test 2

Author: haozhihan

source/module_psi/psi_init.cpp

@@ -26,13 +26,21 @@ PSIInit<T, Device>::PSIInit(const std::string& init_wfc_in,
     this->use_psiinitializer = use_psiinitializer_in;
     this->pw_wfc = pw_wfc_in;
 
-    if (PARAM.inp.init_wfc == "file" || PARAM.inp.device == "gpu")
+    if (PARAM.inp.psi_initializer == true)
     {
-        this->init_psi_method = "old"; // old method;
+        this->init_psi_method = "new";
     }
     else
     {
-        this->init_psi_method = "new"; // new method;
+        if (PARAM.inp.init_wfc == "file" || PARAM.inp.init_wfc == "atomic" || PARAM.inp.device == "gpu"
+            || PARAM.inp.esolver_type == "sdft")
+        {
+            this->init_psi_method = "old"; // old method;
+        }
+        else
+        {
+            this->init_psi_method = "new"; // new method;
+        }
     }
 }
 

source/module_psi/psi_initializer.cpp

@@ -171,7 +171,7 @@ void psi_initializer<T, Device>::random_t(T* psi, const int iw_start, const int
 #endif
                 }
                 // then for each g-component, initialize the wavefunction value
-                #pragma omp parallel for schedule(static, 4096/sizeof(T))
+                // #pragma omp parallel for schedule(static, 4096/sizeof(T))
                 for (int ig = 0; ig < ng; ig++)
                 {
                     // get the correct value of "rr" and "arg" by indexing map "getigl2isz"
@@ -192,7 +192,7 @@ void psi_initializer<T, Device>::random_t(T* psi, const int iw_start, const int
         {
             T* psi_slice = &(psi[iw * this->pw_wfc_->npwk_max * PARAM.globalv.npol]); // get the memory to write directly. For nspin 4, nbasis*2
 
-            #pragma omp parallel for schedule(static, 4096/sizeof(T))
+            // #pragma omp parallel for schedule(static, 4096/sizeof(T))
             for (int ig = 0; ig < ng; ig++)
             {
                 const double rr = std::rand()/double(RAND_MAX); //qianrui add RAND_MAX
@@ -202,7 +202,7 @@ void psi_initializer<T, Device>::random_t(T* psi, const int iw_start, const int
             }
             if(PARAM.globalv.npol == 2) // additionally for nspin 4...
             {
-                #pragma omp parallel for schedule(static, 4096/sizeof(T))
+                // #pragma omp parallel for schedule(static, 4096/sizeof(T))
                 for (int ig = this->pw_wfc_->npwk_max; ig < this->pw_wfc_->npwk_max + ng; ig++)
                 {
                     const double rr = std::rand()/double(RAND_MAX);

source/module_psi/psi_initializer_nao.cpp

@@ -276,7 +276,7 @@ void psi_initializer_nao<T, Device>::proj_ao_onkG(const int ik)
     std::vector<double> qnorm(npw);
     std::vector<ModuleBase::Vector3<double>> q(npw);
 
-    #pragma omp parallel for schedule(static, 4096 / sizeof(double))
+    // #pragma omp parallel for schedule(static, 4096 / sizeof(double))
     for (int ig = 0; ig < npw; ig++)
     {
         q[ig] = this->pw_wfc_->getgpluskcar(ik, ig);
@@ -335,13 +335,13 @@ void psi_initializer_nao<T, Device>::proj_ao_onkG(const int ik)
                                 for (int m = 0; m < 2 * L + 1; m++)
                                 {
                                     const int lm = L * L + m;
-                                    #pragma omp parallel for
+                                    // #pragma omp parallel for
                                     for (int ig = 0; ig < npw; ig++)
                                     {
                                         aux[ig] = sk[ig] * ylm(lm, ig) * Jlfq[ig];
                                     }
 
-                                    #pragma omp parallel for
+                                    // #pragma omp parallel for
                                     for (int ig = 0; ig < npw; ig++)
                                     {
                                         fup = cos(0.5 * alpha) * aux[ig];
@@ -374,7 +374,7 @@ void psi_initializer_nao<T, Device>::proj_ao_onkG(const int ik)
                         for (int m = 0; m < 2 * L + 1; m++)
                         {
                             const int lm = L * L + m;
-                            #pragma omp parallel for
+                            // #pragma omp parallel for
                             for (int ig = 0; ig < npw; ig++)
                             {
                                 (*(this->psig_))(ibasis, ig)