Here is info from Erik Tollerud about the IRATE data format.

The bitbucket project is at https://bitbucket.org/eteq/irate-format
and I've posted a copy of the docs at
http://www.physics.uci.edu/~etolleru/irate-docs/ , in particular
http://www.physics.uci.edu/~etolleru/irate-docs/formatspec.html ,
which details the actual requirements for data to fit in the format.
As far as I can tell, the following steps are needed to make GDF fit
inside the IRATE format:

*move everything except "/simulation_parameters" into a group named "/GridData"

*rename "/simulation_parameters" to "SimulationParameters"

*remove the 'field_types' group (this is not absolutely necessary, but
the convention we had in mind for IRATE is that the dataset names
themselves (e.g. the datasets like /data/gridxxxxxx/density)  serve as
the field definitions.

* The unit information that's in 'field_types' should then be
attributes in either "/GridData" or "/GridData/data" following the
naming scheme e.g. "densityunitcgs" following the unit form given in
the IRATE doc and an additional attribute e.g. "densityunitname"
should be added with the human-readable name of the unit. This unit
information can also live at the dataset level, but it probably makes
more sense to put it instead at the higher level (IRATE supports both
ways of doing it)

* The Cosmology group (as defined in the IRATE specification) must be
added - for simulations that are not technically "cosmological", you
can just use one of the default cosmologies (WMAP7 is a reasonable
choice - there's a function in the IRATE tools that automatically
takes care of all the details for this).

* optional: redo all the group names to follow the CamelCase
convention - that's what we've been using elsewhere in IRATE.  This is
an arbitrary choice, but it would be nice for it to be consistent
throughout the format.

