I assume that if csound is using doubles, the spin and spout buffers accessed 
through the api will also be doubles.
In that case  there must be an API call to get the size of the internal 
samples. 

An argument for using floats is that is much easier to interface csound to 
various plugin APIs and/or intergrate it with other systems if it uses 
floats, since that is pretty much the standard and you can use csounds spin 
and spout directly, which improves latency f.i.

Gerard


-- 
electronic & acoustic musics-- http://www.xs4all.nl/~gml