Re: RFA: invert-speedup-1.diff

[Alexandre Duret-Lutz]

>>> "Raja" == Raja R Harinath <address@hidden> writes:

[...BDD...]

 Raja> I thought of them too.  However, when you convert them back to SOP
 Raja> form to describe to the user, they sometimes may seem more verbose
 Raja> than necessary.  

Isn't Minato-Morreale's algorithm the answer?  (No idea how this
works, the paper [*] is amongst the pile of things I want to read when
I get some spare time.)

 Raja> Also, the size of a ROBDD is related to the ordering of
 Raja> variables, though for normal usage, there shouldn't be a
 Raja> major problem.

In practice I think users often layout their nested if/then/else
blocks in a way that minimize duplications.  So, since BDDs are
also about encoding if/then/else, ordering conditionals the way
they are encountered in Makefile.am seems attractive.

Anyway, as long as the present implementation works fine, it's
probably overkill to think of something else.

[*] http://www5.ocn.ne.jp/~minato.s/SASIMI92.pdf
-- 
Alexandre Duret-Lutz