Releases

Download QMCPACK v3.14.0

Notes

This release focuses on performance improvements to the OpenMP target offload version for GPUs as well as ongoing minor improvements. The new GPU implementation rivals the legacy CUDA version for performance for broad range of problems while offering more functionality, such as three body Jastrow functions. Developers are very interested in feedback from users about the new version and will prioritize developments based on comments received. A new driver_version switch is introduced, currently optional, to disambiguate between the versions and their inputs.

• New global driver_version switch to select between batched and legacy codes. This will become a required input tag in the next major release series of QMCPACK, but remains optional in 3.x versions #3897
• Optimization of block sizes in GPU offload kernels #3910
• GPU Offload of one-body Jastrow ratio calculation in pseudopotential evaluation #3905
• GPU Offload of some Coulomb potential evaluations #3842
• Partial GPU offload of multideterminant evaluation e.g. #3892
• Increased performance via more selective distance table computation #3846
• Improved performance on AMD GPUs via rocSOLVER integration #3756
• HIP build options shown in output #3919
• Documentation improvements, particularly relating to installation.
• Various bug fixes and ongoing cleanup.

NEXUS

• Nexus: proper use of max_seconds in legacy drivers #3877

Download QMCPACK v3.13.0

Notes

This release incorporates support for trial wavefunctions from Quantum ESPRESSO 7.0 and adds GPAW support for the first time. Non-local pseudopotential derivatives are fully supported in the optimizer and recommended in standard calculations.
Numerous minor bug fixes, test and installation improvements have been made. Behind the scenes updates include maturation of the OpenMP target offload implementation and the batched drivers, a partial implementation of fast force calculations, and ongoing modernization of the code. This is a recommended release for all users.

• Support for Quantum ESPRESSO (QE) 7.0 #3683
• Support for GPAW and GPAW to QMCPACK converter #3490
• use_nonlocalpp_deriv is fully supported and preferred in optimization #3785 and others.
• Save and restore of variational parameters during optimization #3640
• Twist attribute takes precedence over twistnum. Twist is preferred specification. #3799
• Fixed inconsistent twist directions between electron gas and spline wavefunctions #1386
• Fixed reported Madelung constant in CoulombPBCAA #3806
• More robust computation of reference ion-ion Coulomb energy #3763
• Expanded test set, including more coverage of plane-wave basis sets and complex molecules #3105, #3822
• More consistent python invocations #3680
• Builds with OpenMP disabled (QMC_OMP=0) again supported #3723
• Modernization of HDF5 usage e.g. #3705
• Minimum supported Intel classic compiler version is 19.1. #3747
• Various minor bug fixes and ongoing code cleanup.

NEXUS

• Nexus: Add --user $USER to squeue command #3796
• Nexus: Add Example and tests for qdens-radial tool #3676
• Nexus: Add Lowdin example #3666
• Nexus: Fixed Nexus 'install' target #3720
• Nexus: Harden Nexus excitation checks #3729
• Nexus: Small fix to excitation checks #3701
• Nexus: Faster configuration time #3706

Download QMCPACK v3.12.0

Notes

This release incorporates several hundred changes to QMCPACK and the supporting ecosystem. It is a recommended release for all users. Note that compilers supporting C++17 and CMake version 3.15 or newer are now required. Changes include newly added support for the DIRAC quantum chemistry code, the RMG-DFT code, and updates for the latest version of Quantum ESPRESSO. Through DIRAC it is now possible to perform highly accurate molecular calculations incorporating spin-orbit with multideterminant trial wavefunctions. Behind the scenes updates include increased checking of inputs, fixes to many edge case bugs, and removal of memory leaks in both QMCPACK and the various converters. In readiness for transition to the new batched drivers that support both CPU and GPU execution, more features are supported and performance improved. Test coverage and robustness is improved in all areas. For developers, tests, sanitizers, and code coverage are now run on Pull Requests using GitHub Actions.

• To aid coexistence of real and complex builds, the qmcpack executable is now named qmcpack_complex for builds with QMC_COMPLEX=1
• Added DIRAC converter and support for MSD wave functions #3510
• Spin-Orbit implementation completed #1770
• Quantum ESPRESSO (QE) v6.8 support #3301
• Support for RMG DFT code #3351
• CMake 3.15 minimum required #3492
• C++17 is required #3348
• CMake CUDA support uses modern FindCUDAToolkit #3460
• Support latest Sphinx-contrib BibTeX 2.x #3176
• One Body Density Matrices supported in batched drivers #3622
• Batched performant Slater matrix inverses #3470
• Safeguards for requesting more orbitals than the input h5 provide #2341
• Implemented One-body spin-dependent Jastrow #3257
• Fixes for low particle counts, such as using a two body Jastrow with more than 2 particle types but only one particle of each type #3137
• ppconvert is built by default #3143
• Documentation on revised input format where SPO sets are created outside the determinant #3456

NEXUS

• Add Density functionality to qdens tool #3541
• Add new qdens-radial tool for radial analysis of densities #3587
• Radial density of requested species only #3099
• Extend structure plotting capabilities for 2D materials #3220
• Support grand-canonical twist averaging #3153
• Extend excitations to allow 'lowest' gap #3628
• Allow singlet/triplet excitation types #2290
• Allow bandstructure plotting with custom k-path #3293
• Generate PySCF inputs without a template #3550
• Add punch extension for GAMESS analysis #3433
• Read pseduopotentials in numhf format (Eric Shirley's numerical HF code) #3097
• Add L2 generation functionality #3079
• Support QMCPACK batched drivers #2901
• Make qdens test more informative #3593
• Resource lock Nexus examples for reliable parallel execution #3585
• Support running tests without mpirun available #3584
• Small fix for custom band plotting #3566
• Improve error handling for bad Jastrow requests #3554
• Fix sizing problem in some single atom workflows #3553
• Fix syntax warnings #3497
• Fix convert4qmc usage #3495
• Verify cif2cell is available before running ntest_nexus_structure #3511
• Fix to add_L2 function in pseudopotential.py #3386
• Expand eshdf features #3334
• Add delay_rank input #3218
• Add max_seconds input #3159
• Add Tref (initial tilematrix) argument to optimal_tilematrix #3141
• Use OS environment by default #3108

Download QMCPACK v3.11.0

Notes

This release includes a large number of refinements to QMCPACK and the supporting ecosystem. These include support for the latest version of Quantum ESPRESSO, new capabilities in AFQMC, space-warp transformation for forces, numerous bug fixes, user-requested feature improvements, and further upgrades to the test system.

• Quantum ESPRESSO (QE) v6.7 support. #2927.
• Detect and automatically use patched version of QE found on the PATH. #2974.
• Support for global max_seconds and STOP file to cleanly halt QMCPACK during a run. #3028.
• Freezing of two-body Jastrow parameters in optimization works. #2814.
• Multideterminant code now works with only alpha determinants (no down electrons). #2698.
• High l-momentum channels as local channels in ECPs work. #2920.
• Space Warp Transformation for ZVZB Forces. #2828.
• Important bug fixes in legacy CUDA implementation causing incorrect energies. #2883.
• Implemented DLA in legacy CUDA. #2887.
• Updates to support CUDA 11.2.1 e.g. #2950.
• AFQMC supports energy estimator with different Hamiltonian (from propagation). #2795.
• Trial wavefunction optimization with spin-orbit supported. #3034.
• ppconvert executable automatically built when configured. #2904.
• Tests added for ppconvert. #2929.
• Fixed SIMD alignment for AVX512 on some systems. #2981.
• Improved wavefunction restart logic in AFQMC. #2942.
• Spin-density supported in batched code. #2840.
• Reduced I/O operations during cmake. #2808.
• Improved detection of unsupported-by-Intel combinations of Intel compilers and libstdc++. #2794.
• Initial support for Andes at OLCF. #3073.
• Deterministic tests expanded in scope and made reliable for more build types and compilers.
• Various minor bug fixes and feature improvements based on user requests for both real-space and AFQMC.
• Improved error handling throughout.
• Numerous performance improvements, expansion of tests, and bug fixes to the batched VMC and DMC codes. Reasonable but not optimal GPU acceleration can now be achieved for spline-based wavefunctions.

NEXUS

• Support AMD nodes on Cori. #2809.
• Interface for RMG code. #2932.
• Added h-channel to list of possible local channels in pseudopotential. #2915.
• Allow non spin-specific occupations in case of noncollinear. #2957.
• More robust handling of QE output when printed eigenvalues touch. #3042.
• Fixed type check for reblock_factors in qmc-fit. #2830.
• Fixed a Jastrow read error/warning, add several QE inputs. #2819.
• Fixed tests on Summit. #2983.
• Fixed module overwrite bug in qmca. #2802.

Download QMCPACK v3.10.0

Notes

This release contains multiple feature improvements and bug fixes. The AFQMC implementation has been significantly enhanced, and an important wavefunction optimization bug fixed in real space QMC.

• The QMCPACK manual is now available at https://qmcpack.readthedocs.io, having been converted to use reStructuredText and the sphinx documentation system.
• Significant improvements to the AFQMC code including HIP support for AMD GPUs, updated documentation, and support for non-collinear calculations and spin-orbit k-point Hamiltonians #2734.
• Improved support for spin-orbit in real-space QMC including documentation #2733.
• Important bug fix for wavefunction optimization in few electron systems such as isolated atoms. The bug would result in slow or no convergence. Thanks to Jaron Krogel and Matus Dubecky for reports and reproducers. #2496.
• Implementation of L2 potentials and evaluation in DMC #1948.
• Consistent with our two year support policy for open source compilers, libraries, and tooling, several version minimums have been increased to either avoid bugs or to utilize new features.
• Clang 7 is the earliest supported Clang compiler. The latest release is recommended.
• Intel 2019 is the earliest supported Intel compiler. The latest release is recommended.
• Future releases of QMCPACK will require C++17. The current minimum is C++14.
• AoS builds are no longer supported. The code has been removed now that the default structures-of-arrays (SoA) build has sufficiently broad capability.
• The default CUDA architecture is set to sm_70 (Volta).
• QMCPACK is built with ENABLE_TIMERS=ON by default #2663
• Various bug fixes to complete the transition to Python 3.
• Ongoing improvements to the OpenMP offload implementation.

NEXUS

• NEXUS manual is now available at https://nexus-workflows.readthedocs.io, having been converted to use the reStructuredText and sphinx documentation system.
• Various small fixes and improvements.