Documentation

Files

Filenames used by GronOR are derived from user-defined identifiers root and calc, such that all integral

and wavefunction files generated using the OpenMolcas program only use the root identifier in the

filename, and individual GronOR calculations use root_calc filenames. This allows different GronOR

calculations differentiated by calc to be carried out with the same OpenMolcas generated files.

In the following table boldface typed files are required in any GronOR calculation.

File Name Description Origin root.one One-electron integral file OpenMolcas & rdtraint root_nnn.two Two-electron file(s) OpenMolcas & rdtraint root.sys Molecular definition file OpenMolcas & rdtraint root_nnn.vec MO vector files(s) OpenMolcas root_nnn.det Determinants file(s) OpenMolcas root_calc.inp Input file for GronOR Editor root_calc.out Output file GronOR root_calc.arx Archive file GronOR root_calc.cpr Checkpoint Restart file GronOR root_calc.cml CML-formatted output file GronOR root_calc.day Dayfile GronOR root_calc.pro Progress file GronOR root_calc.rnk Rank assignment file GronOR root_calc.tst Test file for automated testing GronOR root_calc.xyz XYZ-formatted coordinates GronOR root.log Log file GronOR gronor_nnn.dbg Debug file(s) GronOR The type of file is identified by its extension, as follows: inp: Input for GronOR. out: Output from GronOR. one: One electron integrals provided by OpenMolcas. two: Two-electron integrals provided by OpenMolcas. These intrgrals can be provided over multiple files as a way to reduce repeat file transfers over unreliable connections sym: Symmetry information provided by OpenMolcas. xyz: XYZ-formatted coordinates vec: Vector files provided by OpenMolcas det: Determinant files provided by OpenMolcas arx: Archive file that can be read by the Java tool cml: CML-formatted file for uploading to ioChem-BD. cpr: Checkpoint Restart file. If this file is present it indicates that a previous run aborted on an error or time limit. Restarting the same job will cause this file to be read and the calculation will continue from the last result written to the cpr file. If a calculation finishes correctly, there will not be a cpr file. day: The day file will have periodic entries added as a calculation proceeds. This file can be monitored to see if a job continues to make progress. log: Summary file with results from all jobs run in the current directory that used the same root files. rnk: The rank file will contain information about the ranks used and their assignments and can be used to determine if the job parameters were correctly specified. tst: Test data for automated testing used by developers. pro: A progress file used by developers. dbg: Debug files, generated one per rank, used by developers. Input (Key-Value pair combinations on file root_calc.inp) Keyword Description Format Default Required MEBFs Number of many electron basis functions, integer Yes followed by integer fragment identifiers one line for each fragment Size Number of ranks per worker group integer 1 No Spin Spin multiplicity integer 1 No Task Number of ME contributions per task integer 32 32 No for accelerated and non-accelerated ranks Batch Number of ME contributions calculated in a integer 1 32 No single pass through integrals for accelerated and non-accelerated ranks Threshold Threshold expansion coefficients real 1.0d-5 No Acceleration Number of accelerated ranks per node integer 1 No Threads Number OpenMP threads, non-accelerated ranks integer 1 No Solver Library solver options integer 1 0 1 0 No Acc CUSOLVER SVD 0: no, 1: QR, 2: Jacobi Acc CUSOLVER EVD 0: no, 1: QR, 2: Jacobi Non-acc SVD 0: no, 1: MKL or other Non-acc EVD 0: no, 1: MKL or other Tolerance Tolerance CUSOLVER Jacobi solvers real 1.0e-07 No Values for SVD and EVD Sweep Number Jacobi sweeps integer 15 15 No Values for SVD and EVD Ecorr Dynamic correlation energy correction per state real 0.0d0 No GNWeights Gallup-Norbeck weights integer 0 No 0: off 1: on Print Print option string medium No low: minimal input data and final resuls medium: default input data and final results high: debug: Dayfile Percent complete written to separate file integer 10 No MEBFs: This keyword is the only required keyword. It specifies the number of many-electron basis functions to be constructed from the state wavefunctions specified on the following cards. Each card is a list of states for a single molecule or fragment. The use should give as many of these cards as the number of molecules or fragments to be included. GronOR will automatically determine the number of molecules or fragments from the number of cards provided. Spin: The keyword specifies the total spin state of the combined many-electron basis functions. The default is the singlet state. Therefore, Spin only needs to be specified for higher than singlet spin states. Size: In cases where the memory require to store the two-electron integrals exceeds the available memory on the node, or on an accelerator, the integrals can be divided over multiple ranks such that multiple nodes or multiple accelerators are used for a single copy of the two-electron integrals. The number of ranks to be used in this way is specified by the Size input variable. Acceleration: By default all ranks with access to an accelerator will be assigned to use the accelerator. In cases where the users wants to use accelerated and non-accelerated ranks on a node with a device, the keyword Accelerator specifies the number of ranks on node that will use the accelerator leaving the remaining ranks on that node to use the CPU only. Threads: If GronOR was compiled with the -DOPENMP=ON flag, the keyword Threads can be used to specify the number of OpenMP threads that a rank will use. By default, non-accelerated ranks, even when compiled with for OpenMP threading, will use a single thread. When Threads is not specified, but the environment variable NUM_OMP_THREADS is set, GronOR will use the value specified by the enviroment variable. If Threads is specified, it will override any value of the NUM_OMP_THREADS enviroment variable. Solvers: GronOR uses the most efficient available solvers depending on the chemical system size. In practice it should not be necessary to specify Solvers explicitly. Task: This is the number of matrix element contributions that will be calculated in a single task from the master to worker ranks. Increasing Task will result in fewer communication operations. The default value of 32 appears optimal for most siutations. Batch: This is the number of matrix element contributions calculated with a singke sweep through the two-electron integrals. In practice, this option is only resulting in reduced wall clock time on CPU-only ranks. The valkue of Batch can not exceed the value of Task. The default of 32 in GronOR for non-accelerated ranks appears optimal in most cases. Input for developers (Key-Value pair combinations on root_calc.inp) Keyword Description Format Default Required Timings Wall clock timings analysis integer 0 No Progress Communication events to progress file integer 0 No 0: no progress file 1: matrix elements only 2: all communication events 3: as 1 but accumulated and file saved 4: as 2 but accumulated and file saved Test Flag to write test file integer 0 No Debug Debug output level (one file per rank) integer 0 No MPIbuffer Size two-electron MPI buffer integer 168435456 No

Expert

Option for expert options

integer

0

No

Development Use of development code integer 0 No