deepmodeling · maki49 · Jan 27, 2022 · Jan 29, 2022 · Jan 29, 2022 · Jan 29, 2022
diff --git a/doc/input-main.md b/doc/input-main.md
@@ -63,7 +63,7 @@
 
     - [Variables useful for debugging](#variables-useful-for-debugging)
 
-        [nurse](#nurse) | [t_in_h](#t-in-h) | [vl_in_h](#vl-in-h) | [vnl_in_h](#vnl-in-h) | [test_force](#test-force) | [test_stress](#test-stress) | [colour](#colour) | [new_dm](#new-dm) | [test_just_neighbor](#test-just-neighbor)
+        [nurse](#nurse) | [t_in_h](#t-in-h) | [vl_in_h](#vl-in-h) | [vnl_in_h](#vnl-in-h) | [test_force](#test-force) | [test_stress](#test-stress) | [colour](#colour) | [test_just_neighbor](#test-just-neighbor)
     - [DeePKS](#deepks)
 
         [out_descriptor](#out-descriptor) | [lmax_descriptor](#lmax-descriptor) | [deepks_scf](#deepks-scf) | [model_file](#model-file)
@@ -992,7 +992,7 @@ This part of variables are relevant when using hybrid functionals
 
 - exx_hybrid_type<a id="exx-hybrid-type"></a>
     - *Type*: String
-    - *Description*: Type of hybrid functional used. Options are "hf" (pure Hartree-Fock), "pbe0"(PBE0), "hse" (Note: in order to use HSE functional, LIBXC is required).
+    - *Description*: Type of hybrid functional used. Options are "hf" (pure Hartree-Fock), "pbe0"(PBE0), "hse" (Note: in order to use HSE functional, LIBXC is required). Note also that HSE has been tested while PBE0 has NOT been fully tested yet, and the maxmum parallel cpus for running exx is Nx(N+1)/2, with N being the number of atoms.
 
 
         If set to "no", then no hybrid functional is used (i.e.,Fock exchange is not included.)
@@ -1023,7 +1023,7 @@ adial integration for pseudopotentials, in Bohr.
 
 - exx_pca_threshold<a id="exx-pca-threshold"></a>
     - *Type*: Real
-    - *Description*: To accelerate the evaluation of four-center integrals (ik|jl), the product of atomic orbitals are expanded in the basis of auxiliary basis functions (ABF): &phi;<sub>i</sub>&phi;<sub>j</sub>~C<sup>k</sup><sub>ij</sub>P<sub>k</sub>. The size of the ABF (i.e. number of P<sub>k</sub>) is reduced using principal component analysis. When a large PCA threshold is used, the number of ABF will be reduced, hence the calculations becomes faster. However this comes at the cost of computational accuracy. A relatively safe choice of the value is 1d-3.
+    - *Description*: To accelerate the evaluation of four-center integrals (ik|jl), the product of atomic orbitals are expanded in the basis of auxiliary basis functions (ABF): &phi;<sub>i</sub>&phi;<sub>j</sub>~C<sup>k</sup><sub>ij</sub>P<sub>k</sub>. The size of the ABF (i.e. number of P<sub>k</sub>) is reduced using principal component analysis. When a large PCA threshold is used, the number of ABF will be reduced, hence the calculations becomes faster. However this comes at the cost of computational accuracy. A relatively safe choice of the value is 1d-4.
     - *Default*: 0
 
     [back to top](#input-file)
@@ -1044,21 +1044,21 @@ adial integration for pseudopotentials, in Bohr.
 
 - exx_dm_threshold<a id="exx-dm-threshold"></a>
     - *Type*: Real
-    - *Description*: The Fock exchange can be expressed as &Sigma;<sub>k,l</sub>(ik|jl)D<sub>kl</sub> where D is the density matrix. Smaller values of the density matrix can be truncated to accelerate calculation. The larger the threshold is, the faster the calculation and the lower the accuracy. A relatively safe choice of the value is 1d-3.
+    - *Description*: The Fock exchange can be expressed as &Sigma;<sub>k,l</sub>(ik|jl)D<sub>kl</sub> where D is the density matrix. Smaller values of the density matrix can be truncated to accelerate calculation. The larger the threshold is, the faster the calculation and the lower the accuracy. A relatively safe choice of the value is 1d-4.
     - *Default*: 0
 
     [back to top](#input-file)
 
 - exx_schwarz_threshold<a id="exx-schwarz-threshold"></a>
     - *Type*: Real
-    - *Description*: In practice the four-center integrals are sparse, and using Cauchy-Schwartz inequality, we can find an upper bound of each integral before carrying out explicit evaluations. Those that are smaller than exx_schwarz_threshold will be truncated. The larger the threshold is, the faster the calculation and the lower the accuracy. A relatively safe choice of the value is 1d-4.
+    - *Description*: In practice the four-center integrals are sparse, and using Cauchy-Schwartz inequality, we can find an upper bound of each integral before carrying out explicit evaluations. Those that are smaller than exx_schwarz_threshold will be truncated. The larger the threshold is, the faster the calculation and the lower the accuracy. A relatively safe choice of the value is 1d-5.
     - *Default*: 0
 
     [back to top](#input-file)
 
 - exx_cauchy_threshold<a id="exx-cauchy-threshold"></a>
     - *Type*: Real
-    - *Description*: In practice the Fock exchange matrix is sparse, and using Cauchy-Schwartz inequality, we can find an upper bound of each matrix element before carrying out explicit evaluations. Those that are smaller than exx_cauchy_threshold will be truncated. The larger the threshold is, the faster the calculation and the lower the accuracy. A relatively safe choice of the value is 1d-6.
+    - *Description*: In practice the Fock exchange matrix is sparse, and using Cauchy-Schwartz inequality, we can find an upper bound of each matrix element before carrying out explicit evaluations. Those that are smaller than exx_cauchy_threshold will be truncated. The larger the threshold is, the faster the calculation and the lower the accuracy. A relatively safe choice of the value is 1d-7.
     - *Default*: 0
 
     [back to top](#input-file)
@@ -1567,17 +1567,6 @@ This part of variables are used to control berry phase and wannier90 interfacae
 
     [back to top](#input-file)
 
-- new_dm<a id="new-dm"></a>
-
-    - *Type*: Integer
-    - *Description*: Controls output of some debug information related to our density matrix data-structures.
-        - 1: show no debug information
-        - 2: only show key debug information
-        - 3: show all detail debug information
-    - *Default*: 1
-
-    [back to top](#input-file)
-
 - test_just_neighbor<a id="test-just-neighbor"></a>
     - *Type*: Boolean
     - *Description*: If set to 1, then only perform the neighboring atoms search.

diff --git a/source/Makefile b/source/Makefile
@@ -47,8 +47,8 @@ HONG_MEM = ${HONG_FFTW} -D__FP -D_MCD_CHECK -DWIN32 -DMCD_VERBOSE
 
 #!!!!!!!!!!!!!!!!!!!! CHANE HERE IF YOU LIKE !!!!!!!!!!!!!!
 #! change series version or parallel version~~~
-HONG=${HONG_MPI_SELINV_20210523}
-#HONG=${HONG_SER_SELINV}
+#HONG=${HONG_MPI_SELINV_20210523}
+HONG=${HONG_SER_SELINV}
 #!!!!!!!!!!!!!!!!!!!! CHANE HERE IF YOU LIKE !!!!!!!!!!!!!!
 
 #==========================
@@ -70,6 +70,7 @@ $(OBJS_COMMON)\
 main.o\
 
 FP_OBJS=$(patsubst %.o, ${OBJ_DIR}/%.o, ${FP_OBJS_0})
+FP_OBJS_serial=$(patsubst %.o, ${OBJ_DIR_serial}/%.o, ${FP_OBJS_0})
 PDIAG_OBJS=$(patsubst %.o, ${OBJ_DIR}/%.o, ${OBJS_PDIAG})
 PDIAG_MR=$(patsubst %.o, ${OBJ_DIR}/%.o, ${PDIAG_MR_0})
 
@@ -81,7 +82,7 @@ fp_mpi :
 	@ make -j $(NP) parallel
 
 fp : 
-	@ make init
+	@ make init_serial
 	@ make -j $(NP) serial
 
 fp_gbd : 
@@ -101,11 +102,16 @@ init :
 	@ if [ ! -d $(OBJ_DIR)/README ]; then echo "This directory contains all of the .o files" > $(OBJ_DIR)/README; fi
 	@ if [ ! -d ../bin ]; then mkdir ../bin; fi
 
+init_serial :
+	@ if [ ! -d $(OBJ_DIR_serial) ]; then mkdir $(OBJ_DIR_serial); fi
+	@ if [ ! -d $(OBJ_DIR_serial)/README ]; then echo "This directory contains all of the .o files" > $(OBJ_DIR_serial)/README; fi
+	@ if [ ! -d ../bin ]; then mkdir ../bin; fi
+
 parallel : ${FP_OBJS} ${PDIAG_OBJS} ${PDIAG_MR} ${HEADERS}
 	${CPLUSPLUS_MPI} ${OPTS} ${OPTS_MPI} $(FP_OBJS) ${PDIAG_OBJS} ${PDIAG_MR} ${LIBS} -o  ../bin/${VERSION}.mpi
 
-serial : ${FP_OBJS} ${HEADERS} 
-	${CPLUSPLUS} ${OPTS} $(FP_OBJS) ${LIBS} -o ${VERSION}.fp.x 
+serial : ${FP_OBJS_serial} ${HEADERS} 
+	${CPLUSPLUS} ${OPTS} $(FP_OBJS_serial) ${LIBS} -o ${VERSION}.fp.x 
 
 serial_gdb : ${FP_OBJS} ${HEADERS}
 	${CPLUSPLUS} ${OPTS_GDB} $(FP_OBJS) ${LIBS} -o ${VERSION}.fp_gdb.x
@@ -124,8 +130,14 @@ ${OBJ_DIR}/%.o:%.cpp
 ${OBJ_DIR}/%.o:%.f
 	${FORTRAN} -c ${HONG} $< -o $@	 
 
+${OBJ_DIR_serial}/%.o:%.cpp
+	${CPLUSPLUS} ${OPTS} ${OPTS_MPI} -c ${HONG} $< -o $@
+${OBJ_DIR_serial}/%.o:%.f
+	${FORTRAN} -c ${HONG} $< -o $@
+
 .PHONY:clean
 clean:
 	@ if [ -d $(OBJ_DIR) ]; then rm -rf $(OBJ_DIR); fi
+	@ if [ -d $(OBJ_DIR_serial) ]; then rm -rf $(OBJ_DIR_serial); fi
 	@ if [ -d ../bin ]; then rm -rf ../bin; fi
 	@ cd src_pw; make clean
diff --git a/source/Makefile.Objects b/source/Makefile.Objects
@@ -176,7 +176,6 @@ FORCE_k.o\
 parallel_orbitals.o \
 global_fp.o \
 pdiag_double.o \
-pdiag_basic.o \
 pdiag_common.o \
 diag_scalapack_gvx.o \
 subgrid_oper.o \
@@ -213,7 +212,7 @@ exx_abfs-screen-cauchy.o \
 exx_lcao.o \
 exx_opt_orb.o \
 exx_opt_orb-print.o \
-wfc_dm_2d.o \
+dm_2d.o \
 unk_overlap_lcao.o \
 cal_r_overlap_R.o \
 bessel_basis.o\

diff --git a/source/Makefile.vars b/source/Makefile.vars
@@ -6,34 +6,40 @@ CPLUSPLUS      = icpc
 CPLUSPLUS_MPI = mpiicpc
 #CPLUSPLUS_MPI = /public/intel2017/impi/2017.1.132/intel64/bin/mpiicpc
 
-LAPACK_DIR    = $(MKLROOT)
-#LAPACK_DIR = /public/intel2017/compilers_and_libraries_2017.1.132/linux/mkl
-#LAPACK_DIR = $(MKLROOT)
-#LAPACK_DIR    = /public/intel2017/mkl
-
-FFTW_DIR = /home/mohan/1_Software/impi_fftw-3.3.8
-#FFTW_DIR = /home/qianrui/intelcompile/impi_fftw
-#FFTW_DIR       = /public/udata/xiaohui/software/fftw2
-#FFTW_DIR       =/opt/fftw/3.3.6-p12/intel/2017.update4
-#FFTW_DIR      = /public/fftw-3.3.8
-
-BOOST_DIR = /home/mohan/1_Software/impi_boost-1.70.0
-#BOOST_DIR = /home/qianrui/intelcompile/impi_boost
-#BOOST_DIR      = /public/udata/xiaohui/software/boost_1_39_0
-#BOOST_DIR      = /opt/boost/1.64.0
-
-ELPA_DIR = /home/mohan/1_Software/impi_elpa-16.05.005
-#ELPA_DIR = /home/qianrui/intelcompile/impi_elpa
-#ELPA_DIR   = /public/udata/xiaohui/ELPA-2016.05.004
-#ELPA_DIR = /opt/elpa/intel_2017_update4
-
-CEREAL_DIR = /home/mohan/1_Software/ABACUS_Github/cereal/cereal/
-#CEREAL_DIR = /home/qianrui/intelcompile/cereal/
-
-#DeePKS
-LIBTORCH_DIR = /home/fortneu49/soft/libtorch
-LIBNPY_DIR = /home/fortneu49/soft/libnpy
-#\DeePKS
+#==========================
+# LIB and INCLUDE
+#==========================
+HONG_LAPACK     = -D__MKL -DMKL_ILP64
+LAPACK_DIR    	= $(MKLROOT)
+LAPACK_INCLUDE  = -I${LAPACK_DIR}/include
+LAPACK_LIB_DIR  = ${LAPACK_DIR}/lib/intel64
+LAPACK_LIB      = -L${LAPACK_LIB_DIR} -Wl,--start-group -lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core -lmkl_scalapack_lp64 -lmkl_blacs_intelmpi_lp64 -Wl,--end-group -Wl,-rpath=${LAPACK_LIB_DIR}
+
+HONG_FFTW       = -D__FFTW3
+FFTW_DIR		= /home/liyuanbo/soft/fftw
+FFTW_INCLUDE    = -I${FFTW_DIR}/include
+FFTW_LIB_DIR    = ${FFTW_DIR}/lib
+FFTW_LIB        = -L${FFTW_LIB_DIR} -lfftw3 -Wl,-rpath=${FFTW_LIB_DIR}
+
+ELPA_DIR		= /home/liyuanbo/soft/elpa-2021.05.002/build
+ELPA_INCLUDE    = -I${ELPA_DIR}/include/elpa-2021.05.002
+ELPA_LIB_DIR	= ${ELPA_DIR}/lib
+ELPA_LIB     	= -L${ELPA_LIB_DIR} -lelpa -Wl,-rpath=${ELPA_LIB_DIR}
+
+#LIBXC_INCLUDE   = -I${LIBXC_DIR}/include
+#LIBXC_LIB_DIR   = ${LIBXC_DIR}/lib
+#LIBXC_LIB       = -L${LIBXC_LIB_DIR} -lxc -Wl,-rpath=${LIBXC_LIB_DIR}
+
+CEREAL_DIR 		= /home/liyuanbo/soft/cereal/
+CEREAL_INCLUDE 	= -I${CEREAL_DIR}/include
+
+#============DeePKS================
+#LIBTORCH_INCLUDE=  -isystem ${LIBTORCH_DIR}/include -isystem ${LIBTORCH_DIR}/include/torch/csrc/api/include
+#LIBTORCH_LIB_DIR= ${LIBTORCH_DIR}/lib
+#LIBTORCH_LIB 	= -L${LIBTORCH_LIB_DIR} -ltorch -lc10 -Wl,-rpath,${LIBTORCH_LIB_DIR} -Wl,--no-as-needed,"${LIBTORCH_LIB_DIR}/libtorch_cpu.so" -Wl,--as-needed ${LIBTORCH_LIB_DIR}/libc10.so -lpthread -Wl,--no-as-needed,"${LIBTORCH_LIB_DIR}/libtorch.so" -Wl,--as-needed 
+
+#CNPY_INCLUDE	= -I${LIBNPY_DIR}
 
 OBJ_DIR = obj
+OBJ_DIR_serial = obj_serial
 NP      = 14
diff --git a/source/driver.cpp b/source/driver.cpp
@@ -104,9 +104,12 @@ void Driver::atomic_world(void)
 	}
 #ifdef __LCAO
 	else if(GlobalV::BASIS_TYPE=="lcao")
-	{
-		Run_lcao::lcao_line(p_ensolver);
-	}
+    {
+        use_ensol = "ksdft_lcao";
+        ModuleEnSover::init_esolver(p_ensolver, use_ensol);
+        Run_lcao::lcao_line(p_ensolver);
+        ModuleEnSover::clean_esolver(p_ensolver);
+    }
 #endif
 
 	ModuleBase::timer::finish( GlobalV::ofs_running );

diff --git a/source/input.cpp b/source/input.cpp
@@ -420,7 +420,6 @@ void Input::Default(void)
 
 	cell_factor = 1.2; //LiuXh add 20180619
 
-	new_dm=1; // Shen Yu add 2019/5/9
 	GlobalV::mulliken=0;// qi feng add 2019/9/10
 
 //----------------------------------------------------------			//Peize Lin add 2020-04-04
@@ -1578,10 +1577,6 @@ bool Input::Read(const std::string &fn)
 		{
 			read_value(ifs, cell_factor);
 		}
-		else if (strcmp("newdm", word) == 0)
-		{
-			read_value(ifs, new_dm);
-		}
 		else if (strcmp("test_just_neighbor", word) == 0)
 		{
 			read_value(ifs, test_just_neighbor);
@@ -2283,7 +2278,6 @@ void Input::Bcast()
 
 		//Parallel_Common::bcast_int( epsilon0_choice );
     Parallel_Common::bcast_double( cell_factor); //LiuXh add 20180619
-    Parallel_Common::bcast_int( new_dm ); // Shen Yu add 2019/5/9
     Parallel_Common::bcast_bool( restart_save ); // Peize Lin add 2020.04.04
     Parallel_Common::bcast_bool( restart_load ); // Peize Lin add 2020.04.04
 

diff --git a/source/input.h b/source/input.h
@@ -415,17 +415,6 @@ class Input
 	bool input_error;
     double cell_factor; //LiuXh add 20180619
 
-//==========================================================
-// new DM algorithm test
-// add by Shen Yu @ 2019/5/9
-// newDM values:
-//  0: not use new DM algorithm;
-//  1: use new DM algorithm and show no debug information
-//  2: use new DM algorithm and only show key debug information
-//  3: use new DM algorithm and show all detail debug information
-//==========================================================
-    int new_dm;
-
 //==========================================================
 //    DFT+U       Xin Qu added on 2020-10-29
 //==========================================================

diff --git a/source/input_conv.cpp b/source/input_conv.cpp
@@ -15,7 +15,9 @@
 #include "src_pw/occupy.h"
 #include "module_base/global_function.h"
 #include "module_base/global_variable.h"
+#ifdef __EXX
 #include "src_ri/exx_abfs-jle.h"
+#endif
 #ifdef __LCAO
 #include "module_orbital/ORB_read.h"
 #include "src_lcao/ELEC_evolve.h"
@@ -446,6 +448,7 @@ void Input_Conv::Convert(void)
 //----------------------------------------------------------
 // about exx, Peize Lin add 2018-06-20
 //----------------------------------------------------------
+#ifdef __MPI // liyuanbo 2022/2/23
 #ifdef __LCAO
 	if (INPUT.exx_hybrid_type == "no")
 	{
@@ -503,10 +506,9 @@ void Input_Conv::Convert(void)
 		Exx_Abfs::Jle::tolerence = INPUT.exx_opt_orb_tolerence;
 	}
 #endif
-
+#endif
 	GlobalC::ppcell.cell_factor = INPUT.cell_factor; // LiuXh add 20180619
 
-	//    NEW_DM=INPUT.new_dm;  // Shen Yu add 2019/5/9
 
 	//----------------------------------------------------------
 	// main parameters / electrons / spin ( 2/16 )
@@ -545,10 +547,10 @@ void Input_Conv::Convert(void)
 	GlobalC::en.out_dos = INPUT.out_dos;
 	GlobalC::en.out_band = INPUT.out_band;
 #ifdef __LCAO
-	GlobalC::LOC.out_dm = INPUT.out_dm;
-	GlobalC::ParaO.out_hs = INPUT.out_hs;
-	GlobalC::ParaO.out_hsR = INPUT.out_hs2; // LiuXh add 2019-07-16
-	GlobalC::ParaO.out_lowf = INPUT.out_lowf;
+	Local_Orbital_Charge::out_dm = INPUT.out_dm;
+	Pdiag_Double::out_hs = INPUT.out_hs;
+	Pdiag_Double::out_hsR = INPUT.out_hs2; // LiuXh add 2019-07-16
+	Pdiag_Double::out_lowf = INPUT.out_lowf;
 #endif
 
 	GlobalC::en.dos_emin_ev = INPUT.dos_emin_ev;

diff --git a/source/input_update.cpp b/source/input_update.cpp
@@ -180,8 +180,8 @@ bool Update_input::Read(const std::string &fn)
 				{
 					out_dm = 10000;
 #ifdef __FP
-					this->change(GlobalV::ofs_warning,"out_dm",GlobalC::LOC.out_dm, out_dm);
-					GlobalC::LOC.out_dm=out_dm;
+					this->change(GlobalV::ofs_warning,"out_dm",Local_Orbital_Charge::out_dm, out_dm);
+					Local_Orbital_Charge::out_dm=out_dm;
 #endif
 				}
 			}
@@ -202,10 +202,10 @@ bool Update_input::Read(const std::string &fn)
         {
             read_value(ifs, out_dm);
 #ifdef __FP
-			if(out_dm!=GlobalC::LOC.out_dm)
+			if(out_dm!=Local_Orbital_Charge::out_dm)
 			{
-				this->change(GlobalV::ofs_warning,"out_dm",GlobalC::LOC.out_dm,out_dm);
-				GlobalC::LOC.out_dm = this->out_dm;
+				this->change(GlobalV::ofs_warning,"out_dm",Local_Orbital_Charge::out_dm,out_dm);
+				Local_Orbital_Charge::out_dm = this->out_dm;
 			}
 #endif
         }
@@ -225,10 +225,10 @@ bool Update_input::Read(const std::string &fn)
             read_value(ifs, out_lowf);
 #ifdef __FP
 			//if(out_lowf!=out_lowf)
-			if(out_lowf!=GlobalC::ParaO.out_lowf)		// Peize Lin change out_lowf to GlobalC::ParaO.out_lowf at 2020.01.31
+			if(out_lowf!=Pdiag_Double::out_lowf)		// Peize Lin change out_lowf to GlobalC::ParaO.out_lowf at 2020.01.31
 			{
-				this->change(GlobalV::ofs_warning,"out_lowf",GlobalC::ParaO.out_lowf,out_lowf);
-				GlobalC::ParaO.out_lowf = this->out_lowf;
+				this->change(GlobalV::ofs_warning,"out_lowf",Pdiag_Double::out_lowf,out_lowf);
+				Pdiag_Double::out_lowf = this->out_lowf;
 			}
 #endif
         }
@@ -273,8 +273,8 @@ void Update_input::Bcast()
     Parallel_Common::bcast_double( GlobalV::FORCE_THR);
 #ifdef __LCAO
     Parallel_Common::bcast_double( Force_Stress_LCAO::force_invalid_threshold_ev);
-    Parallel_Common::bcast_int( GlobalC::LOC.out_dm );
-    Parallel_Common::bcast_int( GlobalC::ParaO.out_lowf );
+    Parallel_Common::bcast_int( Local_Orbital_Charge::out_dm );
+    Parallel_Common::bcast_int( Pdiag_Double::out_lowf );
 #endif
     Parallel_Common::bcast_double( GlobalV::DRHO2 );
     Parallel_Common::bcast_int( GlobalV::NITER );

diff --git a/source/module_base/math_bspline.cpp b/source/module_base/math_bspline.cpp
@@ -32,18 +32,12 @@ namespace ModuleBase
         }
     }
 
-    void Bspline::cleanp()
-    {
-        delete[] bezier;
-        bezier = NULL;
-    }
-
     double Bspline::bezier_ele(int n)
     {
         return this->bezier[n];
     }
 
-    void Bspline::getbslpine(double x)
+    void Bspline::getbspline(double x)
     {
         bezier[0] = 1.0;
         for(int k = 1 ; k <= norder ; ++k)