Apr 20, 2017 R1 Public Release

New public release with:

  • RI-MP2 energy and gradients in LIBCCHEM
    • RYSQ Integral (CPU)
      • 2-, 3-center ERIs
      • 2-, 3-center DERIs
      • overlap DERIs (Hermite polynomials)
    • RYSQ bug fixes
      • roots have been modified to match GAMESS
      • shell.py prints correct normalizaton constants
      • shell.py modified for 2- and 3-center integrals
    • rimp2/rizapt energies (CPU/GPU)
    • rimp2 gradients (CPU/GPU)
      • need libint version 2 (C++11 API) for complete result
      • Libint is used to generate kinetic and nuclear attraction derivatives
    • New input group $CCHEMRI (see INPUT.DOC)
    • See $GMS_DIR/libcchem/tests/rimp2Tests for input examples
  • per-fragment-type EFP charge-transfer (CT) cutoff ($EFPARM)
    • Per-fragment charge-transfer cutoff input is coded.
    • Added charge-transfer cutoff on per-fragment-pair basis.
    • New input group $EFPARM (see INPUT.DOC)
    • See $GMS_DIR/tests/solvent/efp-ctcut for input examples

Along with a few minor fixes:

  • address race condition causing scf convergence issues in libcchem (9f7ed0c)
    • Address convergence issues in SCF which appeared when using large basis-sets
  • altered large memory transfer defaults when using Intel MPI (833acc6)
    • Set I_MPI_SHM_LMT=shm in rungms as default for Intel MPI
  • catch cases where user request MULT=0 or less (6eb5f2b)
    • Added extra logic to prevent runs from proceeding if user requests MULT=0 or MULT<0
  • removed extraneous REWIND (07da449)
    • Removed extraneous rewind in efinp.src preventing parallel jobs when using GNU compilers and/or non-Intel MPI

results matching ""

    No results matching ""