2023.03

[2023.03] - 2023-10-27

Known Issues

• GCC 9 and below as well as GCC 11.1.0 are unsupported due to compilation issues. See prior releases for more details.
• NO_QUAD_PRECISION macro is no longer set by FMS, the ENABLE_QUAD_PRECISION macro has replaced prior usage of NO_QUAD_PRECISION. -DENABLE_QUAD_PRECISION should be set if quad precision is to be used, otherwise FMS will not use quad precision reals where applicable.

Added

• UNIT_TESTS: New unit tests have been created or and existing ones expanded on for any modules utilizing mixed precision support.

Changed

• MIXED PRECISION: Most subroutines and functions in FMS have been updated to simultaneously accept both 4 byte and 8 byte reals as arguments. This deprecates the --enable-mixed-mode option, which enabled similar functionality but was limited to certain directories and was not enabled by default. To facilitate easier testing of these code changes, the CMake precision options for default real size were left in (along with an equivalent --disable-r8-default flag for autotools). The resulting libraries will support mixed-precision real kinds regardless of default real size. It should also be noted that many routines that accept real arguments have been moved to include files along with headers in order to be compiled with both kinds. Most module level variables were explicitly declared as r8_kind for these updates.
• Some type/module changes were made to facilitate mixed precision support. They are intended to have minimal impact to other codebases:
  • COUPLER_TYPES: In coupler_types.F90, coupler_nd_field_type and coupler_nd_values_type have been renamed to indicate real kind value: coupler_nd_real4/8_field_type and coupler_nd_real4/8_values_type. The bc field within coupler_nd_bc_type was modified to use r8_kind within the value and field types, and an additional field added bc_r4 to use r4_kind values.
  • TRIDIAGONAL: Module state between r4 and r8 calls are distinct (ie. subsequent calls will only be affected by calls of the same precision). This behaviour can be changed via the save_both_kinds optional argument to tri_invert.
• CODE_STYLE: has been updated to reflect the formatting used for the mixed precision support updates.

Fixed

• DIAG_MANAGER: Tile number (ie. tileX) will now be added to filenames for sub-regional diagnostics.
• MPP: Bug affecting non-intel compilers coming from uninitialized pointer in the nest_domain_type
• MPP: Bug fix for unallocated field causing seg faults in mpp_check_field
• FMS2_IO: Fixed segfault occuring from use of cray pointer remapping along with mpp_scatter/gather
• TEST_FMS: Added various fixes for different compilers within test programs for fms2_io, mpp, diag_manager, parser, and sat_vapor_pres.
• INTERPOLATOR: Deallocates fields in the type that were previously left out in interpolator_end

Removed

• CPP MACROS:
  • no_4byte_reals was removed and will not set any additional macros if used. no_8byte_integers is still functional.
  • NO_QUAD_PRECISION was removed. It was conditionally set if ENABLE_QUAD_PRECISION was undefined. ENABLE_QUAD_PRECISION should be used in model components instead (logic is flipped)
  • use_netCDF was set by autotools previously but wasn't consistently used in the code. FMS should always be compiled with netcdf installed so this was removed with the exception of its use in deprecated IO modules.
• DRIFTERS: The drifters subdirectory has been deprecated. It will only be compiled if using the -Duse_drifters CPP flag.

Tag Commit Hashes

• 2023.03-beta1 06b94a7
• 2023.03-alpha3 b25a7c5
• 2023.03-alpha2 9983ce3
• 2023.03-alpha1 a46bd94