On Thu, Sep 23, 2010 at 10:20 AM, Josh Parmenter
<josh@xxxxxxxxxxxxxxxxx> wrote:
it would mean it is allocated dynamically and should just cause a failed allocation (rather then a crash). Also - DelayL (and their ilk) round up to the next power of two. Allocating otherwise doesn't (I believe).
BufDelay* and cousins truncate the max delay to the next power of two below the buffer size -- see the archives. That was done for performance reasons. It might be nice to have slower but more memory-efficient versions of these units since we cannot assume that performance is the top priority in all cases. (I didn't fully agree with the argument for performance at the time, but didn't have a use case to explain why. Etienne has provided one nicely!)
hjh