64-BitBrainstorm_2026/AMSS-NCKU
8
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add targeted point-to-point Interp_Points overload for surface_integral

Browse Source 
Instead of broadcasting all interpolated point data to every MPI rank,
the new overload sends each point only to the one rank that needs it
for integration, reducing communication volume by ~nprocs times.

The consumer rank is computed deterministically using the same Nmin/Nmax
work distribution formula used by surface_integral callers. Two active
call sites (surf_Wave and surf_MassPAng with MPI_COMM_WORLD) now use
the new overload. Other callers (ShellPatch, Comm_here variants, etc.)
remain unchanged.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
CGH0S7
2026-02-10 19:18:56 +08:00
parent 50e2a845f8
commit d06d5b4db8
3 changed files with 289 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files

42
AMSS_NCKU_source/surface_integral.C
View File

@@ -220,16 +220,9 @@ void surface_integral::surf_Wave(double rex, int lev, cgh *GH, var *Rpsi4, var *
pox[2][n] = rex * nz_g[n];
}
double *shellf;
shellf = new double[n_tot * InList];
GH->PatL[lev]->data->Interp_Points(DG_List, n_tot, pox, shellf, Symmetry);
int mp, Lp, Nmin, Nmax;
mp = n_tot / cpusize;
Lp = n_tot - cpusize * mp;
if (Lp > myrank)
{
Nmin = myrank * mp + myrank;
@@ -241,6 +234,11 @@ void surface_integral::surf_Wave(double rex, int lev, cgh *GH, var *Rpsi4, var *
Nmax = Nmin + mp - 1;
}
double *shellf;
shellf = new double[n_tot * InList];
GH->PatL[lev]->data->Interp_Points(DG_List, n_tot, pox, shellf, Symmetry, Nmin, Nmax);
//|~~~~~> Integrate the dot product of Dphi with the surface normal.
double *RP_out, *IP_out;
@@ -2386,25 +2384,9 @@ void surface_integral::surf_MassPAng(double rex, int lev, cgh *GH, var *chi, var
pox[2][n] = rex * nz_g[n];
}
double *shellf;
shellf = new double[n_tot * InList];
// we have assumed there is only one box on this level,
// so we do not need loop boxes
GH->PatL[lev]->data->Interp_Points(DG_List, n_tot, pox, shellf, Symmetry);
double Mass_out = 0;
double ang_outx, ang_outy, ang_outz;
double p_outx, p_outy, p_outz;
ang_outx = ang_outy = ang_outz = 0.0;
p_outx = p_outy = p_outz = 0.0;
const double f1o8 = 0.125;
int mp, Lp, Nmin, Nmax;
mp = n_tot / cpusize;
Lp = n_tot - cpusize * mp;
if (Lp > myrank)
{
Nmin = myrank * mp + myrank;
@@ -2416,6 +2398,20 @@ void surface_integral::surf_MassPAng(double rex, int lev, cgh *GH, var *chi, var
Nmax = Nmin + mp - 1;
}
double *shellf;
shellf = new double[n_tot * InList];
// we have assumed there is only one box on this level,
// so we do not need loop boxes
GH->PatL[lev]->data->Interp_Points(DG_List, n_tot, pox, shellf, Symmetry, Nmin, Nmax);
double Mass_out = 0;
double ang_outx, ang_outy, ang_outz;
double p_outx, p_outy, p_outz;
ang_outx = ang_outy = ang_outz = 0.0;
p_outx = p_outy = p_outz = 0.0;
const double f1o8 = 0.125;
double Chi, Psi;
double Gxx, Gxy, Gxz, Gyy, Gyz, Gzz;
double gupxx, gupxy, gupxz, gupyy, gupyz, gupzz;