v3.9.0.22

What's Changed

Feature

• Feature: Add a new parameter out_freq_td to control output frequency in RT-TDDFT by @AsTonyshment in #6796
• Feature: add exx_cs_inv_thr to get inverse Coulomb by @AroundPeking in #6793

Fix

• Fix: indptr overflow in output big HR. by @ErjieWu in #6803
• Fix: Add a threshold to avoid E_gap(k)=0 for insulators by @LKFEIYI in #6802
• Fix: Update NEP interface to match NEP_CPU's update by @MoseyQAQ in #6804
• fix: GPU memory leak in deriv_drhoc_scc by @dzzz2001 in #6809
• Fix:: replace deprecated CUDA_VERSION and handle CUDA 13+ architecture requirements by @Flying-dragon-boxing in #6810

Performance

• perf(TDDFT): Add CUDA acceleration for snap_psibeta_half function (Useful information about largely improves the snap_psibeta_half function) by @dzzz2001 in #6808

Refactor

• Simplify Lmax in Center2_Orb and Matrix_Orbs by @linpeize in #6806
• Simplify Lmax and ORB_gaunt_table in Center2_Orb and Matrix_Orbs by @linpeize in #6819

Others

• Update version to 3.9.0.22. by @ErjieWu in #6820

New Contributors

• @LKFEIYI made their first contribution in #6802
• @AroundPeking made their first contribution in #6793

Full Changelog: v3.9.0.21...v3.9.0.22

v3.9.0.21

What's Changed

Fix

• Fix: Fix an extremely serious bug that causes incorrect results in GPU RT-TDDFT calculation by @AsTonyshment in #6767
• Fix energy jumps occur in rt-TDDFT calculations by @ESROAMER in #6779
• [Fix] Report and exit for unsupported DSP KPAR configuration by @Cstandardlib in #6771
• [Fix]Enable abacus to be compiled with CXX 17 & CUDA 13 by @ZhouXY-PKU in #6777

Feature

• Add average iterative diagonalization step output by @Cstandardlib in #6795

Refactor

• Refactor and Optimize snap_psibeta_half_tddft by @dzzz2001 in #6765
• [Refactor] Support different CUDA versions in one single cuda_compat.h by @Cstandardlib in #6770

Test

• Test: Update the parameter and reference results of 23_OF_LPS by @sunliang98 in #6774
• Test: Update CI/CD tests for RT-TDDFT and add hybrid gauge and parameter td_dt tests for CPU and GPU by @AsTonyshment in #6776
• [Test] Use WARNING_QUIT instead of exit, Add and refactor unittests for ParaGlobal by @Cstandardlib in #6781

Others

• Build(deps): Bump actions/upload-artifact from 5 to 6 by @dependabot[bot] in #6783
• Update version to v3.9.0.21 by @xuan112358 in #6797

Full Changelog: v3.9.0.20...v3.9.0.21

v3.9.0.20

What's Changed

Fix

• Fix: remove gamma_only_local as para of write_rhog by @YuLiu98 in #6733
• [Fix] Remove useless code in CMakeLists by @Critsium-xy in #6732
• filter input non-ASCII char before process by @goodchong in #6739
• Fix: Increase threshold for 187_PW_MD_SDFT_ALL_GPU test by @Qianruipku in #6742
• Fix: Change int to long long in save_sparse to avoid overflow. by @ErjieWu in #6743
• Fix: EXX PW occupation number issue in nspin==2 && support for CUDA-Aware MPI marco by @Flying-dragon-boxing in #6750
• Fix CUDA GPU detection crash when running with device=cpu on machines without GPU by @Copilot in #6759

Feature

• [Feature]A new -i/-I/--info argument to show details of configuration and compilation by @ZhouXY-PKU in #6734
• Feature: support init_chg dm to restart with DMR by @dyzheng in #6753
• Feature: tools for selective dynamics with ABACUS + Phonopy by @YuLiu98 in #6755
• Feature: Support RT-TDDFT EDM (energy-density matrix) calculation on GPU by @AsTonyshment in #6762

Refactor

• refactor the call of initialization of charge density by @mohanchen in #6744
• Update some names of deepks tests and refactor some ofdft codes by @mohanchen in #6754

Others

• Build(deps): Bump actions/upload-artifact from 4 to 5 by @dependabot[bot] in #6729
• Build(deps): Bump actions/checkout from 5 to 6 by @dependabot[bot] in #6728
• Bump version 3.9.0.20 by @Flying-dragon-boxing in #6764

New Contributors

• @Copilot made their first contribution in #6759

Full Changelog: v3.9.0.19...v3.9.0.20

v3.9.0.19

What's Changed

Feature

• [Feature] Allow using multiple nodes to caculate on DSP hardware by @Critsium-xy in #6708
• Feature: add xc output info by @linpeize in #6715

Fix

• Add iostream include to cusolver.h by @ieiue in #6696
• Fix the search radius for calculating angular momentum matrices by @maki49 in #6705
• Fix: fix a bug caused by intel compiler by @dzzz2001 in #6703
• [Fix]An Intel Build test for abacus with PEXSI support(PART II) by @ZhouXY-PKU in #6707
• Fix: Bug in EXX PW, nspin=2 due to wrong k-point index by @Flying-dragon-boxing in #6721
• Update Cereal URL in FindCereal.cmake by @ZhouXY-PKU in #6722
• Update LibRI URL in FindLibRI.cmake to use codeload for fetching by @kluonj in #6723
• Fix: Fix coverage test by @sunliang98 in #6725

Refactor

• Remove GlobalC::dftu by @mohanchen in #6694
• Reducing the size of esolver_ks_lcao by @mohanchen in #6697
• Refactor&Feature: Package some parameters in DeePKS for simplification & update HR precalc output. by @ErjieWu in #6706
• Changed the positions of the template class declarations in rdmft_tools.cpp by @ZhouXY-PKU in #6712
• [Refactor] Using a INPUT parameter to adjust dsp cards count by @Critsium-xy in #6717
• Refactor: Support RT-TDDFT EDM calculation in Tensor by @AsTonyshment in #6726

Doc

• Toolchain and docs modification by @QuantumMisaka in #6718
• modify the description of the BFGS parameter by @19hello in #6710

Version Update

• update version to v3.9.0.19 by @19hello in #6720

Full Changelog: v3.9.0.18...v3.9.0.19

LTSv3.10.1

Bug Fix:

• add check and update code format in exx by @linpeize in #6244
• add exx nscf file check by @linpeize in #6289
• fix bug in Pseudopot_upf::set_empty_element() by @linpeize in #6424
• Cherry-picked #5775 to resolve the single-precision version error. by @zgn-26714 in #6418
• Fix: DFT+U nscf calculation of nspin=1 and output onsite.dm with out_chg>=0 by @dyzheng in #6550
• Fix: stress error with nscf by @dyzheng in #6556
• fix: the Nan value when rho is negative in ELF by @pxlxingliang in #6574
• cherry-pick #6233: Resolve MPI communication issue (R coordinate mismatch problem during sparse matrix generation) by @zgn-26714 in #6555
• refactor: output the latest structure to STRU_NOW.cif in relax/cell-relax by @pxlxingliang in #6587
• Fix: screen output redundant by @dyzheng in #6621
• Fix: do not overwrite the input k-point weights in ibz_kpoint by @maki49 in #6552
• Update supported shorthand names of xc functionals from develop branch by @ahxbcn in #6629
• Fix: Use B3LYP does not enter Exx loop by @ahxbcn in #6663
• fix the input parameter in examples/spin_polarized/ATOM/INPUT by @pxlxingliang in #6661
• Add: D3 zero-damping parameters for PBEsol (issue#6646) (#6660) by @dyzheng in #6670
• Fix the wrong HF Coulomb at nspin=2 by @maki49 in #6672
• Fix nspin=2 bug in write_Vxc by @maki49 in #6702

Performance:

• Feature: Implement cal_force_op for sincos parallel by @jieli-matrix in #6265
• Removed the temporary variable DMRGint_full when transitioning from 2D block parallelism to serial in Hcontainer. by @zgn-26714 in #6487
• Removed the temporary variable hRGintCd when transitioning from 2D block parallelism to serial in Hcontainer. by @zgn-26714 in #6488
• Added the serial version incorporating code changes from #6487 and #6488. by @zgn-26714 in #6511
• Fixed memory leaks and output bugs according to #6515 and #6518. by @zgn-26714 in #6585
• cherry-pick #6392 to support GPU version of cal_force_cc under LCAO basis set by @zgn-26714 in #6583
• Refactor: decrease memory cost of get_s (#5979) by @dyzheng in #6627
• Feature: add use_k_continuity method for initializing psi with PW base by @dyzheng in #6724

Build

• [LTS] update toolchain version to 202502 by @QuantumMisaka in #6203
• CI & CD : Trigger devcontainer build for LTS branches and tags by @jieli-matrix in #6210
• Cherry-pick #6337 to resolve the issue where the LTS branch doesn't compile on CUDA Version >= 12.9 (issue #6443). by @zgn-26714 in #6584
• [LTS] cherry-pick toolchain-202503 update to LTS by @QuantumMisaka in #6630
• [LTS][develop][Toolchain] fix elpa-gpu installation problem in toolchain 202503 by @QuantumMisaka in #6631
• [LTS] Version update to v3.10.1 by @dyzheng in #6651
• Fix: update dockerfile to install LTS version by @dyzheng in #6666

Full Changelog: LTSv3.10.0...v3.10.1

v3.9.0.18

What's Changed

Feature

• Feature: EXX PW supports k-point parallelism by @Flying-dragon-boxing in #6648
• Feature: EXX PW supports a ecut different from ecutrho by @Flying-dragon-boxing in #6650
• Feature: D3 zero-damping parameters for PBEsol (issue#6646) by @kirk0830 in #6660
• Add new blas/lapack interfaces by @Cstandardlib in #6658

Fix

• Fix: include elecstate_lcao in edm.cpp by @ZhouXY-PKU in #6644
• Fix: fix declarations of some BLAS and LAPACK functions. by @dzzz2001 in #6637
• Fix: remove rwork from ssyevx function declaration by @kluonj in #6647
• Fix: set the maximal cutoff radius of orbital as the lower bound of search_radius in neighboring search tasks carried in out_mat_l by @kirk0830 in #6653
• Fix: Add template disambiguator for certain Tensor access in ESolver_KS_LCAO_TDDFT by @AsTonyshment in #6678
• Fix: Fix the compilation error of unit tests under source_base by @sunliang98 in #6679
• Fix compilation error on Sunway by @Critsium-xy in #6683
• Fix: Resolve the compilation error in the Intel oneAPI 2025.3 container. by @sunliang98 in #6677
• Fix DM in PEXSI by @mohanchen in #6689
• [Fix]An Intel Build test for abacus with PEXSI support(PART I) by @ZhouXY-PKU in #6688

Refactor

• Refactor: refactor stress_ewa by @dzzz2001 in #6669
• Refactor: RT-TDDFT ESolver ESolver_KS_LCAO_TDDFT by @AsTonyshment in #6668
• Refactor: Simplify functions in sparse_format. by @ErjieWu in #6606
• Refactor: Modify the call to Parallel_Orbital in RT-TDDFT and remove unnecessary code in ctrl_output_td by @AsTonyshment in #6674
• Refactor: change the readin directory of deepks_projdm.dat by @ErjieWu in #6673
• Refactor energy-density matrix in RT-TDDFT to replace raw ScaLAPACK and BLAS calls with ScalapackConnector and BlasConnector interfaces by @AsTonyshment in #6687
• Remove DM from elecstate by @mohanchen in #6675
• Update ESolver, simplify the ESolver_KS_LCAO class by @mohanchen in #6636

Test

• Test: Add an integrate test for out_chg under OFDFT. by @sunliang98 in #6640
• Test: Fix two bugs of test scripts by @sunliang98 in #6642
• Test: Fix a bug about MKL_SCALAPACK in CMakeLists.txt by @sunliang98 in #6645

Doc

• Doc: Fix a minor LaTeX rendering issue of parameter td_ttype in doc by @AsTonyshment in #6692

New Contributors

• @kluonj made their first contribution in #6647
• @ieiue made their first contribution in #6690

Full Changelog: v3.9.0.17...v3.9.0.18

v3.9.0.17

What's Changed

Feature

• [Feature] Add current dependent potential for TD-OFDFT by @lyb9812 in #6593
• [Feature] Add NEP as esolver (Useful information to know that NEP has been interfaced with ABACUS) by @MoseyQAQ in #6603
• Toolchain 202503 Version Update by @QuantumMisaka in #6615
• [Feature] Add float copy function in blas_connector by @Critsium-xy in #6625

Fix

• Update update_cell.cpp to avoid numerical precision error issues. by @ZhouXY-PKU in #6594
• Fix: Resolve the output of E_bandgap by @sunliang98 in #6590
• Fix the unmatched timer tick of DiagoIterAssist::diag_subspace by @Cstandardlib in #6598
• Fix: Fix cal_bandgap for parallelism of k-points and band. by @sunliang98 in #6602
• [develop][Toolchain] fix elpa-gpu installation problem in toolchain by @QuantumMisaka in #6632
• [Toolchain] Fix AOCL linking problem and other ver-202503 patch by @QuantumMisaka in #6623
• Fix a output bug when using MPI in DeePKS. by @ErjieWu in #6626
• fix relax_method parameter bug by @19hello in #6605

Refactor

• Refactor: Use namespace instead of class for LapackConnector by @Cstandardlib in #6581
• Add an interface for psi in esolver_ks_pw by @mohanchen in #6599
• Delete update_pot() in ESolver by @mohanchen in #6614
• Refactor ESolver continues, I move charge mixing codes to module_charge by @mohanchen in #6616
• remove vector in bfgs and remove const number by @19hello in #6613
• Update initialization of pw_wfc in esolver_ks.cpp, also some old gint has been deleted, E_bandgap has been changed to E_gap(k), struct efermi has been changed to Efermi by @mohanchen in #6617
• Refactor: Unify standard/generalized eig driver of LAPACK by @Cstandardlib in #6610
• Refactor: move Exx_Abfs::Abfs_Index to Element_Basis_Index by @linpeize in #6622
• Update ESolver psi and delete some gint codes by @mohanchen in #6619
• Refactor: remove old gint module code by @dzzz2001 in #6635

Test

• Modify BFGS test code by @19hello in #6575
• Test: Fix the script for testing deepks_bandgap>1. by @ErjieWu in #6604

Doc

• Docs: Update documentation of dft_functional by @ahxbcn in #6591

Version Update

• Update version to 3.9.0.17. by @ZhouXY-PKU in #6638

New Contributors

• @MoseyQAQ made their first contribution in #6603

Full Changelog: v3.9.0.16...v3.9.0.17

v3.9.0.16

What's Changed

Feature

• [Feature] Adapt ABACUS to newest version of mtblas and mtfft by @Critsium-xy in #6548
• [Feature] TD-OFDFT by @lyb9812 in #6538
• Add subspace diagonalization for orthogonal input by @Cstandardlib in #6559
• Performance: Optimizing EXX PW memory usage by calculating potential on the fly by @Flying-dragon-boxing in #6560
• Feature: add ABFS_ORBITAL and ABFS_JLES_ORBITAL in class Exx_Opt_Orb by @linpeize in #6561

Fix

• Fix the errors related to the "atomicAdd" function when building with libtorch-gpu. by @ZhouXY-PKU in #6551
• Fix: Fix the errors in building abacus with libtorch-gpu by @sunliang98 in #6554
• Fix: EXX PW GPU by @Flying-dragon-boxing in #6564
• Fix: restore out_chg for OFDFT by @sunliang98 in #6570
• Fixed DFT-1/2 memory bug and added integration testing for DFT-1/2. by @Wuming-HUST in #6573
• [Fix] Fix problems in the latest develop version causing compilation error on DSP hardware. by @Critsium-xy in #6577

Refactor

• Refactor: rename dngvd op by @Cstandardlib in #6542
• Move some output information out of ESolver_KS_PW, not an easy task, actually. by @mohanchen in #6541
• Refactor: rename dngvd files by @Cstandardlib in #6543
• Removed the temporary variable hr_Gint_full_ when transitioning from 2D block parallelism to serial in Hcontainer (develop) by @zgn-26714 in #6510
• change BFGS name and make lattice_change_cg and ions_move_cg shorter by @19hello in #6517
• Make Esolver simpler by @mohanchen in #6557
• Make ESolver simpler again by @mohanchen in #6565
• Clean and Reconstruct ESolver --- 3rd time by @mohanchen in #6567
• Update DeePKS and do some reconstructions by @mohanchen in #6576
• The United Connector of LAPACK by @Cstandardlib in #6579

Version Update

• Update version.h by @Critsium-xy in #6580

New Contributors

• @ZhouXY-PKU made their first contribution in #6551

Full Changelog: v3.9.0.15...v3.9.0.16

v3.9.0.15

What's Changed

Feature

• Feature: add DFT-1/2 and shell DFT-1/2, only support PW esolvers_ks_pw. (Useful information to support DFT-1/2 method with pw basis set) by @Wuming-HUST in #6490

Fix

• Fix : add overlap output for deepks_out_freq_elec by @xuan112358 in #6505
• Fix: correct an output in LR hsolver by @maki49 in #6503
• fix: psi error for large system by @dyzheng in #6501
• Fix a minor bug in v_xc timer by @AsTonyshment in #6514
• Fix: fix output bug by @dzzz2001 in #6518
• Fix: fix memory leak in gint module by @dzzz2001 in #6515
• Fix: replace fflush with flush by @dzzz2001 in #6521
• Fix: Plug memory leak in dav_subspace by @Cstandardlib in #6532

Refactor

• Refactor: change cal_proj() to cal_mul() in Exx_Opt_Orb by @linpeize in #6506
• Refactor: EXX PW Code Structure by @Flying-dragon-boxing in #6431
• [Refactor] Move module_fft to source_base by @Critsium-xy in #6508
• Refactor: Simplify EXX related code in ESolver_KS_PW by @Flying-dragon-boxing in #6530
• Move some codes out of esolver_fp by @mohanchen in #6527
• 1. Improve output accuracy of Exx_Opt_Orb by @linpeize in #6509
• Update output formats, prepare for ABACUS 3.11 version by @mohanchen in #6522
• Add some timers related to FFT by @mohanchen in #6537

Test

• Tests: Update DeePKS UT to make sure it covers some checks. by @ErjieWu in #6480
• Tests: Activate PW/GPU test for out_wfc_pw by @Cstandardlib in #6523

Docs

• docs: Added keyword descriptions for the DFT-1/2 method. by @Wuming-HUST in #6528

Version Update

• Bump version to v3.9.0.15 by @Cstandardlib in #6540

New Contributors

• @Wuming-HUST made their first contribution in #6490

Full Changelog: v3.9.0.14...v3.9.0.15

v3.9.0.14

What's Changed

New feature

• Feature: Add support for ML EXX in training script. by @sunliang98 in #6479
• Feature: Add deepks_out_base to support simultaneous output of numpy files of base and target functionals by @xuan112358 in #6483
• Feature: support NVTX profiling via timer_enable_nvtx flag by @wangtianxiang in #6495

Fix

• Fix: Name conflict in pyabacus. by @ErjieWu in #6466
• Fix: Value error of vdrpre in DeePKS. by @ErjieWu in #6476
• Fix: resolve compile error with USE_ELPA=OFF + BUILD_TESTING=ON and switch to nvtx3 headers when CUDA_VERSION >= 12090 by @wangtianxiang in #6497
• [Bug] Fix dsp compilation error by @Critsium-xy in #6499
• Fix: Fix crash in Debug build with multi-GPU due to forced cudaSetDevice(0) by @wangtianxiang in #6498

Refactor

• Delete useless codes in SIAB directory by @mohanchen in #6467
• Refactor: remove the INPUTw input script and transfer parameters to INPUT (SIAB refactor 1/3) by @kirk0830 in #6475
• add output information about device count by @dzzz2001 in #6484
• Update out_dmk and out_dmr by @mohanchen in #6481
• [Refactor] Move hardware initializer out from esolver code by @Critsium-xy in #6494
• Build(deps): Bump actions/setup-python from 5 to 6 by @dependabot[bot] in #6492
• Removed the temporary variable DMRGint_full when transitioning from 2D block parallelism to serial in Hcontainer(develop) by @zgn-26714 in #6489

Tests&Docs

• Test: Add test for out_alllog to ensure the output log filenames are correct by @AsTonyshment in #6482

Perf

• Perf: Optimize Davidson by fusing operators, offloading CPU computation to GPU, and reducing memory transfers by @wangtianxiang in #6493

Version Updates

• Update version to 3.9.0.14. by @ErjieWu in #6504

New Contributors

• @wangtianxiang made their first contribution in #6495

Full Changelog: v3.9.0.13...v3.9.0.14