cherry-pick #6392 to support GPU version of cal_force_cc under LCAO basis set (#6583)

* Perf: support GPU version of cal_force_cc with LCAO basis (#6392)

* support GPU version of cal_force_cc with LCAO basis

* fix a bug

* Apply the changes related to force calculation from the develop branch, fixing the memory error bug that occurs during direct cherry-pick.

* Reverted the process of removing PAW.

* fix force calc

---------

Co-authored-by: dzzz2001 <[email protected]>

Author: zgn-26714

source/module_hamilt_lcao/hamilt_lcaodft/FORCE_STRESS.cpp

@@ -22,7 +22,7 @@
 #include "module_hamilt_lcao/hamilt_lcaodft/operator_lcao/nonlocal_new.h"
 
 template <typename T>
-Force_Stress_LCAO<T>::Force_Stress_LCAO(Record_adj& ra, const int nat_in) : RA(&ra), f_pw(nat_in), nat(nat_in)
+Force_Stress_LCAO<T>::Force_Stress_LCAO(Record_adj& ra, const int nat_in) : RA(&ra), nat(nat_in)
 {
 }
 template <typename T>
@@ -861,24 +861,39 @@ void Force_Stress_LCAO<T>::calForcePwPart(UnitCell& ucell,
                                           const Structure_Factor& sf)
 {
     ModuleBase::TITLE("Force_Stress_LCAO", "calForcePwPart");
-    //--------------------------------------------------------
-    // local pseudopotential force:
-    // use charge density; plane wave; local pseudopotential;
-    //--------------------------------------------------------
-    f_pw.cal_force_loc(ucell, fvl_dvl, rhopw, locpp.vloc, chr);
-    //--------------------------------------------------------
-    // ewald force: use plane wave only.
-    //--------------------------------------------------------
-    f_pw.cal_force_ew(ucell, fewalds, rhopw, &sf); // remain problem
+#ifdef __CUDA
+    if(PARAM.inp.device == "gpu")
+    {
+        Forces<double, base_device::DEVICE_GPU> f_pw(nat);
+
+        //--------------------------------------------------------
+        // local pseudopotential force:
+        // use charge density; plane wave; local pseudopotential;
+        //--------------------------------------------------------
+        f_pw.cal_force_loc(ucell, fvl_dvl, rhopw, locpp.vloc, chr);
+        //--------------------------------------------------------
+        // ewald force: use plane wave only.
+        //--------------------------------------------------------
+        f_pw.cal_force_ew(ucell, fewalds, rhopw, &sf); // remain problem
+
+        //--------------------------------------------------------
+        // force due to core correlation.
+        //--------------------------------------------------------
+        f_pw.cal_force_cc(fcc, rhopw, chr, locpp.numeric, ucell);
+        //--------------------------------------------------------
+        // force due to self-consistent charge.
+        //--------------------------------------------------------
+        f_pw.cal_force_scc(fscc, rhopw, vnew, vnew_exist, locpp.numeric, ucell);
+    } else
+#endif
+    {
+        Forces<double, base_device::DEVICE_CPU> f_pw(nat);
+        f_pw.cal_force_loc(ucell, fvl_dvl, rhopw, locpp.vloc, chr);
+        f_pw.cal_force_ew(ucell, fewalds, rhopw, &sf); // remain problem
+        f_pw.cal_force_cc(fcc, rhopw, chr, locpp.numeric, ucell);
+        f_pw.cal_force_scc(fscc, rhopw, vnew, vnew_exist, locpp.numeric, ucell);
+    }
 
-    //--------------------------------------------------------
-    // force due to core correlation.
-    //--------------------------------------------------------
-    f_pw.cal_force_cc(fcc, rhopw, chr, locpp.numeric, ucell);
-    //--------------------------------------------------------
-    // force due to self-consistent charge.
-    //--------------------------------------------------------
-    f_pw.cal_force_scc(fscc, rhopw, vnew, vnew_exist, locpp.numeric, ucell);
     return;
 }
 

source/module_hamilt_lcao/hamilt_lcaodft/FORCE_STRESS.h

@@ -60,7 +60,6 @@ class Force_Stress_LCAO
     Record_adj* RA;
     Force_LCAO<T> flk;
     Stress_Func<double> sc_pw;
-    Forces<double> f_pw;
 
     void forceSymmetry(const UnitCell& ucell, 
                        ModuleBase::matrix& fcs,