARM_DESC(tcp_friendliness, "turn on/off tcp friendliness"); @@ -151,65 +156,13 @@ static u32 cubic_root(u64 x) return (u32)end; } -static inline u32 bictcp_K(u32 dist, u32 srtt) -{ - u64 d64; - u32 d32; - u32 count; - u32 result; - - /* calculate the "K" for (wmax-cwnd) = c/rtt * K^3 - so K = cubic_root( (wmax-cwnd)*rtt/c ) - the unit of K is bictcp_HZ=2^10, not HZ - - c = bic_scale >> 10 - rtt = (tp->srtt >> 3 ) / HZ - - the following code has been designed and tested for - cwnd < 1 million packets - RTT < 100 seconds - HZ < 1,000,00 (corresponding to 10 nano-second) - - */ - - /* 1/c * 2^2*bictcp_HZ */ - d32 = (1 << (10+2*BICTCP_HZ)) / bic_scale; - d64 = (__u64)d32; - - /* srtt * 2^count / HZ - 1) to get a better accuracy of the following d32, - the larger the "count", the better the accuracy - 2) and avoid overflow of the following d64 - the larger the "count", the high possibility of overflow - 3) so find a "count" between bictcp_hz-3 and bictcp_hz - "count" may be less than bictcp_HZ, - then d64 becomes 0. that is OK - */ - d32 = srtt; - count = 0; - while (((d32 & 0x80000000)==0) && (count < BICTCP_HZ)){ - d32 = d32 << 1; - count++; - } - d32 = d32 / HZ; - - /* (wmax-cwnd) * (srtt>>3 / HZ) / c * 2^(3*bictcp_HZ) */ - d64 = (d64 * dist * d32) >> (count+3-BICTCP_HZ); - - /* cubic root */ - d64 = cubic_root(d64); - - result = (u32)d64; - return result; -} - /* * Compute congestion window to use. */ static inline void bictcp_update(struct bictcp *ca, u32 cwnd) { - u64 d64; - u32 d32, t, srtt, bic_target, min_cnt, max_cnt; + u64 offs; + u32 delta, t, bic_target, min_cnt, max_cnt; ca->ack_cnt++; /* count the number of ACKs */ @@ -220,8 +173,6 @@ static inline void bictcp_update(struct bictcp *ca, u32 cwnd) ca->last_cwnd = cwnd; ca->last_time = tcp_time_stamp; - srtt = (HZ << 3)/10; /* use real time-based growth function */ - if (ca->epoch_start == 0) { ca->epoch_start = tcp_time_stamp; /* record the beginning of an epoch */ ca->ack_cnt = 1; /* start counting */ @@ -231,7 +182,11 @@ static inline void bictcp_update(struct bictcp *ca, u32 cwnd) ca->bic_K = 0; ca->bic_origin_point = cwnd; } else { - ca->bic_K = bictcp_K(ca->last_max_cwnd-cwnd, srtt); + /* Compute new K based on + * (wmax-cwnd) * (srtt>>3 / HZ) / c * 2^(3*bictcp_HZ) + */ + ca->bic_K = cubic_root(cube_factor + * (ca->last_max_cwnd - cwnd)); ca->bic_origin_point = ca->last_max_cwnd; } } @@ -239,9 +194,9 @@ static inline void bictcp_update(struct bictcp *ca, u32 cwnd) /* cubic function - calc*/ /* calculate c * time^3 / rtt, * while considering overflow in calculation of time^3 - * (so time^3 is done by using d64) + * (so time^3 is done by using 64 bit) * and without the support of division of 64bit numbers - * (so all divisions are done by using d32) + * (so all divisions are done by using 32 bit) * also NOTE the unit of those veriables * time = (t - K) / 2^bictcp_HZ * c = bic_scale >> 10 @@ -255,18 +210,16 @@ static inline void bictcp_update(struct bictcp *ca, u32 cwnd) << BICTCP_HZ) / HZ; if (t < ca->bic_K) /* t - K */ - d32 = ca->bic_K - t; + offs = ca->bic_K - t; else - d32 = t - ca->bic_K; + offs = t - ca->bic_K; - d64 = (u64)d32; - d32 = (bic_scale << 3) * HZ / srtt; /* 1024*c/rtt */ - d64 = (d32 * d64 * d64 * d64) >> (10+3*BICTCP_HZ); /* c/rtt * (t-K)^3 */ - d32 = (u32)d64; + /* c/rtt * (t-K)^3 */ + delta = (cube_rtt_scale * offs * offs * offs) >> (10+3*BICTCP_HZ); if (t < ca->bic_K) /* below origin*/ - bic_target = ca->bic_origin_point - d32; + bic_target = ca->bic_origin_point - delta; else /* above origin*/ - bic_target = ca->bic_origin_point + d32; + bic_target = ca->bic_origin_point + delta; /* cubic function - calc bictcp_cnt*/ if (bic_target > cwnd) { @@ -288,16 +241,16 @@ static inline void bictcp_update(struct bictcp *ca, u32 cwnd) /* TCP Friendly */ if (tcp_friendliness) { - u32 scale = 8*(BICTCP_BETA_SCALE+beta)/3/(BICTCP_BETA_SCALE-beta); - d32 = (cwnd * scale) >> 3; - while (ca->ack_cnt > d32) { /* update tcp cwnd */ - ca->ack_cnt -= d32; + u32 scale = beta_scale; + delta = (cwnd * scale) >> 3; + while (ca->ack_cnt > delta) { /* update tcp cwnd */ + ca->ack_cnt -= delta; ca->tcp_cwnd++; } if (ca->tcp_cwnd > cwnd){ /* if bic is slower than tcp */ - d32 = ca->tcp_cwnd - cwnd; - max_cnt = cwnd / d32; + delta = ca->tcp_cwnd - cwnd; + max_cnt = cwnd / delta; if (ca->cnt > max_cnt) ca->cnt = max_cnt; } @@ -428,6 +381,34 @@ static struct tcp_congestion_ops cubictcp = { static int __init cubictcp_register(void) { BUG_ON(sizeof(struct bictcp) > ICSK_CA_PRIV_SIZE); + + /* Precompute a bunch of the scaling factors that are used per-packet + * based on SRTT of 100ms + */ + + beta_scale = 8*(BICTCP_BETA_SCALE+beta)/ 3 / (BICTCP_BETA_SCALE - beta); + + cube_rtt_scale = (bic_scale << 3) / 10; /* 1024*c/rtt */ + + /* calculate the "K" for (wmax-cwnd) = c/rtt * K^3 + * so K = cubic_root( (wmax-cwnd)*rtt/c ) + * the unit of K is bictcp_HZ=2^10, not HZ + * + * c = bic_scale >> 10 + * rtt = 100ms + * + * the following code has been designed and tested for + * cwnd < 1 million packets + * RTT < 100 seconds + * HZ < 1,000,00 (corresponding to 10 nano-second) + */ + + /* 1/c * 2^2*bictcp_HZ * srtt */ + cube_factor = 1ull << (10+3*BICTCP_HZ); /* 2^40 */ + + /* divide by bic_scale and by constant Srtt (100ms) */ + do_div(cube_factor, bic_scale * 10); + return tcp_register_congestion_control(&cubictcp); } -- cgit v1.1 From 9eb2d627190a8afe4b9276b24615a9559504fa60 Mon Sep 17 00:00:00 2001 From: Stephen Hemminger Date: Wed, 21 Dec 2005 19:32:36 -0800 Subject: [TCP] cubic: use Newton-Raphson Replace cube root algorithim with a faster version using Newton-Raphson. Surprisingly, doing the scaled div64_64 is faster than a true 64 bit division on 64 bit CPU's. Signed-off-by: Stephen Hemminger Signed-off-by: David S. Miller --- net/ipv4/tcp_cubic.c | 93 ++++++++++++++++++++++------------------------------ 1 file changed, 39 insertions(+), 54 deletions(-) (limited to 'net/ipv4/tcp_cubic.c') diff --git a/net/ipv4/tcp_cubic.c b/net/ipv4/tcp_cubic.c index 44fd408..31a4986 100644 --- a/net/ipv4/tcp_cubic.c +++ b/net/ipv4/tcp_cubic.c @@ -52,6 +52,7 @@ MODULE_PARM_DESC(bic_scale, "scale (scaled by 1024) value for bic function (bic_ module_param(tcp_friendliness, int, 0644); MODULE_PARM_DESC(tcp_friendliness, "turn on/off tcp friendliness"); +#include /* BIC TCP Parameters */ struct bictcp { @@ -93,67 +94,51 @@ static void bictcp_init(struct sock *sk) tcp_sk(sk)->snd_ssthresh = initial_ssthresh; } -/* 65536 times the cubic root */ -static const u64 cubic_table[8] - = {0, 65536, 82570, 94519, 104030, 112063, 119087, 125367}; +/* 64bit divisor, dividend and result. dynamic precision */ +static inline u_int64_t div64_64(u_int64_t dividend, u_int64_t divisor) +{ + u_int32_t d = divisor; + + if (divisor > 0xffffffffULL) { + unsigned int shift = fls(divisor >> 32); + + d = divisor >> shift; + dividend >>= shift; + } + + /* avoid 64 bit division if possible */ + if (dividend >> 32) + do_div(dividend, d); + else + dividend = (uint32_t) dividend / d; + + return dividend; +} /* - * calculate the cubic root of x - * the basic idea is that x can be expressed as i*8^j - * so cubic_root(x) = cubic_root(i)*2^j - * in the following code, x is i, and y is 2^j - * because of integer calculation, there are errors in calculation - * so finally use binary search to find out the exact solution + * calculate the cubic root of x using Newton-Raphson */ -static u32 cubic_root(u64 x) +static u32 cubic_root(u64 a) { - u64 y, app, target, start, end, mid, start_diff, end_diff; - - if (x == 0) - return 0; + u32 x, x1; - target = x; - - /* first estimate lower and upper bound */ - y = 1; - while (x >= 8){ - x = (x >> 3); - y = (y << 1); - } - start = (y*cubic_table[x])>>16; - if (x==7) - end = (y<<1); - else - end = (y*cubic_table[x+1]+65535)>>16; - - /* binary search for more accurate one */ - while (start < end-1) { - mid = (start+end) >> 1; - app = mid*mid*mid; - if (app < target) - start = mid; - else if (app > target) - end = mid; - else - return mid; - } + /* Initial estimate is based on: + * cbrt(x) = exp(log(x) / 3) + */ + x = 1u << (fls64(a)/3); - /* find the most accurate one from start and end */ - app = start*start*start; - if (app < target) - start_diff = target - app; - else - start_diff = app - target; - app = end*end*end; - if (app < target) - end_diff = target - app; - else - end_diff = app - target; + /* + * Iteration based on: + * 2 + * x = ( 2 * x + a / 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); - if (start_diff < end_diff) - return (u32)start; - else - return (u32)end; + return x; } /* -- cgit v1.1