Back to the main page of this manual | Input reference of CP2K version 2023.1 |
CP2K_INPUT /
GLOBAL
ALLTOALL_SGL {Logical} |
|
All-to-all communication (FFT) should use single precision | |
This keyword cannot be repeated and it expects precisely one logical. | |
The lone keyword behaves as a switch to
.TRUE. |
|
Default value:
.FALSE. |
BLACS_GRID {Keyword} |
|
how to distribute the processors on the 2d grid needed by BLACS (and thus SCALAPACK) | |
This keyword cannot be repeated and it expects precisely one keyword. | |
Default value:
SQUARE |
|
List of valid keywords:
|
BLACS_REPEATABLE {Logical} |
|
Use a topology for BLACS collectives that is guaranteed to be repeatable on homogeneous architectures | |
This keyword cannot be repeated and it expects precisely one logical. | |
The lone keyword behaves as a switch to
.TRUE. |
|
Default value:
.FALSE. |
CALLGRAPH {Keyword} |
|
At the end of the run write a callgraph to file, which contains detailed timing informations. This callgraph can be viewed e.g. with the open-source program kcachegrind. | |
This keyword cannot be repeated and it expects precisely one keyword. | |
Default value:
NONE |
|
List of valid keywords:
|
CALLGRAPH_FILE_NAME {String} |
|
Name of the callgraph file, which is written at the end of the run. If not specified the project name will be used as filename. | |
This keyword cannot be repeated and it expects precisely one string. |
ECHO_ALL_HOSTS {Logical} |
|
Echo a list of hostname and pid for all MPI processes. | |
This keyword cannot be repeated and it expects precisely one logical. | |
The lone keyword behaves as a switch to
.TRUE. |
|
Default value:
.FALSE. |
ECHO_INPUT {Logical} |
|
If the input should be echoed to the output with all the defaults made explicit | |
This keyword cannot be repeated and it expects precisely one logical. | |
The lone keyword behaves as a switch to
.TRUE. |
|
Default value:
.FALSE. |
ELPA_KERNEL {Keyword} |
|
Specifies the kernel to be used when ELPA is in use | |
This keyword cannot be repeated and it expects precisely one keyword. | |
Default value:
AUTO |
|
List of valid keywords:
|
ELPA_NEIGVEC_MIN {Integer} |
|
Minimum number of eigenvectors for the use of the eigensolver from the ELPA library. The eigensolver from the ScaLAPACK library is used as fallback for all smaller cases | |
This keyword cannot be repeated and it expects precisely one integer. | |
Default value:
16 |
ELPA_QR {Logical} |
|
For ELPA, enable a blocked QR step when reducing the input matrix to banded form in preparation for the actual diagonalization step. See implementation paper for more details. Requires ELPA version 201505 or newer, automatically deactivated otherwise. If true, QR is activated only when the the size of the diagonalized matrix is suitable. Print key PRINT_ELPA is useful in determining which matrices are suitable for QR. Might accelerate the diagonalization of suitable matrices. | |
This keyword cannot be repeated and it expects precisely one logical. | |
The lone keyword behaves as a switch to
.TRUE. |
|
Default value:
.FALSE. |
ELPA_QR_UNSAFE {Logical} |
|
For ELPA, disable block size limitations when used together with ELPA_QR. Keyword relevant only with ELPA versions 201605 or newer. Use keyword with caution, as it might result in wrong eigenvalues with some matrix orders/block sizes when the number of MPI processes is varied. If the print key PRINT_ELPA is active the validity of the eigenvalues is checked against values calculated without ELPA QR. | |
This keyword cannot be repeated and it expects precisely one logical. | |
The lone keyword behaves as a switch to
.TRUE. |
|
Default value:
.FALSE. |
ENABLE_MPI_IO {Logical} |
|
Enable MPI parallelization for all supported I/O routines Currently, only cube file writer/reader routines use MPI I/O. Disabling this flag might speed up calculations dominated by I/O. | |
This keyword cannot be repeated and it expects precisely one logical. | |
The lone keyword behaves as a switch to
.TRUE. |
|
Default value:
.TRUE. |
EPS_CHECK_DIAG {Real} |
|
Check that the orthonormality of the eigenvectors after a diagonalization fulfills the specified numerical accuracy. A negative threshold value disables the check. | |
This keyword cannot be repeated and it expects precisely one real. | |
Default value:
-1.00000000E+000 |
EXTENDED_FFT_LENGTHS {Logical} |
|
Use fft library specific values for the allows number of points in FFTs. The default is to use the internal FFT lengths. For external fft libraries this may create an error at the external library level, because the length provided by cp2k is not supported by the external library. In this case switch on this keyword to obtain, with certain fft libraries, lengths matching the external fft library lengths, or larger allowed grids, or grids that more precisely match a given cutoff. IMPORTANT NOTE: in this case, the actual grids used in CP2K depends on the FFT library. A change of FFT library must therefore be considered equivalent to a change of basis, which implies a change of total energy. | |
This keyword cannot be repeated and it expects precisely one logical. | |
The lone keyword behaves as a switch to
.TRUE. |
|
Default value:
.FALSE. |
FFTW_PLAN_TYPE {Keyword} |
|
FFTW can have improved performance if it is allowed to plan with explicit measurements which strategy is best for a given FFT. While a plan based on measurements is generally faster, differences in machine load will lead to different plans for the same input file, and thus numerics for the FFTs will be slightly different from run to run.PATIENT planning is recommended for long ab initio MD runs. | |
This keyword cannot be repeated and it expects precisely one keyword. | |
Default value:
ESTIMATE |
|
List of valid keywords:
|
|
This keyword cites the following reference: [Frigo2005] |
FFTW_WISDOM_FILE_NAME {String} |
|
The name of the file that contains wisdom (pre-planned FFTs) for use with FFTW3. Using wisdom can significantly speed up the FFTs (see the FFTW homepage for details). Note that wisdom is not transferable between different computer (architectures). Wisdom can be generated using the fftw-wisdom tool that is part of the fftw installation. cp2k/tools/cp2k-wisdom is a script that contains some additional info, and can help to generate a useful default for /etc/fftw/wisdom or particular values for a given simulation. | |
This keyword cannot be repeated and it expects precisely one string. | |
Default value:
/etc/fftw/wisdom |
FFT_POOL_SCRATCH_LIMIT {Integer} |
|
Limits the memory usage of the FFT scratch pool, potentially reducing efficiency a bit | |
This keyword cannot be repeated and it expects precisely one integer. | |
Default value:
15 |
FLUSH_SHOULD_FLUSH {Logical} |
|
Flush output regularly, enabling this option might degrade performance significantly on certain machines. | |
This keyword cannot be repeated and it expects precisely one logical. | |
The lone keyword behaves as a switch to
.TRUE. |
|
Default value:
.TRUE. |
OUTPUT_FILE_NAME {String} |
|
Name of the output file. Relevant only if automatically started (through farming for example). If empty uses the project name as basis for it. | |
This keyword cannot be repeated and it expects precisely one string. |
PREFERRED_DGEMM_LIBRARY {Keyword} |
|
Specifies the DGEMM library to be used. If not available, the BLAS routine is used. This keyword affects some DGEMM calls in the WFC code and turns on their acceleration with SpLA. This keyword affects only local DGEMM calls, not the calls to PDGEMM (see keyword FM%TYPE_OF_MATRIX_MULTIPLICATION). | |
This keyword cannot be repeated and it expects precisely one keyword. | |
Default value:
BLAS |
|
List of valid keywords:
|
PREFERRED_DIAG_LIBRARY {Keyword} |
|
Specifies the diagonalization library to be used. If not available, the ScaLAPACK library is used | |
This keyword cannot be repeated and it expects precisely one keyword. | |
Default value:
ELPA |
|
List of valid keywords:
|
|
This keyword cites the following reference: [Marek2014] |
PREFERRED_FFT_LIBRARY {Keyword} |
|
Specifies the FFT library which should be preferred. If it is not available, use FFTW3 if this is linked in, if FFTW3 is not available use FFTSG. Improved performance with FFTW3 can be obtained specifying a proper value for FFTW_PLAN_TYPE. Contrary to earlier CP2K versions, all libraries will result in the same grids, i.e. the subset of grids which all FFT libraries can transform. See EXTENDED_FFT_LENGTHS if larger FFTs or grids that more precisely match a given cutoff are needed, or older results need to be reproduced. FFTW3 is often (close to) optimal, and well tested with CP2K. | |
This keyword cannot be repeated and it expects precisely one keyword. | |
Default value:
FFTW3 |
|
List of valid keywords:
|
|
This keyword cites the following reference: [Frigo2005] |
PRINT_LEVEL {Keyword} |
|
How much output is written out. | |
This keyword cannot be repeated and it expects precisely one keyword. | |
Default value:
MEDIUM |
|
List of valid keywords:
|
|
Alias names for this keyword: IOLEVEL |
PROGRAM_NAME {Keyword} |
|
Which program should be run | |
This keyword cannot be repeated and it expects precisely one keyword. | |
Default value:
CP2K |
|
List of valid keywords:
|
|
Alias names for this keyword: PROGRAM |
PROJECT_NAME {Word} |
|
Name of the project (used to build the name of the trajectory, and other files generated by the program) | |
This keyword cannot be repeated and it expects precisely one word. | |
Default value:
PROJECT |
|
Alias names for this keyword: PROJECT |
RUN_TYPE {Keyword} |
|
Type of run that you want to perform Geometry optimization, md, montecarlo,... | |
This keyword cannot be repeated and it expects precisely one keyword. | |
Default value:
ENERGY_FORCE |
|
List of valid keywords:
|
|
This keyword cites the following references: [Ceriotti2014] [Schonherr2014] |
SAVE_MEM {Logical} |
|
Some sections of the input structure are deallocated when not needed, and reallocated only when used. This reduces the required maximum memory | |
This keyword cannot be repeated and it expects precisely one logical. | |
The lone keyword behaves as a switch to
.TRUE. |
|
Default value:
.FALSE. |
SEED {Integer} ... |
|
Initial seed for the global (pseudo)random number generator to create a stream of normally Gaussian distributed random numbers. Exactly 1 or 6 positive integer values are expected. A single value is replicated to fill up the full seed array with 6 numbers. | |
This keyword cannot be repeated and it expects a list of integers. | |
Default value:
2000 |
TRACE {Logical} |
|
If a debug trace of the execution of the program should be written | |
This keyword cannot be repeated and it expects precisely one logical. | |
The lone keyword behaves as a switch to
.TRUE. |
|
Default value:
.FALSE. |
TRACE_MASTER {Logical} |
|
For parallel TRACEd runs: only the master node writes output. | |
This keyword cannot be repeated and it expects precisely one logical. | |
The lone keyword behaves as a switch to
.TRUE. |
|
Default value:
.TRUE. |
TRACE_MAX {Integer} |
|
Limit the total number a given subroutine is printed in the trace. Accounting is not influenced. | |
This keyword cannot be repeated and it expects precisely one integer. | |
Default value:
2147483647 |
TRACE_ROUTINES {Word} ... |
|
A list of routines to trace. If left empty all routines are traced. Accounting is not influenced. | |
This keyword cannot be repeated and it expects a list of words. |
WALLTIME {String} |
|
Maximum execution time for this run. Time in seconds or in HH:MM:SS. | |
This keyword cannot be repeated and it expects precisely one string. | |
Alias names for this keyword: WALLTI |
Back to the main page of this manual or the CP2K home page | (Last update: 31.12.2022) |