Re: [Haskell-cafe] [Haskell] Can Haskell use short floats; i.e. 16 bit floats to save space?
moving to Haskell-Cafe ... Am 01.03.2014 07:09, schrieb KC:
Can Haskell use short floats; i.e. 16 bit floats to save space?
You could wrap an Int16 in a newtype and implement your own floating point arithmetic with some bit manipulation and make that instances of Num, Fractional or numeric-prelude or yap classes.
Am 01.03.2014 07:09, schrieb KC:
Can Haskell use short floats; i.e. 16 bit floats to save space?
GHC pads all basic data types to one word in its in-memory representation, so no. It's possible that the boxed vectors in the "vector" package has a more compact representation for arrays of Floats. -- Johan
Did you mean unboxed vectors? Storable vectors will store an Int16 in 16
bits, so that's another approach.
On Fri, Feb 28, 2014 at 11:56 PM, Johan Tibell
Am 01.03.2014 07:09, schrieb KC:
Can Haskell use short floats; i.e. 16 bit floats to save space?
GHC pads all basic data types to one word in its in-memory representation, so no. It's possible that the boxed vectors in the "vector" package has a more compact representation for arrays of Floats.
-- Johan
_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
participants (3)
-
Henning Thielemann -
Johan Tibell -
John Lato