merge Develop to update_MD

.github/workflows/test.yml

@@ -4,9 +4,7 @@ on:
   pull_request_target:
     branches:
       - develop
-      - reconstruction
-      - newelpa
-
+      - update_MD
 jobs:
 
   start-runner:

CMakeLists.txt

@@ -47,7 +47,7 @@ if(USE_OPENMP)
   find_package(OpenMP REQUIRED)
   target_link_libraries(${ABACUS_BIN_NAME} OpenMP::OpenMP_CXX)
   add_compile_options(${OpenMP_CXX_FLAGS})
-  add_compile_definitions(__OPENMP)
+  #add_compile_definitions(_OPENMP)
 endif()
 
 set(CMAKE_CXX_STANDARD 11)
@@ -135,7 +135,7 @@ if(ENABLE_DEEPKS)
   find_package(Torch REQUIRED)
   include_directories(${TORCH_INCLUDE_DIRS})
   target_link_libraries(${ABACUS_BIN_NAME} ${TORCH_LIBRARIES})
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${TORCH_CXX_FLAGS}")
+  add_compile_options(${TORCH_CXX_FLAGS})
 
   find_path(libnpy_SOURCE_DIR
     npy.hpp

Dockerfile.gnu

@@ -39,8 +39,8 @@ RUN cd /tmp \
     && cd /tmp && rm -rf fftw-3.3.9 && rm fftw-3.3.9.tar.gz
 
 RUN cd /tmp \
-    && wget https://download.pytorch.org/libtorch/cpu/libtorch-shared-with-deps-1.9.0%2Bcpu.zip --no-check-certificate --quiet \
-    && unzip libtorch-shared-with-deps-1.9.0+cpu.zip \
+    && wget https://download.pytorch.org/libtorch/cpu/libtorch-cxx11-abi-shared-with-deps-1.9.1%2Bcpu.zip --no-check-certificate --quiet \
+    && unzip libtorch-cxx11-abi-shared-with-deps-1.9.1+cpu.zip \
     && cp -r libtorch/include /usr/local \
     && cp -r libtorch/lib /usr/local \
     && cp -r libtorch/share /usr/local \

source/Makefile

@@ -32,11 +32,11 @@ HONG_SER = -D__FP -D__LCAO  ${HONG_FFTW}
 HONG_SER_SELINV = -D__FP -D__LCAO ${HONG_FFTW} -D__SELINV
 HONG_GDB = -g -D__FP -D__LCAO  ${HONG_FFTW}
 #(2)mpi
-HONG_MPI = -D__FP -D__LCAO  ${HONG_FFTW} -D__MPI -D__OPENMP 
+HONG_MPI = -D__FP -D__LCAO  ${HONG_FFTW} -D__MPI
 # mohan comment out 2021-02-06, add -DUSE_LIBXC=0 if you want to use LIBXC
 #HONG_MPI_SELINV = -D__FP ${HONG_FFTW} ${HONG_LAPACK} -D__MPI -D__SELINV -DMETIS -DEXX_DM=3 -DEXX_H_COMM=2 -DUSE_LIBXC=0 -DTEST_EXX_LCAO=0 -DTEST_EXX_RADIAL=1 -DUSE_CEREAL_SERIALIZATION
-HONG_MPI_SELINV = -D__FP ${HONG_FFTW} ${HONG_LAPACK} -D__MPI -D__LCAO  -D__OPENMP -D__SELINV -DMETIS -DEXX_DM=3 -DEXX_H_COMM=2 -DTEST_EXX_LCAO=0 -DTEST_EXX_RADIAL=1 -DUSE_CEREAL_SERIALIZATION
-HONG_MPI_SELINV_20210523 = -D__FP ${HONG_FFTW} ${HONG_LAPACK} -D__LCAO -D__MPI -D__OPENMP -D__SELINV -DMETIS -DEXX_DM=3 -DEXX_H_COMM=2 -DTEST_EXX_LCAO=0 -DTEST_EXX_RADIAL=1 -DUSE_CEREAL_SERIALIZATION -D__EXX
+HONG_MPI_SELINV = -D__FP ${HONG_FFTW} ${HONG_LAPACK} -D__MPI -D__LCAO -D__SELINV -DMETIS -DEXX_DM=3 -DEXX_H_COMM=2 -DTEST_EXX_LCAO=0 -DTEST_EXX_RADIAL=1 -DUSE_CEREAL_SERIALIZATION
+HONG_MPI_SELINV_20210523 = -D__FP ${HONG_FFTW} ${HONG_LAPACK} -D__LCAO -D__MPI -D__SELINV -DMETIS -DEXX_DM=3 -DEXX_H_COMM=2 -DTEST_EXX_LCAO=0 -DTEST_EXX_RADIAL=1 -DUSE_CEREAL_SERIALIZATION -D__EXX
 HONG_DEEPKS = ${HONG_MPI_SELINV_20210523} -D__DEEPKS	#caoyu add 2021-07-15 , use it in DeePKS. Add LIBTORCH_DIR and LIBNPY_DIR, modify ${LIBS} and build with std=c++14 !!
 #(3)memory
 #(3)memory

source/Makefile.Objects

@@ -164,6 +164,8 @@ LCAO_nnr.o \
 LCAO_diago.o\
 LCAO_evolve.o\
 LCAO_descriptor.o\
+LCAO_descriptor_old.o\
+LCAO_descriptor_new.o\
 ylm.o\
 FORCE_STRESS.o\
 FORCE_gamma.o\
@@ -176,6 +178,7 @@ global_fp.o \
 pdiag_double.o \
 pdiag_basic.o \
 pdiag_common.o \
+diag_scalapack_gvx.o \
 subgrid_oper.o \
 conv_coulomb_pot.o \
 conv_coulomb_pot_k.o \
@@ -225,6 +228,7 @@ parallel_reduce.o\
 parallel_pw.o\
 ft.o\
 parallel_grid.o\
+parallel_deepks.o \
 
 OBJS_FIRST_PRINCIPLES=$(OBJS_MAIN)\
 $(OBJS_PW)\

source/input_conv.cpp

@@ -629,12 +629,13 @@ void Input_Conv::Convert(void)
 	}
 
 	ModuleBase::timer::tick("Input_Conv","Convert");
-    return;
-
 	//-----------------------------------------------
 	//caoyu add for DeePKS
 	//-----------------------------------------------
+	#ifdef __DEEPKS
 	GlobalV::out_descriptor = INPUT.out_descriptor; 
 	GlobalV::deepks_scf = INPUT.deepks_scf;
+	#endif
 
+    return;
 }

source/module_base/complexmatrix.cpp

@@ -414,20 +414,25 @@ ComplexMatrix conj(const ComplexMatrix &m)
 }
 
 // Peize Lin add 2021.09.08
-std::ostream & ComplexMatrix::print( std::ostream & os, const double threshold_norm, const double threshold_imag ) const
+std::ostream & ComplexMatrix::print( std::ostream & os, const double threshold_abs, const double threshold_imag ) const
 {
 	for( int ir=0; ir!=this->nr; ++ir )
 	{
 		for( int ic=0; ic!=this->nc; ++ic )
-			if(std::norm((*this)(ir,ic))>threshold_norm)
+		{
+			const std::complex<double> & data = (*this)(ir,ic);
+			if(std::abs(data)>threshold_abs)
 			{
-				if(std::imag((*this)(ir,ic))>threshold_imag)
-					os<<(*this)(ir,ic)<<"\t";
+				if(std::abs(std::imag(data))>threshold_imag)
+					os<<data<<"\t";
 				else
-					os<<std::real((*this)(ir,ic))<<"\t";
+					os<<std::real(data)<<"\t";
 			}
 			else
+			{
 				os<<0<<"\t";
+			}
+		}
 		os<<std::endl;
 	}	
 	return os;

source/module_base/complexmatrix.h

@@ -56,7 +56,7 @@ class ComplexMatrix
 	void zero_out(void);
 	void set_as_identity_matrix(void);
 
-	std::ostream & print( std::ostream & os, const double threshold_norm=0.0, const double threshold_imag=0.0 ) const;		// Peize Lin add 2021.09.08
+	std::ostream & print( std::ostream & os, const double threshold_abs=0.0, const double threshold_imag=0.0 ) const;		// Peize Lin add 2021.09.08
 };
 
 ComplexMatrix operator+(const ComplexMatrix &m1,  const ComplexMatrix &m2);

source/module_base/mathzone_add1.cpp

@@ -21,7 +21,9 @@
 typedef fftw_complex FFTW_COMPLEX;
 
 //#include <unistd.h>
+#ifdef _OPENMP
 #include <omp.h>
+#endif
 
 namespace ModuleBase
 {
@@ -980,7 +982,9 @@ void Mathzone_Add1::Cubic_Spline_Interpolation
 {	
 	ModuleBase::timer::tick("Mathzone_Add1","Cubic_Spline_Interpolation");
 
+#ifdef _OPENMP
 	#pragma omp parallel for schedule(static)
+#endif
 	for(int m = 0; m < rsize ; m++)
 	{
 		int klo = 0;

source/module_orbital/ORB_atomic_lm.cpp

@@ -4,7 +4,10 @@
 #include "../module_base/timer.h"
 #include "../module_base/math_integral.h"
 #include "../module_base/math_sphbes.h"
+
+#ifdef _OPENMP
 #include <omp.h>
+#endif
 
 Numerical_Orbital_Lm::Numerical_Orbital_Lm()
 {
@@ -224,7 +227,7 @@ void Numerical_Orbital_Lm::extra_uniform(const double &dr_uniform_in, const bool
 
 	// do interpolation here to make grid more dense
 
-#ifdef __OPENMP
+#ifdef _OPENMP
 	#pragma omp parallel for schedule(static)
 #endif
 	for (int ir = 0; ir < this->nr_uniform; ir++)
@@ -268,8 +271,8 @@ void Numerical_Orbital_Lm::extra_uniform(const double &dr_uniform_in, const bool
 		default: 
 			//GlobalV::ofs_warning << " The angular momentum larger than 4 (g orbitals) may be error about eggbox. " << std::endl;
 			//GlobalV::ofs_warning << " Check file " << __FILE__ << " line " << __LINE__ <<std::endl;
-			std::cout << " The angular momentum larger than 4 (g orbitals) may be error about eggbox. " << std::endl;
-			std::cout << " Check file " << __FILE__ << " line " << __LINE__ <<std::endl;
+			//std::cout << " The angular momentum larger than 4 (g orbitals) may be error about eggbox. " << std::endl;
+			//std::cout << " Check file " << __FILE__ << " line " << __LINE__ <<std::endl;
 			ModuleBase::Mathzone_Add1::SplineD2 (ModuleBase::GlobalFunc::VECTOR_TO_PTR(r_radial), ModuleBase::GlobalFunc::VECTOR_TO_PTR(psi), nr, 0.0, 0.0, y2); break;
 	}
 
@@ -545,7 +548,7 @@ void Numerical_Orbital_Lm::cal_kradial_sbpool(void)
 	{
 		r_tmp[ir] *= (ir&1) ? four_three : two_three;
 	}
-#ifdef __OPENMP
+#ifdef _OPENMP
 	#pragma omp parallel for schedule(static)
 #endif
 	for (int ik = 0; ik < nk; ik++)