7.28 global_settings

• type: action command.
• sequence: should precede run_setup.
• function: change global settings.
• Command syntax, including use of equations and subcommands, is discussed in 7.2. One way to use the global_settings command is in a configuration file, which can be supplied via the -configuration option or ELEGANT_CONFIGURATION environment variable; an example of using this would be to set MPI I/O options an a per-system basis. N.B.: unlike other commands, values given for parameters in global_settings commands become the new default for subsequent invocations of the command during the same run. Hence multiple commands can be used to set individual values without overriding previously-given settings.

• inhibit_fsync — By default, elegant forces file synchronization across a network file system to ensure that users see up-to-date files as soon as possible. In cases where a great deal of output is generated, this can degrade performance. Setting this parameter to 1 will turn off synchronization until the end of the run.
• allow_overwriting — By default, elegant will overwrite existing output files. If this parameter is set to 0, it will instead exit if an output file already exists.
• echo_namelists — By default, elegant echoes all namelist input to the terminal. If this parameter is set to 0, this output will be inhibited.
• SR_gaussian_limit — By default, elegant uses a 3-σ cutoff for the gaussian random numbers used in simulation of synchrotron radiation from CSBEND, CSRCSBEND, KQUAD, KSEXT, and SREFFECTS. This parameter allows changing the cutoff.
• inhibit_seed_permutation — If nonzero, randomization of the user-supplied random number seed is not performed. This feature is useful in that it provides a higher degree of apparent randomness, in that small changes in the seed result in very different random sequences.
• log_file — By default, elegant writes status information to the terminal. If a filename is supplied for this parameter, the output will instead go to the file. On Linux and Unix, using /dev/null will result in the output being discarded.
• error_log_file — By default, elegant writes error messages to the terminal. If a filename is supplied for this parameter, the output will instead go to the file. On Linux and Unix, using /dev/null will result in the output being discarded.
• share_tracking_based_matrices — If non-zero, then the matrices determined by tracking for various elements (e.g., BRAT, BGGEXP, CCBEND) are computed only once for a set of identical elements, then shared. This can save considerable computation time.
• mpi_randomization_mode — Controls how the random numbers are seeded on multiple processors
  • 1 — This is the original default, which showed issues in some simulations. The seed on the i^th processor is s₀ + 2 * i.
  • 2 — The seed on the i^th processor is s₀ + 2 * i².
  • 3 — This is the new default. The seed on the i^th processor is s₀ + i * (i + 1).
  • 4 — The seed on the i^th processor is s₀+R_i, where R_i is the i^th random integer returned by the system rand() function.
• exact_normalized_emittance — By default, elegant uses an approximate computation for the normalized emittance, namely, ϵ_n = ϵ⟨βγ⟩, where ϵ is the geometric emittance computed from the trace-space coordinates. If this variable is set to a non-zero value, elegant instead uses a slower but more accurate method, namely, using the momentum coordinates. [43]. The results will show up in the sigma and final output files, if these are requested in the run_setup command.
• mpi_io_force_file_sync — If non-zero, Pelegant will perform a file synchronization after writing each row of an SDDS file. This can significantly degrade performance, but can solve problems on some filesystems that result in corrupted files or files in which zeros appear in place of the expected data.
• mpi_io_read_buffer_size — If non-zero, Pelegant will change the read buffer size to the given value. May allow improving read performance, but should be used with care.
• mpi_io_write_buffer_size — If non-zero, Pelegant will change the write buffer size to the given value. May allow improving write performance, but should be used with care.
• usleep_mpi_io_kludge — If non-zero, Pelegant will sleep for the given number of microseconds after writing each row of an SDDS file. This can degrade performance, but can solve problems on some filesystems that result in corrupted files or files in which zeros appear in place of the expected data. It may give better performance than setting mpi_io_force_file_sync=1. A value of 100 is suggested as a starting point, but this will be highly system-dependent.
• parallel_tracking_based_matrices — If non-zero, then the matrices determined by tracking for various elements (e.g., BRAT, BGGEXP, CCBEND) are computed using parallel resources in Pelegant. This can save considerable wall clock time. N.B.: This is set to zero when using parallel_optimization_setup.
• share_tracking_based_matrices — If non-zero, then the matrices determined by tracking for various elements (e.g., BRAT, BGGEXP, CCBEND) are computed only once for a set of identical elements, then shared.
• tracking_based_matrices_store_limit — The maximum number of matrices that will be stored for possible sharing.
• tracking_matrix_step_factor — The default step size for tracking-based matrices is 5 × 10^-5 (in the appropriate units for each corodinate). This can be increased or decreased by supplying a value for tracking_matrix_step_factor.
• tracking_matrix_points — By default, five grid points are used in each dimension for tracking-based matrix determination. This can be increased by setting tracking_matrix_points to a larger, odd value, at the expense of longer running time. (The run time scales approximately as the sixth power of this value.)
• tracking_matrix_step_size — Sets the step sizes, in each of the six coordinates, used for tracking-based matrix determination
• warning_limit — Allows setting a limit on the number of similar warnings that are printed in detail. Set to 0 to see no detailed warnings, only the summary; this is not recommended. Set to -1 to see all detailed warnings; this is suggested for debugging only, since it may create large log files and reduce performance.
• malign_method — Allows globally setting the default misalignment method for those elements that lack the MALIGN_METHOD parameter.. Five values are recognized:
  • 0 — Indicates the original misalignment method that is entrance-centered and includes three position offsets and roll (TILT).
  • 1 — Indicates the new method [58] in entrance-centered mode in the linear approximation. This includes the possibility of pitch and yaw errors on elements that support it.
  • 2 — Indicates the new method in body-centered mode in the linear approximation. This includes the possibility of pitch and yaw errors on elements that support it.
  • 3 — Indicates the new method [58] in entrance-centered mode using an exact form. This includes the possibility of pitch and yaw errors on elements that support it.
  • 4 — Indicates the new method in body-centered mode using an exact form. This includes the possibility of pitch and yaw errors on elements that support it.
• slope_limit, coord_limit — By default, elegant limits the maximum slopes (x^′ and y^′) and coordinates (x and y) in order to avoid potential overflows and unreasonable values. These parameters allow users to change those limits.