Merge remote-tracking branch 'upstream/develop' into develop

Author: Flying-dragon-boxing

docs/advanced/input_files/input-main.md

@@ -732,7 +732,7 @@ If only one value is set (such as `kspacing 0.5`), then kspacing values of a/b/c
 
 [back to top](#full-list-of-input-keywords)
 
-## Variables related to input files
+## Input files
 
 These variables are used to control parameters related to input files.
 
@@ -1656,16 +1656,16 @@ These variables are used to control the geometry relaxation.
 
 [back to top](#full-list-of-input-keywords)
 
-## Variables related to output information
+## Output information
 
 These variables are used to control the output of properties.
 
 ### out_freq_ion
 
 - **Type**: Integer
-- **Description**: After self-consistent-field calculations, control the interval of ionic movements for printing properties. These properties cover charge density, local potential, electrostatic potential, Hamiltonian matrix, overlap matrix, density matrix, Mulliken population analysis and so on.
+- **Description**: Control the interval to print information every few ion steps. These properties cover charge density, local potential, electrostatic potential, Hamiltonian matrix, overlap matrix, density matrix, Mulliken population analysis and so on.
 - **Default**: 0
-- **Note**: If you want to use out_freq_elec, please set out_freq_ion to 1, otherwise out_freq_elec is useless
+- **Note**: The integer indicates to print information every 'out_freq_ion' ion steps. 
 
 ### out_freq_elec
 
@@ -1679,11 +1679,11 @@ These variables are used to control the output of properties.
 - **Description**:
   The first integer controls whether to output the charge density on real space grids:
   - 1: Output the charge density (in Bohr^-3) on real space grids into the density files in the folder `OUT.${suffix}`. The files are named as:
-    - nspin = 1: `chgs1.cube`;
+    - nspin = 1: `chg.cube`;
     - nspin = 2: `chgs1.cube`, and `chgs2.cube`;
     - nspin = 4: `chgs1.cube`, `chgs2.cube`, `chgs3.cube`, and `chgs4.cube`;
   Note that by using the Meta-GGA functional, additional files containing the kinetic energy density will be output with the following names:
-    - nspin = 1: `taus1.cube`;
+    - nspin = 1: `tau.cube`;
     - nspin = 2: `taus1.cube`, and `taus2.cube`;
     - nspin = 4: `taus1.cube`, `taus2.cube`, `taus3.cube`, and `taus4.cube`;
   - 2: On top of 1, also output the initial charge density files with a suffix name as '_ini', such as `taus1_ini.cube`, etc.

python/pyabacus/CONTRIBUTING.md

@@ -189,7 +189,7 @@ list(APPEND _diago
     ${HSOLVER_PATH}/diago_david.cpp
     ${HSOLVER_PATH}/diag_const_nums.cpp
     ${HSOLVER_PATH}/diago_iter_assist.cpp
-    ${HSOLVER_PATH}/kernels/dngvd_op.cpp
+    ${HSOLVER_PATH}/kernels/hegvd_op.cpp
     ${HSOLVER_PATH}/kernels/bpcg_kernel_op.cpp
     ${BASE_PATH}/kernels/math_kernel_op.cpp
     ${BASE_PATH}/kernels/math_kernel_op_vec.cpp

python/pyabacus/src/hsolver/CMakeLists.txt

@@ -9,7 +9,7 @@ list(APPEND _diago
     ${HSOLVER_PATH}/diago_pxxxgvx.cpp
 
 
-    ${HSOLVER_PATH}/kernels/dngvd_op.cpp
+    ${HSOLVER_PATH}/kernels/hegvd_op.cpp
     ${HSOLVER_PATH}/kernels/bpcg_kernel_op.cpp
     # dependency
     ${BASE_PATH}/kernels/math_kernel_op.cpp

python/pyabacus/src/hsolver/py_diago_cg.hpp

@@ -145,7 +145,7 @@ class PyDiagoCG
             std::copy(hpsi_ptr, hpsi_ptr + nvec * ld_psi, hpsi_out.data<std::complex<double>>());
         };
 
-        auto subspace_func = [] (const ct::Tensor& psi_in, ct::Tensor& psi_out) { /*do nothing*/ };
+        auto subspace_func = [](const ct::Tensor& psi_in, ct::Tensor& psi_out, const bool S_orth) { /*do nothing*/ };
 
         auto spsi_func = [this] (const ct::Tensor& psi_in, ct::Tensor& spsi_out) {
             const auto ndim = psi_in.shape().ndim();

python/pyabacus/src/hsolver/py_diago_dav_subspace.hpp

@@ -144,7 +144,6 @@ class PyDiagoDavSubspace
             dav_ndim, 
             tol, 
             max_iter, 
-            need_subspace, 
             comm_info,
             diag_subspace,
             nb2d

source/CMakeLists.txt

@@ -35,7 +35,7 @@ list(APPEND device_srcs
   source_pw/module_pwdft/kernels/meta_op.cpp
   source_pw/module_stodft/kernels/hpsi_norm_op.cpp
   source_basis/module_pw/kernels/pw_op.cpp
-  source_hsolver/kernels/dngvd_op.cpp
+  source_hsolver/kernels/hegvd_op.cpp
   source_hsolver/kernels/bpcg_kernel_op.cpp
   source_estate/kernels/elecstate_op.cpp
 
@@ -70,7 +70,7 @@ if(USE_CUDA)
     source_pw/module_stodft/kernels/cuda/hpsi_norm_op.cu
     source_pw/module_pwdft/kernels/cuda/onsite_op.cu
     source_basis/module_pw/kernels/cuda/pw_op.cu
-    source_hsolver/kernels/cuda/dngvd_op.cu
+    source_hsolver/kernels/cuda/hegvd_op.cu
     source_hsolver/kernels/cuda/bpcg_kernel_op.cu
     source_estate/kernels/cuda/elecstate_op.cu
 
@@ -101,7 +101,7 @@ if(USE_ROCM)
     source_pw/module_pwdft/kernels/rocm/onsite_op.hip.cu
     source_pw/module_stodft/kernels/rocm/hpsi_norm_op.hip.cu
     source_basis/module_pw/kernels/rocm/pw_op.hip.cu
-    source_hsolver/kernels/rocm/dngvd_op.hip.cu
+    source_hsolver/kernels/rocm/hegvd_op.hip.cu
     source_hsolver/kernels/rocm/bpcg_kernel_op.hip.cu
     source_estate/kernels/rocm/elecstate_op.hip.cu
 

source/Makefile.Objects

@@ -266,6 +266,7 @@ OBJS_ESOLVER=esolver.o\
     esolver_lj.o\
     esolver_dp.o\
     esolver_of.o\
+    esolver_of_tddft.o\
     esolver_of_tool.o\
     esolver_of_interface.o\
     pw_others.o\
@@ -360,6 +361,7 @@ OBJS_HAMILT_OF=kedf_tf.o\
     kedf_xwm.o\
     kedf_lkt.o\
     kedf_manager.o\
+    evolve_ofdft.o\
 
 OBJS_HAMILT_LCAO=hamilt_lcao.o\
     operator_lcao.o\
@@ -395,7 +397,7 @@ OBJS_HSOLVER=diago_cg.o\
     hsolver_lcaopw.o\
     hsolver_pw_sdft.o\
     diago_iter_assist.o\
-    dngvd_op.o\
+    hegvd_op.o\
     bpcg_kernel_op.o\
     diag_const_nums.o\
     diag_hs_para.o\
@@ -579,6 +581,8 @@ OBJS_IO=input_conv.o\
     output_log.o\
     output_mat_sparse.o\
     ctrl_output_lcao.o\
+    ctrl_output_fp.o\
+    ctrl_output_pw.o\
     para_json.o\
     abacusjson.o\
     general_info.o\

source/source_base/kernels/dsp/dsp_connector.cpp

@@ -6,9 +6,9 @@
 extern "C"
 {
 #define complex_double ignore_complex_double
-#include <mt_hthread_blas.h> // MTBLAS_TRANSPOSE etc
+#include <mt_hthread_blas.h> // include faster mtblas kernels
 #undef complex_double
-#include <mtblas_interface.h> // gemm
+#include <mtblas_interface.h> // include normal mtblas kernels that automatically operate memory, but slower.
 }
 namespace mtfunc
 {
@@ -22,45 +22,42 @@ void dspDestoryHandle(int id)
 {
     hthread_dev_close(id);
     std::cout << " ** DSP closed on cluster " << id << " **" << std::endl;
-} // Close dsp cluster at the end
+} // Close dsp cluster at the end of the program
 
-MTBLAS_TRANSPOSE convertBLASTranspose(const char* blasTrans)
+// MTBlas secretly removed its MTBLAS_TRANSPOSE data type and used the original CBLAS_TRANSPOSE. So this function is modified.
+
+CBLAS_TRANSPOSE convertBLASTranspose(const char* blasTrans)
 {
     switch (blasTrans[0])
     {
     case 'N':
     case 'n':
-        return MtblasNoTrans;
+        return CblasNoTrans;
     case 'T':
     case 't':
-        return MtblasTrans;
+        return CblasTrans;
     case 'C':
     case 'c':
-        return MtblasConjTrans;
+        return CblasConjTrans;
     default:
         std::cout << "Invalid BLAS transpose parameter!! Use default instead." << std::endl;
-        return MtblasNoTrans;
+        return CblasNoTrans;
     }
-} // Used to convert normal transpost char to mtblas transpose flag
+} // Used to convert normal transpost char to cblas transpose flag
 
 void* malloc_ht(size_t bytes, int cluster_id)
 {
-    // std::cout << "MALLOC " << cluster_id;
     void* ptr = hthread_malloc((int)cluster_id, bytes, HT_MEM_RW);
-    // std::cout << ptr << " SUCCEED" << std::endl;;
     return ptr;
-}
+} // Malloc on dsp. Used to replace original malloc
+
 
-// Used to replace original malloc
 
 void free_ht(void* ptr)
 {
-    // std::cout << "FREE " << ptr;
     hthread_free(ptr);
-    // std::cout << " FREE SUCCEED" << std::endl;
-}
+} // Free on dsp. Used to replace original free
 
-// Used to replace original free
 
 void sgemm_mt_(const char* transa,
                const char* transb,
@@ -77,7 +74,7 @@ void sgemm_mt_(const char* transa,
                const int* ldc,
                int cluster_id)
 {
-    mtblas_sgemm(MTBLAS_ORDER::MtblasColMajor,
+    mtblas_sgemm(CBLAS_ORDER::CblasColMajor,
                  convertBLASTranspose(transa),
                  convertBLASTranspose(transb),
                  *m,
@@ -109,7 +106,7 @@ void dgemm_mt_(const char* transa,
                const int* ldc,
                int cluster_id)
 {
-    mtblas_dgemm(MTBLAS_ORDER::MtblasColMajor,
+    mtblas_dgemm(CBLAS_ORDER::CblasColMajor,
                  convertBLASTranspose(transa),
                  convertBLASTranspose(transb),
                  *m,
@@ -141,7 +138,7 @@ void zgemm_mt_(const char* transa,
                const int* ldc,
                int cluster_id)
 {
-    mtblas_zgemm(MTBLAS_ORDER::MtblasColMajor,
+    mtblas_zgemm(CBLAS_ORDER::CblasColMajor,
                  convertBLASTranspose(transa),
                  convertBLASTranspose(transb),
                  *m,
@@ -173,7 +170,7 @@ void cgemm_mt_(const char* transa,
                const int* ldc,
                int cluster_id)
 {
-    mtblas_cgemm(MTBLAS_ORDER::MtblasColMajor,
+    mtblas_cgemm(CBLAS_ORDER::CblasColMajor,
                  convertBLASTranspose(transa),
                  convertBLASTranspose(transb),
                  *m,
@@ -207,7 +204,7 @@ void sgemm_mth_(const char* transa,
                 const int* ldc,
                 int cluster_id)
 {
-    mt_hthread_sgemm(MTBLAS_ORDER::MtblasColMajor,
+    mt_hthread_sgemm(CBLAS_ORDER::CblasColMajor,
                      convertBLASTranspose(transa),
                      convertBLASTranspose(transb),
                      *m,
@@ -239,7 +236,7 @@ void dgemm_mth_(const char* transa,
                 const int* ldc,
                 int cluster_id)
 {
-    mt_hthread_dgemm(MTBLAS_ORDER::MtblasColMajor,
+    mt_hthread_dgemm(CBLAS_ORDER::CblasColMajor,
                      convertBLASTranspose(transa),
                      convertBLASTranspose(transb),
                      *m,
@@ -275,7 +272,7 @@ void zgemm_mth_(const char* transa,
     *alp = *alpha;
     std::complex<double>* bet = (std::complex<double>*)malloc_ht(sizeof(std::complex<double>), cluster_id);
     *bet = *beta;
-    mt_hthread_zgemm(MTBLAS_ORDER::MtblasColMajor,
+    mt_hthread_zgemm(CBLAS_ORDER::CblasColMajor,
                      convertBLASTranspose(transa),
                      convertBLASTranspose(transb),
                      *m,
@@ -314,7 +311,7 @@ void cgemm_mth_(const char* transa,
     std::complex<float>* bet = (std::complex<float>*)malloc_ht(sizeof(std::complex<float>), cluster_id);
     *bet = *beta;
 
-    mt_hthread_cgemm(MTBLAS_ORDER::MtblasColMajor,
+    mt_hthread_cgemm(CBLAS_ORDER::CblasColMajor,
                      convertBLASTranspose(transa),
                      convertBLASTranspose(transb),
                      *m,

source/source_base/module_container/ATen/kernels/cuda/lapack.cu

@@ -88,7 +88,7 @@ struct lapack_potrf<T, DEVICE_GPU> {
 };
 
 template <typename T>
-struct lapack_dnevd<T, DEVICE_GPU> {
+struct lapack_heevd<T, DEVICE_GPU> {
     using Real = typename GetTypeReal<T>::type;
     void operator()(
         const char& jobz,
@@ -97,12 +97,12 @@ struct lapack_dnevd<T, DEVICE_GPU> {
         const int& dim,
         Real* eigen_val)
     {
-        cuSolverConnector::dnevd(cusolver_handle, jobz, uplo, dim, Mat, dim, eigen_val);
+        cuSolverConnector::heevd(cusolver_handle, jobz, uplo, dim, Mat, dim, eigen_val);
     }
 };
 
 template <typename T>
-struct lapack_dngvd<T, DEVICE_GPU> {
+struct lapack_hegvd<T, DEVICE_GPU> {
     using Real = typename GetTypeReal<T>::type;
     void operator()(
         const int& itype,
@@ -113,7 +113,7 @@ struct lapack_dngvd<T, DEVICE_GPU> {
         const int& dim,
         Real* eigen_val)
     {
-        cuSolverConnector::dngvd(cusolver_handle, itype, jobz, uplo, dim, Mat_A, dim, Mat_B, dim, eigen_val);
+        cuSolverConnector::hegvd(cusolver_handle, itype, jobz, uplo, dim, Mat_A, dim, Mat_B, dim, eigen_val);
     }
 };
 
@@ -175,15 +175,15 @@ template struct lapack_potrf<double, DEVICE_GPU>;
 template struct lapack_potrf<std::complex<float>,  DEVICE_GPU>;
 template struct lapack_potrf<std::complex<double>, DEVICE_GPU>;
 
-template struct lapack_dnevd<float,  DEVICE_GPU>;
-template struct lapack_dnevd<double, DEVICE_GPU>;
-template struct lapack_dnevd<std::complex<float>,  DEVICE_GPU>;
-template struct lapack_dnevd<std::complex<double>, DEVICE_GPU>;
+template struct lapack_heevd<float,  DEVICE_GPU>;
+template struct lapack_heevd<double, DEVICE_GPU>;
+template struct lapack_heevd<std::complex<float>,  DEVICE_GPU>;
+template struct lapack_heevd<std::complex<double>, DEVICE_GPU>;
 
-template struct lapack_dngvd<float,  DEVICE_GPU>;
-template struct lapack_dngvd<double, DEVICE_GPU>;
-template struct lapack_dngvd<std::complex<float>,  DEVICE_GPU>;
-template struct lapack_dngvd<std::complex<double>, DEVICE_GPU>;
+template struct lapack_hegvd<float,  DEVICE_GPU>;
+template struct lapack_hegvd<double, DEVICE_GPU>;
+template struct lapack_hegvd<std::complex<float>,  DEVICE_GPU>;
+template struct lapack_hegvd<std::complex<double>, DEVICE_GPU>;
 
 template struct lapack_getrf<float,  DEVICE_GPU>;
 template struct lapack_getrf<double, DEVICE_GPU>;