Merge pull request #4482 from rbberger/dihedral_multi_harmonic_kokkos

add dihedral multi/harmonic/kk

doc/src/Commands_bond.rst

@@ -128,7 +128,7 @@ OPT.
    * :doc:`harmonic (iko) <dihedral_harmonic>`
    * :doc:`helix (o) <dihedral_helix>`
    * :doc:`lepton (o) <dihedral_lepton>`
-   * :doc:`multi/harmonic (o) <dihedral_multi_harmonic>`
+   * :doc:`multi/harmonic (ko) <dihedral_multi_harmonic>`
    * :doc:`nharmonic (o) <dihedral_nharmonic>`
    * :doc:`opls (iko) <dihedral_opls>`
    * :doc:`quadratic (o) <dihedral_quadratic>`

doc/src/dihedral_multi_harmonic.rst

@@ -4,7 +4,7 @@
 dihedral_style multi/harmonic command
 =====================================
 
-Accelerator Variants: *multi/harmonic/omp*
+Accelerator Variants: *multi/harmonic/kk*, *multi/harmonic/omp*
 
 Syntax
 """"""

src/KOKKOS/dihedral_multi_harmonic_kokkos.cpp

@@ -0,0 +1,541 @@
+// clang-format off
+/* ----------------------------------------------------------------------
+   LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
+   https://www.lammps.org/, Sandia National Laboratories
+   LAMMPS development team: developers@lammps.org
+
+   Copyright (2003) Sandia Corporation.  Under the terms of Contract
+   DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains
+   certain rights in this software.  This software is distributed under
+   the GNU General Public License.
+
+   See the README file in the top-level LAMMPS directory.
+------------------------------------------------------------------------- */
+
+/* ----------------------------------------------------------------------
+   Contributing author: Richard Berger (LANL)
+------------------------------------------------------------------------- */
+
+#include "dihedral_multi_harmonic_kokkos.h"
+
+#include "atom_kokkos.h"
+#include "atom_masks.h"
+#include "comm.h"
+#include "error.h"
+#include "force.h"
+#include "memory_kokkos.h"
+#include "neighbor_kokkos.h"
+
+#include <cmath>
+
+using namespace LAMMPS_NS;
+
+static constexpr double TOLERANCE = 0.05;
+static constexpr double SMALL = 0.001;
+
+/* ---------------------------------------------------------------------- */
+
+template<class DeviceType>
+DihedralMultiHarmonicKokkos<DeviceType>::DihedralMultiHarmonicKokkos(LAMMPS *lmp) : DihedralMultiHarmonic(lmp)
+{
+  kokkosable = 1;
+  atomKK = (AtomKokkos *) atom;
+  neighborKK = (NeighborKokkos *) neighbor;
+  execution_space = ExecutionSpaceFromDevice<DeviceType>::space;
+  datamask_read = X_MASK | F_MASK | Q_MASK | ENERGY_MASK | VIRIAL_MASK;
+  datamask_modify = F_MASK | ENERGY_MASK | VIRIAL_MASK;
+
+  k_warning_flag = DAT::tdual_int_scalar("Dihedral:warning_flag");
+  d_warning_flag = k_warning_flag.view<DeviceType>();
+  h_warning_flag = k_warning_flag.h_view;
+
+  centroidstressflag = CENTROID_NOTAVAIL;
+}
+
+/* ---------------------------------------------------------------------- */
+
+template<class DeviceType>
+DihedralMultiHarmonicKokkos<DeviceType>::~DihedralMultiHarmonicKokkos()
+{
+  if (!copymode) {
+    memoryKK->destroy_kokkos(k_eatom,eatom);
+    memoryKK->destroy_kokkos(k_vatom,vatom);
+  }
+}
+
+/* ---------------------------------------------------------------------- */
+
+template<class DeviceType>
+void DihedralMultiHarmonicKokkos<DeviceType>::compute(int eflag_in, int vflag_in)
+{
+  eflag = eflag_in;
+  vflag = vflag_in;
+
+  ev_init(eflag,vflag,0);
+
+  // reallocate per-atom arrays if necessary
+
+  if (eflag_atom) {
+    if ((int)k_eatom.extent(0) < maxeatom) {
+    memoryKK->destroy_kokkos(k_eatom,eatom);
+    memoryKK->create_kokkos(k_eatom,eatom,maxeatom,"dihedral:eatom");
+    d_eatom = k_eatom.view<DeviceType>();
+    } else Kokkos::deep_copy(d_eatom,0.0);
+  }
+  if (vflag_atom) {
+    if ((int)k_vatom.extent(0) < maxvatom) {
+    memoryKK->destroy_kokkos(k_vatom,vatom);
+    memoryKK->create_kokkos(k_vatom,vatom,maxvatom,"dihedral:vatom");
+    d_vatom = k_vatom.view<DeviceType>();
+    } else Kokkos::deep_copy(d_vatom,0.0);
+  }
+
+  k_a1.template sync<DeviceType>();
+  k_a2.template sync<DeviceType>();
+  k_a3.template sync<DeviceType>();
+  k_a4.template sync<DeviceType>();
+  k_a5.template sync<DeviceType>();
+
+  x = atomKK->k_x.view<DeviceType>();
+  f = atomKK->k_f.view<DeviceType>();
+  neighborKK->k_dihedrallist.template sync<DeviceType>();
+  dihedrallist = neighborKK->k_dihedrallist.view<DeviceType>();
+  int ndihedrallist = neighborKK->ndihedrallist;
+  nlocal = atom->nlocal;
+  newton_bond = force->newton_bond;
+
+  h_warning_flag() = 0;
+  k_warning_flag.modify_host();
+  k_warning_flag.template sync<DeviceType>();
+
+  copymode = 1;
+
+  // loop over neighbors of my atoms
+
+  EV_FLOAT ev;
+
+  if (evflag) {
+    if (newton_bond) {
+      Kokkos::parallel_reduce(Kokkos::RangePolicy<DeviceType, TagDihedralMultiHarmonicCompute<1,1> >(0,ndihedrallist),*this,ev);
+    } else {
+      Kokkos::parallel_reduce(Kokkos::RangePolicy<DeviceType, TagDihedralMultiHarmonicCompute<0,1> >(0,ndihedrallist),*this,ev);
+    }
+  } else {
+    if (newton_bond) {
+      Kokkos::parallel_for(Kokkos::RangePolicy<DeviceType, TagDihedralMultiHarmonicCompute<1,0> >(0,ndihedrallist),*this);
+    } else {
+      Kokkos::parallel_for(Kokkos::RangePolicy<DeviceType, TagDihedralMultiHarmonicCompute<0,0> >(0,ndihedrallist),*this);
+    }
+  }
+
+  // error check
+
+  k_warning_flag.template modify<DeviceType>();
+  k_warning_flag.sync_host();
+  if (h_warning_flag())
+    error->warning(FLERR,"Dihedral problem");
+
+  if (eflag_global) energy += ev.evdwl;
+  if (vflag_global) {
+    virial[0] += ev.v[0];
+    virial[1] += ev.v[1];
+    virial[2] += ev.v[2];
+    virial[3] += ev.v[3];
+    virial[4] += ev.v[4];
+    virial[5] += ev.v[5];
+  }
+
+  if (eflag_atom) {
+    k_eatom.template modify<DeviceType>();
+    k_eatom.sync_host();
+  }
+
+  if (vflag_atom) {
+    k_vatom.template modify<DeviceType>();
+    k_vatom.sync_host();
+  }
+
+  copymode = 0;
+}
+
+template<class DeviceType>
+template<int NEWTON_BOND, int EVFLAG>
+KOKKOS_INLINE_FUNCTION
+void DihedralMultiHarmonicKokkos<DeviceType>::operator()(TagDihedralMultiHarmonicCompute<NEWTON_BOND,EVFLAG>, const int &n, EV_FLOAT& ev) const {
+
+  // The f array is atomic
+  Kokkos::View<F_FLOAT*[3], typename DAT::t_f_array::array_layout,typename KKDevice<DeviceType>::value,Kokkos::MemoryTraits<Kokkos::Atomic|Kokkos::Unmanaged> > a_f = f;
+
+  const int i1 = dihedrallist(n,0);
+  const int i2 = dihedrallist(n,1);
+  const int i3 = dihedrallist(n,2);
+  const int i4 = dihedrallist(n,3);
+  const int type = dihedrallist(n,4);
+
+  // 1st bond
+
+  const F_FLOAT vb1x = x(i1,0) - x(i2,0);
+  const F_FLOAT vb1y = x(i1,1) - x(i2,1);
+  const F_FLOAT vb1z = x(i1,2) - x(i2,2);
+
+  // 2nd bond
+
+  const F_FLOAT vb2x = x(i3,0) - x(i2,0);
+  const F_FLOAT vb2y = x(i3,1) - x(i2,1);
+  const F_FLOAT vb2z = x(i3,2) - x(i2,2);
+
+  const F_FLOAT vb2xm = -vb2x;
+  const F_FLOAT vb2ym = -vb2y;
+  const F_FLOAT vb2zm = -vb2z;
+
+  // 3rd bond
+
+  const F_FLOAT vb3x = x(i4,0) - x(i3,0);
+  const F_FLOAT vb3y = x(i4,1) - x(i3,1);
+  const F_FLOAT vb3z = x(i4,2) - x(i3,2);
+
+  // c0 calculation
+
+  const F_FLOAT sb1 = 1.0 / (vb1x * vb1x + vb1y * vb1y + vb1z * vb1z);
+  const F_FLOAT sb2 = 1.0 / (vb2x * vb2x + vb2y * vb2y + vb2z * vb2z);
+  const F_FLOAT sb3 = 1.0 / (vb3x * vb3x + vb3y * vb3y + vb3z * vb3z);
+
+  const F_FLOAT rb1 = sqrt(sb1);
+  const F_FLOAT rb3 = sqrt(sb3);
+
+  F_FLOAT c0 = (vb1x * vb3x + vb1y * vb3y + vb1z * vb3z) * rb1 * rb3;
+
+  // 1st and 2nd angle
+
+  F_FLOAT b1mag2 = vb1x * vb1x + vb1y * vb1y + vb1z * vb1z;
+  F_FLOAT b1mag = sqrt(b1mag2);
+  F_FLOAT b2mag2 = vb2x * vb2x + vb2y * vb2y + vb2z * vb2z;
+  F_FLOAT b2mag = sqrt(b2mag2);
+  F_FLOAT b3mag2 = vb3x * vb3x + vb3y * vb3y + vb3z * vb3z;
+  F_FLOAT b3mag = sqrt(b3mag2);
+
+  F_FLOAT ctmp = vb1x * vb2x + vb1y * vb2y + vb1z * vb2z;
+  F_FLOAT r12c1 = 1.0 / (b1mag * b2mag);
+  F_FLOAT c1mag = ctmp * r12c1;
+
+  ctmp = vb2xm * vb3x + vb2ym * vb3y + vb2zm * vb3z;
+  F_FLOAT r12c2 = 1.0 / (b2mag * b3mag);
+  F_FLOAT c2mag = ctmp * r12c2;
+
+  // cos and sin of 2 angles and final c
+
+  F_FLOAT sin2 = MAX(1.0 - c1mag * c1mag, 0.0);
+  F_FLOAT sc1 = sqrt(sin2);
+  if (sc1 < SMALL) sc1 = SMALL;
+  sc1 = 1.0 / sc1;
+
+  sin2 = MAX(1.0 - c2mag * c2mag, 0.0);
+  F_FLOAT sc2 = sqrt(sin2);
+  if (sc2 < SMALL) sc2 = SMALL;
+  sc2 = 1.0 / sc2;
+
+  F_FLOAT s1 = sc1 * sc1;
+  F_FLOAT s2 = sc2 * sc2;
+  F_FLOAT s12 = sc1 * sc2;
+  F_FLOAT c = (c0 + c1mag * c2mag) * s12;
+
+  // error check
+
+  if ((c > 1.0 + TOLERANCE || c < (-1.0 - TOLERANCE)) && !d_warning_flag())
+    d_warning_flag() = 1;
+
+  if (c > 1.0) c = 1.0;
+  if (c < -1.0) c = -1.0;
+
+  // force & energy
+  // p = sum (i=1,5) a_i * c**(i-1)
+  // pd = dp/dc
+
+  F_FLOAT p = d_a1[type] + c * (d_a2[type] + c * (d_a3[type] + c * (d_a4[type] + c * d_a5[type])));
+  F_FLOAT pd = d_a2[type] + c * (2.0 * d_a3[type] + c * (3.0 * d_a4[type] + c * 4.0 * d_a5[type]));
+
+  E_FLOAT edihedral = 0.0;
+  if (eflag) edihedral = p;
+
+  const F_FLOAT a = pd;
+  c = c * a;
+  s12 = s12 * a;
+  const F_FLOAT a11 = c * sb1 * s1;
+  const F_FLOAT a22 = -sb2 * (2.0 * c0 * s12 - c * (s1 + s2));
+  const F_FLOAT a33 = c * sb3 * s2;
+  const F_FLOAT a12 = -r12c1 * (c1mag * c * s1 + c2mag * s12);
+  const F_FLOAT a13 = -rb1 * rb3 * s12;
+  const F_FLOAT a23 = r12c2 * (c2mag * c * s2 + c1mag * s12);
+
+  const F_FLOAT sx2 = a12 * vb1x + a22 * vb2x + a23 * vb3x;
+  const F_FLOAT sy2 = a12 * vb1y + a22 * vb2y + a23 * vb3y;
+  const F_FLOAT sz2 = a12 * vb1z + a22 * vb2z + a23 * vb3z;
+
+  F_FLOAT f1[3],f2[3],f3[3],f4[3];
+  f1[0] = a11 * vb1x + a12 * vb2x + a13 * vb3x;
+  f1[1] = a11 * vb1y + a12 * vb2y + a13 * vb3y;
+  f1[2] = a11 * vb1z + a12 * vb2z + a13 * vb3z;
+
+  f2[0] = -sx2 - f1[0];
+  f2[1] = -sy2 - f1[1];
+  f2[2] = -sz2 - f1[2];
+
+  f4[0] = a13 * vb1x + a23 * vb2x + a33 * vb3x;
+  f4[1] = a13 * vb1y + a23 * vb2y + a33 * vb3y;
+  f4[2] = a13 * vb1z + a23 * vb2z + a33 * vb3z;
+
+  f3[0] = sx2 - f4[0];
+  f3[1] = sy2 - f4[1];
+  f3[2] = sz2 - f4[2];
+
+  // apply force to each of 4 atoms
+
+  if (NEWTON_BOND || i1 < nlocal) {
+    a_f(i1,0) += f1[0];
+    a_f(i1,1) += f1[1];
+    a_f(i1,2) += f1[2];
+  }
+
+  if (NEWTON_BOND || i2 < nlocal) {
+    a_f(i2,0) += f2[0];
+    a_f(i2,1) += f2[1];
+    a_f(i2,2) += f2[2];
+  }
+
+  if (NEWTON_BOND || i3 < nlocal) {
+    a_f(i3,0) += f3[0];
+    a_f(i3,1) += f3[1];
+    a_f(i3,2) += f3[2];
+  }
+
+  if (NEWTON_BOND || i4 < nlocal) {
+    a_f(i4,0) += f4[0];
+    a_f(i4,1) += f4[1];
+    a_f(i4,2) += f4[2];
+  }
+
+  if (EVFLAG)
+    ev_tally(ev,i1,i2,i3,i4,edihedral,f1,f3,f4,
+             vb1x,vb1y,vb1z,vb2x,vb2y,vb2z,vb3x,vb3y,vb3z);
+}
+
+template<class DeviceType>
+template<int NEWTON_BOND, int EVFLAG>
+KOKKOS_INLINE_FUNCTION
+void DihedralMultiHarmonicKokkos<DeviceType>::operator()(TagDihedralMultiHarmonicCompute<NEWTON_BOND,EVFLAG>, const int &n) const {
+  EV_FLOAT ev;
+  this->template operator()<NEWTON_BOND,EVFLAG>(TagDihedralMultiHarmonicCompute<NEWTON_BOND,EVFLAG>(), n, ev);
+}
+
+/* ---------------------------------------------------------------------- */
+
+template<class DeviceType>
+void DihedralMultiHarmonicKokkos<DeviceType>::allocate()
+{
+  DihedralMultiHarmonic::allocate();
+
+  int n = atom->ndihedraltypes;
+  k_a1 = DAT::tdual_ffloat_1d("DihedralMultiHarmonic::a1",n+1);
+  k_a2 = DAT::tdual_ffloat_1d("DihedralMultiHarmonic::a2",n+1);
+  k_a3 = DAT::tdual_ffloat_1d("DihedralMultiHarmonic::a3",n+1);
+  k_a4 = DAT::tdual_ffloat_1d("DihedralMultiHarmonic::a4",n+1);
+  k_a5 = DAT::tdual_ffloat_1d("DihedralMultiHarmonic::a5",n+1);
+
+  d_a1 = k_a1.template view<DeviceType>();
+  d_a2 = k_a2.template view<DeviceType>();
+  d_a3 = k_a3.template view<DeviceType>();
+  d_a4 = k_a4.template view<DeviceType>();
+  d_a5 = k_a5.template view<DeviceType>();
+}
+
+/* ----------------------------------------------------------------------
+   set coeffs for one type
+------------------------------------------------------------------------- */
+
+template<class DeviceType>
+void DihedralMultiHarmonicKokkos<DeviceType>::coeff(int narg, char **arg)
+{
+  DihedralMultiHarmonic::coeff(narg, arg);
+
+  int n = atom->ndihedraltypes;
+  for (int i = 1; i <= n; i++) {
+    k_a1.h_view[i] = a1[i];
+    k_a2.h_view[i] = a2[i];
+    k_a3.h_view[i] = a3[i];
+    k_a4.h_view[i] = a4[i];
+    k_a5.h_view[i] = a5[i];
+  }
+
+  k_a1.modify_host();
+  k_a2.modify_host();
+  k_a3.modify_host();
+  k_a4.modify_host();
+  k_a5.modify_host();
+}
+
+/* ----------------------------------------------------------------------
+   proc 0 reads coeffs from restart file, bcasts them
+------------------------------------------------------------------------- */
+
+template<class DeviceType>
+void DihedralMultiHarmonicKokkos<DeviceType>::read_restart(FILE *fp)
+{
+  DihedralMultiHarmonic::read_restart(fp);
+
+  int n = atom->ndihedraltypes;
+  for (int i = 1; i <= n; i++) {
+    k_a1.h_view[i] = a1[i];
+    k_a2.h_view[i] = a2[i];
+    k_a3.h_view[i] = a3[i];
+    k_a4.h_view[i] = a4[i];
+    k_a5.h_view[i] = a5[i];
+  }
+
+  k_a1.modify_host();
+  k_a2.modify_host();
+  k_a3.modify_host();
+  k_a4.modify_host();
+  k_a5.modify_host();
+}
+
+/* ----------------------------------------------------------------------
+   tally energy and virial into global and per-atom accumulators
+   virial = r1F1 + r2F2 + r3F3 + r4F4 = (r1-r2) F1 + (r3-r2) F3 + (r4-r2) F4
+          = (r1-r2) F1 + (r3-r2) F3 + (r4-r3 + r3-r2) F4
+          = vb1*f1 + vb2*f3 + (vb3+vb2)*f4
+------------------------------------------------------------------------- */
+
+template<class DeviceType>
+//template<int NEWTON_BOND>
+KOKKOS_INLINE_FUNCTION
+void DihedralMultiHarmonicKokkos<DeviceType>::ev_tally(EV_FLOAT &ev, const int i1, const int i2, const int i3, const int i4,
+                        F_FLOAT &edihedral, F_FLOAT *f1, F_FLOAT *f3, F_FLOAT *f4,
+                        const F_FLOAT &vb1x, const F_FLOAT &vb1y, const F_FLOAT &vb1z,
+                        const F_FLOAT &vb2x, const F_FLOAT &vb2y, const F_FLOAT &vb2z,
+                        const F_FLOAT &vb3x, const F_FLOAT &vb3y, const F_FLOAT &vb3z) const
+{
+  E_FLOAT edihedralquarter;
+  F_FLOAT v[6];
+
+  // The eatom and vatom arrays are atomic
+  Kokkos::View<E_FLOAT*, typename DAT::t_efloat_1d::array_layout,typename KKDevice<DeviceType>::value,Kokkos::MemoryTraits<Kokkos::Atomic|Kokkos::Unmanaged> > v_eatom = k_eatom.view<DeviceType>();
+  Kokkos::View<F_FLOAT*[6], typename DAT::t_virial_array::array_layout,typename KKDevice<DeviceType>::value,Kokkos::MemoryTraits<Kokkos::Atomic|Kokkos::Unmanaged> > v_vatom = k_vatom.view<DeviceType>();
+
+  if (eflag_either) {
+    if (eflag_global) {
+      if (newton_bond) ev.evdwl += edihedral;
+      else {
+        edihedralquarter = 0.25*edihedral;
+        if (i1 < nlocal) ev.evdwl += edihedralquarter;
+        if (i2 < nlocal) ev.evdwl += edihedralquarter;
+        if (i3 < nlocal) ev.evdwl += edihedralquarter;
+        if (i4 < nlocal) ev.evdwl += edihedralquarter;
+      }
+    }
+    if (eflag_atom) {
+      edihedralquarter = 0.25*edihedral;
+      if (newton_bond || i1 < nlocal) v_eatom[i1] += edihedralquarter;
+      if (newton_bond || i2 < nlocal) v_eatom[i2] += edihedralquarter;
+      if (newton_bond || i3 < nlocal) v_eatom[i3] += edihedralquarter;
+      if (newton_bond || i4 < nlocal) v_eatom[i4] += edihedralquarter;
+    }
+  }
+
+  if (vflag_either) {
+    v[0] = vb1x*f1[0] + vb2x*f3[0] + (vb3x+vb2x)*f4[0];
+    v[1] = vb1y*f1[1] + vb2y*f3[1] + (vb3y+vb2y)*f4[1];
+    v[2] = vb1z*f1[2] + vb2z*f3[2] + (vb3z+vb2z)*f4[2];
+    v[3] = vb1x*f1[1] + vb2x*f3[1] + (vb3x+vb2x)*f4[1];
+    v[4] = vb1x*f1[2] + vb2x*f3[2] + (vb3x+vb2x)*f4[2];
+    v[5] = vb1y*f1[2] + vb2y*f3[2] + (vb3y+vb2y)*f4[2];
+
+    if (vflag_global) {
+      if (newton_bond) {
+        ev.v[0] += v[0];
+        ev.v[1] += v[1];
+        ev.v[2] += v[2];
+        ev.v[3] += v[3];
+        ev.v[4] += v[4];
+        ev.v[5] += v[5];
+      } else {
+        if (i1 < nlocal) {
+          ev.v[0] += 0.25*v[0];
+          ev.v[1] += 0.25*v[1];
+          ev.v[2] += 0.25*v[2];
+          ev.v[3] += 0.25*v[3];
+          ev.v[4] += 0.25*v[4];
+          ev.v[5] += 0.25*v[5];
+        }
+        if (i2 < nlocal) {
+          ev.v[0] += 0.25*v[0];
+          ev.v[1] += 0.25*v[1];
+          ev.v[2] += 0.25*v[2];
+          ev.v[3] += 0.25*v[3];
+          ev.v[4] += 0.25*v[4];
+          ev.v[5] += 0.25*v[5];
+        }
+        if (i3 < nlocal) {
+          ev.v[0] += 0.25*v[0];
+          ev.v[1] += 0.25*v[1];
+          ev.v[2] += 0.25*v[2];
+          ev.v[3] += 0.25*v[3];
+          ev.v[4] += 0.25*v[4];
+          ev.v[5] += 0.25*v[5];
+        }
+        if (i4 < nlocal) {
+          ev.v[0] += 0.25*v[0];
+          ev.v[1] += 0.25*v[1];
+          ev.v[2] += 0.25*v[2];
+          ev.v[3] += 0.25*v[3];
+          ev.v[4] += 0.25*v[4];
+          ev.v[5] += 0.25*v[5];
+        }
+      }
+    }
+
+    if (vflag_atom) {
+      if (newton_bond || i1 < nlocal) {
+        v_vatom(i1,0) += 0.25*v[0];
+        v_vatom(i1,1) += 0.25*v[1];
+        v_vatom(i1,2) += 0.25*v[2];
+        v_vatom(i1,3) += 0.25*v[3];
+        v_vatom(i1,4) += 0.25*v[4];
+        v_vatom(i1,5) += 0.25*v[5];
+      }
+      if (newton_bond || i2 < nlocal) {
+        v_vatom(i2,0) += 0.25*v[0];
+        v_vatom(i2,1) += 0.25*v[1];
+        v_vatom(i2,2) += 0.25*v[2];
+        v_vatom(i2,3) += 0.25*v[3];
+        v_vatom(i2,4) += 0.25*v[4];
+        v_vatom(i2,5) += 0.25*v[5];
+      }
+      if (newton_bond || i3 < nlocal) {
+        v_vatom(i3,0) += 0.25*v[0];
+        v_vatom(i3,1) += 0.25*v[1];
+        v_vatom(i3,2) += 0.25*v[2];
+        v_vatom(i3,3) += 0.25*v[3];
+        v_vatom(i3,4) += 0.25*v[4];
+        v_vatom(i3,5) += 0.25*v[5];
+      }
+      if (newton_bond || i4 < nlocal) {
+        v_vatom(i4,0) += 0.25*v[0];
+        v_vatom(i4,1) += 0.25*v[1];
+        v_vatom(i4,2) += 0.25*v[2];
+        v_vatom(i4,3) += 0.25*v[3];
+        v_vatom(i4,4) += 0.25*v[4];
+        v_vatom(i4,5) += 0.25*v[5];
+      }
+    }
+  }
+}
+
+/* ---------------------------------------------------------------------- */
+
+namespace LAMMPS_NS {
+template class DihedralMultiHarmonicKokkos<LMPDeviceType>;
+#ifdef LMP_KOKKOS_GPU
+template class DihedralMultiHarmonicKokkos<LMPHostType>;
+#endif
+}
+

src/KOKKOS/dihedral_multi_harmonic_kokkos.h

@@ -0,0 +1,102 @@
+/* -*- c++ -*- ----------------------------------------------------------
+   LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
+   https://www.lammps.org/, Sandia National Laboratories
+   LAMMPS development team: developers@lammps.org
+
+   Copyright (2003) Sandia Corporation.  Under the terms of Contract
+   DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains
+   certain rights in this software.  This software is distributed under
+   the GNU General Public License.
+
+   See the README file in the top-level LAMMPS directory.
+------------------------------------------------------------------------- */
+
+#ifdef DIHEDRAL_CLASS
+// clang-format off
+DihedralStyle(multi/harmonic/kk,DihedralMultiHarmonicKokkos<LMPDeviceType>);
+DihedralStyle(multi/harmonic/kk/device,DihedralMultiHarmonicKokkos<LMPDeviceType>);
+DihedralStyle(multi/harmonic/kk/host,DihedralMultiHarmonicKokkos<LMPHostType>);
+// clang-format on
+#else
+
+// clang-format off
+#ifndef LMP_DIHEDRAL_MULTI_HARMONIC_KOKKOS_H
+#define LMP_DIHEDRAL_MULTI_HARMONIC_KOKKOS_H
+
+#include "dihedral_multi_harmonic.h"
+#include "kokkos_type.h"
+
+namespace LAMMPS_NS {
+
+template<int NEWTON_BOND, int EVFLAG>
+struct TagDihedralMultiHarmonicCompute{};
+
+template<class DeviceType>
+class DihedralMultiHarmonicKokkos : public DihedralMultiHarmonic {
+ public:
+  typedef DeviceType device_type;
+  typedef EV_FLOAT value_type;
+  typedef ArrayTypes<DeviceType> AT;
+
+  DihedralMultiHarmonicKokkos(class LAMMPS *);
+  ~DihedralMultiHarmonicKokkos() override;
+  void compute(int, int) override;
+  void coeff(int, char **) override;
+  void read_restart(FILE *) override;
+
+  template<int NEWTON_BOND, int EVFLAG>
+  KOKKOS_INLINE_FUNCTION
+  void operator()(TagDihedralMultiHarmonicCompute<NEWTON_BOND,EVFLAG>, const int&, EV_FLOAT&) const;
+
+  template<int NEWTON_BOND, int EVFLAG>
+  KOKKOS_INLINE_FUNCTION
+  void operator()(TagDihedralMultiHarmonicCompute<NEWTON_BOND,EVFLAG>, const int&) const;
+
+  //template<int NEWTON_BOND>
+  KOKKOS_INLINE_FUNCTION
+  void ev_tally(EV_FLOAT &ev, const int i1, const int i2, const int i3, const int i4,
+                          F_FLOAT &edihedral, F_FLOAT *f1, F_FLOAT *f3, F_FLOAT *f4,
+                          const F_FLOAT &vb1x, const F_FLOAT &vb1y, const F_FLOAT &vb1z,
+                          const F_FLOAT &vb2x, const F_FLOAT &vb2y, const F_FLOAT &vb2z,
+                          const F_FLOAT &vb3x, const F_FLOAT &vb3y, const F_FLOAT &vb3z) const;
+
+  DAT::tdual_efloat_1d k_eatom;
+  DAT::tdual_virial_array k_vatom;
+
+ protected:
+
+  class NeighborKokkos *neighborKK;
+
+  typename AT::t_x_array_randomread x;
+  typename AT::t_f_array f;
+  typename AT::t_int_2d dihedrallist;
+  typename ArrayTypes<DeviceType>::t_efloat_1d d_eatom;
+  typename ArrayTypes<DeviceType>::t_virial_array d_vatom;
+
+  int nlocal,newton_bond;
+  int eflag,vflag;
+
+  DAT::tdual_int_scalar k_warning_flag;
+  typename AT::t_int_scalar d_warning_flag;
+  HAT::t_int_scalar h_warning_flag;
+
+  DAT::tdual_ffloat_1d k_a1;
+  DAT::tdual_ffloat_1d k_a2;
+  DAT::tdual_ffloat_1d k_a3;
+  DAT::tdual_ffloat_1d k_a4;
+  DAT::tdual_ffloat_1d k_a5;
+
+  typename AT::t_ffloat_1d d_a1;
+  typename AT::t_ffloat_1d d_a2;
+  typename AT::t_ffloat_1d d_a3;
+  typename AT::t_ffloat_1d d_a4;
+  typename AT::t_ffloat_1d d_a5;
+
+  void allocate() override;
+};
+
+}
+
+#endif
+#endif
+

src/MOLECULE/dihedral_multi_harmonic.cpp

@@ -43,7 +43,7 @@ DihedralMultiHarmonic::DihedralMultiHarmonic(LAMMPS *_lmp) : Dihedral(_lmp)
 
 DihedralMultiHarmonic::~DihedralMultiHarmonic()
 {
-  if (allocated) {
+  if (allocated && !copymode) {
     memory->destroy(setflag);
     memory->destroy(a1);
     memory->destroy(a2);

src/MOLECULE/dihedral_multi_harmonic.h

@@ -38,7 +38,7 @@ class DihedralMultiHarmonic : public Dihedral {
  protected:
   double *a1, *a2, *a3, *a4, *a5;
 
-  void allocate();
+  virtual void allocate();
 };
 
 }    // namespace LAMMPS_NS