Refactor: use less memory and optimize performance to calculate force and stress in pw base

source/Makefile.Objects

@@ -576,6 +576,8 @@ OBJS_SRCPW=H_Ewald_pw.o\
     fp_energy.o\
     forces.o\
     forces_us.o\
+    forces_nl.o\
+    fs_nonlocal_tools.o\
     force_op.o\
     stress_op.o\
     wf_op.o\

source/module_base/module_device/cuda/memory_op.cu

@@ -269,6 +269,10 @@ template struct delete_memory_op<float, base_device::DEVICE_GPU>;
 template struct delete_memory_op<double, base_device::DEVICE_GPU>;
 template struct delete_memory_op<std::complex<float>, base_device::DEVICE_GPU>;
 template struct delete_memory_op<std::complex<double>, base_device::DEVICE_GPU>;
+template struct delete_memory_op<float*, base_device::DEVICE_GPU>;
+template struct delete_memory_op<double*, base_device::DEVICE_GPU>;
+template struct delete_memory_op<std::complex<float>*, base_device::DEVICE_GPU>;
+template struct delete_memory_op<std::complex<double>*, base_device::DEVICE_GPU>;
 
 } // namespace memory
 } // end of namespace base_device

source/module_base/module_device/memory_op.cpp

@@ -141,10 +141,13 @@ template struct delete_memory_op<float, base_device::DEVICE_CPU>;
 template struct delete_memory_op<double, base_device::DEVICE_CPU>;
 template struct delete_memory_op<std::complex<float>, base_device::DEVICE_CPU>;
 template struct delete_memory_op<std::complex<double>, base_device::DEVICE_CPU>;
+template struct delete_memory_op<float*, base_device::DEVICE_CPU>;
+template struct delete_memory_op<double*, base_device::DEVICE_CPU>;
+template struct delete_memory_op<std::complex<float>*, base_device::DEVICE_CPU>;
+template struct delete_memory_op<std::complex<double>*, base_device::DEVICE_CPU>;
 
 #if !(defined(__CUDA) || defined(__ROCM))
 
-
 template <typename FPTYPE>
 struct resize_memory_op<FPTYPE, base_device::DEVICE_GPU>
 {

source/module_base/module_device/rocm/memory_op.hip.cu

@@ -265,6 +265,10 @@ template struct delete_memory_op<float, base_device::DEVICE_GPU>;
 template struct delete_memory_op<double, base_device::DEVICE_GPU>;
 template struct delete_memory_op<std::complex<float>, base_device::DEVICE_GPU>;
 template struct delete_memory_op<std::complex<double>, base_device::DEVICE_GPU>;
+template struct delete_memory_op<float*, base_device::DEVICE_GPU>;
+template struct delete_memory_op<double*, base_device::DEVICE_GPU>;
+template struct delete_memory_op<std::complex<float>*, base_device::DEVICE_GPU>;
+template struct delete_memory_op<std::complex<double>*, base_device::DEVICE_GPU>;
 
 } // namespace memory
 } // end of namespace base_device

source/module_esolver/esolver_ks_pw.cpp

@@ -914,6 +914,7 @@ void ESolver_KS_PW<T, Device>::cal_stress(ModuleBase::matrix& stress)
     }
     ss.cal_stress(stress,
                   GlobalC::ucell,
+                  &GlobalC::ppcell,
                   this->pw_rhod,
                   &GlobalC::ucell.symm,
                   &this->sf,

source/module_esolver/esolver_sdft_pw.cpp

@@ -253,7 +253,9 @@ void ESolver_SDFT_PW::cal_stress(ModuleBase::matrix& stress)
                   pw_wfc,
                   this->psi,
                   this->stowf,
-                  pelec->charge);
+                  pelec->charge,
+                  &GlobalC::ppcell,
+                  GlobalC::ucell);
 }
 
 void ESolver_SDFT_PW::after_all_runners()

source/module_hamilt_pw/hamilt_pwdft/CMakeLists.txt

@@ -8,6 +8,7 @@ list(APPEND objects
     operator_pw/meta_pw.cpp
     operator_pw/velocity_pw.cpp
     operator_pw/operator_pw.cpp
+    forces_nl.cpp
     forces.cpp
     forces_us.cpp
     stress_func_cc.cpp
@@ -32,6 +33,7 @@ list(APPEND objects
     global.cpp
     parallel_grid.cpp
     elecond.cpp
+    fs_nonlocal_tools.cpp
 )
 
 add_library(

source/module_hamilt_pw/hamilt_pwdft/VNL_in_pw.cpp

@@ -495,7 +495,6 @@ void pseudopot_cell_vnl::getvnl(Device* ctx, const int& ik, std::complex<FPTYPE>
                  npw,
                  this->wfcpw->npwk_max,
                  this->nhm,
-                 GlobalV::NQX,
                  this->tab.getBound2(),
                  this->tab.getBound3(),
                  atom_na,