fields Module

Does the related namelist exist in the target input file?

Used to hold the specific index of numbered namelists

Do we want to skip the read step in init?

Do we want to skip the broadcast step in init?

Do we want to skip the smaart defaults in init?

Used with field_option='local'. If true and x/y are distributed then in time advance only update local part of field in operations like phinew=phinew+phi etc.

Writes files containing the field response matrix after initialisation. This currently works for field_option='implicit' or 'local'. We write to netcdf files by default but fall back to fortran unformatted (binary) files (which are not really portable) in the absence of netcdf.

Set to true to use an allreduce (on mp_comm) in field calculation (field_option='local' only) rather than a reduction on a sub-communicator followed by a global broadcast. Typically a little faster than default performance but may depend on 1 implementation.

Set to true, along with field_local_allreduce and intspec_sub , to replace the allreduce used in the field calculation with an allreduce on a sub-communicator followed by a reduction on a "perpendicular" communicator. Typically a bit faster than default and scales slightly more efficiently. Note if this option is active only proc0 has knowledge of the full field arrays. Other processors know the full field for any supercell (connected x-y domains) for which it has any of the xy indices local in the g_lo layout.

If true then use nonblocking collective operations in the fields_local field calculation. This may or may not improve performance.

Set to true when using field_option='local' to automatically tune and select the best performing minimum block size (in a single supercell) assigned to a single processor. This can improve performance, but is not guaranteed to.

The field_option variable controls which time-advance algorithm is used for the linear terms. Allowed values are:

Set to true to use allgatherv to fetch parts of the field update vector calculated on other procs. When false uses a sum_allreduce instead. This doesn't rely on sub-communicators so should work for any layout and processor count. Note: This only impacts field_option='implicit'

If true then recalculate the fields from the distribution function using get_init_field when restarting a simulation rather than using the values in the restart file.

Used with field_option='local' to set the minimum block size (in a single supercell) assigned to a single processor. Tuning this parameter changes the balance between work parallelisation and communication. As this value is lowered, more communication needs to be done, but more processors get assigned work. This may reduce the time spent in computation at the cost of time spent in communication. The optimal value is likely to depend upon the size of the problem and the number of processors being used. Furthermore it will affect intialisation and advance in different ways. Can be automatically tuned using field_local_tuneminnrow.

Reads files containing the field response matrix and uses to initialise GS2s response matrix rather than using the usual initialisation process.

Delete zonal flows at every timestep.

Sets location in which to store/look for response dump files. We don't currently check that this location exists before attempting to use it, which could cause problems. The default is to save them in the working directory.

Allows customisation of the base filename to be used for response files. If not set then we use run_name derived from the input file name.