Releases

Current release


For questions please use the QMCPACK Google Group.

A list of releases made from GitHub can be browsed at https://github.com/QMCPACK/qmcpack/releases

 

[3.2.0] – 2017-09-21

Download QMCPACK v3.2.0

Notes

This release provides a significant speed increase for many calculations. A C++11 compiler is now required. It is a recommended update.

QMCPACK updates

  • Major speedup for calculations using spline wavefunctions via initial implementation of “Structure of Arrays” data layout and improved algorithms. Enabled via -DENABLE_SOA=1. Benefits all CPU architectures. Many runs are doubled in speed. Not yet available for Gaussian-basis sets or for all observables and QMC methods. See writeup in manual for guidance.
  • A compiler supporting C++11 is now required.
  • DMC respects MaxCPUSecs parameter and will gracefully shut down and not start a new block if there is not sufficient estimated time to complete it.
  • Checkpointing code rewritten for robustness and performance at scale. Parallel as well as serial HDF5 supported and autodetected.
  • Improved beta-release of AFQMC code and documentation.
  • Backflow documentation and optimization tips added.
  • Correlated sampling VMC drivers reactivated.
  • Added carbon graphite performance test similar to CORAL benchmark.
  • Improvements to CMake and CTest usage.
  • Build instructions for NERSC, ALCF, and OLCF machines updated.
  • Latest manual PDF now available at http://docs.qmcpack.org

NEXUS

  • Significantly improved manual entry for “qmca” analysis tool, the main recommended tool for statistical analysis of QMCPACK data.
  • Added time step fitting tool “qfit” for timestep extrapolation. Uses jack-knife statistical technique.
  • Improved density file postprocessing.
  • Support for Makov-Payne corrections.

[3.1.1] – 2017-08-01

Download QMCPACK v3.1.1

Notes

This is a bugfix release and recommended update.

QMCPACK updates

  • Added numerical tolerance to check of jastrow cutoff and Wigner Seitz radius.
  • CMake correctly configures when MPI is not present.
  • Improved support for test coverage measurements.
  • Added unit tests for some estimators.

NEXUS

  • IPython compatible exit handling (from Duy Le)

[3.1.0] – 2017-06-21

Download QMCPACK v3.1.0

Notes

This release incorporates an improved DMC equilibration scheme, numerous bugfixes, small improvements, and significantly improved testing. It is a recommended update.

QMCPACK updates

  • Improved population control during DMC equilibration. Reduces variance on larger runs.
  • Bugfix: Real valued wavefunction GPU code gave incorrect result for some non-gamma twists that could be made real, e.g. X point. Complex code (QMC_COMPLEX=1) was always correct.
  • All particle move VMC and DMC algorithms enabled, tests added.
  • Reptation Monte Carlo (RMC) enabled, tests added.
  • Significantly improved AFQMC implementation.
  • Added NiO based VMC and DMC performance tests and description in manual. Wavefunction files accessed via QMC_DATA.
  • Added DMC tests with locality and t-moves approximations.
  • Added AFQMC tests.
  • Added test of real space QMC restart capabilities.
  • Added tests for several estimators.
  • Added unit test for DMC walker propagation, effective core potentials, and OhmmsPETE.
  • To avoid filesystem limitations, QMC_SYMLINK_TEST_FILES can be set to symlink (1) or copy test files (0).
  • Fixed mixed precision Ceperley force evaluation.
  • Many updated tests to improve statistical reliability. Removed flux estimator from short tests because they were not reliable enough.
  • Tests that rely on non-standard python modules that are not available are skipped.
  • Error trap jastrow factors with cutoff radii larger than Wigner Seitz radius.
  • Bugfix: Prevent users from adding correlation terms on non-existing electron pairs, e.g. up-down correlation terms when only up-spin particles are present.
  • Support for measuring test coverage and performing coverage runs with cmake and ctest.
  • Support for GCC7 and IBM XL (non Blue Gene) compiler.
  • Support selecting GPU microarchitecture via -DCUDA_ARCH=sm_35(default).
  • SummitDev IBM Minsky build recipe (Power8 + NVIDIA Pascal P100 GPUs).
  • Significantly updated optimizer description in manual, including excited state optimization.
  • Added description of using Intel MKL with non-Intel compilers in manual.
  • Added description of MPIEXEC and MPIEXEC_NUMPROCS_FLAG to manual for systems where MPI runner is non-standard.
  • Updated labs with correct pseudopotentials, basis set files.
  • Many updated error messages and warnings.

Known problems

  • AFQMC without MKL will fail, e.g. short-afqmc-N2_vdz-4-1 test fails.

NEXUS updates

  • Improved selection algorithm to obtain optimally tiled supercells.
  • Support for parallel pw2qmcpack workflows.
  • Support for HPC resources at the Leibniz Supercomputing Center.
  • Better consistency checks for the Structure class.
  • Bugfix: forbid job bundling for simulations that depend on each other.
  • Bugfix: correctly select low spin polarization in primitive and tiled (net_spin=”low” option).

[3.0.0] – 2017-01-30

Download QMCPACK v3.0.0

Notes

We are adopting Semantic Versioning with this release. It is the first to be made from the git repository on GitHub, and the first named release since 2016-06-02 and subversion revision 6964.

A potentially severe bug is fixed for periodic wavefunctions in this version, in addition to many usability improvements and bugfixes. All users are strongly recommended to upgrade.

NEXUS updates are listed after QMCPACK updates.

QMCPACK updates

  • IMPORTANT BUGFIX: Real-valued wavefunction code would occasionally make a numerically unstable choice for constructing real-valued periodic wavefunctions, leading to large variances and poor energies. Algorithm for constructing wavefunctions improved.
  • Fully parallel pw2qmcpack.x for QE 5.3, enables conversion of large wavefunctions and use of same parallel setup as pw.x runs.
  • Full testing of Quantum Espresso workflows (pw.x -> pw2qmcpack.x -> qmcpack). Specify directory containing QE binaries via QE_BIN during configuration.
  • Added open boundary conditions tests using QE wavefunctions, as might be used for molecular work. Requires QE_BIN and computes trial wavefunction on the fly.
  • Added DMC, optimizer and additional system tests.
  • Added unit tests using the Catch framework.
  • Plane wave wavefunctions can be evaluated in plane waves, use “pw” as determinantset type. Slow, but useful for checking spline accuracy. Tests added.
  • Complex implementation on GPUs, supports arbitrary twists and complex phase wavefunctions as per CPU code.
  • Flux estimator correct for complex wavefunctions.
  • Mixed precision CPU implementation, activated via -DQMC_MIXED_PRECISION=1.
  • Double precision GPU implementation, complementing existing mixed precision implementation, activated via -DQMC_MIXED_PRECISION=0.
  • GAMESS CI converter improved.
  • C++11 detection and support.
  • Initial release of new optimizer, requires C++11 (contact Eric Neuscamman).
  • Initial release of orbital-based AFQMC code, requires C+11 and MKL (contact Miguel Morales).
  • Fine grained timers implemented, activated via -DENABLE_TIMERS=1.
  • Improved Intel math and vector math library support. MKL and MKL VML more easily supported with GCC as well as Intel compilers.
  • Many code updates to eliminate CLANG warnings.
  • Configure scripts, printed headers, manual updated for git. Git version printed during configure and on standard output.
  • Source files headers updated to consistently show UIUC/NCSA open source license and list development history.
  • Numerous manual updates.
  • Updated QMCPACK tutorial laboratories.
  • Many small bug fixes, improvements and optimizations.

NEXUS updates

  • General
    • Nexus output now tracks time instead of poll number.
    • Reported memory use now includes child processes.
  • Workflow generator
    • Major new capability to generate simple to complex workflows involving QE, VASP, and QMCPACK.
    • Aim is to allow single notebook/worksheet describing all simulation workflows needed in a project.
    • Users can succinctly create any subchain of the workflow: relax->scf->nscf->orbital_conv->qmc.
    • Additional elements can be added to workflow chains over time as needed.
    • Scans of structural parameters and input parameters at any level of the chain are possible.
    • No programming constructs are required (for/if, etc).
    • Directory substructure is automatically generated in the case of scans.
    • Native support for visualizing workflows via pydot is provided.
    • Documentation for this feature is pending.
  • Quantum Espresso workflows
    • Support for vdW functional input.
    • Fixes to SCF->NSCF workflows for QE 5.3.0+.
    • Support for automatic restarts of SCF runs.
    • Native support for workflows involving post-processing tools
    • pp.x, dos.x, bands.x, projwfc.x, cppp.x, pw_export.x supported.
    • Postprocessing and summary of Lowdin charge data from projwfc.x.
  • QMCPACK workflows
    • Fixes for QE/VASP structural relaxation -> QMCPACK workflows.
    • Fixed job bundling of twist averaged runs.
    • Support for partitioned sposet input.
    • Supercomputing environments
    • Native support for several supercomputing environments located at Sandia Nat. Labs.
    • Atomic structure manipulation
    • Ability to find optimal supercells, similar to getSupercell tool.
    • Robustness fixes to tiling operations.
  • Tools
    • qmca
      • Fix for twist averaging with user-provided weights.
    • qmcfit
      • New command line tool for jack-knife fitting of QMCPACK data.
      • Timestep extrapolation currently supported.
      • General binding/equation of state fitting pending.

Older releases


Release date: 06-02-2016

Name: QMCPACK_rev6964_20160602

Revision: 6964

MD5 = 5f6f654116b3f89464697f730dbd285f

Description:

  1. QMCPACK workshop tutorial files included with distribution in labs directory.
  2. Initial pseudopotentials database. BFD pseudopotentials in GAMESS, UPF, and QMCPACK XML formats. Other pseudopotentials will be added in later versions.
  3. Manual updates.
  4. Output cleanup.
  5. Many updates and cleanup for initial support of CLANG C++ compiler and LLVM.
  6. C++11 support improved.
  7. Faster gaussian basis calculations.
  8. Overall improved GPU performance.
  9. GPU spline memory limitation removed via Spline_Size_Limit_MB setting. Uses zero copy memory on CPU for spline data that exceeds this setting. Enables large spline runs on GPUs, but at a performance penalty.
  10. Double precision GPU implementation  (matrix inversions and coulomb interaction are always double).
  11. Flux estimator added to tests and supported by GPU.

 

Release date: 04-04-2016

Name: QMCPACK_rev6854_20160404

Revision: 6854

MD5 = 7bd61d38414ed7bb09f55c204c5a2a8a

Description:

1. Latest Quantum Espresso 5.3 supported. See external_codes/quantum_espresso/README. Patches for previous versions also provided.

2. QMCPACK executable now “qmcpack”, no longer “qmcapp”.

3. CMake build system significantly updated. Run once on most systems, better autodetect on more systems. See README.

4. Preview of new manual, see manual/qmcpack_manual.pdf Has latest/up to date description of installation, tests, quick start. Later sections are work in progress, but all known input  parameters are listed. Final version will also be online in HTML.

5. Build recipes for common systems included in manual: standard Linux, homebrew/Mac, Blue Gene, Crays (CPU & GPU). Tested at OLCF (titan, eos), ALCF (mira, cetus), NERSC (cori,  edison), Linux workstations.

6. Test system included. ctest based system runs numerical checks of Monte Carlo code and some converters. See README.

7. Major NEXUS update including:

– GAMESS->QMCPACK workflows (HF/DFT/CISD/MCSCF, enforcing nuclear cusps on gaussians).
– GAUSSIAN/GAMESS/CRYSTAL basis set conversions.
– GAMESS->QMCPACK pseudopotential format conversion.
– Interface to Pitzer’s ATOMSCF code, sometimes used in pseudopotential development.
– Support for template QMCPACK input files provided by user in addition to input file generation.
– Support for CIF files for structure input.
– Support for spglib to analyze crystal symmetries.
– (Fixed) Ability for automated workflow execution during interactive supercomputer sessions .

8.  Initial version of Fragment Molecular Orbital (FMO QMC) implementation with GAMESS

Minor updates include:

– Optimization and build flags set according to build type, including CUDA flags.
– Many builds now complete without warnings.
– MKL used automatically when using Intel compilers.
– Codebase now supports C++11, but not enabled by default.
– Compatibility with latest CUDA.
– Serial build supported.
– Slurm supported on Crays.
– Improved spline performance.
– Memory usage of determinants and 3-body jastrow function significantly reduced.
– ppconvert supports up to g angular momenta
– QMCPACK is now tested nightly and weekly on an automatic basis.


 

Release date: 12-08-2014

NameQMCPACK_rev6406_20141208

Revision: 6400

MD5 = 7fffe1c8d1c7ab3e6b02a613ec530a70

Description:

  • Important bug fix to Quantum Espresso 5.1 converter. Accommodates a change in array specification made in Quantum Espresso. Wavefunctions previously converted using 5.1 are incorrect.
  • Note: Fix for examples broken link. The links in the example folder for the Water molecule and the Graphite benchmark are unfortunately broken. Please download the file below and replace it in ~/qmcpack/examples/sc12_qmc

getfiles.sh

These links will be updated in the next release.


 

Release date: 11-10-2014

Name: QMCPACK_rev6400_20141110

Revision: 6400

MD5 = 4911a15ba4b652a5c7d482861c4e9c52

Description:

  • PPconvert included. See QMCTools/ppconvert/README
  • Automated patch and download of Quantum Espresso 5.1. See external_codes/quantum_espresso/README
  • Large particle count supported on CUDA GPUs
  • “Project suite” has been renamed “Nexus”. Many improvements including to GAMESS implementation.
  • Significant improvements in reptation Monte Carlo implementation
  • CMake toolchain updates for several systems
  • Builds on OS X Yosemite using Accelerate framework
  • Many minor bug fixes and extensions including:
    • g(r) in open boundary conditions
    • Abort when no rcut is given for bspline Jastrows in isolated systems
    • K-space Jastrow correctly optimizes

 


Release date: 07-30-2014

Name: QMCPACK_rev6334_20140730

Revision: 6334

MD5 = 3c36ebbe039dc16e4c187e2b79962562

Description:

  • First Release