Stabilize and wire BSSN-EScalar C path

AMSS_NCKU_source/Parallel.C

@@ -5,6 +5,42 @@
 #include "misc.h"
 #include "parameters.h"
 
+namespace
+{
+enum { MAX_DATA_PACKER_VARS = 64 };
+
+int expand_var_list_pack_info(MyList<var> *src_list, MyList<var> *dst_list,
+                              int *src_sgfn, int *dst_sgfn, double **src_soa)
+{
+  int count = 0;
+  MyList<var> *src_it = src_list;
+  MyList<var> *dst_it = dst_list;
+
+  while (src_it && dst_it)
+  {
+    if (count >= MAX_DATA_PACKER_VARS)
+    {
+      cout << "Parallel::data_packer: too many variables in communication list." << endl;
+      MPI_Abort(MPI_COMM_WORLD, 1);
+    }
+    src_sgfn[count] = src_it->data->sgfn;
+    dst_sgfn[count] = dst_it->data->sgfn;
+    src_soa[count] = src_it->data->SoA;
+    count++;
+    src_it = src_it->next;
+    dst_it = dst_it->next;
+  }
+
+  if (src_it || dst_it)
+  {
+    cout << "error in short data packer, var lists does not match." << endl;
+    MPI_Abort(MPI_COMM_WORLD, 1);
+  }
+
+  return count;
+}
+}
+
 int Parallel::partition1(int &nx, int split_size, int min_width, int cpusize, int shape) // special for 1 diemnsion
 {
   nx = Mymax(1, shape / min_width);
@@ -3730,21 +3766,10 @@ int Parallel::data_packer(double *data, MyList<Parallel::gridseg> *src, MyList<P
   if (!src || !dst)
     return size_out;
 
-  MyList<var> *varls, *varld;
-
-  varls = VarLists;
-  varld = VarListd;
-  while (varls && varld)
-  {
-    varls = varls->next;
-    varld = varld->next;
-  }
-
-  if (varls || varld)
-  {
-    cout << "error in short data packer, var lists does not match." << endl;
-    MPI_Abort(MPI_COMM_WORLD, 1);
-  }
+  int src_sgfn[MAX_DATA_PACKER_VARS];
+  int dst_sgfn[MAX_DATA_PACKER_VARS];
+  double *src_soa[MAX_DATA_PACKER_VARS];
+  const int var_count = expand_var_list_pack_info(VarLists, VarListd, src_sgfn, dst_sgfn, src_soa);
 
   int type; /* 1 copy, 2 restrict, 3 prolong */
   if (src->data->Bg->lev == dst->data->Bg->lev)
@@ -3756,44 +3781,58 @@ int Parallel::data_packer(double *data, MyList<Parallel::gridseg> *src, MyList<P
 
   while (src && dst)
   {
-    if ((dir == PACK && dst->data->Bg->rank == rank_in && src->data->Bg->rank == myrank) ||
-        (dir == UNPACK && src->data->Bg->rank == rank_in && dst->data->Bg->rank == myrank))
-    {
-      varls = VarLists;
-      varld = VarListd;
-      while (varls && varld)
+    const bool rank_match =
+        (dir == PACK && dst->data->Bg->rank == rank_in && src->data->Bg->rank == myrank) ||
+        (dir == UNPACK && src->data->Bg->rank == rank_in && dst->data->Bg->rank == myrank);
+
+    if (rank_match)
     {
+      const int segment_size = dst->data->shape[0] * dst->data->shape[1] * dst->data->shape[2];
+      int offset = size_out;
+
       if (data)
       {
         if (dir == PACK)
+        {
           switch (type)
           {
-              // attention must be paied to the difference between src's llb,uub and dst's llb,uub
             case 1:
-              f_copy(DIM, dst->data->llb, dst->data->uub, dst->data->shape, data + size_out,
-                     src->data->Bg->bbox, src->data->Bg->bbox + dim, src->data->Bg->shape, src->data->Bg->fgfs[varls->data->sgfn],
-                     dst->data->llb, dst->data->uub);
+              for (int iv = 0; iv < var_count; iv++, offset += segment_size)
+                f_copy(DIM, dst->data->llb, dst->data->uub, dst->data->shape, data + offset,
+                       src->data->Bg->bbox, src->data->Bg->bbox + dim, src->data->Bg->shape,
+                       src->data->Bg->fgfs[src_sgfn[iv]], dst->data->llb, dst->data->uub);
               break;
             case 2:
-              f_restrict3(DIM, dst->data->llb, dst->data->uub, dst->data->shape, data + size_out,
-                          src->data->Bg->bbox, src->data->Bg->bbox + dim, src->data->Bg->shape, src->data->Bg->fgfs[varls->data->sgfn],
-                          dst->data->llb, dst->data->uub, varls->data->SoA, Symmetry);
+              for (int iv = 0; iv < var_count; iv++, offset += segment_size)
+                f_restrict3(DIM, dst->data->llb, dst->data->uub, dst->data->shape, data + offset,
+                            src->data->Bg->bbox, src->data->Bg->bbox + dim, src->data->Bg->shape,
+                            src->data->Bg->fgfs[src_sgfn[iv]], dst->data->llb, dst->data->uub,
+                            src_soa[iv], Symmetry);
               break;
             case 3:
-              f_prolong3(DIM, src->data->Bg->bbox, src->data->Bg->bbox + dim, src->data->Bg->shape, src->data->Bg->fgfs[varls->data->sgfn],
-                         dst->data->llb, dst->data->uub, dst->data->shape, data + size_out,
-                         dst->data->llb, dst->data->uub, varls->data->SoA, Symmetry);
+              for (int iv = 0; iv < var_count; iv++, offset += segment_size)
+                f_prolong3(DIM, src->data->Bg->bbox, src->data->Bg->bbox + dim, src->data->Bg->shape,
+                           src->data->Bg->fgfs[src_sgfn[iv]], dst->data->llb, dst->data->uub,
+                           dst->data->shape, data + offset, dst->data->llb, dst->data->uub,
+                           src_soa[iv], Symmetry);
+              break;
+            default:
+              break;
           }
-          if (dir == UNPACK) // from target data to corresponding grid
-            f_copy(DIM, dst->data->Bg->bbox, dst->data->Bg->bbox + dim, dst->data->Bg->shape, dst->data->Bg->fgfs[varld->data->sgfn],
-                   dst->data->llb, dst->data->uub, dst->data->shape, data + size_out,
-                   dst->data->llb, dst->data->uub);
         }
-        size_out += dst->data->shape[0] * dst->data->shape[1] * dst->data->shape[2];
-        varls = varls->next;
-        varld = varld->next;
+        else
+        {
+          for (int iv = 0; iv < var_count; iv++, offset += segment_size)
+            f_copy(DIM, dst->data->Bg->bbox, dst->data->Bg->bbox + dim, dst->data->Bg->shape,
+                   dst->data->Bg->fgfs[dst_sgfn[iv]], dst->data->llb, dst->data->uub,
+                   dst->data->shape, data + offset, dst->data->llb, dst->data->uub);
         }
       }
+
+      size_out = offset + ((!data) ? segment_size * var_count : 0);
+      if (data)
+        size_out = offset;
+    }
     dst = dst->next;
     src = src->next;
   }
@@ -3819,21 +3858,10 @@ int Parallel::data_packermix(double *data, MyList<Parallel::gridseg> *src, MyLis
   if (!src || !dst)
     return size_out;
 
-  MyList<var> *varls, *varld;
-
-  varls = VarLists;
-  varld = VarListd;
-  while (varls && varld)
-  {
-    varls = varls->next;
-    varld = varld->next;
-  }
-
-  if (varls || varld)
-  {
-    cout << "error in short data packer, var lists does not match." << endl;
-    MPI_Abort(MPI_COMM_WORLD, 1);
-  }
+  int src_sgfn[MAX_DATA_PACKER_VARS];
+  int dst_sgfn[MAX_DATA_PACKER_VARS];
+  double *src_soa[MAX_DATA_PACKER_VARS];
+  const int var_count = expand_var_list_pack_info(VarLists, VarListd, src_sgfn, dst_sgfn, src_soa);
 
   int type; /* 1 copy, 2 restrict, 3 prolong */
   if (src->data->Bg->lev == dst->data->Bg->lev)
@@ -3851,31 +3879,42 @@ int Parallel::data_packermix(double *data, MyList<Parallel::gridseg> *src, MyLis
 
   while (src && dst)
   {
-    if ((dir == PACK && dst->data->Bg->rank == rank_in && src->data->Bg->rank == myrank) ||
-        (dir == UNPACK && src->data->Bg->rank == rank_in && dst->data->Bg->rank == myrank))
-    {
-      varls = VarLists;
-      varld = VarListd;
-      while (varls && varld)
+    const bool rank_match =
+        (dir == PACK && dst->data->Bg->rank == rank_in && src->data->Bg->rank == myrank) ||
+        (dir == UNPACK && src->data->Bg->rank == rank_in && dst->data->Bg->rank == myrank);
+
+    if (rank_match)
     {
+      const int segment_size =
+          (src->data->shape[0] + 2 * ghost_width) *
+          (src->data->shape[1] + 2 * ghost_width) *
+          (src->data->shape[2] + 2 * ghost_width);
+      int offset = size_out;
+
       if (data)
       {
         if (dir == PACK)
-            f_prolongcopy3(DIM, src->data->Bg->bbox, src->data->Bg->bbox + dim, src->data->Bg->shape, src->data->Bg->fgfs[varls->data->sgfn],
-                           dst->data->llb, dst->data->uub, src->data->shape, data + size_out,
-                           src->data->llb, src->data->uub, varls->data->SoA, Symmetry);
-          if (dir == UNPACK) // from target data to corresponding grid
-            f_prolongmix3(DIM, dst->data->Bg->bbox, dst->data->Bg->bbox + dim, dst->data->Bg->shape, dst->data->Bg->fgfs[varld->data->sgfn],
-                          src->data->llb, src->data->uub, src->data->shape, data + size_out,
-                          dst->data->llb, dst->data->uub, varls->data->SoA, Symmetry, dst->data->illb, dst->data->iuub);
+        {
+          for (int iv = 0; iv < var_count; iv++, offset += segment_size)
+            f_prolongcopy3(DIM, src->data->Bg->bbox, src->data->Bg->bbox + dim, src->data->Bg->shape,
+                           src->data->Bg->fgfs[src_sgfn[iv]], dst->data->llb, dst->data->uub,
+                           src->data->shape, data + offset, src->data->llb, src->data->uub,
+                           src_soa[iv], Symmetry);
         }
-        // the symmetry problem should be dealt in prolongcopy3,
-        // so we always have ghost_width for both sides
-        size_out += (src->data->shape[0] + 2 * ghost_width) * (src->data->shape[1] + 2 * ghost_width) * (src->data->shape[2] + 2 * ghost_width);
-        varls = varls->next;
-        varld = varld->next;
+        else
+        {
+          for (int iv = 0; iv < var_count; iv++, offset += segment_size)
+            f_prolongmix3(DIM, dst->data->Bg->bbox, dst->data->Bg->bbox + dim, dst->data->Bg->shape,
+                          dst->data->Bg->fgfs[dst_sgfn[iv]], src->data->llb, src->data->uub,
+                          src->data->shape, data + offset, dst->data->llb, dst->data->uub,
+                          src_soa[iv], Symmetry, dst->data->illb, dst->data->iuub);
         }
       }
+
+      size_out = offset + ((!data) ? segment_size * var_count : 0);
+      if (data)
+        size_out = offset;
+    }
     dst = dst->next;
     src = src->next;
   }

AMSS_NCKU_source/bssnEScalar_class.C

@@ -133,6 +133,9 @@ void bssnEScalar_class::Initialize()
     } 
    
   GH = new cgh(0, ngfs, Symmetry, pname, checkrun, ErrorMonitor);
+  ConstraintRefreshLevels = new int[GH->levels];
+  for (int il = 0; il < GH->levels; il++)
+    ConstraintRefreshLevels[il] = 0;
   if (checkrun)
     CheckPoint->readcheck_cgh(PhysTime, GH, myrank, nprocs, Symmetry);
   else
@@ -165,6 +168,14 @@ void bssnEScalar_class::Initialize()
     PhysTime = StartTime;
     Setup_Black_Hole_position();
   }
+
+  // BSSN-EScalar currently uses the uncached communication fallback paths.
+  sync_cache_pre = 0;
+  sync_cache_cor = 0;
+  sync_cache_rp_coarse = 0;
+  sync_cache_rp_fine = 0;
+  sync_cache_restrict = 0;
+  sync_cache_outbd = 0;
 }
 
 //================================================================================================
@@ -230,6 +241,9 @@ void bssnEScalar_class::Read_Ansorg()
     }
     int BH_NM;
     double *Porg_here;
+    double *pmom_local;
+    double *spin_local;
+    double *mass_local;
     // read parameter from file
     {
       const int LEN = 256;
@@ -271,9 +285,9 @@ void bssnEScalar_class::Read_Ansorg()
     }
 
     Porg_here = new double[3 * BH_NM];
-    Pmom = new double[3 * BH_NM];
-    Spin = new double[3 * BH_NM];
-    Mass = new double[BH_NM];
+    pmom_local = new double[3 * BH_NM];
+    spin_local = new double[3 * BH_NM];
+    mass_local = new double[BH_NM];
     // read parameter from file
     {
       const int LEN = 256;
@@ -308,7 +322,7 @@ void bssnEScalar_class::Read_Ansorg()
         if (sgrp == "BSSN" && sind < BH_NM)
         {
           if (skey == "Mass")
-            Mass[sind] = atof(sval.c_str());
+            mass_local[sind] = atof(sval.c_str());
           else if (skey == "Porgx")
             Porg_here[sind * 3] = atof(sval.c_str());
           else if (skey == "Porgy")
@@ -316,17 +330,17 @@ void bssnEScalar_class::Read_Ansorg()
           else if (skey == "Porgz")
             Porg_here[sind * 3 + 2] = atof(sval.c_str());
           else if (skey == "Spinx")
-            Spin[sind * 3] = atof(sval.c_str());
+            spin_local[sind * 3] = atof(sval.c_str());
           else if (skey == "Spiny")
-            Spin[sind * 3 + 1] = atof(sval.c_str());
+            spin_local[sind * 3 + 1] = atof(sval.c_str());
           else if (skey == "Spinz")
-            Spin[sind * 3 + 2] = atof(sval.c_str());
+            spin_local[sind * 3 + 2] = atof(sval.c_str());
           else if (skey == "Pmomx")
-            Pmom[sind * 3] = atof(sval.c_str());
+            pmom_local[sind * 3] = atof(sval.c_str());
           else if (skey == "Pmomy")
-            Pmom[sind * 3 + 1] = atof(sval.c_str());
+            pmom_local[sind * 3 + 1] = atof(sval.c_str());
           else if (skey == "Pmomz")
-            Pmom[sind * 3 + 2] = atof(sval.c_str());
+            pmom_local[sind * 3 + 2] = atof(sval.c_str());
         }
       }
       inf.close();
@@ -362,7 +376,7 @@ void bssnEScalar_class::Read_Ansorg()
                                       cg->fgfs[Sfx0->sgfn], cg->fgfs[Sfy0->sgfn], cg->fgfs[Sfz0->sgfn],
                                       cg->fgfs[dtSfx0->sgfn], cg->fgfs[dtSfy0->sgfn], cg->fgfs[dtSfz0->sgfn],
                                       cg->fgfs[Sphi0->sgfn], cg->fgfs[Spi0->sgfn],
-                                      Mass, Porg_here, Pmom, Spin, BH_NM);
+                                      mass_local, Porg_here, pmom_local, spin_local, BH_NM);
           }
           if (BL == Pp->data->ble)
             break;
@@ -404,7 +418,7 @@ void bssnEScalar_class::Read_Ansorg()
                                        cg->fgfs[Sfx0->sgfn], cg->fgfs[Sfy0->sgfn], cg->fgfs[Sfz0->sgfn],
                                        cg->fgfs[dtSfx0->sgfn], cg->fgfs[dtSfy0->sgfn], cg->fgfs[dtSfz0->sgfn],
                                        cg->fgfs[Sphi0->sgfn], cg->fgfs[Spi0->sgfn],
-                                       Mass, Porg_here, Pmom, Spin, BH_NM);
+                                       mass_local, Porg_here, pmom_local, spin_local, BH_NM);
         }
         if (BL == Pp->data->ble)
           break;
@@ -415,6 +429,9 @@ void bssnEScalar_class::Read_Ansorg()
 #endif
 
     delete[] Porg_here;
+    delete[] pmom_local;
+    delete[] spin_local;
+    delete[] mass_local;
     // dump read_in initial data
     //   for(int lev=0;lev<GH->levels;lev++) Parallel::Dump_Data(GH->PatL[lev],StateList,0,PhysTime,dT);
   }
@@ -455,6 +472,9 @@ void bssnEScalar_class::Read_Pablo()
     }
     int BH_NM;
     double *Porg_here;
+    double *pmom_local;
+    double *spin_local;
+    double *mass_local;
     // read parameter from file
     {
       const int LEN = 256;
@@ -496,9 +516,9 @@ void bssnEScalar_class::Read_Pablo()
     }
 
     Porg_here = new double[3 * BH_NM];
-    Pmom = new double[3 * BH_NM];
-    Spin = new double[3 * BH_NM];
-    Mass = new double[BH_NM];
+    pmom_local = new double[3 * BH_NM];
+    spin_local = new double[3 * BH_NM];
+    mass_local = new double[BH_NM];
     // read parameter from file
     {
       const int LEN = 256;
@@ -533,7 +553,7 @@ void bssnEScalar_class::Read_Pablo()
         if (sgrp == "BSSN" && sind < BH_NM)
         {
           if (skey == "Mass")
-            Mass[sind] = atof(sval.c_str());
+            mass_local[sind] = atof(sval.c_str());
           else if (skey == "Porgx")
             Porg_here[sind * 3] = atof(sval.c_str());
           else if (skey == "Porgy")
@@ -541,17 +561,17 @@ void bssnEScalar_class::Read_Pablo()
           else if (skey == "Porgz")
             Porg_here[sind * 3 + 2] = atof(sval.c_str());
           else if (skey == "Spinx")
-            Spin[sind * 3] = atof(sval.c_str());
+            spin_local[sind * 3] = atof(sval.c_str());
           else if (skey == "Spiny")
-            Spin[sind * 3 + 1] = atof(sval.c_str());
+            spin_local[sind * 3 + 1] = atof(sval.c_str());
           else if (skey == "Spinz")
-            Spin[sind * 3 + 2] = atof(sval.c_str());
+            spin_local[sind * 3 + 2] = atof(sval.c_str());
           else if (skey == "Pmomx")
-            Pmom[sind * 3] = atof(sval.c_str());
+            pmom_local[sind * 3] = atof(sval.c_str());
           else if (skey == "Pmomy")
-            Pmom[sind * 3 + 1] = atof(sval.c_str());
+            pmom_local[sind * 3 + 1] = atof(sval.c_str());
           else if (skey == "Pmomz")
-            Pmom[sind * 3 + 2] = atof(sval.c_str());
+            pmom_local[sind * 3 + 2] = atof(sval.c_str());
         }
       }
       inf.close();
@@ -598,7 +618,7 @@ void bssnEScalar_class::Read_Pablo()
                                         cg->fgfs[Sfx0->sgfn], cg->fgfs[Sfy0->sgfn], cg->fgfs[Sfz0->sgfn],
                                         cg->fgfs[dtSfx0->sgfn], cg->fgfs[dtSfy0->sgfn], cg->fgfs[dtSfz0->sgfn],
                                         cg->fgfs[Sphi0->sgfn], cg->fgfs[Spi0->sgfn],
-                                        Mass, Porg_here, Pmom, Spin, BH_NM);
+                                        mass_local, Porg_here, pmom_local, spin_local, BH_NM);
             }
             if (BL == Pp->data->ble)
               break;
@@ -662,7 +682,7 @@ void bssnEScalar_class::Read_Pablo()
                                          cg->fgfs[Sfx0->sgfn], cg->fgfs[Sfy0->sgfn], cg->fgfs[Sfz0->sgfn],
                                          cg->fgfs[dtSfx0->sgfn], cg->fgfs[dtSfy0->sgfn], cg->fgfs[dtSfz0->sgfn],
                                          cg->fgfs[Sphi0->sgfn], cg->fgfs[Spi0->sgfn],
-                                         Mass, Porg_here, Pmom, Spin, BH_NM);
+                                         mass_local, Porg_here, pmom_local, spin_local, BH_NM);
           }
           if (BL == Pp->data->ble)
             break;
@@ -686,6 +706,9 @@ void bssnEScalar_class::Read_Pablo()
 #endif
 
     delete[] Porg_here;
+    delete[] pmom_local;
+    delete[] spin_local;
+    delete[] mass_local;
     if (flag && myrank == 0)
       MPI_Abort(MPI_COMM_WORLD, 1);
     // dump read_in initial data
@@ -739,7 +762,7 @@ void bssnEScalar_class::Step(int lev, int YN)
                      cg->fgfs[Ayy0->sgfn], cg->fgfs[Ayz0->sgfn], cg->fgfs[Azz0->sgfn]);
 #endif
 
-        if (f_compute_rhs_bssn_escalar(cg->shape, TRK4, cg->X[0], cg->X[1], cg->X[2],
+        if (f_compute_rhs_bssn_escalar_c(cg->shape, TRK4, cg->X[0], cg->X[1], cg->X[2],
                                        cg->fgfs[phi0->sgfn], cg->fgfs[trK0->sgfn],
                                        cg->fgfs[gxx0->sgfn], cg->fgfs[gxy0->sgfn], cg->fgfs[gxz0->sgfn], 
                                        cg->fgfs[gyy0->sgfn], cg->fgfs[gyz0->sgfn], cg->fgfs[gzz0->sgfn],
@@ -1081,7 +1104,7 @@ void bssnEScalar_class::Step(int lev, int YN)
                          cg->fgfs[Ayy->sgfn], cg->fgfs[Ayz->sgfn], cg->fgfs[Azz->sgfn]);
 #endif
 
-          if (f_compute_rhs_bssn_escalar(cg->shape, TRK4, cg->X[0], cg->X[1], cg->X[2],
+          if (f_compute_rhs_bssn_escalar_c(cg->shape, TRK4, cg->X[0], cg->X[1], cg->X[2],
                                          cg->fgfs[phi->sgfn], cg->fgfs[trK->sgfn],
                                          cg->fgfs[gxx->sgfn], cg->fgfs[gxy->sgfn], cg->fgfs[gxz->sgfn], 
                                          cg->fgfs[gyy->sgfn], cg->fgfs[gyz->sgfn], cg->fgfs[gzz->sgfn],
@@ -1858,7 +1881,7 @@ void bssnEScalar_class::Interp_Constraint()
           if (myrank == cg->rank)
           {
             if (lev > 0)
-              f_compute_rhs_bssn_escalar(cg->shape, TRK4, cg->X[0], cg->X[1], cg->X[2],
+              f_compute_rhs_bssn_escalar_c(cg->shape, TRK4, cg->X[0], cg->X[1], cg->X[2],
                                          cg->fgfs[phi0->sgfn], cg->fgfs[trK0->sgfn],
                                          cg->fgfs[gxx0->sgfn], cg->fgfs[gxy0->sgfn], cg->fgfs[gxz0->sgfn], 
                                          cg->fgfs[gyy0->sgfn], cg->fgfs[gyz0->sgfn], cg->fgfs[gzz0->sgfn],
@@ -2078,7 +2101,7 @@ void bssnEScalar_class::Constraint_Out()
             if (myrank == cg->rank)
             {
               if (lev > 0)
-                f_compute_rhs_bssn_escalar(cg->shape, TRK4, cg->X[0], cg->X[1], cg->X[2],
+                f_compute_rhs_bssn_escalar_c(cg->shape, TRK4, cg->X[0], cg->X[1], cg->X[2],
                                            cg->fgfs[phi0->sgfn], cg->fgfs[trK0->sgfn],
                                            cg->fgfs[gxx0->sgfn], cg->fgfs[gxy0->sgfn], cg->fgfs[gxz0->sgfn], 
                                            cg->fgfs[gyy0->sgfn], cg->fgfs[gyz0->sgfn], cg->fgfs[gzz0->sgfn],

AMSS_NCKU_source/bssn_class.C

@@ -299,6 +299,28 @@ bssn_class::bssn_class(double Couranti, double StartTimei, double TotalTimei,
   MPI_Comm_size(MPI_COMM_WORLD, &nprocs);
   MPI_Comm_rank(MPI_COMM_WORLD, &myrank);
 
+  // Derived classes override Initialize(), so ownership-sensitive members must
+  // be in a known state before any specialized setup path runs.
+  GH = 0;
+  SH = 0;
+  PhysTime = 0.0;
+  BH_num = 0;
+  BH_num_input = 0;
+  Porg0 = 0;
+  Porgbr = 0;
+  Porg = 0;
+  Porg1 = 0;
+  Porg_rhs = 0;
+  Mass = 0;
+  Pmom = 0;
+  Spin = 0;
+  sync_cache_pre = 0;
+  sync_cache_cor = 0;
+  sync_cache_rp_coarse = 0;
+  sync_cache_rp_fine = 0;
+  sync_cache_restrict = 0;
+  sync_cache_outbd = 0;
+
   // setup Monitors
   {
     stringstream a_stream;
@@ -986,13 +1008,21 @@ void bssn_class::Initialize()
     Setup_Black_Hole_position();
   }
 
-  // Initialize sync caches (per-level, for predictor and corrector)
+  // BSSN-EScalar uses the uncached communication fallback paths.
+  sync_cache_pre = 0;
+  sync_cache_cor = 0;
+  sync_cache_rp_coarse = 0;
+  sync_cache_rp_fine = 0;
+  sync_cache_restrict = 0;
+  sync_cache_outbd = 0;
+#if (ABEtype != 1)
   sync_cache_pre = new Parallel::SyncCache[GH->levels];
   sync_cache_cor = new Parallel::SyncCache[GH->levels];
   sync_cache_rp_coarse = new Parallel::SyncCache[GH->levels];
   sync_cache_rp_fine = new Parallel::SyncCache[GH->levels];
   sync_cache_restrict = new Parallel::SyncCache[GH->levels];
   sync_cache_outbd = new Parallel::SyncCache[GH->levels];
+#endif
 }
 
 //================================================================================================
@@ -1007,6 +1037,13 @@ void bssn_class::Initialize()
 
 bssn_class::~bssn_class()
 {
+#if (ABEtype == 1)
+  if (myrank == 0)
+  {
+    cout << "[dtor] begin" << endl;
+    cout.flush();
+  }
+#endif
 #ifdef With_AHF
   AHList->clearList();
   AHDList->clearList();
@@ -1041,6 +1078,13 @@ bssn_class::~bssn_class()
   ConstraintList->clearList();
 
   delete[] ConstraintRefreshLevels;
+#if (ABEtype == 1)
+  if (myrank == 0)
+  {
+    cout << "[dtor] lists cleared" << endl;
+    cout.flush();
+  }
+#endif
 
   delete phio;
   delete trKo;
@@ -1218,6 +1262,13 @@ bssn_class::~bssn_class()
   delete Cons_Gx;
   delete Cons_Gy;
   delete Cons_Gz;
+#if (ABEtype == 1)
+  if (myrank == 0)
+  {
+    cout << "[dtor] core vars freed" << endl;
+    cout.flush();
+  }
+#endif
 
 #ifdef Point_Psi4
   delete phix;
@@ -1249,33 +1300,71 @@ bssn_class::~bssn_class()
   // Destroy sync caches before GH
   if (sync_cache_pre)
   {
+#if (ABEtype != 1)
     for (int i = 0; i < GH->levels; i++)
       sync_cache_pre[i].destroy();
+#endif
     delete[] sync_cache_pre;
   }
   if (sync_cache_cor)
   {
+#if (ABEtype != 1)
     for (int i = 0; i < GH->levels; i++)
       sync_cache_cor[i].destroy();
+#endif
     delete[] sync_cache_cor;
   }
   if (sync_cache_rp_coarse)
   {
+#if (ABEtype != 1)
     for (int i = 0; i < GH->levels; i++)
       sync_cache_rp_coarse[i].destroy();
+#endif
     delete[] sync_cache_rp_coarse;
   }
   if (sync_cache_rp_fine)
   {
+#if (ABEtype != 1)
     for (int i = 0; i < GH->levels; i++)
       sync_cache_rp_fine[i].destroy();
+#endif
     delete[] sync_cache_rp_fine;
   }
+  if (sync_cache_restrict)
+  {
+#if (ABEtype != 1)
+    for (int i = 0; i < GH->levels; i++)
+      sync_cache_restrict[i].destroy();
+#endif
+    delete[] sync_cache_restrict;
+  }
+  if (sync_cache_outbd)
+  {
+#if (ABEtype != 1)
+    for (int i = 0; i < GH->levels; i++)
+      sync_cache_outbd[i].destroy();
+#endif
+    delete[] sync_cache_outbd;
+  }
+#if (ABEtype == 1)
+  if (myrank == 0)
+  {
+    cout << "[dtor] caches freed" << endl;
+    cout.flush();
+  }
+#endif
 
   delete GH;
 #ifdef WithShell
   delete SH;
 #endif
+#if (ABEtype == 1)
+  if (myrank == 0)
+  {
+    cout << "[dtor] grids freed" << endl;
+    cout.flush();
+  }
+#endif
 
   for (int i = 0; i < BH_num; i++)
   {
@@ -1295,6 +1384,13 @@ bssn_class::~bssn_class()
   delete[] Mass;
   delete[] Spin;
   delete[] Pmom;
+#if (ABEtype == 1)
+  if (myrank == 0)
+  {
+    cout << "[dtor] puncture arrays freed" << endl;
+    cout.flush();
+  }
+#endif
 
   delete ErrorMonitor;
   delete Psi4Monitor;
@@ -1303,8 +1399,22 @@ bssn_class::~bssn_class()
   delete ConVMonitor;
   delete TimingMonitor;
   delete Waveshell;
+#if (ABEtype == 1)
+  if (myrank == 0)
+  {
+    cout << "[dtor] monitors freed" << endl;
+    cout.flush();
+  }
+#endif
 
   delete CheckPoint;
+#if (ABEtype == 1)
+  if (myrank == 0)
+  {
+    cout << "[dtor] checkpoint freed" << endl;
+    cout.flush();
+  }
+#endif
 }
 
 //================================================================================================

AMSS_NCKU_source/bssn_escalar_rhs_c.C

@@ -0,0 +1,169 @@
+#include "macrodef.h"
+#include "bssn_rhs.h"
+#include "share_func.h"
+#include "tool.h"
+#include <vector>
+
+namespace
+{
+    // Reuse the temporary workspace across block calls to avoid repeated heap churn
+    // in the EScalar wrapper. MPI ranks execute this path sequentially, so a single
+    // process-local buffer is sufficient here.
+    std::vector<double> g_escalar_tmp_store;
+}
+
+#ifdef fortran1
+#define f_frpotential frpotential
+#endif
+#ifdef fortran2
+#define f_frpotential FRPOTENTIAL
+#endif
+#ifdef fortran3
+#define f_frpotential frpotential_
+#endif
+
+extern "C"
+{
+    void f_frpotential(int *, double *, double *, double *);
+}
+
+int f_compute_rhs_bssn_escalar_c(int *ex, double &T,
+                                 double *X, double *Y, double *Z,
+                                 double *chi, double *trK,
+                                 double *dxx, double *gxy, double *gxz, double *dyy, double *gyz, double *dzz,
+                                 double *Axx, double *Axy, double *Axz, double *Ayy, double *Ayz, double *Azz,
+                                 double *Gamx, double *Gamy, double *Gamz,
+                                 double *Lap, double *betax, double *betay, double *betaz,
+                                 double *dtSfx, double *dtSfy, double *dtSfz,
+                                 double *Sphi, double *Spi,
+                                 double *chi_rhs, double *trK_rhs,
+                                 double *gxx_rhs, double *gxy_rhs, double *gxz_rhs, double *gyy_rhs, double *gyz_rhs, double *gzz_rhs,
+                                 double *Axx_rhs, double *Axy_rhs, double *Axz_rhs, double *Ayy_rhs, double *Ayz_rhs, double *Azz_rhs,
+                                 double *Gamx_rhs, double *Gamy_rhs, double *Gamz_rhs,
+                                 double *Lap_rhs, double *betax_rhs, double *betay_rhs, double *betaz_rhs,
+                                 double *dtSfx_rhs, double *dtSfy_rhs, double *dtSfz_rhs,
+                                 double *Sphi_rhs, double *Spi_rhs,
+                                 double *rho, double *Sx, double *Sy, double *Sz,
+                                 double *Sxx, double *Sxy, double *Sxz, double *Syy, double *Syz, double *Szz,
+                                 double *Gamxxx, double *Gamxxy, double *Gamxxz, double *Gamxyy, double *Gamxyz, double *Gamxzz,
+                                 double *Gamyxx, double *Gamyxy, double *Gamyxz, double *Gamyyy, double *Gamyyz, double *Gamyzz,
+                                 double *Gamzxx, double *Gamzxy, double *Gamzxz, double *Gamzyy, double *Gamzyz, double *Gamzzz,
+                                 double *Rxx, double *Rxy, double *Rxz, double *Ryy, double *Ryz, double *Rzz,
+                                 double *ham_Res, double *movx_Res, double *movy_Res, double *movz_Res,
+                                 double *Gmx_Res, double *Gmy_Res, double *Gmz_Res,
+                                 int &Symmetry, int &Lev, double &eps, int &co)
+{
+    const int nx = ex[0], ny = ex[1], nz = ex[2];
+    const int all = nx * ny * nz;
+
+    const size_t workspace_size = size_t(all) * 17;
+    if (g_escalar_tmp_store.size() < workspace_size)
+        g_escalar_tmp_store.resize(workspace_size);
+
+    double *tmp_ptr = g_escalar_tmp_store.data();
+    auto alloc_tmp = [&](int n = 1) -> double *
+    {
+        double *ptr = tmp_ptr;
+        tmp_ptr += size_t(all) * n;
+        return ptr;
+    };
+
+    double *chix = alloc_tmp(), *chiy = alloc_tmp(), *chiz = alloc_tmp();
+    double *Kx = alloc_tmp(), *Ky = alloc_tmp(), *Kz = alloc_tmp();
+    double *fxx = alloc_tmp(), *fxy = alloc_tmp(), *fxz = alloc_tmp();
+    double *fyy = alloc_tmp(), *fyz = alloc_tmp(), *fzz = alloc_tmp();
+    double *Lapx = alloc_tmp(), *Lapy = alloc_tmp(), *Lapz = alloc_tmp();
+    double *V = alloc_tmp(), *dVdSphi = alloc_tmp();
+
+    const double ZEO = 0.0, ONE = 1.0, TWO = 2.0, HALF = 0.5;
+    const double SSS[3] = {1.0, 1.0, 1.0};
+
+    fderivs(ex, chi, chix, chiy, chiz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
+    fderivs(ex, Lap, Lapx, Lapy, Lapz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
+    fderivs(ex, Sphi, Kx, Ky, Kz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
+    fdderivs(ex, Sphi, fxx, fxy, fxz, fyy, fyz, fzz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
+
+    f_frpotential(ex, Sphi, V, dVdSphi);
+
+    for (int i = 0; i < all; ++i)
+    {
+        const double alpn1 = Lap[i] + ONE;
+        const double chin1 = chi[i] + ONE;
+        const double gxx = dxx[i] + ONE;
+        const double gyy = dyy[i] + ONE;
+        const double gzz = dzz[i] + ONE;
+        const double det = gxx * gyy * gzz + gxy[i] * gyz[i] * gxz[i] + gxz[i] * gxy[i] * gyz[i]
+                         - gxz[i] * gyy * gxz[i] - gxy[i] * gxy[i] * gzz - gxx * gyz[i] * gyz[i];
+        const double gupxx = (gyy * gzz - gyz[i] * gyz[i]) / det;
+        const double gupxy = -(gxy[i] * gzz - gyz[i] * gxz[i]) / det;
+        const double gupxz = (gxy[i] * gyz[i] - gyy * gxz[i]) / det;
+        const double gupyy = (gxx * gzz - gxz[i] * gxz[i]) / det;
+        const double gupyz = -(gxx * gyz[i] - gxy[i] * gxz[i]) / det;
+        const double gupzz = (gxx * gyy - gxy[i] * gxy[i]) / det;
+
+        Sphi_rhs[i] = alpn1 * Spi[i];
+
+        Spi_rhs[i] = gupxx * fxx[i] + gupyy * fyy[i] + gupzz * fzz[i]
+                   + TWO * (gupxy * fxy[i] + gupxz * fxz[i] + gupyz * fyz[i])
+                   - ((Gamx[i] + (gupxx * chix[i] + gupxy * chiy[i] + gupxz * chiz[i]) / TWO / chin1) * Kx[i]
+                   +  (Gamy[i] + (gupxy * chix[i] + gupyy * chiy[i] + gupyz * chiz[i]) / TWO / chin1) * Ky[i]
+                   +  (Gamz[i] + (gupxz * chix[i] + gupyz * chiy[i] + gupzz * chiz[i]) / TWO / chin1) * Kz[i]);
+
+        Spi_rhs[i] = Spi_rhs[i] * alpn1
+                   + gupxx * Lapx[i] * Kx[i] + gupxy * Lapx[i] * Ky[i] + gupxz * Lapx[i] * Kz[i]
+                   + gupxy * Lapy[i] * Kx[i] + gupyy * Lapy[i] * Ky[i] + gupyz * Lapy[i] * Kz[i]
+                   + gupxz * Lapz[i] * Kx[i] + gupyz * Lapz[i] * Ky[i] + gupzz * Lapz[i] * Kz[i];
+
+        Spi_rhs[i] = Spi_rhs[i] * chin1 + alpn1 * (trK[i] * Spi[i] - dVdSphi[i]);
+
+        rho[i] = chin1 * ((gupxx * Kx[i] * Kx[i] + gupyy * Ky[i] * Ky[i] + gupzz * Kz[i] * Kz[i]) * HALF
+               + gupxy * Kx[i] * Ky[i] + gupxz * Kx[i] * Kz[i] + gupyz * Ky[i] * Kz[i])
+               + Spi[i] * Spi[i] * HALF + V[i];
+        Sx[i] = -Spi[i] * Kx[i];
+        Sy[i] = -Spi[i] * Ky[i];
+        Sz[i] = -Spi[i] * Kz[i];
+
+        const double pressure = (rho[i] - Spi[i] * Spi[i]) / chin1;
+        Sxx[i] = Kx[i] * Kx[i] - pressure * gxx;
+        Sxy[i] = Kx[i] * Ky[i] - pressure * gxy[i];
+        Sxz[i] = Kx[i] * Kz[i] - pressure * gxz[i];
+        Syy[i] = Ky[i] * Ky[i] - pressure * gyy;
+        Syz[i] = Ky[i] * Kz[i] - pressure * gyz[i];
+        Szz[i] = Kz[i] * Kz[i] - pressure * gzz;
+    }
+
+    if (f_compute_rhs_bssn(ex, T, X, Y, Z,
+                           chi, trK,
+                           dxx, gxy, gxz, dyy, gyz, dzz,
+                           Axx, Axy, Axz, Ayy, Ayz, Azz,
+                           Gamx, Gamy, Gamz,
+                           Lap, betax, betay, betaz,
+                           dtSfx, dtSfy, dtSfz,
+                           chi_rhs, trK_rhs,
+                           gxx_rhs, gxy_rhs, gxz_rhs, gyy_rhs, gyz_rhs, gzz_rhs,
+                           Axx_rhs, Axy_rhs, Axz_rhs, Ayy_rhs, Ayz_rhs, Azz_rhs,
+                           Gamx_rhs, Gamy_rhs, Gamz_rhs,
+                           Lap_rhs, betax_rhs, betay_rhs, betaz_rhs,
+                           dtSfx_rhs, dtSfy_rhs, dtSfz_rhs,
+                           rho, Sx, Sy, Sz,
+                           Sxx, Sxy, Sxz, Syy, Syz, Szz,
+                           Gamxxx, Gamxxy, Gamxxz, Gamxyy, Gamxyz, Gamxzz,
+                           Gamyxx, Gamyxy, Gamyxz, Gamyyy, Gamyyz, Gamyzz,
+                           Gamzxx, Gamzxy, Gamzxz, Gamzyy, Gamzyz, Gamzzz,
+                           Rxx, Rxy, Rxz, Ryy, Ryz, Rzz,
+                           ham_Res, movx_Res, movy_Res, movz_Res,
+                           Gmx_Res, Gmy_Res, Gmz_Res,
+                           Symmetry, Lev, eps, co))
+        return 1;
+
+    lopsided_kodis(ex, X, Y, Z, Sphi, Sphi_rhs, betax, betay, betaz, Symmetry, SSS, eps);
+    lopsided_kodis(ex, X, Y, Z, Spi, Spi_rhs, betax, betay, betaz, Symmetry, SSS, eps);
+
+    for (int i = 0; i < all; ++i)
+    {
+        if (Sphi_rhs[i] != Sphi_rhs[i] || Spi_rhs[i] != Spi_rhs[i] || rho[i] != rho[i])
+            return 1;
+    }
+
+    return 0;
+}

AMSS_NCKU_source/bssn_rhs.h

@@ -67,6 +67,27 @@ extern "C"
                                int &, int &, double &, int &);
 }
 
+int f_compute_rhs_bssn_escalar_c(int *, double &, double *, double *, double *,                                                      // ex,T,X,Y,Z
+                                 double *, double *,                                                                                 // chi, trK
+                                 double *, double *, double *, double *, double *, double *,                                         // gij
+                                 double *, double *, double *, double *, double *, double *,                                         // Aij
+                                 double *, double *, double *,                                                                       // Gam
+                                 double *, double *, double *, double *, double *, double *, double *,                               // Gauge
+                                 double *, double *,                                                                                 // Sphi, Spi
+                                 double *, double *,                                                                                 // chi, trK
+                                 double *, double *, double *, double *, double *, double *,                                         // gij
+                                 double *, double *, double *, double *, double *, double *,                                         // Aij
+                                 double *, double *, double *,                                                                       // Gam
+                                 double *, double *, double *, double *, double *, double *, double *,                               // Gauge
+                                 double *, double *,                                                                                 // Sphi, Spi
+                                 double *, double *, double *, double *, double *, double *, double *, double *, double *, double *, // stress-energy
+                                 double *, double *, double *, double *, double *, double *,                                         // Christoffel
+                                 double *, double *, double *, double *, double *, double *,                                         // Christoffel
+                                 double *, double *, double *, double *, double *, double *,                                         // Christoffel
+                                 double *, double *, double *, double *, double *, double *,                                         // Ricci
+                                 double *, double *, double *, double *, double *, double *, double *,                               // constraint violation
+                                 int &, int &, double &, int &);
+
 extern "C"
 {
         int f_compute_rhs_bssn_ss(int *, double &, double *, double *, double *,                                                      // ex,T,rho,sigma,R

AMSS_NCKU_source/makefile

@@ -86,8 +86,8 @@ ifeq ($(USE_CXX_KERNELS),0)
 # Fortran mode: no C rewrite files; bssn_rhs.o is included via F90FILES below
 CFILES =
 else
-# C++ mode (default): C rewrite of bssn_rhs and helper kernels
-CFILES = bssn_rhs_c.o fderivs_c.o fdderivs_c.o kodiss_c.o lopsided_c.o lopsided_kodis_c.o
+# C++ mode (default): C rewrite of bssn/bssn-escalar rhs and helper kernels
+CFILES = bssn_rhs_c.o bssn_escalar_rhs_c.o fderivs_c.o fdderivs_c.o kodiss_c.o lopsided_c.o lopsided_kodis_c.o
 endif
 
 ## RK4 kernel switch (independent from USE_CXX_KERNELS)