oldk1331 wrote:
> > More precisely, instead of building
> > list and passing it to constructor build desired object
> > directly.
>
> Building the desired object through list might be faster
> than build directly -- for example, enqueue! checks if its
> argument is empty every time, that slows thing down.
Main cost is due to memory allocation. Array version
can allocate big hunk of memory, but list-based one
pays costs of several allocations.
> > OTOH in other cases signifcant savings are
> > obtained by sharing substructures. But sharing works
> > well only if we avoid destructive operations.
>
> But like you said in the other thread, Spad is pretty imperative,
> thus destructive operations are unavoidable, thus Spad
> (and Lisp) provides both destructive operations and non
> destructive operation. That gives more reason to have
> a destructive version of constructor.
Well, we provide destructive operations not for love of
impeartive things but due to speed. ATM I see no case
where destructive constructor would give us important
speedup.
> > Concerning idea of destructive constructor: memory is
> > freed by garbage collector, so all we can do to free
> > memory is to avoid keeping references to unneeded things.
> > This means that destructive constructor can give any
> > gain only in very special cases.
>
> For some data structure, I think a destructive constructor
> takes O(1) instead of O(n): for example Stack (because of
> the direct assign). For other data structure, a destructive
> constructor can save half time and half peak memory --
> copy is O(n), we can save that step.
Sure, destructive constructor could be faster than
normal constructor. But I see no case when:
- we care about speed of constructor
- destructive constructor gives us best space
To put it differently, if you say "see we have this
important calculation and destructive constructor
gives us better speed than alternatives", then
I will not oppose. But I oppose intrducing destructive
constructor just because it makes some artifical
benchmark faster.
--
Waldek Hebisch
--
You received this message because you are subscribed to the Google Groups
"FriCAS - computer algebra system" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/fricas-devel.
For more options, visit https://groups.google.com/d/optout.
