Floating point processing

[James McCartney]

On the PowerPC single precision operations are faster, but it depends on
the specific chip.
There is an frsp instruction for conversion from double to single precision
when there is an implicit or explicit cast to single precision.
You don't have to convert the other way because all values are held in
registers as double precision, even if you are doing single precision ops.
The register loads and stores convert precision automatically from or
to the address precision. Double precision values are 64 bits on PPC.


   --- james mccartney     james at clyde.as.utexas.edu   james at lcsaudio.com
If you have a PowerMac check out SuperCollider, a real time synth program:
http://www.audiosynth.com