[PRL] laziness
Jeffrey Palm
jpalm at ccs.neu.edu
Wed Jun 23 06:07:18 EDT 2004
David A. Herman wrote:
> 2) Will it "automatically" buy me better performance?
>
> In my multiple-traversal program, which is going very slowly, if I make
> all the constructors into DELAY-s and the observers into FORCE-s, will my
> program suddenly be faster? In light of (1), I would think its overall
> performance should be the same, although ostensibly the big gain is when
> you delay computations that never end up needing to be performed. In a
> program transformation, though, don't you end up doing all the
> computations anyway? So is it just DELAY-ing the inevitable?
I would think delaying any computation could help your memory
performance, at least. Assuming that performing a computation adds to
the heap, if you can delay this until absolutely necessary you'll avoid
keeping these results around. Of course, this depends on what it costs
-- memory-wise -- to delay your traversals, because that could be more
expensive.
Jeff
--
Jeffrey Palm --> http://www.ccs.neu.edu/home/jpalm
More information about the PRL
mailing list