Fix(paged): fix PAGED_GPU buffer overflow, invalid D2H copy, and psi_norm abort

- psi_prepare.cpp: use nbands_l instead of nbands_start in memcpy to
  prevent buffer overflow when nbands_start > nbands
- psi_prepare.cpp: skip invalid D2H copy to GPU pointer in dav solver
  path for PAGED_GPU mode (data already in kspw_psi CPU buffer)
- diago_dav_subspace.cpp: replace WARNING_QUIT with convergence
  handling when psi_norm <= 1e-12 for both GPU and CPU paths;
  when all new directions have near-zero norms, mark subspace as
  converged instead of aborting

Author: dyzheng

source/source_hsolver/diago_dav_subspace.cpp

@@ -440,13 +440,23 @@ void Diago_DavSubspace<T, Device>::cal_grad(const HPsiFunc& hpsi_func,
         {
             if (psi_norm_host[i] <= 1.0e-12)
             {
-                std::cout << "Diago_DavSubspace::cal_grad: psi_norm <= 0 for band " << i << std::endl;
-                std::cout << "This may be due to npwx < nbands: the number of plane waves is less than" << std::endl;
-                std::cout << "the number of bands, leading to a rank-deficient problem." << std::endl;
-                std::cout << "Please increase ecutwfc or reduce nbands." << std::endl;
-                delmem_real_h_op()(psi_norm_host);
-                delmem_real_op()(psi_norm);
-                ModuleBase::WARNING_QUIT("cal_grad", "psi_norm <= 0");
+                std::cout << "Diago_DavSubspace::cal_grad: norm of new direction for band " << i
+                          << " is too small, treating as converged." << std::endl;
+            }
+        }
+        {
+            bool all_zero = true;
+            for (int i = 0; i < notconv; i++)
+            {
+                if (psi_norm_host[i] > 1.0e-12)
+                {
+                    all_zero = false;
+                    break;
+                }
+            }
+            if (all_zero && notconv > 0)
+            {
+                this->notconv = 0;
             }
         }
         delmem_real_h_op()(psi_norm_host);
@@ -470,12 +480,23 @@ void Diago_DavSubspace<T, Device>::cal_grad(const HPsiFunc& hpsi_func,
         {
             if (psi_norm[i] <= 1.0e-12)
             {
-                std::cout << "Diago_DavSubspace::cal_grad: psi_norm <= 0 for band " << i << std::endl;
-                std::cout << "This may be due to npwx < nbands: the number of plane waves is less than" << std::endl;
-                std::cout << "the number of bands, leading to a rank-deficient problem." << std::endl;
-                std::cout << "Please increase ecutwfc or reduce nbands." << std::endl;
-                delmem_real_h_op()(psi_norm);
-                ModuleBase::WARNING_QUIT("cal_grad", "psi_norm <= 0");
+                std::cout << "Diago_DavSubspace::cal_grad: norm of new direction for band " << i
+                          << " is too small, treating as converged." << std::endl;
+            }
+        }
+        {
+            bool all_zero = true;
+            for (int i = 0; i < notconv; i++)
+            {
+                if (psi_norm[i] > 1.0e-12)
+                {
+                    all_zero = false;
+                    break;
+                }
+            }
+            if (all_zero && notconv > 0)
+            {
+                this->notconv = 0;
             }
         }
         delmem_real_h_op()(psi_norm);

source/source_psi/psi_prepare.cpp

@@ -179,7 +179,7 @@ void PSIPrepare<T, Device>::initialize_psi(Psi<std::complex<double>>* psi,
                 if (kspw_psi->get_storage_mode() == psi::PsiStorageMode::PAGED_GPU)
                 {
                     T* dst_cpu = kspw_psi->get_cpu_pointer(ik);
-                    std::memcpy(dst_cpu, psi_cpu->get_pointer(), sizeof(T) * nbands_start * nbasis);
+                    std::memcpy(dst_cpu, psi_cpu->get_pointer(), sizeof(T) * nbands_l * nbasis);
                     kspw_psi->load_k_to_gpu(ik);
                 }
                 else
@@ -219,8 +219,6 @@ void PSIPrepare<T, Device>::initialize_psi(Psi<std::complex<double>>* psi,
                 {
                     if (kspw_psi->get_storage_mode() == psi::PsiStorageMode::PAGED_GPU)
                     {
-                        base_device::memory::synchronize_memory_op<T, base_device::DEVICE_CPU, Device>()(
-                            kspw_psi->get_pointer(), psi_device->get_pointer(), nbands_l * nbasis);
                     }
                     else
                     {