[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[netCDF #IST-525664]: Half-precision floating point format

As far as I know, HDF5 does not support 16 bit floating point.
If the goal is to save space, then one approach is to create
a special filter that converts between 16 bit in the file and 32 bit
in memory.

> Organization: EUMETSAT
> Package Version:
> Operating System:
> Hardware:
> Description of problem: This is not really a request, it is just curiosity.
> 
> Regarding floating-points data formats, currently netCDF implements the 
> standard single and double precision.
> 
> I am wondering if netCDF considered the implementation of the half-precision 
> floating point format (aka 2-byte floating point, binary16, etc).
> https://en.wikipedia.org/wiki/Half-precision_floating-point_format
> 
> This data type has a resolution that could be good enough for many purposes, 
> and at the same time it keeps simplicity (no need to bother with 2-byte 
> integers and "add_offset" and "scale_factor" attributes; or no need to define 
> single precision format together with the "significant number of digits" and 
> compress).
> 
> Perhaps the key point is whether this is already implemented in HDF5, and 
> this is not clear to me.
> 
> Anyway, my question is: Is half-precision floating point data format 
> considered for implementation in netCDF?
> 
> Thank you.
> 
> 

=Dennis Heimbigner
  Unidata


Ticket Details
===================
Ticket ID: IST-525664
Department: Support netCDF
Priority: Normal
Status: Open
===================
NOTE: All email exchanges with Unidata User Support are recorded in the Unidata 
inquiry tracking system and then made publicly available through the web.  If 
you do not want to have your interactions made available in this way, you must 
let us know in each email you send to us.