On Tue, 8 Feb 2005 20:33:52 +0200, Radu Grigore <radugrigore@gmail.com> wrote:
> On Tue, 08 Feb 2005 09:12:32 -0800 (PST), skaller
> <skaller@users.sourceforge.net> wrote:
> 
> > However it isn't clear (to me) that just picking an arbitrary
> > assignment to split into two using a temporary
> > actually minimises the number of temporaries.
> 
> I'm mildly convinced that it works.

I was obviously too tired yesterday: there are lots of mistakes in my argument.

Actually here is an example on which your algorithm uses more
assignments than necessary:

a = b+c
b = c
c = a+b

The solution is
t = c
c = a+b
a = b+t
b = t

-- 
regards,
 radu
http://rgrig.idilis.ro/