int updateReg(unsigned int A[], unsigned int E[], unsigned int m[], int n) {
// Compute the register value E[n], having already computed the others.
// Pass parameters as: updateReg (a1, e1, m1, n) or updateReg(a2, e2, m2, n)
E[n] = E[n-4] + S1(E[n-1]) + ch(E[n-1], E[n-2], E[n-3]) + A[n-4] + k[n] + m[n];
A[n] = -A[n-4] + S0(A[n-1]) + mj(A[n-1]), A[n-2], A[n-3]) + E[n];
}
void sg::cd_system::cd_update(int pid, int id, unsigned ft, bool lock, int index /* = 0 */)
{
Json::Value jCdInfo;
jCdInfo["id"] = id * 10 + index;
jCdInfo["ft"] = ft;
jCdInfo["lc"] = lock;
Json::Value respJson;
respJson["msg"][0u] = jCdInfo;
string respond_str = respJson.toStyledString();
//respond_str = commom_sys.tighten(respond_str);
na::msg::msg_json mj(sg::protocol::g2c::cd_cdInfo_update_resp, respond_str);
player_mgr.send_to_online_player(pid, mj);
}
// So, first, let us write a function to update a particular bit of A and E depending on M. n - step number
int updateBit (int bitPos, int step, unsigned int A[], unsigned int E[], unsigned int m[], char mChar[][32], char aChar[][32], char eChar[][32]) {
// Let us build the bitPos - bit of A and E.
int bitE, bitA, bitEC, bitAC;
int chBit1 = extractBit(E, step - 1, bitPos);
int chBit2 = extractBit(E, step - 2, bitPos);
int chBit3 = extractBit(E, step - 3, bitPos);
int chBit = ch(chBit1, chBit2, chBit3);
int constBit = extractBit(k, step, bitPos);
int wordBit = extractBit(m, step, bitPos);
int EBit_4 = extractBit(E, step - 4, bitPos);
int ABit_4 = extractBit(A, step - 4, bitPos);
int sBit1 = extractBit(E, step - 1, (bitPos - 6) % 32);
int sBit2 = extractBit(E, step - 1, (bitPos - 11) % 32);
int sBit3 = extractBit(E, step - 1, (bitPos - 25) % 32);
int sBit = sBit1 ^ sBit2 ^ sBit3;
int bitECarry = extractBit(ECarry, step, bitPos-1);
int bitE = EBit_4 ^ sBit ^ chBit ^ ABit_4 ^ constBit ^ wordBit ^ bitECarry;
int bitEC = (EBit_4 + sBit + chBit + ABit_4 + constBit + wordBit + bitECarry) >> 1;
// Dropped the last bit to generate the carry.
int SBit1 = extractBit(A, step - 1, (bitPos - 2) % 32);
int SBit2 = extractBit(A, step - 1, (bitPos - 13) % 32);
int SBit3 = extractBit(A, step - 1, (bitPost - 22) % 32);
int SBit = SBit1 ^ SBit2 ^ SBit3;
int bitACarry = extractBit(ACarry, step, bitPos - 1);
int mjBit1 = extractBit(A, step - 1, bitPos);
int mjBit2 = extractBit(A, step - 2, bitPos);
int mjBit3 = extractBit(A, step - 3, bitPos);
int mjBit = mj(mjBit1, mjBit2, mjBit3);
int bitA = (-ABit_4) ^ SBit ^ bitE ^ mjBit;
int bitACarry = ((-ABit_4) + SBit + bitE + mjBit) >> 1;
}
void sg::cd_system::modelData_update_client(int player_id)
{
Json::Value jCdList;
building_sub_sys.collect_cd_info(player_id, jCdList);
train_system.collect_cd_info(player_id, jCdList);
building_sys.collect_cd_info(player_id, jCdList);
equipment_sys.collect_cd_info(player_id, jCdList);
player_mgr.collect_cd_info(player_id, jCdList);
Json::Value respJson;
respJson["msg"][0u] = jCdList;
string respond_str = respJson.toStyledString();
//respond_str = commom_sys.tighten(respond_str);
na::msg::msg_json mj(sg::protocol::g2c::cd_modelData_update_resp, respond_str);
player_mgr.send_to_online_player(player_id, mj);
}
void cnbint(
int i,
const double pos[][3],
const double vel[][3],
const double mass[],
int nnb,
int list[],
double f[3],
double fdot[3]){
const int NBMAX = 512;
assert(nnb <= NBMAX);
float xbuf[NBMAX] __attribute__ ((aligned(16)));
float ybuf[NBMAX] __attribute__ ((aligned(16)));
float zbuf[NBMAX] __attribute__ ((aligned(16)));
float vxbuf[NBMAX] __attribute__ ((aligned(16)));
float vybuf[NBMAX] __attribute__ ((aligned(16)));
float vzbuf[NBMAX] __attribute__ ((aligned(16)));
float mbuf[NBMAX] __attribute__ ((aligned(16)));
assert((unsigned long)xbuf % 16 == 0);
double xi = pos[i][0];
double yi = pos[i][1];
double zi = pos[i][2];
float vxi = vel[i][0];
float vyi = vel[i][1];
float vzi = vel[i][2];
for(int k=0; k<nnb; k++){
int j = list[k];
#if 1
int jj = list[k+4];
__builtin_prefetch(pos[jj]);
__builtin_prefetch(pos[jj]+2);
__builtin_prefetch(vel[jj]);
__builtin_prefetch(vel[jj]+2);
__builtin_prefetch(&mass[jj]);
#endif
#if 0
xbuf[k] = pos[j][0] - xi;
ybuf[k] = pos[j][1] - yi;
zbuf[k] = pos[j][2] - zi;
vxbuf[k] = vel[j][0] - vxi;
vybuf[k] = vel[j][1] - vyi;
vzbuf[k] = vel[j][2] - vzi;
mbuf[k] = mass[j];
#else
double xj = pos[j][0];
double yj = pos[j][1];
double zj = pos[j][2];
float vxj = vel[j][0];
float vyj = vel[j][1];
float vzj = vel[j][2];
float mj = mass[j];
xj -= xi;
yj -= yi;
zj -= zi;
vxj -= vxi;
vyj -= vyi;
vzj -= vzi;
xbuf[k] = xj;
ybuf[k] = yj;
zbuf[k] = zj;
vxbuf[k] = vxj;
vybuf[k] = vyj;
vzbuf[k] = vzj;
mbuf[k] = mj;
#endif
}
for(int k=nnb; k%4; k++){
xbuf[k] = 16.0f;
ybuf[k] = 16.0f;
zbuf[k] = 16.0f;
vxbuf[k] = 0.0f;
vybuf[k] = 0.0f;
vzbuf[k] = 0.0f;
mbuf[k] = 0.0f;
}
v4df ax(0.0), ay(0.0), az(0.0);
v4sf jx(0.0), jy(0.0), jz(0.0);
for(int k=0; k<nnb; k+=4){
v4sf dx(xbuf + k);
v4sf dy(ybuf + k);
v4sf dz(zbuf + k);
v4sf dvx(vxbuf + k);
v4sf dvy(vybuf + k);
v4sf dvz(vzbuf + k);
v4sf mj(mbuf + k);
v4sf r2 = dx*dx + dy*dy + dz*dz;
v4sf rv = dx*dvx + dy*dvy + dz*dvz;
rv *= v4sf(-3.0f);
// v4sf rinv1 = r2.rsqrt() & v4sf(mask);
#if 1
v4sf rinv1 = r2.rsqrt();
#else
v4sf rinv1 = v4sf(v4df(1.0) / v4df(r2).sqrt());
#endif
v4sf rinv2 = rinv1 * rinv1;
rv *= rinv2;
v4sf rinv3 = mj * rinv1 * rinv2;
dx *= rinv3; ax += v4df(dx);
dy *= rinv3; ay += v4df(dy);
dz *= rinv3; az += v4df(dz);
dvx *= rinv3; jx += dvx;
dvy *= rinv3; jy += dvy;
dvz *= rinv3; jz += dvz;
dx *= rv; jx += dx;
dy *= rv; jy += dy;
dz *= rv; jz += dz;
}
f[0] = ax.sum();
f[1] = ay.sum();
f[2] = az.sum();
fdot[0] = (v4df(jx)).sum();
fdot[1] = (v4df(jy)).sum();
fdot[2] = (v4df(jz)).sum();
assert(f[0] == f[0]);
assert(f[1] == f[1]);
assert(f[2] == f[2]);
assert(fdot[0] == fdot[0]);
assert(fdot[1] == fdot[1]);
assert(fdot[2] == fdot[2]);
}
void cnbint(
int i,
const double pos[][3],
const double vel[][3],
const double mass[],
int nnb,
int list[],
double f[3],
double fdot[3]){
const int NBMAX = 512;
assert(nnb <= NBMAX);
double xbuf[NBMAX] __attribute__ ((aligned(16)));
double ybuf[NBMAX] __attribute__ ((aligned(16)));
double zbuf[NBMAX] __attribute__ ((aligned(16)));
float vxbuf[NBMAX] __attribute__ ((aligned(16)));
float vybuf[NBMAX] __attribute__ ((aligned(16)));
float vzbuf[NBMAX] __attribute__ ((aligned(16)));
float mbuf[NBMAX] __attribute__ ((aligned(16)));
for(int k=0; k<nnb; k++){
int j = list[k];
xbuf[k] = pos[j][0];
ybuf[k] = pos[j][1];
zbuf[k] = pos[j][2];
vxbuf[k] = vel[j][0];
vybuf[k] = vel[j][1];
vzbuf[k] = vel[j][2];
mbuf[k] = mass[j];
}
for(int k=nnb; k%4; k++){
xbuf[k] = 16.0;
ybuf[k] = 16.0;
zbuf[k] = 16.0;
vxbuf[k] = 0.0f;
vybuf[k] = 0.0f;
vzbuf[k] = 0.0f;
mbuf[k] = 0.0f;
}
v4df xi(pos[i][0]);
v4df yi(pos[i][1]);
v4df zi(pos[i][2]);
v4sf vxi(vel[i][0]);
v4sf vyi(vel[i][1]);
v4sf vzi(vel[i][2]);
v4df ax(0.0), ay(0.0), az(0.0);
v4sf jx(0.0), jy(0.0), jz(0.0);
for(int k=0; k<nnb; k+=4){
v4df xj(xbuf + k);
v4df yj(ybuf + k);
v4df zj(zbuf + k);
v4sf vxj(vxbuf + k);
v4sf vyj(vybuf + k);
v4sf vzj(vzbuf + k);
v4sf mj(mbuf + k);
v4sf dx = v4sf::_v4sf(xj - xi);
v4sf dy = v4sf::_v4sf(yj - yi);
v4sf dz = v4sf::_v4sf(zj - zi);
v4sf dvx = vxj - vxi;
v4sf dvy = vyj - vyi;
v4sf dvz = vzj - vzi;
v4sf r2 = dx*dx + dy*dy + dz*dz;
v4sf rv = dx*dvx + dy*dvy + dz*dvz;
rv *= v4sf(-3.0f);
// v4sf rinv1 = r2.rsqrt() & v4sf(mask);
v4sf rinv1 = r2.rsqrt();
v4sf rinv2 = rinv1 * rinv1;
rv *= rinv2;
v4sf rinv3 = mj * rinv1 * rinv2;
dx *= rinv3; ax += v4df(dx);
dy *= rinv3; ay += v4df(dy);
dz *= rinv3; az += v4df(dz);
dvx *= rinv3; jx += dvx;
dvy *= rinv3; jy += dvy;
dvz *= rinv3; jz += dvz;
dx *= rv; jx += dx;
dy *= rv; jy += dy;
dz *= rv; jz += dz;
}
f[0] = ax.sum();
f[1] = ay.sum();
f[2] = az.sum();
fdot[0] = (v4df(jx)).sum();
fdot[1] = (v4df(jy)).sum();
fdot[2] = (v4df(jz)).sum();
assert(f[0] == f[0]);
assert(f[1] == f[1]);
assert(f[2] == f[2]);
assert(fdot[0] == fdot[0]);
assert(fdot[1] == fdot[1]);
assert(fdot[2] == fdot[2]);
}
