GAMESS - Release Notes

GAMESS - Release Notes

September 30, 2020 R2 Public Release

GAMESS

• OpenMP-threaded RI-CC method (#413) (D. Datta of Iowa State University)
  • This method is currently only available for the OpenMP-threaded version of GAMESS.
  • Only CCSD and CCSD(T) calculations can run using the RI approximation for two-electron repulsion integrals and a hybrid MPI/OpenMP parallelization model
  • See new CCERI flag in $CCINP
  • See new $RICC input group
    • This group controls the calculation of ground state energies for closed-shell molecules (SCFTYP=RHF) using the coupled-cluster method within the singles and doubles (CCSD) truncation scheme or CCSD augmented with an approximate triples correction: CCSD(T). These calculations run in parallel using a hybrid MPI/OpenMP parallelization model and employ the resolution-of-the-identity (RI) approximation for two-electron repulsion integrals (2ERIs).
• ENTRY statement replacement:
  • in mcqud.src, mcqdpt.src, mcpgrd.src, gugdm.src, gugdm2.src, and gugem.src (#335) (L. Carrington of EP Analytics)

Along with a few minor fixes:

GAMESS

• Set RORB to zero when NROT=0 (#400) (C. Friedl of Johannes Kepler University Linz)
• Fix $GMS_DIR/tests/sformas/eth-sfcepa.inp (#391) (J. Mato of University of Colorado Denver)
• Address FPE in standard exams related to opimtization level of dgeev.src and dftgrd.src when compiled with PGI 19.10 (#393) (C. Friedl of Johannes Kepler University Linz)

Software related changes:

GAMESS

• Performance regression metrics (#415) (S. Leang of EP Analytics)

  • Updated parser definition
    • as a result several validation files had to be regenerated
  • New flag for checkgms.py --test_path TEST_PATH to specify a directory path for checking
  • New flag -m {validation,regression,both}, --mode {validation,regression,both} to allow users to select between performing validation testing, regression testing, or both
  • Flags for -m regression, --mode regression
    • -a ARCH, --arch ARCH - architecture string used for performance regression
    • -s SYSTEM, --system SYSTEM - HPC system string used for performance regression
    • -n NCPUS, --ncpus NCPUS - number of GAMESS compute processes for performance regression
    • -g NACCELERATORS, --naccelerators NACCELERATORS - number of accelerators for performance regression
  • Performance data is stored with .perf extension
  • Example usage: ./checkgms.py --mode regression --system "Summit" --arch "Power9 with V100" --ncpus 40 --naccelerators 1
  • Running ./checkgms.py --mode regression without any additional flags will show you what baselines are available and what flags need to be passed.
  • Performance tolerance is currently set at +/- 5%

• Production code does not set GMS_FPE_FLAGS (#393) (S. Leang of EP Analytics)

• serial-debug support in misc/automation/rungms and tests/runtest.py (#388) (C. Friedl of Johannes Kepler University Linz)

  • tests/runtest.py now has the ability to print the output of the test runs directly to console via the new argument -i or --print_to_stdout.
  • For a build with the option GMS_DDI_COMM=serial-debug: When using tests/runtest.py -i misc/automation/rungms now automatically runs the tests under the control of gdb. In the case of using tests/runtest.py without -i, the tests are executed normally (without gdb) and the output is written to the .log files in their standard location.

• Removal of actvte.x (#383) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)

• Use standard functions in place of compiler extensions (#334) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)

  • IAND, IOR, IEOR
  • DATE_AND_TIME, SYSTEM_CLOCK, CPU_TIME
  • GET_ENVINROMENT_VARIABLE
  • FLUSH
  • Instead of C calls MEMGET and MEMREL the static allocatable array is used (defined in FAST_MEMORY in unport.src) with the hack (function get_ptr_addr in unport.src) for pointer arithmetic (Fortran 2003).

• Add --pre and --post flags to runtest.py (#389) (S. Leang of EP Analytics)

• IBM64 Summit modifications (#406) (S. Leang of EP Analytics)

  • 64-bit ESSL support
  • Force auto deallocation

• Build improvements (#393) (S. Leang of EP Analytics) (F. P. Rotzinger):

  • Floating point exception flags are not set in production.
  • Add "-Wno-maybe-uninitialized" flag for GNU compiler builds in production.

• Array handling mod_shell_tools for IBM XL compilers (#373) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)

June 30, 2020 R1 Public Release

GAMESS

• The nuclear-electronic orbital (NEO) method package in GAMESS has been updated to include NEO-DFT capability with electron-proton correlation methods [1,2]. The epc17 functional is now available as a treatment for electron-proton correlation in a NEO-DFT calculation. The capability to calculate proton density on a set of grid points provided by the user has also been added. The documents NEO.DOC and NEOINP.DOC in the qmnuc/ folder have been updated to reflect these changes, but a brief summary follows. Electron-proton treatment is controlled by the NEODFT flag. Currently, the only available treatment is NEODFT=CS4 for epc17. Parameters for the functional [1,2] are specified by YANGA, YANGB, and YANGC. The flag IPRDEN=1 designates a run that will calculate proton density at specific Cartesian points (given by the user in $POINTS, see GAMESS input manual). All new input flags belong to the $NEO group. (#183) (P. Schneider of Yale University / Hammes-Schiffer Group)
  • [1] Y. Yang, K. R. Brorsen, T. Culpitt, M. V. Pak, S. Hammes-Schiffer, Development of a practical multicomponent density functional for electron-proton correlation to produce accurate proton densities, J. Chem. Phys., 147, 114113 (2017).
  • [2] K. R. Brorsen, Y. Yang, S. Hammes-Schiffer, Multicomponent density functional theory: impact of nuclear quantum effects on proton affinities and geometries, J. Phys. Chem. Lett., 8, 3488-3493 (2017).
• MPI/OpenMP implementation for QM-EFP2 energy calculation. More specifically, the implementation is for QM-EFP2 exchange repulsion electron repulsion integrals (ERI) and corresponding Fock contribution construction, as well as QM-EFP2 Schwartz inequality screening subroutine. A QM-EFP2 test folder is created inside $GMS_DIR/tests/solvent (look in $GMS_DIR/tests/solvent/qm-efp2/exrep/mpiomp). The pre-existing $GMS_DIR/tests/exrep folder was moved inside $GMS_DIR/tests/solvent. (#207) (P. Xu and T. Sattasathuchana of Iowa State University)
• QM-EFP improvements (#216) (P. Xu and T. Sattasathuchana of Iowa State University)
  • Use dynamic memory allocation for QM-EFP2 dispersion (EFAIDISP).
  • Post-HF methods, MP2 and Coupled-Cluster, now are interfaced with EFP2
  • New tests added to $GMS_DIR/tests/exrep
• FMO 5.4 (#233) (D. G. Fedorov of the National Institute of Advanced Industrial Science and Technology Japan)
  • New Features
    • Multiple-level GDDI parallelization (A. Gaenko, S. Pruitt).
    • Analytic FMO-PCM Hessian (H. Nakata).
    • Long-range corrected DFTB (V. Q. Vuong and Y. Nishimoto).
    • New (so called partial) screening model for PIEDA/PCM.
    • Minimum energy crossing with FMO (D. Kaliakin).
    • Molecular electrostatic potential (MEP) with PCM (V. Mironov).
  • Improvements
    • Improved error diagnostic for host lookup in DDI.
    • Acceleration improvements for FMO3.
    • Dimer data stored in FMO for later reuse.
    • Some improvements in SCC of DFTB (Y. Nishimoto).
    • Some non-minimal basis sets are enabled for HF-3c (K and Ca).
    • Acceleration for the old (local) screening in FMO/PCM.
    • FMO-MP2 runs enabled for fragments without valence electrons (Ca2+ etc).
    • RUNTYP=OPTIMIZE for FMO now does not stop when PROJCT fails.
    • Parallelized nuclear contributions to MEP.
    • FMO-DFT and HF-3c with PIEDA now print all components clearly.
    • Cleanup of FMO output.
  • Fixes
    • Array declaration issue in FMO (a compiler error for some compilers).
    • Trap undeclared SCFTYP in open-shell FMO.
    • FMO/FDD/PCM had a bug with RESPPC.
    • MODGAM=8 did not work with FMO-DFTB/PCM (Y. Nishimoto).
    • Coordinate units were wrong for printing tesserae data in FMO/PCM.
    • RI-MP2 ignored SCSPT option.
    • FMO/AP/PCM has a major memory bug.
    • Total binding for modmol options did not add solvent screening in FMO/PCM.
    • FMO-MP2/AP had a major bug for dimer calculations.
    • Divisions by zero in DFTB are removed. (Y. Nishimoto).
    • FMO-UDFT incorrectly calculated the electron count.
    • FMO-TDDFT had a major bug for iexcit(2)=1.
    • TDDFT overwrote exchange fraction.
    • Pople integrals used undeclared variables.
    • LAPACK KDIAG=5 was fixed.
• FMO 5.4 patch (#268) (D. G. Fedorov of the National Institute of Advanced Industrial Science and Technology Japan)
  • Patch DDI to stop warnings during compilation (define prototypes).
  • Enable the first row of transition metals in HF-3c.
  • Describe FMO/FRET in the manual.
  • Remove symlink for DFTB parameters folder
  • Rename DFTB parameters folder
  • Adjust DFTB test inputs
• Additional of revised M11 functional for DFT energy and gradients DFTTYP=REVM11. (#244) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
  • REVM11 reference
• Interface to LIBXC for ground-state DFT (#274) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
  • Requires GCC compiler > 5.0
  • Build process
    • ./config (answer yes to LIBXC build option)
    • ./tools/libxc/download-libxc.csh
    • make ddi
    • make libxc -j$(nproc)
    • make modules
    • make -j$(nproc)
  • See $LIBXC for usage in docs-input.txt
• MPI and MPI+OpenMP implementation of QM-EFP2 electron repulsion integral schemes (Rotated axis (sp and spd) and ERIC) (#310) (P. Xu and T. Sattasathuchana of Iowa State University).
  • QM-EFP ERI scheme can be controlled by INTTYP in $CONTRL:
    • INTTYP=best (default)
    • INTTYP=rotaxis
    • INTTYP=Eric
    • INTTYP=rysquad
  • Tests
    • MPI implementation: $GMS_DIR/tests/solvent/qm-efp2/exrep/
      • h2o_acetone_efpqm.rys.inp
      • h2o_acetone_efpqm.rotaxis.inp
      • h2o_acetone_efpqm.eric.inp
    • MPI+OpenMP implementation: $GMS_DIR/tests/solvent/qm-efp2/exrep/mpiomp/
      • h2o_meoh_qmefp.rys.inp
      • h2o_meoh_qmefp.rotaxis.inp
      • h2o_meoh_qmefp.eric.inp
• Spin-flip Occupation Restricted Multiple Active Space (ORMAS) energy and gradients (#316) (J. Mato of Iowa State University)
  • CITYP=SFORM or CITYP=SFDET when ALDET code is used.
  • Test inputs in $GMS_DIR/tests/sformas
  • J. Mato, M. S. Gordon, A general spin-complete spin-flip configuration interaction method, Phys. Chem. Chem. Phys., 20, 2615-2626 (2018)
  • J. Mato, M. S. Gordon, Analytic Gradients For The Spin-flip ORMAS-CI Method: Optimizing Minima, Saddle Points, And Conical Intersections, J. Phys. Chem. A, 123, 1260-1272 (2019)
• Nudge Elastic Band Release 3 (#319) (S. Koseki of Osaka Prefecture University, Japan)
  • See tools/neb/README.NEB.md for more information
• Updates to OpenMP threaded RI-MP2 code (#330) (#361) (B. Q. Pham of Iowa State University)
  • Converts energy code from F77 to F90
  • Optimize (E)FMO/RIMP2 performance for systems there the biggest dimer fits into single process memory
• EFMO dispersion tweaks (#372) (A. Gunina of Iowa State University)
• Initial interface for LibAccInt (#378) (B. Westheimer of Iowa State University)

LIBCCHEM

• Decoupling of LIBCCHEM RI-MP2 and DF-HF. The LibCChem RI-MP2 code can be run using different GAMESS HF drivers. (#371) (D. Poole of Iowa State University)

Along with a few minor fixes:

GAMESS

• Resolve floating point exceptions (FPE) in GAMESS:
  • Numerical exception in DFTB: tests/dftb/parallel/exam04.inp (#192) (Y. Nishimoto of Kyoto University)
  • Overflow when running $GMS_DIR/tests/dftb/raman.inp (#220) (C. Friedl of Johannes Kepler University Linz)
  • Overflow in RPBE and RPBEX (#224) (C. Friedl of Johannes Kepler University Linz)
  • FPE in quanpo.src by clearing ENALL(1:42) before GSUMF(2410,ENALL,42) (#228) (C. Friedl of Johannes Kepler University Linz)
  • Issues uncovered with -finit-real=snan (#245) (#248) (#249) (#250) (#278) (C. Friedl of Johannes Kepler University Linz)
  • FPE in H0 (impacts: M08-HXC, M08-HX, M08-SOC, M08-SO, M11-C, M11, M11-LC, M11-L, MN12-L, MN12-SX, MN15, MN15-L, REVM11) (#285) (C. Friedl of Johannes Kepler University Linz)
  • Issues associated with un-initialized variables (#347) (#342) (C. Friedl of Johannes Kepler University Linz)
• Fix scaling parameters in tddfun.src (#236) (S. Leang of EP Analytics)
• Fix segfaults for Intel 18/19 in efmo.src (#201) (I. Rostov of National Computational Infrastructure at the Australian National University)
• Additional TD-DFT tests added to $GMS_DIR/tests/tddft/parallel (#237) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Fix H, I orbital output if AIMPAC=.T. (#211) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Adjust variable array sizing found in basg3x.src, dftbhs.src, ecp.src, efmogrd.src, and qfmm.src (#253) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Common block alignments (#246) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Increase parameter limits in mx_limits.src: mxcpuefp, mxefmopts, mxefmoppts (#256) (S. Leang of EP Analytics)
• Resolve issue with FMO Hessian calculations (#267) (V. Mironov of Lomonosov Moscow State University)
• Resolve issue with DC-RHF when OpenMP is enabled. (#321) (C. Bertoni of Argonne National Laboratory)
• Resolve OOB issues in dftbsk.src, efinp.src, and gmcpt.src (#327) (T. Sattasathuchana of Iowa State University)
• Resolve issues with MC-PDFT code (#312) (#356) (A. O. Lykhin of University of Minnesota Twin Cities)
• Resolve UHF/UDFT wavefunction output (#331) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Resolve ATTRIBUTES directive discrepancy in mod1eprimitives.src (#328) and omp_int1.src (#345) (C. Bertoni of Argonne National Laboratory)
• Fix memory allocation issue in inputb.src (#351) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Re-enabled kinetic bond order and bond order printing during a QUAO procedure (#376) (J. L. Galvez Vallejo of Iowa State University)

LIBCCHEM

• No fixes to report.

Software related changes:

GAMESS

• Ability to write arrays as numpy files (#193) (C. Friedl of Johannes Kepler University Linz)
• Update bin/create-install-info.py to support --ddi_comp==serial-debug (#209) (C. Friedl of Johannes Kepler University Linz)
• Update rungms to check of LD_LIBRARY_PATH exists in the user's environment (#212) (C. Friedl of Johannes Kepler University Linz)
• ENTRY statement replacement in dftgrd.src (#215) (L. Carrington of EP Analytics)
• Define MATHLIBS for cray-xc targets (#227) (C. Bertoni of Argonne National Laboratory)
• Update documentation for DAREAD and IODA (#221) (C. Bertoni of Argonne National Laboratory)
• Add .gitattributes for proper syntax highlighting in .src files (#230) and C-shell scripts (#282) (#311) (C. Friedl of Johannes Kepler University Linz)
• Remove CR LF endings in tddxce.src (#238) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Intel 2019 support:
  • Compilers (#255) (S. Leang of EP Analytics)
  • Intel MPI (#279) (C. Bertoni of Argonne National Laboratory)
• Add DDI stubs for serial and serial-debug compilations. (#263) (S. Leang of EP Analytics)
• Address linking issue with OpenMP builds of GAMESS (#290) (C. Bertoni of Argonne National Laboratory)
• Adjust line-endings (dos2unix) for several files (#297) (S. Leang of EP Analytics)
• path/to/file.inp support for rungms (#307) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Remove EQUIVALENCE statements in GAMESS source files (#308) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Common block to modules for HERMIT, WERMIT, and XYZCHI (#284) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Common block to modules to non-threaded DFT code (#318) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences):
  • CM05; CM06; CM08; CSOGGA; FUNLIB; METGGA; CBECKE; SCTAX; CLYPC; CVWNC; CP86CF; CPBE0; CPW91C; CX3LYP; CPW91L; CTPSSH; SCPBES; CPKZB; SCAEDF; SCAOPC; SCLGIL; SCPFRE; SCPZ81; SCRPBE; B97TYP; SLPBEC, NLRC; DFTDH; DFTCAM.
  • DFTEXC was split to two modules: XDERIX and CDERIX.
• Patches from NCI ANU for issues encountered using Fujitsu fortran compiler (#258) (I. Rostov of National Computational Infrastructure at the Australian National University)
• Makefile changes (#322) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Resolve build issues for mac64 targets (#323) (E. Guidez of University of Colorado Denver)
• Update Travis-CI to used GNU 4.9.4 as baseline GNU compiler version for GAMESS testing (#324) (S. Leang of EP Analytics)
• Update GCC build support (#328) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Enable the running of serial tests calculations in parallel using multi-threading (#341) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Clean up config script (#357) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• Build fixes/enhancements for ibm64 target (Summit)
  • Use of environmental variables in config (#360) (S. Leang of EP Analytics)
  • Resolve compiling issues (#333) (#326) (S. Leang of EP Analytics) (#374) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
  • Address field separator (#301) (S. Leang of EP Analytics)
• Build enhancements (#346) (S. Leang of EP Analytics)
  • Added GMS_HPC_SYSTEM_TARGET to enable conditional paths in GAMESS build and launch scripts for specific systems (e.g., generic (default), onyx, theta, summit, hokulea, qcengine)
  • Added --system_target flag to create-install-info.py
  • Added my_ipcrm script to $GMS_DIR/bin to help clean up user's semaphores (do not use when running multiple GAMESS calculations on same compute nodes)
• Software documentation changes (#379) (S. Leang of EP Analytics):
  • INTRO.DOC -> docs-intro.txt
  • INPUT.DOC -> docs-input.txt
  • TESTS.DOC -> docs-tests.txt
  • REFS.DOC -> docs-references.txt
  • PROG.DOC -> docs-prog.txt
  • IRON.DOC -> docs-hardware.txt
  • Attribution format has changed. Split list for individuals and their affiliated organization listed in alphabetical order. GAMESS banner has also slightly changed.

LIBCCHEM

• Link C++ bindings to LIBCCHEM builds using OpenMPI and removed -lcublas_device flag from linking. (#367) (K. Keipert of NVIDIA)
• LIBCCHEM build improvements (#368) (D. Poole of Iowa State University)
  • Remove -march=native flag in CMake build script
  • Stub additional routines in ga.src

September 30, 2019 R2 Public Release

GAMESS

• Multiconfiguration Pair-Density Functional Theory (MCPDFT). MCPDFT combines MCSCF and DFT approaches [1,2]. In MC-PDFT, an MCSCF wave function is used to calculate the kinetic energy, electron–nuclear attraction the Coulomb classical electron-electron repulsion contributions to the total energy; the remaining exchange−correlation energy comes from an on-top density functional. Various attempts have been made to combine MCSCF and DFT methodology to get an affordable yet accurate method to treat strongly correlated systems. Most of the previous methods encountered two problems: 1) double counting of electron correlation, as MCSCF contains some amount of dynamic correlation which is counted again if one simply adds a correlation energy computed from a correlation functional to the MCSCF energy, and 2) the incompatibility of available functionals developed for Kohn-Sham theory with the spin densities of a multiconfigurational wave function. An attempt has been made in MC-PDFT method to avoid these two problems. The first problem is avoided by using only the kinetic, electron-nuclear and Coulomb interaction terms of the MCSCF energy, with the remaining part of the energy calculated through on-top energy functional. The second problem is solved by introducing translated Kohn-Sham density functionals that depend on the on-top pair density and the total density in contrast to the alpha and the beta density dependence of KS-DFT functionals [1,3]. See PDFTYP in the $CONTRL group. (#182) (Andrew Sand and Prachi Sharma of University of Minnesota)
  • [1] Li Manni, G.; Carlson, R. K.; Luo, S.; Ma, D.; Olsen, J.; Truhlar, D. G.; and Gagliardi, L. J. Chem. Theory Comput. 2014, 10(9), 3669-3680.
  • [2] Gagliardi, L.; Truhlar, D. G.; Li Manni, G; Carlson, R. K.; Hoyer, C. E.; Bao, J. L. Acc. Chem. Res. 2016, 50(1), 66-73
  • [3] Carlson, R. K.; Truhlar, D. G.; Gagliardi, L. J. Chem. Theory Comput. 2015, 11( 9), 4077-4085.
• New keyword TRNSD in $TDDFT group that allows property computations on the transition density for the transition from the ground state to the state IROOT. This allows one - for example - to calculate PDCs from the electric potential of the transitions density as proposed by the TrESP method [1]. (#174) (C. Friedl of Johannes Kepler University Linz)
  • [1] Madjet, M. E.; Abdurahman, A.; Renger, J. Phys. Chem. B 2006, 110 (34), 17268–17281.

Along with a few minor fixes:

GAMESS

• Fix parallel FMO gradient issue (#202) (C. Bertoni of Argonne National Laboratory)
• Fix issue in shared-memory Fock matrix code (#198) (#204) (V. Mironov of Lomonosov Moscow State University)
• Fix SLB bug in FMO OMP (#197) (V. Mironov of Lomonosov Moscow State University)
• Fix floating point execption trap in EFMO for ./tests/efmo/exam03-serial (#195) (C. Bertoni of Argonne National Laboratory)
• Change TDTYP to TDDFT in $FMO group. (#180) (C. Bertoni of Argonne National Laboratory)

Software related changes:

GAMESS

• Updated Windows 64 support (#190) (S. Leang of EP Analytics)
• New GMS_DDI_COMM options: serial and serial-debug. Useful for serial calculations and/or debugging. (#189) (#174) (C. Friedl of Johannes Kepler University Linz)
• Aggregation of array limits into Fortran module mx_limits. (#178) (C. Bertoni of Argonne National Laboratory)

LIBCCHEM

• Improve IBM Power 64 (Summit) support in LIBCCHEM. (#172) (D. Poole of Iowa State University)
• Modularization of LIBCCHEM build process. (#169) (D. Poole of Iowa State University)

June 30, 2019 R1 Patch 1 Release

Along with a few minor fixes: GAMESS

• Fixed GAMESS release version output
• Fixed memory leak when using Intel 2017 & 2018 and running large FMO jobs (#173) (V. Mironov of Lomonosov Moscow State University)
• Fixed PGI 19.4 Community Edition compiler support with new DFT code (#170) (V. Mironov of Lomonosov Moscow State University)
• Fixed math library issue with efpmodule when compiling on IBM systems (Summit and Ascent) (#167) (D. Poole of Iowa State University)
• Fixed numerical discrepancies with GGA energies in new DFT code (#165) (V. Mironov of Lomonosov Moscow State University)

June 30, 2019 R1 Public Release

GAMESS

• Conversion of EFP related common blocks to modules (#155) (P. Xu and of T. Sattasathuchana of Iowa State University)
• PGI toolchain support (#152) (D. Poole of Iowa State University)
  • The -tp=px flag was added to all files when pgfortran is requested, to remove architecture-specific PGI-compiled binaries
  • DDI is now compiled with the PGI C compiler when pgfortran is requested
  • PGI LAPACK support was added to lked
• HF, DFT, FMO OpenMP Update (#151) (V. Mironov of Lomonosov Moscow State University)
  • Multithreaded FMO
  • Multithreaded one-electron integrals
  • Multithreaded PCM (partial)
  • DFT code improvements
  • DFT code improvements:
    • new driver for grid DFT XC integration with multithreading support;
    • support for SSF (by Stratmann, Scuseria and Frisch) molecular grid partitioning scheme (in addition to Becke's one);
    • various partitioning functions: Becke's, SSF, Erf1 (from NWChem), smoothstep;
    • use BLAS to update Kohn-Sham matrix;
    • option to ignore grid weight contribution to the DFT gradient;
    • two new types of radial quadrature: Mura-Knowles and Treutler-Ahlrichs;
    • few new spherical grids of various symmetry (octahedral with and without inversion, icosahedral) for tests;
    • disable grid switching when $DFT switch=0.0 instead of crashing;
    • cleanup and bug fixes in legacy DFT code, change layout of some arrays from C to Fortran style;
  • Various other improvements:
    • thread-safe ERIC integrals;
    • cleanup and bug fixes in older OpenMP code;
    • remove unnecessary matrix copying in eigen.src when using LAPACK;
• ARM Clang compiler support for linux32 and linux64 targets (#149) (C. Bertoni of Argonne National Laboratory)
• New DFT functionals: APF, HCTH, HCTH14, HCTH76, HCTH407P, HCTH407P, CAMQTP00, CAMQTP01, RM06X, RM06C, REVM06, RM06-LX, RM06-LC, REVM06-L. See $DFT in INPUT.DOC (#148) (I. S. Gerasimov of A.N.Nesmeyanov Institute of Organoelement Compounds of Russian Academy of Sciences)
• AMD software suite support: BLIS BLAS library, LibFLAME LAPACK library, AMD Optimizing C Compiler (AOCC), LibM mathematical functions library. (#143) (D. Poole of Iowa State University)
• Nudged Elastic Band (NEB) Methods. See $NEB in INPUT.DOC (#145) (N Sawada, M. Haruta, T. Asada, S. Koseki of Osaka Prefecture University, Japan)
• QM/EFP2 update (coulomb, exchange-repulsion and Dispersion terms) (#112) (P. Xu of Iowa State University)
• Improve CC file handling through 2GB file chunking. (#111) (J. L. G. Vallejo of Iowa State University) (J. Boatz of Air Force Research Laboratory)
• Enable EFMO/CCSD(T) calculations (#108) (C. Bertoni of Argonne National Laboratory)

LIBCCHEM

• PGI toolchain support (#152) (D. Poole of Iowa State University)
  • PGI compiler can now be used to compile LIBCCHEM
• Closed- and Open-shell RI-energy improvements for CPU and GPU. (#122) (L. Roskop of Iowa State University)
• Update LIBCCHEM to use C++11 standard (#121) (J. L. G. Vallejo of Iowa State University)
• CMake capabililities for LIBCCHEM (#110) (L. Roskop of Iowa State University)

GAMESS + LIBCCHEM

• Singularity support added to launch Singularity container builds of GAMESS/GAMESS+LIBCCHEM. Run ./config and select singularity as target. (S. Leang of EP Analytics)

Software related changes:

• Improve GAMESS build process. Users can now perform the following commands to build GAMESS after running ./config:
  • make ddi
  • make modules
  • make -j

Sept 30, 2018 R3 Public Release

New public release with:

GAMESS

• New general features

  • Analytic Raman intensities for DFTB (Y. Nishimoto).
  • Replica-exchange umbrella sampling (S. Ito).
  • Subgroups in GDDI (A. Gaenko, S. Pruitt)
  • Density of states (DOS).
  • Assign blocks of tasks for DLB (instead of individual tasks).
  • DFTB input simplified (can omit parameter file names).
  • CMIRS Version 1.1 (D. Chipman)
  • New version of QMCPACK-GAMESS interface that includes AFO-type extension of the QMC-EFMO method to fragmentations across covalent bonds
  • OpenMP threaded RI-MP2 gradients (B. Pham)
    • Set "setenv GMS_OPENMP true" in install.info
    • Set "GMS_OPENMP = true" in Makefile
    • Set $INTGRL INTOMP=1 $END in input file
    • Set $MP2 CODE=RIMP2 $END in input file
  • Improved harmonic restraints
    • Allow angle and dihedral restraint types
    • See IHMCON and SHMCON in INPUT documentation

• New FMO features

  • FMO-DFTB/AFO (Y. Nishimoto)
  • ESP-PC approximation for FMO Hessians (H. Nakata).
  • FMO/LCMO(X) and FMO3/F(X).
  • PIEDA/DFT and PIEDA/DFTB.
  • Fluctuation analysis for FMO/MD.
  • FMO/SMD solvation model.
  • Numerical gradient for FMO.
  • Any number of atoms in FMO/MD.
  • Solvent dipole contributions for FMO/PCM.
  • Direct matrix inversion in FMO/PCM.
  • Charge difference printed in FMO-DFTB showing convergence.

• Changes of option defaults

  • The defaults of DAMPEX and HUBDER for DFTB3 were changed for 3ob-3-1 parameters.
  • Umbrella sampling MD's option IUSTYP=1 was changed to agree with other programs; the old way is available with IUSTYP=-1.

LIBCCHEM

• New features (L. Roskop)
  • Kinetic energy and electron nuclear attraction derivatives added
  • RYSQ library: libint is no longer required for density-fitting-perturbation-theory (DF-PT) gradients
  • DF-ZAPT added (multi-GPU and CPU)
  • DF-MP2 now has multi-GPU support
  • closed- and open-shell DF-SCF added (multi-GPU and CPU)
  • New $LCCHEM subgroup added to for LIBCCHEM control

Along with a few minor fixes:

GAMESS

• For systems without electrons (Na+ etc), DFTB gradient had a bug.
• DFTB did not use the origin in computing dipoles.
• FMO1/PCM had a bug with IDISP=1.
• FMO/FDD/PCM did not work with OPTFMO.
• FMO/F did not work with PCM.
• In FMO/PCM[1(n)], virial components were not computed properly.
• Quadrupole and octupole moments had a parallel bug.
• FMO0 generated wrong F30 file.
• Stability analysis did not work in multilayer FMO.
• FMO/PCM gradient had a bug with ESP-PC.
• Charges were not reused in FMO-DFTB/MD.
• F40, stored in DDI memory, had a bug with Raman.
• In DFTB, d-orbitals resulted in NaN.
• FMO/FDD/PCM had a memory bug in the gradient.
• MD did not print incremental energy for JEVERY>10.
• FMO/SA did not work properly for correlated methods.
• SCS-MP2 energies were wrong for FMO-MP2/AB runs.
• RUNTYP=FMOHESS did not work with DDI-stored F40.
• DFTB Hessian did not work with DISP=SK(HP).
• In the Jan, 2017 changes, FMO gradient may have been affected.
• EOM/CCSD3A was not checked properly.
• FMO-MP2 erroneously aborted by checking consistency in $DATA.
• HW ECP did not work with FMO.
• Better memory management for ORMAS runs (J. Ivanic).

Software related changes:

• VB2000 is not longer compiled by default. To enable VB2000 modify compall and lked.

Aug 02, 2018 R2 Public Release

New public release with:

• New Michigan State University Coupled-Cluster Theory Methods: CCSD3A/CCT3
  • You must run config and answer "yes" to the CCSD3A/CCT3 build question.
  • CCTYP (works for both RHF and ROHF references)
  • CCSD3A
    • CC calculation with singles, doubles, and active-space triples, designated as CCSDt. When all orbitals in the MO basis are set to be active, CCSDt becomes full CCSDT. See $CCINP.
  • CCT3
    • In addition to CCSDt (see CCTYP=CCSD3A), compute the CC(t;3) noniterative energy correction due to the missing triples not captured by CCSDt.
  • See $GMSDIR/tests/cc/cc-ccsd3a-*.inp and $GMSDIR/tests/cc/cc-cct3-*.inp for example input files.
  • More information is provided in INPUTS.DOC
• New Minnesota Density Functionals: MN12/MN15
  • Available for ground-state energy and gradients.
  • Available for excited-state energy.
  • DFTTYP
  • MN12-L
  • MN12-SX
  • MN15
  • MN15-L

Along with a few minor fixes:

• Chunking I/O was added to locpol.src to address large DAF files during MAKEFP runs. (#61)
• Minimize OMP warnings (#67)
• Fixed DGEMM handing in efpaul.src

Software related changes:

• Added doxygen config file for developers (requires doxygen v.1.8.15+).
  • cd $GMS_DIR
  • doxygen Doxyfile
• Cavium ThunderX2 (ARM64) support through aarch64 conditionals during compilation.
  • Uses 'linux64' target. Conditionals within build scripts will pick up on prescence of aarch64 architecture.
• Running 'make' after successfully creaing an 'install.info' will now create the 'object' directory if it does not already exist.
• Added gfortran 5.5 support in config.
• Changed the maximum number of SMP procs (affects 'sockets' builds of GAMESS)
• Build alternative to 'config' or 'compall' is provided in $GMS_DIR/bin/create-install-info.py (requires Python 3 and jinja2 module).
  • cd $GMS_DIR
  • bin/create-install.info.py --help
  • Simplest build example: linux64 target, gfortran 4.8.5, use built in GAMESS BLAS library, use sockets for DDI communication:
    • bin/create-install-info.py --fortran_version=4.8.5
    • make ddi
    • make -j nproc
• Alternative to $GMSDIR/rungms provided in $GMSDIR/machines/xeon_phi/rungms.interactive
  • Copy into your $GMS_DIR
  • Edit 'target' in rungms.interactive
  • Use rungms.interactive as you would rungms
• Improved BLAS handling in LIBCCHEM

Feb 14, 2018 R1 Public Release

New public release with:

• This adds code for the analytical gradient for EFMO.
  • It is turned on with the IEFMOG flag in $FMO in the input file (see INPUT.DOC).
  • Described in C. Bertoni and M. S. Gordon, "Analytic Gradients for the Effective Molecular Orbital Method" Journal of Chemical Theory and Computation 12 (10), 4743-4767 (2016).
• New target cray-xc added.
  • GMS_PHI flag in install.info changed from true|false to none|knl|knc

Along with a few minor fixes:

• Modifications to comp, lked and config (#51)(#50):
  • Removed flags EFPOPENMP, MAKEFPOPENMP and RIMP2_OPENMP from install.info
  • Threaded codes are now controlled by the flag GMS_OPENMP in install.info

Nov 11, 2017 R3 Public Release

• OpenMP two-electron code (#28)
  • To build the OpenMP threaded two-electron code set GMS_OPENMP flag to true in install.info and Makefile after running config.
  • See the INTOMP flag within the $INTGRL group for more information.
• OpenMP HF algorithm with Fock matrix shared among threads (#28)
  • See the SHFOCK flag within the $INTGRL group for more information.

• See repository https://github.com/gms-bbg/gamess-papers/tree/mv/SC17 for more information on the above changes

• GAMESS Release Notes:

  • https://gamess.gitbooks.io/gamess-release-notes/content

Oct 7, 2017 Hot Fix (A)

Along with a few minor fixes:

• Address issue with using 'compall' for building GAMESS (#39)

Sept 30, 2017 R2 Public Release

New public release with:

• OpenMP threaded RI-MP2 energy (#31)
• OpenMP threaded EFP Charge Transfer (#2)
• OpenMP threaded MAKEFP (#35)
• QMC-EFMO interface between QMCPACK and GAMESS (#36)

Along with a few minor fixes:

• Address issue with SP shells for 2- and 3- center ERI in LIBCCHEM (#29)
• Address issue with multi-pole moment calculations when running in parallel (#25)
• Address issue with using 'compall' for building GAMESS (#39)

Platform updates:

• Added support for GNU compilers up to version 7.2 (Linux, OSX)