Merge branch 'develop' into convert_vtxc_v

Author: PeizeLin

.github/workflows/coverage.yml

@@ -15,10 +15,10 @@ jobs:
         uses: actions/checkout@v4
       - name: Install Requirements for Coverage Testing
         run: |
-          apt update && apt install -y lcov
+          apt update && apt install -y lcov gpg
       - name: Building
         run: |
-          cmake -B build -DENABLE_DEEPKS=ON -DENABLE_LIBXC=ON -DBUILD_TESTING=ON -DENABLE_COVERAGE=ON
+          cmake -B build -DENABLE_COVERAGE=ON -DBUILD_TESTING=ON -DENABLE_DEEPKS=ON -DENABLE_LIBXC=ON -DENABLE_LIBRI=ON -DENABLE_PAW=ON -DENABLE_GOOGLEBENCH=ON -DENABLE_RAPIDJSON=ON
           cmake --build build -j`nproc`
           cmake --install build
       - name: Testing

docs/advanced/input_files/input-main.md

@@ -39,7 +39,6 @@
     - [pw\_diag\_thr](#pw_diag_thr)
     - [pw\_diag\_nmax](#pw_diag_nmax)
     - [pw\_diag\_ndim](#pw_diag_ndim)
-    - [diago\_full\_acc](#diago_full_acc)
     - [erf\_ecut](#erf_ecut)
     - [fft\_mode](#fft_mode)
     - [erf\_height](#erf_height)
@@ -779,12 +778,6 @@ These variables are used to control the plane wave related parameters.
 - **Description**: Only useful when you use `ks_solver = dav` or `ks_solver = dav_subspace`. It indicates dimension of workspace(number of wavefunction packets, at least 2 needed) for the Davidson method. A larger value may yield a smaller number of iterations in the algorithm but uses more memory and more CPU time in subspace diagonalization.
 - **Default**: 4
 
-### diago_full_acc
-
-- **Type**: bool
-- **Description**: Only useful when you use `ks_solver = dav_subspace`. If `TRUE`, all the empty states are diagonalized at the same level of accuracy of the occupied ones. Otherwise the empty states are diagonalized using a larger threshold (10-5) (this should not affect total energy, forces, and other ground-state properties).
-- **Default**: false
-
 ### erf_ecut
 
 - **Type**: Real
@@ -925,7 +918,7 @@ calculations.
   - **cg**: cg method.
   - **bpcg**: bpcg method, which is a block-parallel Conjugate Gradient (CG) method, typically exhibits higher acceleration in a GPU environment.
   - **dav**: the Davidson algorithm.
-  - **dav_subspace**: subspace Davidson algorithm
+  - **dav_subspace**: Davidson algorithm without orthogonalization operation, this method is the most recommended for efficiency. `pw_diag_ndim` can be set to 2 for this method.
 
   For atomic orbitals basis,
 

docs/advanced/install.md

@@ -93,9 +93,9 @@ cmake -B build -DUSE_CUDA=1 -DCMAKE_CUDA_COMPILER=${path to cuda toolkit}/bin/nv
 
 ## Build math library from source
 
-> Note: This flag is **enabled by default**. It will get better performance than the standard implementation on `gcc` and `clang`. But it **will be disabled** when using `Intel Compiler` since the math functions will get wrong results and the performance is also unexpectly poor.
+> Note: We recommend using the latest available compiler sets, since they offer faster implementations of math functions.
 
-To build math functions from source code, instead of using c++ standard implementation, define `USE_ABACUS_LIBM` flag.
+This flag is disabled by default. To build math functions from source code, define `USE_ABACUS_LIBM` flag. It is expected to get a better performance on legacy versions of `gcc` and `clang`.
 
 Currently supported math functions:
  `sin`, `cos`, `sincos`, `exp`, `cexp`

examples/lr-tddft/lcao_Si2/INPUT

@@ -37,4 +37,3 @@ out_alllog	1
 
 nvirt 19
 abs_wavelen_range  100 175
-#diago_full_acc 1

python/pyabacus/src/py_diago_dav_subspace.hpp

@@ -106,7 +106,7 @@ class PyDiagoDavSubspace
         double tol,
         int max_iter,
         bool need_subspace,
-        std::vector<bool> is_occupied,
+        std::vector<double> diag_ethr,
         bool scf_type,
         hsolver::diag_comm_info comm_info
     ) {
@@ -141,7 +141,7 @@ class PyDiagoDavSubspace
             comm_info
         );
 
-        return obj->diag(hpsi_func, psi, nbasis, eigenvalue, is_occupied, scf_type);
+        return obj->diag(hpsi_func, psi, nbasis, eigenvalue, diag_ethr.data(), scf_type);
     }
 
 private:

python/pyabacus/src/py_diago_david.hpp

@@ -132,7 +132,6 @@ class PyDiagoDavid
             const std::complex<double> *psi_in, 
             std::complex<double> *spsi_out, 
             const int nrow, 
-            const int npw, 
             const int nbands
         ) {
             syncmem_op()(this->ctx, this->ctx, spsi_out, psi_in, static_cast<size_t>(nbands * nrow));

python/pyabacus/src/py_hsolver.cpp

@@ -59,8 +59,8 @@ void bind_hsolver(py::module& m)
                 The maximum number of iterations.
             need_subspace : bool
                 Whether to use the subspace function.
-            is_occupied : list[bool]
-                A list of boolean values indicating whether the band is occupied,
+            diag_ethr : list[float]
+                A list of float values indicating the thresholds of each band for the diagonalization,
                 meaning that the corresponding eigenvalue is to be calculated.
             scf_type : bool
                 Whether to use the SCF type, which is used to determine the
@@ -76,7 +76,7 @@ void bind_hsolver(py::module& m)
         "tol"_a, 
         "max_iter"_a, 
         "need_subspace"_a, 
-        "is_occupied"_a, 
+        "diag_ethr"_a, 
         "scf_type"_a, 
         "comm_info"_a)
         .def("set_psi", &py_hsolver::PyDiagoDavSubspace::set_psi, R"pbdoc(

python/pyabacus/src/pyabacus/hsolver/_hsolver.py

@@ -25,7 +25,7 @@ def dav_subspace(
     tol: float = 1e-2,
     max_iter: int = 1000,
     need_subspace: bool = False,
-    is_occupied: Union[List[bool], None] = None,
+    diag_ethr: Union[List[float], None] = None,
     scf_type: bool = False
 ) -> Tuple[NDArray[np.float64], NDArray[np.complex128]]:
     """ A function to diagonalize a matrix using the Davidson-Subspace method.
@@ -52,10 +52,8 @@ def dav_subspace(
         The maximum number of iterations, by default 1000.
     need_subspace : bool, optional
         Whether to use subspace function, by default False.
-    is_occupied : List[bool] | None, optional
-        The list of occupied bands, by default None. This indicates how many eigenvalues 
-        need to be calculated, starting from the smallest eigenvalue. Only the energy levels 
-        occupied by electrons (occupied) need to be calculated.
+    diag_ethr : List[float] | None, optional
+        The list of thresholds of bands, by default None.
     scf_type : bool, optional
         Indicates whether the calculation is a self-consistent field (SCF) calculation. 
         If True, the initial precision of eigenvalue calculation can be coarse. 
@@ -72,8 +70,8 @@ def dav_subspace(
     if not callable(mvv_op):
         raise TypeError("mvv_op must be a callable object.")
 
-    if is_occupied is None:
-        is_occupied = [True] * num_eigs
+    if diag_ethr is None:
+        diag_ethr = [tol] * num_eigs
 
     if init_v.ndim != 1 or init_v.dtype != np.complex128:
         init_v = init_v.flatten().astype(np.complex128, order='C')
@@ -93,7 +91,7 @@ def dav_subspace(
         tol,
         max_iter,
         need_subspace,
-        is_occupied,
+        diag_ethr,
         scf_type,
         comm_info
     )
@@ -113,7 +111,6 @@ def davidson(
     tol: float = 1e-2,
     max_iter: int = 1000,
     use_paw: bool = False,
-    # is_occupied: Union[List[bool], None] = None,
     # scf_type: bool = False
 ) -> Tuple[NDArray[np.float64], NDArray[np.complex128]]:
     """ A function to diagonalize a matrix using the Davidson-Subspace method.

source/Makefile.Objects

@@ -568,12 +568,9 @@ OBJS_LCAO=evolve_elec.o\
       FORCE_STRESS.o\
 	  FORCE_gamma.o\
 	  FORCE_k.o\
-	  fvl_dphi_gamma.o\
-	  fvl_dphi_k.o\
-	  fedm_gamma.o\
-	  fedm_k.o\
-	  ftvnl_dphi_gamma.o\
-	  ftvnl_dphi_k.o\
+      stress_tools.o\
+      edm.o\
+      pulay_force_stress_center2.o\
 	  fvnl_dbeta_gamma.o\
 	  fvnl_dbeta_k.o\
 	  grid_init.o\

source/driver.cpp

@@ -41,7 +41,7 @@ void Driver::init()
 
     // (3) output information
     time_t time_finish = std::time(nullptr);
-    Print_Info::print_time(time_start, time_finish);
+    ModuleIO::print_time(time_start, time_finish);
 
     // (4) close all of the running logs
     ModuleBase::Global_File::close_all_log(GlobalV::MY_RANK, PARAM.inp.out_alllog,PARAM.inp.calculation);