yohai/phigrape
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Removed redundant copying from calc_self_grav and got rid of calc_self_grav_zero

Browse source
This commit is contained in:
Yohai Meiron 2020-03-13 19:36:18 -04:00
parent 2c05355eaa
commit 4b8ef0ca61
1 changed files with 36 additions and 147 deletions
Download patch file Download diff file Expand all files Collapse all files

147
phigrape.cpp
View file

@ -145,6 +145,8 @@ Last redaction : 2019.04.16 12:55
#include <sys/time.h> #include <sys/time.h>
#include <sys/resource.h> #include <sys/resource.h>
#include <algorithm>
/* /*
double aaa; double aaa;
double aaapars[5]; double aaapars[5];
@ -952,101 +954,6 @@ fclose(out);
#endif #endif
void calc_self_grav_zero()
{
/* calc the grav for the active particles */
g6_set_ti(clusterid, time_cur);
ni = n_act;
/* define the local phi, a, adot for these active particles */
for (i=0; i<ni; i+=npipe)
{
nn = npipe;
if (ni-i < npipe) nn = ni - i;
for (ii=0; ii<nn; ii++)
{
index_i[ii] = ind_act[i+ii];
h2_i[ii] = eps2;
for (k=0;k<3;k++)
{
x_i[ii][k] = x_act[i+ii][k]; v_i[ii][k] = v_act[i+ii][k];
} /* k */
p_i[ii] = -1.0;
for (k=0;k<3;k++)
{
a_i[ii][k] = 10.0;
jerk_i[ii][k] = 100.0;
} /* k */
} /* ii */
g6calc_firsthalf(clusterid, n_loc, nn, index_i, x_i, v_i , a_i, jerk_i, p_i, eps2, h2_i);
g6calc_lasthalf(clusterid, n_loc, nn, index_i, x_i, v_i, eps2, h2_i, a_i, jerk_i, p_i);
for (ii=0; ii<nn; ii++)
{
if (ABS(jerk_i[ii][0]) < 1.0E-05) jerk_i[ii][0] = 1.0E-05;
if (ABS(jerk_i[ii][1]) < 1.0E-05) jerk_i[ii][1] = 1.0E-05;
if (ABS(jerk_i[ii][2]) < 1.0E-05) jerk_i[ii][2] = 1.0E-05;
if (ABS(jerk_i[ii][0]) > 100.0) jerk_i[ii][0] = 100.0;
if (ABS(jerk_i[ii][1]) > 100.0) jerk_i[ii][1] = 100.0;
if (ABS(jerk_i[ii][2]) > 100.0) jerk_i[ii][2] = 100.0;
}
g6calc_firsthalf(clusterid, n_loc, nn, index_i, x_i, v_i , a_i, jerk_i, p_i, eps2, h2_i);
g6calc_lasthalf(clusterid, n_loc, nn, index_i, x_i, v_i, eps2, h2_i, a_i, jerk_i, p_i);
for (ii=0; ii<nn; ii++)
{
pot_act_tmp[i+ii] = p_i[ii];
for (k=0;k<3;k++)
{
a_act_tmp[i+ii][k] = a_i[ii][k];
adot_act_tmp[i+ii][k] = jerk_i[ii][k];
} /* k */
} /* ii */
} /* i */
/* Store the value of the local partial force etc... */
for (i=0; i<n_act; i++)
{
pot_act_tmp_loc[i] = pot_act_tmp[i];
for (k=0;k<3;k++)
{
a_act_tmp_loc[i][k] = a_act_tmp[i][k];
adot_act_tmp_loc[i][k] = adot_act_tmp[i][k];
} /* k */
} /* i */
}
void calc_ext_grav_zero() void calc_ext_grav_zero()
{ {
@ -1296,17 +1203,15 @@ for (i=0; i<ni; i++)
} }
void calc_self_grav() void calc_self_grav(double t)
{ {
/* calc the new grav for the active particles */ /* calc the new grav for the active particles */
// TODO possibly redundant copying in this function
#ifdef TIMING #ifdef TIMING
get_CPU_time(&CPU_tmp_real0, &CPU_tmp_user0, &CPU_tmp_syst0); get_CPU_time(&CPU_tmp_real0, &CPU_tmp_user0, &CPU_tmp_syst0);
#endif #endif
g6_set_ti(clusterid, min_t); g6_set_ti(clusterid, t);
ni = n_act; ni = n_act;
@ -1316,30 +1221,11 @@ void calc_self_grav()
nn = npipe; nn = npipe;
if (ni-i < npipe) nn = ni - i; if (ni-i < npipe) nn = ni - i;
for (ii=0; ii<nn; ii++) { for (ii=0; ii<nn; ii++) {
iii = ind_act[i+ii]; h2_i[ii] = eps2; // TODO This should be a global or something
index_i[ii] = iii;
h2_i[ii] = eps2;
for (k=0;k<3;k++) {
x_i[ii][k] = x_act_new[i+ii][k];
v_i[ii][k] = v_act_new[i+ii][k];
} /* k */
p_i[ii] = pot_act_tmp_loc[iii];
for (k=0;k<3;k++) {
a_i[ii][k] = a_act_tmp_loc[iii][k];
jerk_i[ii][k] = adot_act_tmp_loc[iii][k];
} /* k */
} /* ii */ } /* ii */
g6calc_firsthalf(clusterid, n_loc, nn, index_i, x_i, v_i , a_i, jerk_i, p_i, eps2, h2_i); g6calc_firsthalf(clusterid, n_loc, nn, ind_act+i, x_act_new+i, v_act_new+i, a_act_tmp+i, adot_act_tmp+i, pot_act_tmp+i, eps2, h2_i);
g6calc_lasthalf(clusterid, n_loc, nn, index_i, x_i, v_i, eps2, h2_i, a_i, jerk_i, p_i); g6calc_lasthalf( clusterid, n_loc, nn, ind_act+i, x_act_new+i, v_act_new+i, eps2, h2_i, a_act_tmp+i, adot_act_tmp+i, pot_act_tmp+i);
g6_calls++; g6_calls++;
for (ii=0; ii<nn; ii++) {
pot_act_tmp[i+ii] = p_i[ii];
for (k=0;k<3;k++) {
a_act_tmp[i+ii][k] = a_i[ii][k];
adot_act_tmp[i+ii][k] = jerk_i[ii][k];
} /* k */
} /* ii */
} /* i */ } /* i */
/* Store the new value of the local partial force etc... */ /* Store the new value of the local partial force etc... */
@ -1736,7 +1622,7 @@ E_tot_corr = E_tot + E_corr;
E_tot_corr_sd = E_tot + E_corr + E_sd; E_tot_corr_sd = E_tot + E_corr + E_sd;
// if ((Timesteps == 0.0) && (time_cur == 0.0) ) // if ((Timesteps == 0.0) && (time_cur == 0.0) )
if ( (Timesteps == 0.0) ) if (Timesteps == 0.0)
{ {
/* initialize the E_tot_0 + etc... */ /* initialize the E_tot_0 + etc... */
@ -2251,7 +2137,10 @@ int main(int argc, char *argv[])
} /* i */ } /* i */
calc_self_grav_zero(); // NOTE this is where calc_self_grav_zero() used to be. Hopefully the copying from *_act to *_act_new is just a temporary measure.
std::copy(&x_act[0][0], &x_act[0][0]+3*N, &x_act_new[0][0]);
std::copy(&v_act[0][0], &v_act[0][0]+3*N, &v_act_new[0][0]);
calc_self_grav(time_cur);
/* Wait to all processors to finish his works... */ /* Wait to all processors to finish his works... */
MPI_Barrier(MPI_COMM_WORLD); MPI_Barrier(MPI_COMM_WORLD);
@ -2604,7 +2493,7 @@ int main(int argc, char *argv[])
n_act = 0; n_act = 0;
for (i=0; i<N; i++) { for (i=0; i<N; i++) {
if ( (t[i]+dt[i]) == min_t ) { if (t[i]+dt[i] == min_t ) {
ind_act[n_act] = i; ind_act[n_act] = i;
n_act++; n_act++;
} }
@ -2702,7 +2591,7 @@ int main(int argc, char *argv[])
DT_ACT_PRED += (CPU_tmp_user - CPU_tmp_user0); DT_ACT_PRED += (CPU_tmp_user - CPU_tmp_user0);
#endif #endif
calc_self_grav(); calc_self_grav(min_t);
/* Reduce the "global" vectors from "local" on all the nodes */ /* Reduce the "global" vectors from "local" on all the nodes */
@ -2859,7 +2748,7 @@ int main(int argc, char *argv[])
#else #else
#ifdef ADD_BH1 #ifdef ADD_BH1
if ( (ind_act[i] == 0) ) { if (ind_act[i] == 0) {
dt_new = sqrt(eta_bh*(a1abs*a2dot1abs+adot1abs*adot1abs)/(adot1abs*a3dot1abs+a2dot1abs*a2dot1abs)); dt_new = sqrt(eta_bh*(a1abs*a2dot1abs+adot1abs*adot1abs)/(adot1abs*a3dot1abs+a2dot1abs*a2dot1abs));
} else { } else {
dt_new = sqrt(eta*(a1abs*a2dot1abs+adot1abs*adot1abs)/(adot1abs*a3dot1abs+a2dot1abs*a2dot1abs)); dt_new = sqrt(eta*(a1abs*a2dot1abs+adot1abs*adot1abs)/(adot1abs*a3dot1abs+a2dot1abs*a2dot1abs));
@ -2965,9 +2854,9 @@ int main(int argc, char *argv[])
iii = ind_act[i]; iii = ind_act[i];
if ( (tmp_r2 < SEMI_a2*R_INF2) ) { if (tmp_r2 < SEMI_a2*R_INF2) {
if ( (inf_event[iii] == 0) ) { if (inf_event[iii] == 0) {
fprintf(out,"INF1 %.6E %.16E %07d %07d %.6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E % .6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E % .6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E %.6E % .6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E \n", fprintf(out,"INF1 %.6E %.16E %07d %07d %.6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E % .6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E % .6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E %.6E % .6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E \n",
Timesteps, time_cur, i, ind_act[i], Timesteps, time_cur, i, ind_act[i],
@ -2982,7 +2871,7 @@ int main(int argc, char *argv[])
} }
} else { } else {
if ( (inf_event[iii] == 1) ) { if (inf_event[iii] == 1) {
fprintf(out,"INF2 %.6E %.16E %07d %07d %.6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E % .6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E % .6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E %.6E % .6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E \n", fprintf(out,"INF2 %.6E %.16E %07d %07d %.6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E % .6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E % .6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E %.6E % .6E % .6E % .6E % .6E % .6E % .6E % .6E %.6E \n",
Timesteps, time_cur, i, ind_act[i], Timesteps, time_cur, i, ind_act[i],
sqrt(DR2), x_bbhc[0], x_bbhc[1], x_bbhc[2], v_bbhc[0], v_bbhc[1], v_bbhc[2], sqrt(DR2), x_bbhc[0], x_bbhc[1], x_bbhc[2], v_bbhc[0], v_bbhc[1], v_bbhc[2],