LMDz with OpenACC annotations LMDZ/acc

LMDz with OpenACC annotations `LMDZ/acc`#

Description#

This is the plugin to run the model LMDZ. Please see further details on LMDZ itself here

YAML arguments#

The following arguments are used to configure the plugin. pyCIF will return an exception at the initialization if mandatory arguments are not specified, or if any argument does not fit accepted values or type:

Mandatory arguments#

fileexec : str, mandatory: path to DISPERSION executable

Optional arguments#

autoflush : bool, optional, default True: Clean sub-directories of temporary files (mod.bin, etc.)

ensrf_restart_file : str, optional: Use in EnSRF mode to point at relevant restart file

ensrf_datei : str, optional: New start date to use in EnSRF

ensrf_same_restart : bool, optional: Use in EnSRF mode to check if same restart

run_command : str, optional, default “./dispersion.e”: Commnand used to run DISPERSION executable: run_command dispersion.e. By default, the DISPERSION executable will be run with :bash:`./dispersion.e

dir_sources : str, optional, default “”: Path to DISPERSION sources. Overwrites fileexec if specified.

modules : list, optional: Modules to be loaded for compiling and running LMDZ, if incompatible with the modules required to run PyCIF. This option should be set through a platform plugin.

periods : str, optional, default “1MS”: Frequency for sub-simulations

flx_tresol : str, optional, default “1D”: Expected time step between flux values in the executable

physic : bool, optional, default True: Use physics in LMDZ

compressed_phystoke : bool, optional, default True: Use compressed phystoke files

no_trajq : bool, optional, default False: Do not write ‘trajq_*.bin’ files. WARNING: ‘trajq_*.bin’ files are necessary for the adjoint mode

dump_chemical_sink : bool, optional, default False: Write tracers chemical and tracers lifetime sink in NetCDF files.

dump : bool, optional, default False: Dump trajq as netCDF

do_chemistry : bool, optional, default True: Use the chemistry

approx_operator : optional

Approximated operator arguments. The approximated operator is used in the “tangent-linear” mode when all the defined thresholds are reached or in “adjoint” mode when the the parallel observation operator is used. The approximated operator operator considers the atmosphere as well mixed (i.e. uniform) and only compute the chemical losses with a simple newton scheme.

Argument structure:

time_step : str, optional: time step for the newton scheme integration

condition : “and” or “or”, optional, default “and”: condition on which the approximated operator replaces the model, either ‘and’ (when all species threshold are met, default) or ‘or’ (when at least one species threshold is met)

species_threshold : optional

Argument structure:

any_key : optional
species name

Argument structure:

threshold : float, mandatory
concentration threshold in ppm

compiler : “gfortran” or “mpif90” or “nvfortran”, optional: Compiler to use

auto-recompile : bool, optional, default False: Auto-compile executables before running if executables not available

force-recompile : bool, optional, default False: Force compilation even if executables available

force-compile-stderr : bool, optional, default False: Force printing what happened during compiling

compile-clean : bool, optional, default False: Cleaning repositories before compiling; slows the compilation but prevent side-effects from re-compiling.Can be switched of for minor modifications in the code

compile-mode : “PROD” or “DEBUG”, optional, default “PROD”: Compilation mode

“PROD”: production mode; all optimizations are activated
“DEBUG”: check-all and traceback are activated, hence slowing down the execution

compile-acc-target : “multicore” or “gpu” or “none”, optional, default “multicore”: Target for ACC annotations

nproc : int, optional: Force using a given number of processors. By default, LMDz uses all available processors

dont_perturb_species : list, optional: List of species to not perturb in batch sampling

copy_inputs : “any subset of [‘meteo’, ‘prescrconcs’, ‘prodloss3d’, ‘kinetic’]”, optional, default []: Copy (rather creating symbolic links) the input files

Requirements#

The current plugin requires the present plugins to run properly:

Requirement name	Requirement type	Explicit definition	Any valid	Default name	Default version
domain	Domain	False	False	LMDZ	std
flux	DataStream	True	False	LMDZ	sflx
chemistry	Chemistry	False	False	CHIMERE	gasJtab
emis_species	DataStream	True	False	LMDZ	bin
meteo	DataStream	True	False	LMDZ	mass-fluxes-acc
inicond	DataStream	True	False	LMDZ	ic
prescrconcs	DataStream	True	False	LMDZ	prescrconcs
kinetic	DataStream	True	False	LMDZ	photochem
prodloss3d	DataStream	True	False	LMDZ	prodloss3d

YAML template#

Please find below a template for a YAML configuration:

model:
  plugin:
    name: LMDZ
    version: acc
    type: model

  # Mandatory arguments
  fileexec: XXXXX  # str

  # Optional arguments
  autoflush: XXXXX  # bool
  ensrf_restart_file: XXXXX  # str
  ensrf_datei: XXXXX  # str
  ensrf_same_restart: XXXXX  # bool
  run_command: XXXXX  # str
  dir_sources: XXXXX  # str
  modules: XXXXX  # list
  periods: XXXXX  # str
  flx_tresol: XXXXX  # str
  physic: XXXXX  # bool
  compressed_phystoke: XXXXX  # bool
  no_trajq: XXXXX  # bool
  dump_chemical_sink: XXXXX  # bool
  dump: XXXXX  # bool
  do_chemistry: XXXXX  # bool
  approx_operator:
    time_step: XXXXX  # str
    condition: XXXXX  # and|or
    species_threshold:
      any_key:
        threshold: XXXXX  # float
  compiler: XXXXX  # gfortran|mpif90|nvfortran
  auto-recompile: XXXXX  # bool
  force-recompile: XXXXX  # bool
  force-compile-stderr: XXXXX  # bool
  compile-clean: XXXXX  # bool
  compile-mode: XXXXX  # PROD|DEBUG
  compile-acc-target: XXXXX  # multicore|gpu|none
  nproc: XXXXX  # int
  dont_perturb_species: XXXXX  # list
  copy_inputs: XXXXX  # any subset of ['meteo', 'prescrconcs', 'prodloss3d', 'kinetic']