[gclist] Allocator algorithms.

[Henry G. Baker]

> > This makes next fit's fragmentation go through the ceiling---to
> > around 400%
> 
> With what measure of fragmentation? I find choosing a single number
> for fragmentation confusing; it's multi-dimensional.
> 
> Nick Barnes

I have looked at using the concepts of 'entropy' and 'temperature' to
characterize fragmentation.  Entropy is roughly analogous to the amount
of information required to describe the disordered state.  One could
get a quick estimate by setting the information _within_ the blocks
to all zeros, and then gzipping the whole heap.

-- 
Henry Baker
www/ftp directory:
ftp.netcom.com:/pub/hb/hbaker/home.html