Merge pull request #93 from RMGDFT/develop

Develop merge for 6.1 release

CMakeCPackOptions.cmake.in

@@ -2,7 +2,7 @@
 set(CPACK_GENERATOR "DEB")
 
 set(CPACK_PACKAGE_VERSION_MAJOR "6")
-set(CPACK_PACKAGE_VERSION_MINOR "0")
+set(CPACK_PACKAGE_VERSION_MINOR "1")
 set(CPACK_PACKAGE_VERSION_PATCH "0")
 set(COMPLETE_VERSION ${CPACK_PACKAGE_VERSION_MAJOR}.${CPACK_PACKAGE_VERSION_MINOR}.${CPACK_PACKAGE_VERSION_PATCH})
 set(CPACK_PACKAGE_NAME "rmg")

CMakeLists.txt

@@ -28,7 +28,7 @@
 cmake_minimum_required (VERSION 3.17) 
 
 set(RMG_VERSION_MAJOR "6")
-set(RMG_VERSION_MINOR "0")
+set(RMG_VERSION_MINOR "1")
 set(RMG_VERSION_PATCH "0")
 set(COMPLETE_VERSION ${RMG_VERSION_MAJOR}.${RMG_VERSION_MINOR}.${RMG_VERSION_PATCH})
 
@@ -207,6 +207,11 @@ if(CRAY_HOST)
 else()
     find_package(MPI REQUIRED)
 endif()
+
+if(${MPI_CXX_VERSION_MAJOR} LESS 3)
+    message(FATAL_ERROR "RMG requires MPI CXX version 3.0 or greater.")
+endif()
+
 set(RMGLIBS "${MPI_CXX_LIBRARIES}")
 list(APPEND RMGLIBS ${MPI_C_LIBRARIES})
 list(APPEND RMGLIBS ${MPI_Fortran_LIBRARIES})

Headers/State.h

@@ -54,12 +54,27 @@ template <typename StateType> class State {
 
     /** Nuclear potential */
     double *vnuc;
+    double *vnuc_f;   // Ephemeral object that only exists in GetNewRho
+    double *vnuc_c;   // Ephemeral object that only exists in GetNewRho
+
+    /** Nuclear potential energy correction term */
+    double vnuc_correction;
+
+    /** EXC energy correction term */
+    double vxc_correction;
+
+    /** ES energy correction term */
+    double vh_correction;
 
     /** Hartree potential */
     double *vh;
+    double *vh_f;   // Ephemeral object that only exists in GetNewRho
+    double *vh_c;   // Ephemeral object that only exists in GetNewRho
 
     /** Exchange correlation potential */
     double *vxc;
+    double *vxc_f;   // Ephemeral object that only exists in GetNewRho
+    double *vxc_c;   // Ephemeral object that only exists in GetNewRho
 
     /** Total potential */
     double *vtot;

Headers/common_prototypes.h

@@ -19,6 +19,8 @@
 
 template <typename T> T MyConj(T val);
 
+void WritePoscar(FILE *FH, int md_step);
+
 /* Blas wrappers */
 void QMD_daxpy (int n, double alpha, double *x, int incx, double *y, int incy);
 void QMD_dscal (int n, double alpha, double *x, int incx);

Headers/rmg_control.h

@@ -382,7 +382,10 @@ class CONTROL
     /* Prolong operator default order */
     int prolong_order;
     /* Prolong operator mixing parameter */
-    double cmix = 1.0;
+    double cmix = 0.0;
+    /* Flag indicating whether or not to use cmix */
+    bool use_cmix = true;
+
 
     // Flag indicating whether or not to use gpu finite differencing for the hamiltonian
     bool use_gpu_fd;
@@ -721,6 +724,9 @@ class CONTROL
     // Flag to indicate whether or not to use gpu managed memory
     bool gpu_managed_memory;
 
+    // Flag to use new energy correction terms
+    bool use_energy_correction;
+
 #if CUDA_ENABLED || HIP_ENABLED || SYCL_ENABLED
 
 

Headers/species.h

@@ -165,15 +165,11 @@ class SPECIES : Atomic
     std::vector<double> atomic_wave_energy;
 
     /* Finite difference auto tuning factors. One for each atomic orbital (including m-dependence) */
-    std::vector<double> fd_factor1;
-    std::vector<double> fd_fke1;
-    std::vector<double> fd_slopes;
-    std::vector<double> fd_xint;
-    std::vector<double> fd_yint;
-    std::vector<double> pd_slopes;
-    std::vector<double> pd_xint;
-    std::vector<double> pd_yint;
+    std::vector<double> occ_weight;
+    std::vector<double> fd_mins;
     std::vector<double> pd_mins;
+    std::vector<std::array<double, 8> > fd_coeffs;
+    std::vector<std::array<double, 8> > pd_coeffs;
 
     std::vector<std::string> atomic_wave_label;
 

Headers/transition.h

@@ -202,6 +202,7 @@ void Precond_drho(double *);
 template <typename T> void Write_Wfs_forWannier(int kpt_global, Kpoint<T> *kptr, std::vector<bool> exclude_bands, std::string wavefule);
 double GetPlanarAnisotropy(double *density);
 void GetFdFactor(int kidx);
+void SimplePolyFit(const double *x, double *y, const size_t n, const size_t k, double *beta);
 
 #endif
 #endif

Input/LoadUpfPseudo.cpp

@@ -388,6 +388,11 @@ void LoadUpfPseudo(SPECIES *sp)
 
             // UPF stores atomic wavefunctions * r so divide through
             for(int ix = 0;ix < sp->rg_points;ix++) sp->atomic_wave[iwf][ix] /= sp->r[ix];
+            for(int ix = 0;ix < sp->rg_points;ix++)
+            {
+//               if(sp->r[ix] > 8.0) sp->atomic_wave[iwf][ix] *= exp(-(sp->r[ix] - 8.0)*(sp->r[ix] - 8.0));
+            }
+
         }
         ct.max_orbitals = std::max(ct.max_orbitals, sp->num_atomic_waves_m);
         sp->num_orbitals = sp->num_atomic_waves_m;

Input/ReadCommon.cpp

@@ -729,7 +729,16 @@ void ReadCommon(char *cfile, CONTROL& lc, PE_CONTROL& pelc, std::unordered_map<s
     If.RegisterInputKey("afd_cfac", &lc.afd_cfac, 0.0, 10.0, 0.0,
             CHECK_AND_FIX, OPTIONAL,
             "Manual setting for the adaptive finite differencing parameter. ",
-            "afd_cfac must lie in the range (0.0,3.0). Resetting to auto. ", CONTROL_OPTIONS);
+            "afd_cfac must lie in the range (0.0,10.0). Resetting to auto. ", CONTROL_OPTIONS);
+
+    If.RegisterInputKey("adaptive_cmix", &lc.cmix, 0.0, 10.0, 0.0,
+            CHECK_AND_FIX, OPTIONAL,
+            "Manual setting for the adaptive interpolation parameter. ",
+            "adaptive_cmix must lie in the range (0.0,10.0). Resetting to auto. ", CONTROL_OPTIONS);
+
+    If.RegisterInputKey("use_cmix", &lc.use_cmix, true, 
+            "Use adaptive interpolation "
+            "can improve energies/forces for some but not all systems. ", MISC_OPTIONS|EXPERIMENTAL_OPTION);
 
     If.RegisterInputKey("semilocal_projectors", &lc.semilocal_projectors, 6, 40, 10,
             CHECK_AND_FIX, OPTIONAL,
@@ -1228,6 +1237,9 @@ void ReadCommon(char *cfile, CONTROL& lc, PE_CONTROL& pelc, std::unordered_map<s
             "Some AMD and Nvidia GPUs support managed gou memory which is "
             "useful when GPU memory limits are exceeded. ", CONTROL_OPTIONS|EXPERT_OPTION);
 
+    If.RegisterInputKey("use_energy_correction", &lc.use_energy_correction, false, 
+            "Experimental energy correction term ", CONTROL_OPTIONS|EXPERIMENTAL_OPTION);
+
     If.RegisterInputKey("use_numa", &lc.use_numa, true, 
             "Numa stands for Non Uniform Memory Access and means that the main "
             "memory of a computer is organized into seperate distinct banks. "

Input/ReadDynamics.cpp

@@ -166,6 +166,15 @@ void ReadDynamics(char *cfile, CONTROL& lc, std::unordered_map<std::string, Inpu
 
     }
 
+    for (auto& sp : Species)
+    {
+        sp.num_atoms = 0;
+    }
+    for(auto& Atom : Atoms)
+    {
+        Atom.Type->num_atoms +=1;
+    }
+
 }
 
 

Misc/CMakeLists.txt

@@ -1,4 +1,6 @@
 add_library (Misc
+PolyFit.cpp
+WritePoscar.cpp
 OutputBandPlot.cpp
 kind.f90
 kpoint_grid.f90