Wiki

Clone wiki

gyre / Release Notes

4.0 (February 2, 2015)

New Features

  • The GYRE_DIR environment variable now needs to be set, to point to the top-level GYRE directory
  • GYRE no longer needs gfortran 4.10 (aka gfortran 5.0) to compile; it works fine with gfortran 4.9.3
  • Implemented dedicated code paths for real and complex arithmetic, resulting in significant speedups in many cases
  • Added uniform_rot flag and Omega_uni parameter to &model namelist, to enforce rigid rotation on the model
  • Added reconstruct_As flag to &model namelist, controlling numerical reconstruction of the dimensionless Brunt-Vaisala frequency
  • Added rotation_method parameter to &osc namelist, to determine how rotation effects are treated. The possible treatments include the traditional approximation
  • Added inner_bound parameter to &osc namelist, allowing specification of the inner boundary conditions
  • Added r_root_solver and c_root_solver parameters to &num namelist, allowing specification of the real and complex root solvers, respectively
  • Added grid_frame and freq_frame parameters to &scan namelist, allowing specification of the frames in which the frequency grid is distributed and the frequency units are defined, respectively
  • Added 'GRAVITY_DELTA' and 'ACOUSTIC_DELTA' options for the &freq_units parameter in the &scan and &output namelists
  • Added 'MIXED_DELTA' option for the &freq_units parameter in the &scan namelist
  • Renamed ivp_solver_type parameter to ivp_solver
  • Renamed outer_bound_type parameter to outer_bound
  • Renamed inner_bound_type parameter to inner_bound
  • Renamed inertia_norm_type parameter to inertia_norm
  • Renamed variables_type parameter to variable_set
  • Replaced use_banded parameter by matrix_type parameter

3.3 (October 27, 2014)

New Features

  • GYRE now requires gfortran 4.10 to compile (earlier versions of the compiler contain bugs, which result in failed compilation and/or incorrect results)
  • Implemented more-stringent checking of namelist counts
  • Renamed X_n_pg_min and X_n_pg_max parameters to n_pg_min and n_pg_max (as they're no longer experimental). Also, the semantics of these parameters has changed: they now also apply to adiabatic calculations
  • Added prune_modes flag to &output namelist. When set to .TRUE., eigenfunction data for each mode is discarded after the mode is (possibly) written to disk
  • Added support for reading unformatted AMDL-format stellar models
  • Added new deflate_roots flag to &num namelist. When set to .TRUE. (now the default), root deflation allows gyre_nad to avoid converging on the same mode multiple times.
  • Changed the complex root solver in gyre_nad from secant method to a complex-modified Ridders' method; this seems to fix cases of slow convergence.
  • Renamed the FINDIFF_GL2 and FINDIFF_GL4 IVP solvers to COLLOC_GL2 and COLLOC_GL4.
  • Added new FINDIFF IVP solver, which uses Sugimoto's approach to handle thermal instability in non-adiabatic calculations
  • Restricted non-adiabatic calculations to the two (currently) known-to-work IVP solvers: MAGNUS_GL2 and FINDIFF
  • Replaced omega_im output routine with omega_int, which estimates the real & imaginary parts of the eigenfrequency from an integral expression (comparing this against omega provides a good check on the reliability of results)
  • Moved the reduce_order flag from the &num namelist to the &osc namelist
  • Renamed delp output item to lag_P, and other del* items to lag_*
  • Renamed phip output item to eul_phi, and dphip_dx to deul_phi
  • Added eul_P, eul_rho and eul_T
  • Added RESAMP_STRUCT grid operation, which resamples based on changes in the structure variables
  • Changed the units of the dW_dx and W output items

Bug Fixes

  • Fixed calculation of pressure and gravitational perturbation eigenfunctions for radial modes when the reduce_order flag is .TRUE.

3.2 (May 2, 2014)

New Features

  • Split off mode parameters (l, m, X_n_pg_min and X_n_pg_max) from the &osc namelist; they are now specified in their own &mode namelist(s).
  • Added support for the Doppler shift due to (possibly-differential) rotation; set m to a non-zero value to enable this.
  • Added the ability to do non-adiabatic calculations in the MESA interface module.
  • Added new regularize flag to the &model namelist; this regularizes an input model, to try to iron out inconsistencies in hydrostatic equilibrium and mass conservation.
  • Added support for LOSC-format stellar models

Bug Fixes

  • Changed format specifiers for TXT output files, so that 3-digit exponents don't lead to a missing 'E' character.

3.1 (April 14, 2014)

New Features

  • Added new mode_template parameter to &output namelist, as an alternative to the existing mode_prefix for specifying the names of output mode files (if both parameters are set, then mode_template is used and mode_prefix is ignored). Filenames are constructed by performing pattern substitution on the mode_template value; patterns can be one of the following:
    • %J : run index of mode, formatted in fixed-width field; counts from 1 upwards for each mode found during the GYRE run (unique)
    • %j : same as %J, but formatted in variable-width field
    • %L : harmonic degree of mode, formatted in fixed-width field
    • %l : same as %L, but formatted in variable-width field
    • %N : radial order of mode, formatted in fixed-width field
    • %n : same as %N, but formatted in variable-width field
  • Improved handling of stellar models containing duplicate radius points (a warning is now printed, instead of the code stopping).

Bug Fixes

  • Fixed bug which prevented CREATE_MIDPOINT being used for the shooting grid
  • Fixed bug which caused floating exception when l=0 and JCD variables used

3.0 (Feburary 26, 2014)

New Features

  • New op_type option, CREATE_MIDPOINT, in &shoot_grid and &recon_grid namelists. This places grid points at the midpoints of the input grids. This option is key to producing 'good' eigenfunctions when performing non-adiabatic calculations
  • Renamed the &coeffs namelist to &model, and the coeffs_type parameter to model_type
  • Added new &constants namelist, to allow overriding of physical constants
  • New output items: f_T, f_g, psi_T and psi_g, being the non-adiabatic coefficients defined by Dupret et al. (2003). These are evaluated at x = x_ref
  • New &osc parameter: x_ref
  • Added xi_r_ref and xi_h_ref output items, which give the displacement eigenfunctions at x = x_ref
  • Added warnings when scan frequency range goes above/below atmospheric cutoffs

Bug Fixes

  • Removed incorrect factor of 4pi in inertia calculations
  • Removed spurious factor of 4pi in work calculations
  • Refactored code to avoid possible dangling pointer issues

2.4 (December 27, 2013)

New Features

  • Added support for singular outer boundaries

Bug Fixes

  • Fixed crash when reduce_order=.FALSE. and variables_type='JCD'
  • Fixed crash when using op_type='RESAMP_CENTER' with radial modes of polytropic models
  • Fixed some ongoing memory leak issues caused by gfortran PR 57922

2.3 (October 26, 2013)

New Features

  • New 'MIX' option for the variables_type parameter
  • New &osc parameter: inertia_norm_type
  • New output item: freq_units
  • Improved testing framework

Bug Fixes

  • Fixed bugs in Jacobian calculation for radial modes using JCD variables
  • Fixed null pointer issue with radial modes when using DZIEM boundary condition
  • Fixed bug in rotation kernel and beta calculation

2.2 (October 5, 2013)

New Features

  • New experimental &osc parameters: X_n_pg_min and X_n_pg_max
  • Switched to modular IVP solvers

Bug Fixes

  • Removed duplicate cache filling, which should improve initialization times
  • Modified storage of structure coefficients, resulting in a dramatic reduction in memory usage (GYRE shouldn't run out of memory anymore)

2.1.1 (September 19, 2013)

Bug Fixes

  • Fixed bugs in Unno and JCD boundary conditions for radial pulsations
  • Further modified extended-range arithmetic to handle underflow gracefully
  • Modified FGONG file reader to handle r /= 0 central points

2.1 (September 13, 2013)

New Features

  • Support for multiple &num namelists in input files
  • Support for FAMDL-format stellar models
  • New tag system for controlling parameter selection
  • New numerical parameter: variables_type
  • New grid operation: 'CREATE_FROM_FILE'
  • Discriminant output is replaced by a convergence measure chi

Bug Fixes

  • Created workaround for gfortran PR 58007
  • Modified extended-range arithmetic to handle underflow gracefully
  • Changed testing so that return values are explicitly checked
  • Fixed possible division-by-zero in Unno and JCD boundary conditions

2.0 (August 18, 2013)

New Features

  • Support for multiple &osc namelists in input files
  • Support for text ('TXT') format output files
  • Checks for defective Magnus matrices
  • Rearranged source tree
  • Reworked build & test system
  • Takata classification scheme for dipole modes
  • Radial modes now handled by gyre_ad (replacing gyre_rad)
  • API to allow integration with MESA
  • Parallelized model initialization
  • New output items: Yt_1, Yt_2, K, prop_type
  • New &freq_units option: 'PER_DAY'
  • New &num parameters: reduce_order and use_banded
  • New &grid_op option: 'CREATE_UNIFORM'

Bug Fixes

  • Fixed 'no such instruction' compilation issue on newer Macs
  • Created workaround for gfortran PR 57922

Known Issues

  • In some circumstances, eigenfunction reconstruction can fail. The symptom is large numbers of 'Smallest element not in final block' warnings. The fix is being worked on; in the meantime, try setting the use_banded parameter in the &num namelist to .TRUE.

1.1.2 (June 7, 2013)

Bug Fixes

  • Fixed compilation error in gyre_rad
  • Further optimized the system matrix solver

1.1.1 (June 4, 2013)

Bug Fixes

  • Fixed severe slowdown setting up coefficient caches for large grids
  • Fixed memory access bug with HOM and POLY models

1.1 (June 4, 2013)

New Features

  • Changed format of GSM files (non-backwardly compatible); see here
  • Changed format of MESA files (backwardly compatible); see here
  • Created a radial adiabatic code, gyre_rad (see the rad subdirectory)
  • Added many performance enhancements, including
    • real-arithmetic shortcuts
    • coefficient caching
    • calls to the BLAS library
    • improved OpenMP scheduling
  • Added hooks to interface with MESA
  • Added a new RESAMP_THERMAL option to the op_type field of the &shoot_grid and &recon_grid namelists. This can be used to ensure smooth entropy profiles when performing non-adiabatic calculations
  • Added a new data_format field to the &coeffs_type namelist, to specify the Fortran data format when reading MESA- and OSC-format files. This allows reading of OSC-format files written by older versions of MESA, which used the wrong data format.
  • Added monotonicity checks for evolutionary stellar models

1.0.2 (May 21, 2013)

Bug Fixes

  • Added code to check that MESA input files have monotonic mass and radius coordinates

1.0.1 (May 20, 2013)

Bug Fixes

  • Fixed problem with gyre_nad giving incorrect non-adiabatic eigenfrequencies

1.0 (May 17, 2013)

Initial public release of GYRE.

Updated