>Is changing data types on the fly really that expensive? I'd expect it would
>be fairly cheap CPU wise, and could be made really cheap with inline asm if
>you really care.
I don't like including asm, as it automatically makes the code
non-portable. But Benno suggests that its not so expensive ...
>I'd personally think it's better to allow whatever file format, and do any
>processing separately after recording if necessary (besides which, recording
>raw signal is usually best anyway).
I wasn't proposing limitations on the file format in terms of
RIFF/AIFF etc, just the sample width. And sure, all processing happens
subsequently, and preferably non-destructively.
--p
