[Haskell-cafe] Re: [Haskell] How to get the binary respentation of the Int/Double.

[Ryan Ingram]

Actually, this is a good question, at least as relating to floating
point values.  Is there a primitive to view the machine representation
of floats?

I'm thinking of functions like:

reinterpretFloatAsWord :: Float -> Word32
reinterpretWordAsFloat :: Word32 -> Float
reinterpretDoubleAsWord :: Double -> Word64
reinterpretWordAsDouble :: Word64 -> Double

There are decodeFloat and encodeFloat in the RealFloat class, but
these seem to be a bit heavy for inner-loop usage, especially when you
are just going to paste the values back together into a word
immediately anyways.  Are the precision of Float and Double specified
by the standard?

There's actually lots of nice tricks you can do with these functions;
for example, if you know you have a floating point value between 0 and
2^23-1, you can convert it to integer with a floating point add and an
integer mask:

> magic = 0x00800000 :: Float -- 2^23
> magicMask = 0x007fffff :: Word32 -- 2^23 - 1

> fastFloatToInt23 :: Float -> Word32
> fastFloatToInt23 x = reinterpretFloatAsWord (x + magic) .&. magicMask

The add sets the exponent of the float to +23 and arranges it so the
low bit of the float corresponds to "1", rounding if necessary.  Then
you can interpret the result as an integer and mask off the exponent
and sign bit from the float, leaving you with just the integer value.
This relies on the representation of IEEE single-precision floats, of
course.

These sorts of tricks come have in handy when working on graphics
code; often I've found myself wanting to index into an array with the
nearest value after some floating point computation.   Texture mapping
hardware does this conversion all the time, for example, and sometimes
it's useful to emulate that behavior in software.

  -- ryan

On Tue, Oct 28, 2008 at 10:08 AM, Brandon S. Allbery KF8NH
<allbery at ece.cmu.edu> wrote:
> On 2008 Oct 28, at 10:38, haihualin wrote:
>>
>> I am wondering how to get the binary respentation of the Int/Double.
>> So I can save the integer as the binary data so that the C program can
>> read it from the file.
>
>
> The Binary package on Hackage (http://hackage.haskell.org) allows you to
> serialize Haskell data in any form including C-compatible binary output.
>
> --
> brandon s. allbery [solaris,freebsd,perl,pugs,haskell] allbery at kf8nh.com
> system administrator [openafs,heimdal,too many hats] allbery at ece.cmu.edu
> electrical and computer engineering, carnegie mellon university    KF8NH
>
>
> _______________________________________________
> Haskell mailing list
> Haskell at haskell.org
> http://www.haskell.org/mailman/listinfo/haskell
>