diff --git a/src/KOKKOS/atom_vec_kokkos.cpp b/src/KOKKOS/atom_vec_kokkos.cpp
index f2c04bec1b6281b3922513628b2a655e44c9baf7..555bb2e7478e0b11ae64e58f2208842a2124215b 100644
--- a/src/KOKKOS/atom_vec_kokkos.cpp
+++ b/src/KOKKOS/atom_vec_kokkos.cpp
@@ -305,6 +305,277 @@ void AtomVecKokkos::unpack_comm_kokkos(const int &n, const int &first,
}
}
+
+/* ---------------------------------------------------------------------- */
+
+template<class DeviceType,int PBC_FLAG,int TRICLINIC,int DEFORM_VREMAP>
+struct AtomVecKokkos_PackCommVel {
+ typedef DeviceType device_type;
+
+ typename ArrayTypes<DeviceType>::t_x_array_randomread _x;
+ typename ArrayTypes<DeviceType>::t_int_1d _mask;
+ typename ArrayTypes<DeviceType>::t_v_array _v;
+ typename ArrayTypes<DeviceType>::t_xfloat_2d_um _buf;
+ typename ArrayTypes<DeviceType>::t_int_2d_const _list;
+ const int _iswap;
+ X_FLOAT _xprd,_yprd,_zprd,_xy,_xz,_yz;
+ X_FLOAT _pbc[6];
+ X_FLOAT _h_rate[6];
+ const int _deform_vremap;
+
+ AtomVecKokkos_PackCommVel(
+ const typename DAT::tdual_x_array &x,
+ const typename DAT::tdual_int_1d &mask,
+ const typename DAT::tdual_v_array &v,
+ const typename DAT::tdual_xfloat_2d &buf,
+ const typename DAT::tdual_int_2d &list,
+ const int &iswap,
+ const X_FLOAT &xprd, const X_FLOAT &yprd, const X_FLOAT &zprd,
+ const X_FLOAT &xy, const X_FLOAT &xz, const X_FLOAT &yz, const int* const pbc,
+ const double * const h_rate,
+ const int &deform_vremap):
+ _x(x.view<DeviceType>()),
+ _mask(mask.view<DeviceType>()),
+ _v(v.view<DeviceType>()),
+ _list(list.view<DeviceType>()),_iswap(iswap),
+ _xprd(xprd),_yprd(yprd),_zprd(zprd),
+ _xy(xy),_xz(xz),_yz(yz),
+ _deform_vremap(deform_vremap)
+ {
+ const size_t elements = 6;
+ const int maxsend = (buf.template view<DeviceType>().extent(0)*buf.template view<DeviceType>().extent(1))/elements;
+ _buf = typename ArrayTypes<DeviceType>::t_xfloat_2d_um(buf.view<DeviceType>().data(),maxsend,elements);
+ _pbc[0] = pbc[0]; _pbc[1] = pbc[1]; _pbc[2] = pbc[2];
+ _pbc[3] = pbc[3]; _pbc[4] = pbc[4]; _pbc[5] = pbc[5];
+ _h_rate[0] = h_rate[0]; _h_rate[1] = h_rate[1]; _h_rate[2] = h_rate[2];
+ _h_rate[3] = h_rate[3]; _h_rate[4] = h_rate[4]; _h_rate[5] = h_rate[5];
+ }
+
+ KOKKOS_INLINE_FUNCTION
+ void operator() (const int& i) const {
+ const int j = _list(_iswap,i);
+ if (PBC_FLAG == 0) {
+ _buf(i,0) = _x(j,0);
+ _buf(i,1) = _x(j,1);
+ _buf(i,2) = _x(j,2);
+ _buf(i,3) = _v(j,0);
+ _buf(i,4) = _v(j,1);
+ _buf(i,5) = _v(j,2);
+ } else {
+ if (TRICLINIC == 0) {
+ _buf(i,0) = _x(j,0) + _pbc[0]*_xprd;
+ _buf(i,1) = _x(j,1) + _pbc[1]*_yprd;
+ _buf(i,2) = _x(j,2) + _pbc[2]*_zprd;
+ } else {
+ _buf(i,0) = _x(j,0) + _pbc[0]*_xprd + _pbc[5]*_xy + _pbc[4]*_xz;
+ _buf(i,1) = _x(j,1) + _pbc[1]*_yprd + _pbc[3]*_yz;
+ _buf(i,2) = _x(j,2) + _pbc[2]*_zprd;
+ }
+
+ if (DEFORM_VREMAP == 0) {
+ _buf(i,3) = _v(j,0);
+ _buf(i,4) = _v(j,1);
+ _buf(i,5) = _v(j,2);
+ } else {
+ if (_mask(i) & _deform_vremap) {
+ _buf(i,3) = _v(j,0) + _pbc[0]*_h_rate[0] + _pbc[5]*_h_rate[5] + _pbc[4]*_h_rate[4];
+ _buf(i,4) = _v(j,1) + _pbc[1]*_h_rate[1] + _pbc[3]*_h_rate[3];
+ _buf(i,5) = _v(j,2) + _pbc[2]*_h_rate[2];
+ } else {
+ _buf(i,3) = _v(j,0);
+ _buf(i,4) = _v(j,1);
+ _buf(i,5) = _v(j,2);
+ }
+ }
+ }
+ }
+};
+
+/* ---------------------------------------------------------------------- */
+
+int AtomVecKokkos::pack_comm_vel_kokkos(
+ const int &n,
+ const DAT::tdual_int_2d &list,
+ const int & iswap,
+ const DAT::tdual_xfloat_2d &buf,
+ const int &pbc_flag,
+ const int* const pbc)
+{
+ if(commKK->forward_comm_on_host) {
+ sync(Host,X_MASK|V_MASK);
+ if (pbc_flag) {
+ if (deform_vremap) {
+ if (domain->triclinic) {
+ struct AtomVecKokkos_PackCommVel<LMPHostType,1,1,1> f(
+ atomKK->k_x,atomKK->k_mask,
+ atomKK->k_v,
+ buf,list,iswap,
+ domain->xprd,domain->yprd,domain->zprd,
+ domain->xy,domain->xz,domain->yz,pbc,h_rate,deform_vremap);
+ Kokkos::parallel_for(n,f);
+ } else {
+ struct AtomVecKokkos_PackCommVel<LMPHostType,1,0,1> f(
+ atomKK->k_x,atomKK->k_mask,
+ atomKK->k_v,
+ buf,list,iswap,
+ domain->xprd,domain->yprd,domain->zprd,
+ domain->xy,domain->xz,domain->yz,pbc,h_rate,deform_vremap);
+ Kokkos::parallel_for(n,f);
+ }
+ } else {
+ if(domain->triclinic) {
+ struct AtomVecKokkos_PackCommVel<LMPHostType,1,1,0> f(
+ atomKK->k_x,atomKK->k_mask,
+ atomKK->k_v,
+ buf,list,iswap,
+ domain->xprd,domain->yprd,domain->zprd,
+ domain->xy,domain->xz,domain->yz,pbc,h_rate,deform_vremap);
+ Kokkos::parallel_for(n,f);
+ } else {
+ struct AtomVecKokkos_PackCommVel<LMPHostType,1,0,0> f(
+ atomKK->k_x,atomKK->k_mask,
+ atomKK->k_v,
+ buf,list,iswap,
+ domain->xprd,domain->yprd,domain->zprd,
+ domain->xy,domain->xz,domain->yz,pbc,h_rate,deform_vremap);
+ Kokkos::parallel_for(n,f);
+ }
+ }
+ } else {
+ if(domain->triclinic) {
+ struct AtomVecKokkos_PackCommVel<LMPHostType,0,1,0> f(
+ atomKK->k_x,atomKK->k_mask,
+ atomKK->k_v,
+ buf,list,iswap,
+ domain->xprd,domain->yprd,domain->zprd,
+ domain->xy,domain->xz,domain->yz,pbc,h_rate,deform_vremap);
+ Kokkos::parallel_for(n,f);
+ } else {
+ struct AtomVecKokkos_PackCommVel<LMPHostType,0,0,0> f(
+ atomKK->k_x,atomKK->k_mask,
+ atomKK->k_v,
+ buf,list,iswap,
+ domain->xprd,domain->yprd,domain->zprd,
+ domain->xy,domain->xz,domain->yz,pbc,h_rate,deform_vremap);
+ Kokkos::parallel_for(n,f);
+ }
+ }
+ } else {
+ sync(Device,X_MASK|V_MASK);
+ if(pbc_flag) {
+ if(deform_vremap) {
+ if(domain->triclinic) {
+ struct AtomVecKokkos_PackCommVel<LMPDeviceType,1,1,1> f(
+ atomKK->k_x,atomKK->k_mask,
+ atomKK->k_v,
+ buf,list,iswap,
+ domain->xprd,domain->yprd,domain->zprd,
+ domain->xy,domain->xz,domain->yz,pbc,h_rate,deform_vremap);
+ Kokkos::parallel_for(n,f);
+ } else {
+ struct AtomVecKokkos_PackCommVel<LMPDeviceType,1,0,1> f(
+ atomKK->k_x,atomKK->k_mask,
+ atomKK->k_v,
+ buf,list,iswap,
+ domain->xprd,domain->yprd,domain->zprd,
+ domain->xy,domain->xz,domain->yz,pbc,h_rate,deform_vremap);
+ Kokkos::parallel_for(n,f);
+ }
+ } else {
+ if(domain->triclinic) {
+ struct AtomVecKokkos_PackCommVel<LMPDeviceType,1,1,0> f(
+ atomKK->k_x,atomKK->k_mask,
+ atomKK->k_v,
+ buf,list,iswap,
+ domain->xprd,domain->yprd,domain->zprd,
+ domain->xy,domain->xz,domain->yz,pbc,h_rate,deform_vremap);
+ Kokkos::parallel_for(n,f);
+ } else {
+ struct AtomVecKokkos_PackCommVel<LMPDeviceType,1,0,0> f(
+ atomKK->k_x,atomKK->k_mask,
+ atomKK->k_v,
+ buf,list,iswap,
+ domain->xprd,domain->yprd,domain->zprd,
+ domain->xy,domain->xz,domain->yz,pbc,h_rate,deform_vremap);
+ Kokkos::parallel_for(n,f);
+ }
+ }
+ } else {
+ if(domain->triclinic) {
+ struct AtomVecKokkos_PackCommVel<LMPDeviceType,0,1,0> f(
+ atomKK->k_x,atomKK->k_mask,
+ atomKK->k_v,
+ buf,list,iswap,
+ domain->xprd,domain->yprd,domain->zprd,
+ domain->xy,domain->xz,domain->yz,pbc,h_rate,deform_vremap);
+ Kokkos::parallel_for(n,f);
+ } else {
+ struct AtomVecKokkos_PackCommVel<LMPDeviceType,0,0,0> f(
+ atomKK->k_x,atomKK->k_mask,
+ atomKK->k_v,
+ buf,list,iswap,
+ domain->xprd,domain->yprd,domain->zprd,
+ domain->xy,domain->xz,domain->yz,pbc,h_rate,deform_vremap);
+ Kokkos::parallel_for(n,f);
+ }
+ }
+ }
+ return n*6;
+}
+
+/* ---------------------------------------------------------------------- */
+
+template<class DeviceType>
+struct AtomVecKokkos_UnpackCommVel {
+ typedef DeviceType device_type;
+
+ typename ArrayTypes<DeviceType>::t_x_array _x;
+ typename ArrayTypes<DeviceType>::t_v_array _v;
+ typename ArrayTypes<DeviceType>::t_xfloat_2d_const _buf;
+ int _first;
+
+ AtomVecKokkos_UnpackCommVel(
+ const typename DAT::tdual_x_array &x,
+ const typename DAT::tdual_v_array &v,
+ const typename DAT::tdual_xfloat_2d &buf,
+ const int& first):
+ _x(x.view<DeviceType>()),
+ _v(v.view<DeviceType>()),
+ _first(first)
+ {
+ const size_t elements = 6;
+ const int maxsend = (buf.template view<DeviceType>().extent(0)*buf.template view<DeviceType>().extent(1))/elements;
+ buffer_view<DeviceType>(_buf,buf,maxsend,elements);
+ };
+
+ KOKKOS_INLINE_FUNCTION
+ void operator() (const int& i) const {
+ _x(i+_first,0) = _buf(i,0);
+ _x(i+_first,1) = _buf(i,1);
+ _x(i+_first,2) = _buf(i,2);
+ _v(i+_first,0) = _buf(i,3);
+ _v(i+_first,1) = _buf(i,4);
+ _v(i+_first,2) = _buf(i,5);
+ }
+};
+
+/* ---------------------------------------------------------------------- */
+
+void AtomVecKokkos::unpack_comm_vel_kokkos(const int &n, const int &first,
+ const DAT::tdual_xfloat_2d &buf ) {
+ if(commKK->forward_comm_on_host) {
+ sync(Host,X_MASK|V_MASK);
+ modified(Host,X_MASK|V_MASK);
+ struct AtomVecKokkos_UnpackCommVel<LMPHostType> f(atomKK->k_x,atomKK->k_v,buf,first);
+ Kokkos::parallel_for(n,f);
+ } else {
+ sync(Device,X_MASK|V_MASK);
+ modified(Device,X_MASK|V_MASK);
+ struct AtomVecKokkos_UnpackCommVel<LMPDeviceType> f(atomKK->k_x,atomKK->k_v,buf,first);
+ Kokkos::parallel_for(n,f);
+ }
+}
+
/* ---------------------------------------------------------------------- */
int AtomVecKokkos::pack_comm(int n, int *list, double *buf,
diff --git a/src/KOKKOS/atom_vec_kokkos.h b/src/KOKKOS/atom_vec_kokkos.h
index 38ffed9d374b47143d93c39e5d98046c1298cadb..e1d626ac1b0cfb8e3efbcfcffcde56754e8e0c8d 100644
--- a/src/KOKKOS/atom_vec_kokkos.h
+++ b/src/KOKKOS/atom_vec_kokkos.h
@@ -63,11 +63,11 @@ class AtomVecKokkos : public AtomVec {
virtual int
pack_comm_vel_kokkos(const int &n, const DAT::tdual_int_2d &list,
const int & iswap, const DAT::tdual_xfloat_2d &buf,
- const int &pbc_flag, const int pbc[]) { return 0; }
+ const int &pbc_flag, const int pbc[]);
virtual void
unpack_comm_vel_kokkos(const int &n, const int &nfirst,
- const DAT::tdual_xfloat_2d &buf) {}
+ const DAT::tdual_xfloat_2d &buf);
virtual int
unpack_reverse_self(const int &n, const DAT::tdual_int_2d &list,