Firefly and PC GAMESS Version History

8.2.0, build # 10203. Official Firefly binaries released to the public September 19, 2016.

• Available from the Downloads section of the Firefly homepage.

Main new features are:

• Changes for generators-direct (X)MCQDPT2 code for large active spaces.
• Changes for RI-MCQDPT2 and RI-XMCQDPT2 energies for large systems.
• Changes for RI-MP2 energies for RHF, ROHF and UHF references and for double hybrid DFT.
• Changes to allow more parallelism in (X)MCQDPT2 code.
• Changes for faster CASSCF integral transformation for large and very large systems.
• Improvements in Firefly's P2P protocol. P2P libraries are updated to version 2.4.
• Fastdiag libraries are updated to version 3.0 to use newer and faster code.
• Changes to use newer, more precise value of Bohr radius, namely 0.52917721067 Angstrom, throughout all the code. Previous versions of Firefly used older, less precise value of 0.52917724924 Angstrom.
  Note: this transition causes minor differences in computed energies and other properties as compared to older Firefly versions when initial geometry is specified using Angstroms.
• Various minor improvements and bugfixes.

8.1.1, build # 9295. Official Firefly binaries released to the public September 3, 2015.

• Available from the Downloads section of the Firefly homepage at MSU.

This version of Firefly is primary a bugfix release to version 8.1.0. The list of changes is as follows:

• Multiple minor bugfixes.
• Changes to increase overall precision of large QDPT jobs by dramatically reducing accumulated round-off errors.
• Changes to implement DFT-NL.
• Preliminary changes for AVX512.
• Changes for better multi-core performance.

8.1.0, build # 9035. Official Firefly binaries released to the public August 24, 2014.

• Available from the Downloads section of the Firefly homepage at MSU.

This version of Firefly includes implementation of analytical TDHF/TDDFT gradients. Gradients for singlet states are available for all DFT functionals for which TDDFT energy is presently programmed. Frozen core and TDA approximations can be used together with gradients.

Other changes

• Performance-related improvements to MCQDPT and XMCQDPT2 code.
• Multiple minor bugfixes and improvements

8.0.1, build # 8540. Official Firefly binaries released to the public January 1, 2014.

• Available from the Downloads section of the Firefly homepage at MSU. This is primary the bugfix release to Firefly v. 8.0.0. Nonetheless, there are some new features:

Changes to MCQDPT and XMCQDPT2 code

• Changes to allow MCQDPT and XMCQDPT for larger systems (with up to 3000-4000 atomic orbitals and more).
• Changes to allow MCQDPT and XMCQDPT for larger active spaces (with up to 10 millions and more of CSFs in active space after or without CSF selection/prescreening).

Other changes

• Multiple minor bugfixes and improvements

8.0.0, build # 8240. Official Firefly binaries released to the public September 5, 2013.

• Available from the Downloads section of the Firefly homepage at MSU. Main new features are:

System-wide changes

• New hierarchical eXtreme Parallel (xp) parallel execution model
• Improved performance of threaded code with all presently known issues fixed
• Improved memory management which is capable to allocate more memory and to handle it by a more intelligent manner
• Improved support of Lustre and other non-local filesystems with all presently known compatibility issues resolved
• Support of Intel's AVX and AVX2-enabled processors
• Support of AMD Bulldozer processors and FMA4
• New high-performance intelligent numerical gradients mode
• New, updated documentation for Firefly. This documentation combines information taken from the old manual, the various README files, and the forums, and also many new chapters.

Changes to HF and DFT code

• Improved DIIS code with lots of additional modes of operation, configurable using dedicated input options
• Better parallel scalability of DFT code due to faster DIIS
• Improved accuracy of DFT quadratures
• DFT-D empirical dispersion correction
• New GGA and hybrid GGA functionals
• Support of double-hybrid functionals
• New HFX parameter allowing users to modify the fraction of the exact HF exchange used in hybrid functionals. Helpful in TDDFT calculations of Rydberg and charge-transfer states.

Changes to MP2 code

• SCS and SOS MP2
• More intelligent handling of I/O errors while running in parallel

Changes to CUDA-enabled MP4 code

• Changes for CUDA 4.x and 5.x and performance improvements

Changes to CASSCF, MCQDPT and XMCQDPT2 code

• New spin-orbit coupling code
• New high-performance determinant-based CASCI/CASSCF code with better parallel scalability and efficiency, and very modest memory demands
• Multiple functional and performance improvements and extensions to MCQDPT and XMCQDPT module
• Improved conical intersection location code now capable to deal with arbitrary averaging of multiple states in SA-CASSCF

Other changes

• CPCM solvent model
• Fully variational DPCM and CPCM-like extended solvent models
• PCM is enabled for UHF/UDFT and ROHF/RODFT
• Extended restart capabilities
• Improvements and extensions to surface scan module
• Extended limits on maximum number of atoms, shells, and basis functions
• Multiple bugfixes and minor improvements

7.1.G (i.e., 7.1.16), build # 5618. Official Firefly binaries released to the public December 4, 2009.

• Available from the Downloads section of the Firefly homepage at MSU. Main new features are:

• Multiple minor bugfixes and improvements.
• More agressive set of default settings for better performance and stability.
• Improved performance on AMD Phenom/Phenom II/Barcelona/Shanghai/Istanbul processors.
• Improved performance of MP2 Energy/Energy Gradient method=1 code.
• Improved DLC engine.
• Improved code for relaxed PES scans using DLCs.
• Improvements to P2P communication interface.
• Completely redesigned threading model for better multithreaded performance and CUDA interoperability.
• Improved MP4 code with CUDA support.
• Program name changed to Firefly.

7.1.F (i.e., 7.1.15), build # 5211. Official PC GAMESS/Firefly binaries released to the public February 24, 2009.

• Available from the Downloads section of the PC GAMESS/Firefly homepage at MSU. This is mainly the maintenance/bugfix release over PC GAMESS/Firefly version 7.1.E. Main new features and bugfixes are:

• Support of Mac OS X/Intel platform (press to visit our Mac OS X pages).
• Better multithreaded performance on some platforms/hardware.
• Workaround for nasty CPUID bug (press for more information on problem).
• New fully dynamically linked MPICH/NPTL binaries for Linux.
• Faster GVB/MCSCF gradients.
• State-Specific gradient code for State-Averaged MCSCF allows use of PCM solvation model for excited states optimization and location of Conical Intersections.

7.1.E (i.e., 7.1.14), build # 5190. Official PC GAMESS/Firefly binaries released to the public January 11, 2009.

• Available from the Downloads section of the PC GAMESS/Firefly homepage at MSU. This is mainly the maintenance/bugfix release over PC GAMESS/Firefly version 7.1.C. However, it adds some new important features:

• Completely new DLC engine and improved default geometry optimizer.
• Better support of Core i7 (as well as other Nehalem core processors).
• More efficient I/O under Windows Vista/Windows Server 2008 R1.
• Suport of httfix option under Linux.
• New mpich2-linked binaries for Linux.
• Fixes for compatibility with the most recent Linux distributions.

7.1.C (i.e., 7.1.12), build # 5014. Official PC GAMESS/Firefly binaries released to the public October 17, 2008.

• Available from the Downloads section of the PC GAMESS/Firefly homepage at MSU. Accumulates multiple bugfixes, performance improvements, and new functionality introduced since the release of PC GAMESS v. 7.1.5:

• Massive internal engine rewrites.
• Extended P2P interface with support of up to 1024 nodes.
• New eXtended MultiConfiguration QuasiDegenerate Perturbation Theory (XMCQDPT) code (see the related presentation in "Archives" section for details and "Manuals" section for the list of new available options).
• Improved CPHF=AO code for analytical RHF second derivatives with reduced memory requirements and better parallel scalability. Note it is now turned on by default!
• Improved, much more stable and faster SOSCF code available for RHF/ROHF/UHF/GVB/MCSCF wavefunctions optimization (list of new available options is to be published soon in "Manuals" section).
• Improved METHOD=GDIIS geometry optimizations.
• Updated NBO (version 5.G) module.
• Relaxed 1D and 2D PES scans over bonds, angles, or torsions in DLCs (list of new available options is to be published soon in "Manuals" section).
• Added RM1 parametrization to MOPAC code.
• Improved stability of the ALDET's Davidson Ci diagonalization code.
• Improved state-averaged MCSCF gradients code (list of new available options is to be published soon in "Manuals" section).
• Improved MCSCF state-tracking feature (list of new available options is to be published soon in "Manuals" section).
• New code for location of conical intersections (CI) (list of new available options is to be published soon in "Manuals" section).
• Improved DIIS code with reduced memory demands.
• More accurate memory demands estimation for CIS/TD code.
• New, much more stable code to convert internals to cartesians.
• DFT gradients are now slightly more accurate.
• Support of Gaussian-style O3LYP functional.
• Support of Intel Atom processors.
• Initial optimization for Intel Core i7 (codename Nehalem) microarchitecture.
• Improved compatibility with some buggy Linux versions.
• Linux/MPICH version of PC GAMESS now uses ssh by default.

7.1.6 - 7.1.11 (i.e., 7.1.B), internal builds.

7.1.5, build # 4630, December 23, 2007.

• Available as the set of the update patches to the original PC GAMESS version 7.1 build # 4471 from the Downloads section of the PC GAMESS homepage at MSU. Incorporates multiple bugfixes and performance improvements (especially for AMD Phenom/Barcelona and Intel Core 2 processors), as well as some new functionality.

7.1.4, November 2007, internal build.

7.1.3, October 2007, internal build.

7.1.2, October 2007, internal build.

7.1.1, September 2007, internal build.

7.1, Official PC GAMESS binaries released to the public, build # 4471, September 5, 2007.

• This version accumulates all the changes introduced in intermediate releases versions 7.0.2-7.0.7, as well as state-specific analytic gradients for state-averaged MCSCF and completely new state of the art serial and parallel semidirect MP2 gradient program with excellent performance, scalability and very modest memory demands.

7.1 final RC, build # 4400, June 27, 2007.

• The beta test program of the PC GAMESS v. 7.1 final RC is now opened. New PC GAMESS version features completely new state of the art parallel semidirect MP2 gradient program with excellent performance, scalability and very modest memory demands. The updated PC GAMESS binaries are available from MSU for registered PC GAMESS users for testing purposes upon request.

7.0.7, build # 4194, April 23, 2007.

• Dynamically linked Linux binaries of PC GAMESS are now fully nptl compatible and were carefully tested with MPICH, MPICH-GM, MPICH-MX, OpenMPI, Scali MPI, HP-MPI, Intel MPI 3.0 and Infinipath MPI implementations. The updated PC GAMESS binaries are available from MSU for registered PC GAMESS users for testing purposes upon request.

7.0.6, April 2007, internal build.

7.0.5, March 2007, internal build.

7.0.4, build #4102, February 16, 2007, internal build.

• More numerically stable PCM energy and gradient code
• Serious improvements of parallel scalability of ALDET CASCI and CASSCF code The updated PC GAMESS binaries are available from MSU for registered PC GAMESS users for testing purposes upon request.

7.0.3, build #4063, January 21, 2007, internal build.

• Shared memory version of P2P interface (Windows)
• support of OPTX, OLYP, and O3LYP xc functionals
• Parallel PCM with enlarged PCM dimensions The updated PC GAMESS binaries are available from MSU for registered PC GAMESS users for testing purposes upon request.

7.0.2, build #4020, October 8, 2006, internal build.

• Added completely new, much faster MCQDPT2 code developed at MSU. The updated PC GAMESS binaries are available from MSU for registered PC GAMESS users for testing purposes upon request.

7.0.1, build #3970, August 17, 2006

• Better optimization for Intel Core 2 (Woodcrest/Conroe/Merom processors) microarchitecture
• Minor bugfixes and improvements

7.0, build #3910, May 11, 2006

• Initial optimization for Intel Core 2 (Woodcrest/Conroe/Merom processors) microarchitecture
• Time-dependent Hartree-Fock (TDHF) and density functional theory (TDDFT)
• Configuration Interaction Singles (CIS)
• Static and dynamic hyperpolarizabilities via TDDFT
• Large-scale direct and conventional parallel MCSCF
• Semidirect MCQDPT
• Support of general contraction basis sets
• faster PCM
• Improved Linux compatibility
• Unified support of different MPI implementations under Windows

6.5, Release Candidate # 9, January 2005, was available to beta testers only

• Support for spherical basis functions
• Quantum Fast Multipole Method and Linear Exchange for linear scaling HF and DFT
• Faster direct SCF/DFT code
• New 2-e integrals code
• Much faster MCQDPT code with SMP support

6.4, March 2004

• Fast hybrid DFT
• Support of AMD Opteron and Intel Pentium M processors
• Improved performance and stability
• Dynamic load balancing on the top of P2P interface
• Improvements to P2P parallel MP2 energy code
• Better support of SMP including HTT environment
• Several other enhancements and extensions (cube facility ...)

6.3, June 2003

• Optimized binaries for AMD Athlon and Intel Pentium 4 processors
• New parallel mode P2P communication interface
• Efficient parallel MP2 energy code based on P2P model allowing thousands of basis functions
• Various bugfixes and enhancements

6.2, 2001

• Optimized binaries for AMD Athlon
• Cumulative bugfixes and improvements

6.1, 2001, Internal release

6.0, May 2000

• PC GAMESS is now based on updated GAMESS (US) sources
• Pentium/Pentium Pro/Pentium II/Pentium III-optimized native Linux SMP-aware PC GAMESS ELF binaries
• Pentium Pro/Pentium II/Pentium III-optimized OS/2-based SMP & JFS-aware binaries
• Large files are supported via transparent file splitting on all 32-bit file systems (2 or 4 GB limit)
• Enhanced convergence options for solving MP2 CPHF equations
• PG GAMESS can now write out cube files for visualization
• Advanced surface scan interface
• NBO module is incorporated into PC GAMESS
• Support of direct MP3/MP4
• gdiis method for geometry optimization

5.4, 1999

• Changes for faster MP3/MP4
• Changes to better support SMP throughout the code
• OS and CPU autodetection

5.3, 1999, Internal release

• Changes to use updated MKL libraries
• Changes to support parallel execution

5.2, May 1999

• New RHF MP4(SDTQ) (i.e., full MP4) energy module is added
• The SMP scaling properties of MP3/MP4 calculations are improved considerably
• The speed of MP2 gradient calculations for non-abelian symmetry groups is increased
• The speed of Conjugated Gradient solver (used mainly during calculations of analytical second derivatives) is improved significantly
• The Finite Field module is changed to avoid reevaluations of 2-e integrals if possible
• The ECP integrals module is rewritten to avoid numerical instability problems
• It is now possible to run the PC GAMESS under Linux using customized Wine

5.1, Build #1519,October 1998

• RHF MP3/MP4(SDQ) energy modules are added.
• Memory management is improved again (Win32 and OS/2-specific).
• The CPHF solver used during MP2-level geometry optimization/hessian runs is now faster.
• GAMESS bugs concerning PCM-level geometry optimization/hessian runs are fixed.
• The precision of the HF density matrix generated by SOSCF converger is enhanced

5.0, August 1998

• Based on May 8, 1998 GAMESS code
• Faster TDHF, MCQDPT, MP2 energy, and MP2 energy gradient calculations
• The disk usage is reduced for non-FORS GUGA CI jobs
• Improved support of SMP (Windows NT specific)

4.5, Internal Release

4.4, March 1998

• The completely new RHF/ROHF/UHF MP2 energy program
• Additional GUGA CI Hamiltonian packing
• Improved memory management (Windows NT specific)

4.3, October 1997

• Updated MP2, MP2 gradient, and CI gradient codes
• Large direct access files (> 2 GB) are supported under Windows NT
• Ctrl-C and Ctrl-Break signal handling is implemented

4.2, Internal Release

4.1, Build number 1220

• The fast (non-Fortran) file I/O as well as the AO integrals and GUGA CI Hamiltonian packing are implemented
• Large sequential access files (> 2 GB) are supported under Windows NT

4.0, Build number 1080

• First public release based on the original GAMESS sources dated March 18, 1997

Versions 1.0 - 3.0 were used locally at MSU.