v3.8.2

Feature

• Feature: LR-TDDFT for open-shell systems by @maki49 in #5312
• Feature: Porting abacus to DSP hardware (mtblas part) by @Critsium-xy in #5301
• Feature: MPI available on ABACUS DSP version by @Critsium-xy in #5351
• Feature: make SCF converges only if drho and dene is smaller than threshold by @WHUweiqingzhou in #5357
• Feature: make init stochastic WF support GPU by @Qianruipku in #5365
• Feature: move scf_ene_thr before mix_rho by @WHUweiqingzhou in #5375
• Feature: change the default value of ecutwfc based on basis_type by @WHUweiqingzhou in #5390
• Feature: Add Cell for pyabacus by @jieli-matrix in #5247
• Unify the output of BFGS, CG, and FIRE relaxation method by @sunliang98 in #5374

Refactor

• Refactor: replace sto_hchi by HamiltSdftPW::hPsi by @Qianruipku in #5298
• Refactor: remove read_rho and make read_cube independent of esolver_type by @maki49 in #5323
• Refactor: Organize code structure to improve maintainability in pyabacus by @a1henu in #5346
• Refactor: move print_rhofft and print_wfcfft to ModuleIO by @YuLiu98 in #5358
• Refactor IO cube (Useful Information: excellent example for discussions and actions among ABACUS developers) by @maki49 in #5362
• Remove DiagH class in hsolver by @Cstandardlib in #5307
• Remove diagh.h in module_hsolver by @Cstandardlib in #5332
• Remove install_dsp.sh by @Critsium-xy in #5337

Fix

• Fix: cannot compile without OPENMP by @Qianruipku in #5316
• Fix: fix nspin=4 error when device=gpu by @dzzz2001 in #5319
• Fix: PR conflicts of #5298 and #5312 by @Qianruipku in #5329
• Fix::Change the gamma_only_lcao when nspin=4 by @A-006 in #5322
• Fix: boost Davidson by removal of useless scc by @Cstandardlib in #4874
• Fix: do not bcast my_rank by @YuLiu98 in #5339
• Fix: Fix Consts memory leak by directly removing custome consts by @Critsium-xy in #5344
• Fix: make EXX symmetry consistent with the fixed irreducible k-points by @maki49 in #5348
• Change the default setting for symmetry when berry_phase=1 by @ErjieWu in #5349
• Add number of threads output in stdout by @QuantumMisaka in #5281
• Fix: make the symmetry condition in RPA consistent with EXX by @maki49 in #5355
• Fix: force and stress calculation with noncollinear-spin or SOC for PW code by @dyzheng in #5377
• Fix: widen the range of support on STRU file of Molden interface by @kirk0830 in #5388
• Fix: disable four rarely used functionals which involves the laplacian of rho by @kirk0830 in #5392

Doc

• doc: fix typo in kpt.md by @pxlxingliang in #5320
• docs: replace sphinx-rtd-theme with sphinx-book-theme by @njzjz-bot in #5360
• Docs: update the docs about DeePKS by @WHUweiqingzhou in #5385
• Docs: add the docs about DeePTB by @WHUweiqingzhou in #5386

Perf

• Performance: change the position of open mp by @Qianruipku in #5327

Build

• Toolchain: fix build scripts by @QuantumMisaka in #5350

Test

• Test: unit test for read/write_cube by @maki49 in #5384

New Contributors

• @ErjieWu made their first contribution in #5349

Full Changelog: v3.8.1...v3.8.2

v3.8.1

What's Changed

Build(deps)

• Bump pre-commit-ci/lite-action from 1.0.3 to 1.1.0 by @dependabot in #5227

CI

• Align test commands for coverage tests by @caic99 in #5219
• add back missing command for codecov by @caic99 in #5222

Docs

• Modify the berry phase document. by @jingan-181 in #5293
• update default behaviors on building built-in libm by @caic99 in #5218
• update the documentation about pseudopotential and orbitals by @kirk0830 in #5294

Feature

• Becke's partition for multi-center grid integration by @jinzx10 in #5292
• Enable PyTorch backend for DeePMD-kit v3 by @YuLiu98 in #5253
• Support outputting real space wave functions in Gaussian CUBE format by @AsTonyshment in #5140
• update new version of dav_subspace with higher performance by @dyzheng in #5199
• TDDFT now can use ks_solver=cusolver, and velocity gauge is able to use Heaviside external field by @AsTonyshment in #5238

Fix

• Fix a segfault in rhog_symmetry by @maki49 in #5303
• Fix memory bug in XC_Functional_Libxc::cal_gdr() by @PeizeLin in #5208
• Fix minor bug of EXX operator and default value by @maki49 in #5252
• Fix the memory leak in integrated test case 919_OF_out_elf by @sunliang98 in #5211
• fix a deadlock bug in CUDA version of gint by @dzzz2001 in #5210
• fix undefined behavior in cusolver by @dzzz2001 in #5251
• md restart can not read STRU by @YuLiu98 in #5215
• nupdown = 0 for spin-unpolarized calculation by @jinzx10 in #5305
• performance and memory for NSPIN=4 in LCAO base by @dyzheng in #5249
• remove unnecessary gpu memory allocation under LCAO basis by @dzzz2001 in #5258
• revert pre-commit update by @caic99 in #5244

Refactor

• Change the variable type. by @A-006 in #5201
• Change wrapper spsi_func in hsolver-dav by @Cstandardlib in #5205
• Refactor XC_Functional_Libxc::convert_vtxc_v() by @PeizeLin in #5212
• Refactor class DensityMatrix: remove the dependence on K_Vectors and the ambiguity of _nks by @maki49 in #5224
• Refactor hpsi_func in hsolver by @Cstandardlib in #5202
• Remove Base class DiagH in LCAO code by @Cstandardlib in #5239
• Remove DiagH Base class in pw by @Cstandardlib in #5225
• Remove DiagH of lcao PEXSI by @Cstandardlib in #5259
• Seperate BLAS functions' declaration and implementation, and add a device_type flag for blas kernels by @Critsium-xy in #5242
• Sorting out the calculation logic of pexsi in hsolver-lcao by @haozhihan in #5299
• Unify the interfaces of Pulay terms of force and stress by @maki49 in #5130
• Use memory_op to set diag_const_nums by @Critsium-xy in #5246
• refactor solve func in hsolver-lcao class by @haozhihan in #5257
• remove nspin dependence form Gint by @maki49 in #5304
• Use PARAM instead of GlobalV::NBAND,NLocal by @A-006 in #5193

Test

• add cases for wfcinit by @Qianruipku in #5197
• fix some incorrect INPUT files by @dzzz2001 in #5284

Tools

• Toolchain fix stage4 by @QuantumMisaka in #5233
• toolchain bug fix by @QuantumMisaka in #5223

Full Changelog: v3.8.0...v3.8.1

v3.8.0

What's Changed

Feature

• Add davidson function to pyabacus by @a1henu in #5112
• Calculate and output electron localization function (ELF) with KSDFT and OFDFT by @sunliang98 in #5139
• Delley's grid for quadrature on the unit sphere by @jinzx10 in #5131
• RI-Hartree for LR-TDDFT benchmark with FHI-aims by @maki49 in #5084
• Radial quadrature grid by @jinzx10 in #5173
• Space group symmetry for EXX based on LibRI v0.2.1.0 by @maki49 in #5063
• add new init_chg method with wavefunctions by @Qianruipku in #5082

Fix

• Fix docs according to #5133 by @Cstandardlib in #5169
• Fix parallel LR-EXX and add a test case by @maki49 in #5188
• Fix read/write HexxR at nspin=4 by @maki49 in #5142
• Fix serial version compiling bug by @Critsium-xy in #5079
• Fix the undefined behavior in sph_bessel_recursive by @maki49 in #5158
• If the number of NAOs for an angular momentum is 0, filter the NAOs in exx by @PeizeLin in #5152
• Update function calls in pyabacus to align with new function signature in hpsi_func by @a1henu in #5176
• timer not closed in TDDFT code by @AsTonyshment in #5172
• add MPI_Barrier in unittest of CifParser when necessary to avoid unexpected behavoir by @kirk0830 in #5167
• add threshold file and enlarge ecut in exx test cases to prevent CI test failure by randomness by @maki49 in #5096
• bug of cal_ux by @Qianruipku in #5165
• discourage out_mul for PW to prohibit segfault error by @kirk0830 in #5102
• enlarge the threshold of case 281 by @maki49 in #5117
• fix a bug in diago_elpa_native.cpp by @dzzz2001 in #5155
• fix a cuda runtime error introduced by PR #5061 by @dzzz2001 in #5071
• init_chg wfc support npsin = 4 by @Qianruipku in #5166
• let only one rank do file I/O in unittest of orb_io by @kirk0830 in #5185
• minor fixes in LR by @maki49 in #5187
• move back Kpoint when basis_type="pw" by @A-006 in #5129
• optimize the memory usage by psi_initializer by @kirk0830 in #5120
• recover the wannier90 interface for lcao_in_pw case by @kirk0830 in #5177
• segmentation fault when deepks_scf=0 but deepks_out_labels=1 by @xuan112358 in #5090
• support kpar > 1 case to write rhog by @kirk0830 in #5145
• wrong STRU read in md restart case by @YuLiu98 in #5157
• fix cusolvermp compiling error with icpc and update ks_solver doc by @dzzz2001 in #5196

Perf

• change default pw_seed=0 by @WHUweiqingzhou in #5179
• optimize td_current by @dzzz2001 in #5181

Refactor

• Refactor ModuleIO::read/write_cube() by @PeizeLin in #5150
• USE PARAM instead of GlobalV by @A-006 in #5053
• Use PARAM instead of GlobalV::dir* by @A-006 in #5097
• Use PARAM instead of GlobalV::domag* by @A-006 in #5115
• Use PARAM instead of GlobalV::test* by @A-006 in #5065
• Use PARAM instead of deepks* by @A-006 in #5104
• Use PARAM instead of ks_solver,nb2d,search_radius by @A-006 in #5138
• Use Param.inp insetead of GlobalV::MIXING* by @A-006 in #5070
• add CalAtomsInfo to modify parameter by @Qianruipku in #5132
• add namespace XC_Functional_Libxc by @PeizeLin in #5151
• consistent order of hpsi by @Cstandardlib in #5134
• optimize the performanace of psi_initializer with omp by @kirk0830 in #5146
• refactor HsolverPW & HsolverPW_SDFT func by @haozhihan in #5094
• refactor hsolver-lcao func by @haozhihan in #5148
• refactor paw code in HSolverLIP by @haozhihan in #5126
• remove GlobalC::ORB by @jinzx10 in #5085
• remove GlobalC::ORB in module_hamilt_lcao by @jinzx10 in #5068
• remove hsolver base class from module_hsolver by @haozhihan in #5093
• remove my_conj func from hsolver_lcao by @haozhihan in #5128
• remove useless warning for dp by @YuLiu98 in #5153
• use ELPA_WITH_NVIDIA_GPU_VERSION to judge elpa setup gpu by @goodchong in #5163
• Use PARAM instead of nspin,cal_stress by @A-006 in #5136

Tools

• Toochain update: new download url and MPICH/OpenMPI version by @QuantumMisaka in #5088
• Toolchain 202403 by @QuantumMisaka in #5144

Build

• Bump pre-commit-ci/lite-action from 1.0.2 to 1.0.3 by @dependabot in #5110

Docs

• update hpsi_func requirements by @Cstandardlib in #5190

Full Changelog: v3.7.5...v3.8.0

v3.7.5

What's Changed

Feature

• Feature: support automatic binary output of charge density by @kirk0830 in #4991
• add elpa native to solver generate eigen by @goodchong in #4969
• Feature: add fparam and aparam for dp esolver by @YuLiu98 in #5047
• Feature: add rescaling parameter into dp by @Qianruipku in #5050

Performance

• Perf: optimize the openmp strategy in calculate_HR by @dzzz2001 in #5037
• Perf: optimize function snap_psibeta_half_tddft by @dzzz2001 in #5041

Fix

• Build: Fix ENABLE_CNPY Link Errors by @jieli-matrix in #5003
• Fix: bug in charge extrapolation by @YuLiu98 in #5007
• Fix toolchain scalapack problem by @QuantumMisaka in #5014
• Fix HexxR readin directory by @maki49 in #5025
• Fix bug in reallocating HContainter when not all the processors have elements of (0,0) atom pair by @maki49 in #5028
• Fix: close mixing_gg0 after PBE-loop in no-separate-loop EXX process by @maki49 in #5034
• Fix: fix the issue that out_chg -1 not really work by @kirk0830 in #5046
• Fix: Assert the mesh of <PP_BETA> not exceeding pp.mesh. by @sunliang98 in #5049
• fix wrong dp_potential after rescaling by @Qianruipku in #5055

Refactor

• Refactor: refactor set_diagethr func and remove phsol from esolver [version 2] by @haozhihan in #5017
• Refactor:Replace GlobalV::KSPACING with PARAM.inp.kspacing. by @A-006 in #5040
• Fix&Refactor: support naive and complete CIF file I/O interface by @kirk0830 in #5023
• Refactor: remove GlobalC::ORB from module_io by @jinzx10 in #5056
• Refactor: remove useless code about phsol by @haozhihan in #5054

Full Changelog: v3.7.4...v3.7.5

v3.7.4

What's Changed

Feature

• Feature: update LibRI v0.2.0 by @PeizeLin in #4830
• Feature: support control on precison of charge cube output by @kirk0830 in #4958
• Feature: Add diago_dav_subspace module to pyabacus by @a1henu in #4883
• Feature: output T(k) by @jinzx10 in #4982

Fix

• Fix an EXX bug with out_mat_hs2 by @maki49 in #4956
• fix bug in catching properties for out_eband_terms by @maki49 in #4962
• Fix: Fix the I/O problem of Rappe pseudopotential caused by the change of mesh. by @sunliang98 in #4973
• Fix: do not always output HexxR but only for NSCF by @maki49 in #4994
• Fix: enable scf_ene_thr for EXX outer loop by @maki49 in #4993
• Fix compile warnings about after_scf by @maki49 in #4996

Refactor

• Refactor: rescale vel to temperature when read in vel by @YuLiu98 in #4953
• Refactor: reset_diag_ethr && cal_hsolve_error func in hsolver by @haozhihan in #4940
• Refactor: Remove Dependency on Custom Class const_nums in david by @Cstandardlib in #4950
• Refactor: update charge extra to accelerate md by @YuLiu98 in #4954
• Refactor: Remove the inheritance relationship between hsolverLCAO/hsolverLIP and hsolver by @haozhihan in #4960
• Refactor: remove template in Chebyshev by @Qianruipku in #4972
• Performance: heterogeneous operator of npw loop in force_cc. by @grysgreat in #4978

Tools

• Toolchain : libcomm version update by @QuantumMisaka in #4966
• add elpa gpu in docker cuda image by @goodchong in #4980

• update version by @pxlxingliang in #5001

New Contributors

• @a1henu made their first contribution in #4883

Full Changelog: v3.7.3...v3.7.4

v3.7.3

What's Changed

Feature

• (for development) support general realspace projection onto PW basis: SphericalBesselTransformer-based-FourierTransformation by @kirk0830 in #4864
• Support outputting partial charge densities for different k-points and spins separately when using PW basis set by @AsTonyshment in #4829
• automatically use gpu if possible by @caic99 in #4826
• enable SCF can converge if SCF reach a chosen energy threshold by @WHUweiqingzhou in #4898
• enable multi-k calculation for CUDA version of module_gint by @dzzz2001 in #4839
• stop abacus manually according to the keyword in EXIT by @YuLiu98 in #4822

Doc

• real spherical harmonics and basis ordering clarified by @jinzx10 in #4922
• SchmidtOrth in david by @Cstandardlib in #4855

Fix

• Fix a bug in stress op that could cause errors in OpenMP multithreading calculations. by @grysgreat in #4932
• Fix bug due to the forgotten fix_gamma in writing eband terms by @maki49 in #4934
• Fix bug in add_HexxR at nspin=4 by @maki49 in #4893
• Fix the issue of mete-vlocal not being initialized. by @A-006 in #4933
• Fix the issue of some timer instances not being closed, resulting in incorrect time statistics by @AsTonyshment in #4939
• Incorrect negative signs in length gague RT-TDDFT by @AsTonyshment in #4871
• Revert "Move the cal_nnrg file and modify the variable length. (#4886)" by @A-006 in #4921
• a possible bug when dir is auto by @Qianruipku in #4948
• avoid nan in the output of time percentage by @pxlxingliang in #4857
• avoid the redundant file-reading and remove the useless output in write_Vxc by @maki49 in #4902
• correct the behavior of profile1d function in cube_manipulator.py by @kirk0830 in #4892
• dcu error in pw_force calculation by @grysgreat in #4786
• delete the unexpected debug info during scf by @kirk0830 in #4936
• error with DFTU force&stress with NSPIN=4 by @dyzheng in #4825
• fix the bug induced by pseudo_dir and orbital_dir default value by @Qianruipku in #4881
• improve the experience of using molden.py postprocessing tool by both developer and user by @kirk0830 in #4942
• reference to parameter instead of global variable in tool functions in DiagoDavid by @Cstandardlib in #4873

Perf

• optimize the construction of HexxR in NSCF by @maki49 in #4952
• use openmp to accelerate fvnl_dbeta_gamma.cpp by @dzzz2001 in #4814

Refactor

• Add heterogeneous parallel code for the stress loc module. by @grysgreat in #4854
• Continue to move the deepks output files into the OUT folder by @xuan112358 in #4924
• Heterogeneous parallel acceleration hsolver-pw force_scc. by @grysgreat in #4759
• Make Hsolver_sdft a local variable in hamilt_to_density function by @Qianruipku in #4941
• Move pseudo::set_pseudo to Pseudopot::complete_default. by @sunliang98 in #4811
• Move the cal_nnrg file and modify the variable length. by @A-006 in #4886
• Relate to PEXSI-related changes caused by refactor #4767 by @Flying-dragon-boxing in #4858
• Remove INPUT class by @Qianruipku in #4815
• Replace Esolver_OF::mu_ with efermi. by @sunliang98 in #4849
• Replace the pointers in pseudo and Pseudopot_upf with std::vector. by @sunliang98 in #4862
• Use lambda expressions to reduce if conditions. by @grysgreat in #4828
• Use pointers instead of function calls in OpenMP by @A-006 in #4827
• add cal_becp-like example for radial projector-involving PW operators development by @kirk0830 in #4887
• change HsolverPW call manner by @haozhihan in #4795
• change hsolver-lcao call manner in SpinConstrain by @haozhihan in #4946
• change hsolver-lip call manner by @haozhihan in #4931
• change hsolver_lcao call manner by @haozhihan in #4929
• change the order of initializing member variables by @kirk0830 in #4865
• discourage part of the use of c-style array in psi initializer by @kirk0830 in #4866
• merge after_scf by @YuLiu98 in #4889
• move set_is_occupied func from hsolver to elecstate by @haozhihan in #4847
• remove pelec frome hsolverpw solve func by @haozhihan in #4846
• remove do_after_converge by @YuLiu98 in #4910
• remove function mult_psi_DM and mult_psi_DM_new by @dzzz2001 in #4926
• remove pes from hsolver by @haozhihan in #4861
• remove the inheritance of HSolverLR by @maki49 in #4917
• remove uncessary interfaces in module_io by @YuLiu98 in #4835
• remove useless values in Hsolverpw by @haozhihan in #4875
• replace dsytrf & dsytri & gemv by dsysv in Broyden_mixing by @WHUweiqingzhou in #4842

Tools

• Toolchain 202402 by @QuantumMisaka in #4945
• support molden file generation from ABACUS LCAO calculation via NAO2GTO by @kirk0830 in #4937

Full Changelog: v3.7.2...v3.7.3

v3.7.2

Feature

• Feature: LR-TDDFT by @maki49 in #4472
• Feature: enable multiple species in LJ by @YuLiu98 in #4659
• Feature : Add deepks_v_delta, which can help to train DeepKS model with loss term about Hamiltonian, psi and band by @xuan112358 in #4594
• Feature: nspin=2 LR-TDDFT solving singlet/triplet by @maki49 in #4710
• Feature: read a slice of bands from WFC_NAO files & enable custom close-to-HOMO nocc for LR by @maki49 in #4691
• Feature: Support calculation of multi-k point band-decomposed charge density under LCAO basis by @AsTonyshment in #4698
• Feature: output vacuum level when output electrostatic potential by @YuLiu98 in #4799
• Feature: write band energy of T, VL, VNL, Hartree and XC terms separately by @maki49 in #4805
• Feature: Partial function heterogeneous parallel acceleration of force/stress_cc. by @grysgreat in #4555
• Add normalization procedure for partial charge in multi-k calculations when if_separate_k=1 by @AsTonyshment in #4724
• Feature: support k-parallelism diagonalization in lcao basis by @hongriTianqi in #4650
• update out_dos_lcao of '2' for output PDOS by @Chentao168 in #4782

CI

• test: add an option to set thread number in Autotest.sh by @pxlxingliang in #4668
• Test: Unit test for diago_lapack, add a build test and update the document by @Critsium-xy in #4610
• Test: add ccache installation step to cuda.yml by @dzzz2001 in #4688
• test: to make test fail when execute catch_properties.sh failed by @pxlxingliang in #4700
• test: correct the inputs in /examples by @pxlxingliang in #4717
• Test: add return value for mock in tests to suppress compile warning by @Cstandardlib in #4775

Perf

• Perf: optimize function grad_rl_sph_harm by @dzzz2001 in #4671
• Perf: optimize function pow in module_gint by @dzzz2001 in #4680
• Perf: optimize force calculation in module_gint (Useful information about how to take advantages of data locality to improve the code efficiency) by @dzzz2001 in #4727

Fix

• Fix MKL problem in pyabacus example ex_s_rotate.py by @maki49 in #4669
• Fix a possible compile error in module_lr/lr_spectrum.cpp by @maki49 in #4672
• Fix the problematic logic in catch_properties.sh induced by #4472 by @maki49 in #4678
• Fix the exx-md/relax bug induced by LM refactor by @maki49 in #4686
• Fix: change C++14 codes to C++11 by @Qianruipku in #4682
• Fix the missing EXX operator and minor refactor in LR-HSE by @maki49 in #4696
• Fix: fix a typo related to pexsi in fedm_gamma.cpp by @Flying-dragon-boxing in #4702
• Fix the bug causing incorrect Berry phase calculations under the pw basis. by @jingan-181 in #4704
• Fix: error of non-diagonal nonlocal pseudopotential with LCAO by @dyzheng in #4705
• Fix: replace part of GlobalV::var by parameter by @Qianruipku in #4706
• Fix: memory leak of basis set in david by @Cstandardlib in #4731
• Fix: replace most of bare pointers with weak/unique_ptr in module_lr to prevent memory leak (Useful information about how to use smart pointers in C++11)) by @maki49 in #4720
• Fix: DeePKS MD calculation restart by DMK and fixed the lack of DeePKS tests by @dyzheng in #4605
• Fix: eliminate compilaton error due to the #define private public macro in all unittests by @kirk0830 in #4733
• Fix: no Fail of read_input_serial_test.cpp by @Qianruipku in #4749
• Fix: Fix the memory leak cause by kbeta in read_pp.cpp by @sunliang98 in #4758
• Fix: fix a memory leak in module_gint by @dzzz2001 in #4721
• Fix: consider fractional electron number by @Qianruipku in #4770
• Fix: eliminate the implicit type conversion when calculating number of electrons of system and repair bessel_nao_rcut behavior by @kirk0830 in #4788
• Fix a memory leak in LR constructor by @maki49 in #4793
• Fix bugs in deepks_v_delta by @xuan112358 in #4780
• fix: wrong use of bessel_nao_rcut by @Qianruipku in #4792
• Fix the incorrect calculation of partial charge for multi-k-points under both PW and LCAO when symmetry is turned on by @AsTonyshment in #4803

Refactor

• Refactor: adjust cif file output by @hongriTianqi in #4584
• Refactor: remove EXX in LM and finally remove LM by @maki49 in #4679
• Refactor: remove comm_2D from Parallel_2D (Useful Information: BLACS encapsulates MPI and maintains its own internal data structure for MPI communicators) by @jinzx10 in #4658
• delete redundant setting in STRU by @Satinelamp in #4689
• Refactor: replace most of input varibles by parameter by @Qianruipku in #4693
• Refactor david: transitionally replace the Psi-type variables basis and psi by @Cstandardlib in #4701
• Refactor david: remove HamiltPW::sPsi inside DiagoDavid by @Cstandardlib in #4708
• Refactor: reorganized HSolverPW<T, Device>::solve function in HSolverPW by @haozhihan in #4675
• Refactor: Remove useless variables from read_pp.h. by @sunliang98 in #4711
• Remove redundant code about david in hsolver by @Cstandardlib in #4713
• Refactor: separate parallel_comm from parallel_global by @haozhihan in #4715
• Refactor ESolver, separate out some functions that will be moved to other places later. by @mohanchen in #4812
• Refactor: enable const md_para as md_line passing parameter by @YuLiu98 in #4808
• Refactor david: complete removal of Psi and Hamilt datatypes by @Cstandardlib in #4722
• Refactor hsolver: redesign david interface to be consistent with dav_subspace by @Cstandardlib in #4726
• Refactor: Remove most usage of pdiagh in hsolver_lcao.cpp by @Critsium-xy in #4742
• Refactor: remove the explicit use of std::make_unique use in unittest by @kirk0830 in #4745
• Remove init_basis_lcao from ESolver_LCAO to LCAO_domain （Useful information about how to clean up ESolver) by @mohanchen in #4746
• Refactor for the DeePKS module by @mohanchen in #4762
• Refactor david: add docs and standardize the code by @Cstandardlib in #4734
• Refactor: Remove usage of pdiag_parak in hsolver_lcao by @Critsium-xy in #4767
• Reduce the size of LCAO_deepks_torch.cpp by @mohanchen in #4773
• Style: add braces to avoid dangling else in module_io/restart.cpp by @Cstandardlib in #4774
• Refactor: refactor set_diagethr, cal_hsolerror, cal_hsolerror func in hsolver class by @haozhihan in #4769
• Refactor: move GPU deallocation code in david by @Cstandardlib in #4794
• Refactor DeepKS (the fourth one) by @mohanchen in #4784
• Refactor: reorder INPUT parameters by @Qianruipku in https://github.com/deepmodeling/a...

v3.7.1

CI

• Remove unused workflow by @caic99 #4512
• build docs only for main repo by @caic99 #4306

Fix

• Move Vxc output from after_scf to after_all_runners by @maki49 #4497
• Remove the dependency of write_dmr on globalV::NLOCAL by @pxlxingliang #4500
• Fix no STRU_ION_1 file output by @hongriTianqi #4506
• Fix init_Table_Spherical_Bessel with correct k & r mesh multiple by @jinzx10 #4516
• Fix the bug which SCF restart for unconverged cases by @WHUweiqingzhou #4538
• Fix compile failure of main.cpp by @Qianruipku #4638
• Fix MKL scalapack linking by @caic99 #4644
• Fix DeePKS MD/Relax/Cell-relax calculation convergence by @dyzheng #4575
• Fix oneAPI runtime error when build with libtorch by @caic99 #4613
• Fix esolver.cpp by @pxlxingliang #4612
• Fix misunderstanding absolute path as annotation by @Qianruipku #4629
• Fix missing head of running.log by @Qianruipku #4630
• Fix the memory problem in berryphase by @pxlxingliang #4631
• Fix wrong running_*.log by @Qianruipku #4648
• Fix MKL finding problem in pyabacus by @maki49 #4656
• Fix some minor security risks with ModuleESolver::init_esolver() by @PeizeLin #4590
• Fix some minor security risks with ModuleIO::read_wfc_nao by @PeizeLin #4558
• Address the atomicAdd error by @denghuilu #4234
• Fix cusolver error under multi-k situation by @dzzz2001 #4278
• Fix makefile cannot compile tests in module_pw by @Qianruipku #4270
• Fix orb_matrix file is not safely written/renamed by @kirk0830 #4299
• Fix the bug in dav gpu code and add more integrate test for GPU by @haozhihan #4265

Refactor

• Remove pdiagh pointer and redundant code by @Cstandardlib #4520
• Change nkstot_ibz and kvec_d_ibz from members into local variables by @maki49 #4515
• Remove some variables from LCAO_Matrix class by @WHUweiqingzhou #4519
• Remove BasedOnStyle: Microsoft from clang-format by @haozhihan #4522
• Remove the useless loc in veff by @pxlxingliang #4532
• Removal of ORB_table_phi by @jinzx10 #4503
• Remove check_structure() from UnitCell by @sunliang98 #4499
• Remove some members from LCAO_Matrix class by @WHUweiqingzhou #4546
• Replace the interface of out_dm by using DensityMatrix.DMK instead of loc.DM by @pxlxingliang #4509
• Remove initDiagh function in hsolver by @Cstandardlib #4527
• Fully refactored original obsolete lapack hsolver by @Critsium-xy #4581
• Separate Class diag_comm_info and Class const_nums by @haozhihan #4582
• Remove additional dependencies of parallel_reduce.h by @haozhihan #4585
• Remove Hloc, Hloc2, Sloc, Sloc2 in LCAO_Matrix by @dyzheng #4574
• Remove some variables from LCAO_Matrix class by @WHUweiqingzhou #4568
• Remove the useless variables and functions in unkOverlap_lcao by @pxlxingliang #4569
• Remove redundant Local_Orbital_Charge in class Veff and DeepKS by @PeizeLin #4572
• Use alicloud image sync by @caic99 #4567
• Fix the wrong type of ibz2bz by @maki49 #4577
• Login to use alicloud image sync by @caic99 #4573
• Fix get_S failed when smearing_method is gauss by @hongriTianqi #4571
• Correct the warning message about ntype by @pxlxingliang #4562
• Split the function into different functionalities by @A-006 #4566
• Remove module_device additional dependencies by @haozhihan #4579

• update version to 3.7.1 by @dyzheng in #4662

Full Changelog: v3.7.0...v3.7.1

v3.7.0

What's Changed

Feature

• Feature: add the output of energy for all esolver types in abacus.json by @pxlxingliang in #4413
• Feature: add command line option --check-input to check input only by @hongriTianqi in #4424
• Feature: support nupdown = 0.0 calculation by @hongriTianqi in #4425
• Feature: add function get_nbasis in class HContainer by @pxlxingliang in #4447
• Add INPUT parameter bands_to_print for get_wf calculation by @AsTonyshment in #4466

Fix

• Fix: BLACS lib in unit test by @jinzx10 in #4338
• Fix: report error when fr-uspp used in linear calculations by @YuLiu98 in #4348
• Fix: address memory leak with single precision by @denghuilu in #4492
• Fix: correct the dimension of wfc_k_grid for allocating memory in istate_envelope function by @kirk0830 in #4494
• fix the output format in aveElecStatPot.py by @xdzhu in #4367
• Fix: support cross-boundary structures in orbital generation by @jinzx10 in #4376
• Fix: matrix bug in uspp by @YuLiu98 in #4397
• Fix: no inversion symmetry, reset symmetry of SOC calculation to -1 by @dyzheng in #4375
• Fix: improve the precision according to issue#4183 by @kirk0830 in #4460
• Fix: remove the warning_quit when there are more than 50 basis of one atom by @kirk0830 in #4461
• Fix: correct function read_abacus_lowf by @kirk0830 in #4433
• fix: fix the incorrect usage of macro #elif by @pxlxingliang in #4478
• Fix: bug in mulliken output by @hongriTianqi in #4463

Test

• Test: fix clang-format settings by @haozhihan in #4382
• Test: compare the relative difference for Onsager.txt by @Qianruipku in #4378
• test: update the path of SR.csr in examples/matrix_hs/out_s_multik/run.sh by @pxlxingliang in #4352
• Test: use origin python in test.yml to support numpy by @dyzheng in #4383
• Test: accelerate integrate tests by @YuLiu98 in #4358
• Test: fix precommit failure by @caic99 in #4426
• test: support reading threshold from file 'threshold' in integrate test by @pxlxingliang in #4450
• CI: modify clang-tidy setting by @haozhihan in #4488
• CI: enable clang-tidy in CI by pre-commit-config by @haozhihan in #4384
• CI: add clang-tidy readability-braces-around-statements setting by @haozhihan in #4390

Perf

• Perf: Optimize hsolver GPU code (useful information of GPU optimization: __syncwarp() should be used instead of __syncthreads()) by @OldDriver233 in #4295
• Perf: remove unnecessary sync for cuda ops by @caic99 in #4361
• Perf: reduce data transmission in the GPU code of module_gint by @dzzz2001 in #4356
• Perf: use less memory and optimize performance to calculate force and stress in pw base by @dyzheng in #4047

Docs

• doc: correct the typo in input-main.md by @pxlxingliang in #4423
• Docs: add docs for --version and --check-input by @hongriTianqi in #4428

Refactor

• Refactor: redesign the constructor and member variables of davsubspace class by @haozhihan in #4330
• Update before_all_runners in ESolver by @mohanchen in #4334
• Add and Remove some functions in klist.h etc by @Haerxile in #4257
• Update write_potential and "others" function in ESolver by @mohanchen in #4336
• Update the 'calculation' description in the documents and fix a small bug in elecstate_energy.cpp, update set_force and set_stress in LCAO_matrix by @mohanchen in #4340
• refactor: separate the print_psi from wfc_2d_to_grid and remove the invalid call by @pxlxingliang in #4268
• refactor: remove the usused LOC in getForceStress by @pxlxingliang in #4343
• Modify the average electrostatic potential script （Useful Information：How to plot electrostatic potential） by @xdzhu in #4350
• Refactor: replace cudamalloc and cudafree with cuda_mem_wrapper by @dzzz2001 in #4347
• Refactor: Remove DiagoIterAssist dependencies in DiagoDavid by @Cstandardlib in #4346
• CI: add pre-commit.ci to apply clang-format and clang-tidy fixes by @caic99 in #4225
• Using the standard CUDA <<<...>>> syntax to avoid compilation errors. by @LiuXiaohui123321 in #4362
• Refactor: add alternatives to functions LOWF::gamma_file and LOWF::allocate_k by @kirk0830 in #4331
• Refactor: change R_index in class AtomPair to type ModuleBase::vector3<int> by @DylanWRh in #4243
• Refactor: remove Kubo-Greenwood and DOS functions from esolver by @Qianruipku in #4355
• Add header file when using memset() function to avoid compi… by @LiuXiaohui123321 in #4380
• Refactor: replace pointer to Diago_DavSubspace with object by @Cstandardlib in #4386
• Refactor: delete some useless files in module_psi/kernel/ folder by @haozhihan in #4391
• bug: warning_quit when reading WFC file failed. by @pxlxingliang in #4388
• Refactor istate_charge.cpp using new DensityMatrix class by @AsTonyshment in #4392
• Refactor: remove unused class Output_Interface by @hongriTianqi in #4394
• Update LCAO_matrix and related functions and delete LCAO_gen_fixedH by @mohanchen in #4400
• Add several warning quits in DensityMatrix and Gint when parameters are not initialized by @AsTonyshment in #4389
• delete 31 double arrays in LCAO_matrix, add a new class named force_stress_arrays.h used in force and stress calcualtions by @mohanchen in #4407
• Refactor: Initial removal of Hamilt from diag_once func in Davsubspace by @haozhihan in #4398
• Refactor: Mulliken output codes with more flexible io interfaces by @hongriTianqi in #4359
• Refactor: extracted transpose parameter checking routine of gemm/v by @OldDriver233 in #4279
• Adding comments and modifying properties in the PW module by @DoctorDra in #4405
• Refactor: remove useless function from wavefunc by @Qianruipku in #4408
• Refactor: move snap_psibeta_half_tddft to module_tddft by @jinzx10 in #4414
• Refactor: Initial removal of psi::Psi<T, Device> from Diago_DavSubspace by @haozhihan in #4416
• Use Vector to replace new and delete in gint，spilt gint_tools and change Makefile,remove ORB in gint by @A-006 in #4353
• Refactor: 1. change "enum calculation_type" to "enum class calculation_type" for safety by @PeizeLin in #4319
• Refactor: delete some variables of LCAO_Matrix by @WHUweiqingzhou in #4415
• refactor: output sparse DMR by Output_HContainer instead of output_dm1.h/cpp by @pxlxingliang in #4374
• Refactor: replace pointer to Diago classes with object by @Cstandardlib in #4409
• Build: disable build libm by default by @caic99 in https://github.com/deepmodeling/a...

v3.6.5

What's Changed

Feature

• Feature : set deepks_equiv as input variable by @wenfei-li in #4195
• Feature: Input and output vector potential in tddft calculation by @ESROAMER in #4173
• Feature: a better looking of SCF iteration information on screen by @kirk0830 in #4185
• Feature : printing initial charge density by @wenfei-li in #4254

Fix

• Fix: address the atomicAdd error by @denghuilu in #4234
• Fix: replace cal_stress in GPU relax calculation，use NUM_STREAM in INPUT file and fix memory leak in force calculation by @A-006 in #4200
• Fix: add one whitespaces between each column of scf stdout by @kirk0830 in #4252
• Fix: fix the bug in dav gpu code and add more integrate test for GPU by @haozhihan in #4265
• Fix: makefile cannot compile tests in module_pw by @Qianruipku in #4270
• Fix: fix cusolver error under multi-k situation by @dzzz2001 in #4278
• Fix: present orb_matrix file is not safely written/renamed by @kirk0830 in #4299

Perf

• Perf: split code_gen.cpp to reduce compilation time by @dzzz2001 in #4210
• Perf: optimize single-process performence of cusolver by @dzzz2001 in #4191
• Perf: optimize single-process performence of cusolver by @dzzz2001 in #4191
• Perf: Incorporate OpenMP in the force and rho computations on the GPU by @dzzz2001 in #4280
• Perf: optimize psir_dot function in gint_rho_gpu.cu by @dzzz2001 in #4326

Refactor

• Refactor: Remove unused function Gint_Gamma::vl_grid_to_2D and related variables by @DylanWRh in #4229
• Refactor: rename diago_blas to diago_scalapack by @haozhihan in #4233
• Refactor: Modify class BaseMatrix by @DylanWRh in #4239
• Refactor: remove all globalV in diago_david by @haozhihan in #4211
• Refactor: simplified Parallel_2d interface by @jinzx10 in #4237
• Refactor: Replace modify the functions form cu files to cpp files by @A-006 in #4245
• Refactor: Encapsulation of K_point modules by @jwsyzy in #4240
• Refactor LCAO force: merge gamma and k by @maki49 in #4263
• Refactor: Remove Unused chg_extrap = dm INPUT Parameter Code by @AsTonyshment in #4285
• Refactor: Remove redundant function and modify comments in the PW module by @yang2022i in #4271
• Refactor: eliminate most use of LOWF instance in esolver - step 1 by @kirk0830 in #4273
• Refactor Separate the coupled GPU task functions. by @A-006 in #4281
• Refactor: Improve the parallel part of the K-point module by @jwsyzy in #4289
• Refactor: delete GlobalV::CURRENT_SPIN in code by @dyzheng in #4301
• Refactor: removal of GlobalC::UOT by @jinzx10 in #4322

Test

• Test: Modify the input file of examples/interface_wannier90 to run the Wannier90 interface correctly. by @jingan-181 in #4236
• test: modify the script to read energy from running_wcf_xx.log in example/bsse by @pxlxingliang in #4259
• CI: build docs only for main repo by @caic99 in #4306

Doc

• Docs: Doxygen Documentation Deployment by @jinzx10 in #4264

Full Changelog: v3.6.4...v3.6.5