[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: idl2matlab translate-o-matic

davidf@dfanning.com (David Fanning) writes:

> David McClain (dmcclain@azstarnet.com) writes:
> I'm sorry, but I think this *completely* misses the point.
> Cleaning up variables is one thing, but checking for *every*
> pointer reference at the end of every program module that
> exits would bring just about any program--never mind IDL--
> to a complete stand-still. It shouldn't be done. I applaud
> the folks at RSI for dismissing the idea out of hand.

Your're correct in the letter of what your wrote, but dead wrong in spirit.
Doing a full GC at every program module -- FULL, that is, chasing down and
verifying every reference -- would not be a good thing.  No garbage
collector I've known of has done this.  But, if you look at the beautiful
and efficient GCs written since 1978 or so, you'll see that they impose
perhaps 1-2% run-time overhead.  I've spent many years working on a system
(a LispM derivative called the TI Explorer) which was written entirely
(yes, *entirely*, from basic OS on up) in Lisp and had a GC running
constantly.  Once nice thing about these system was called the "modeline"
which displayed the currently active module: GC was hardly ever displayed
there.  More modern programming systems, such as Caml, Scheme, and the
like, have very, very good GCs, and, frankly, I'd rather be using a system
like that than worry about needing to explicity allocate and deallocate
objects, as in C/C++, etc., and having to trace down obscure memory leaks.

So, to summarize, GCs are not inherently bad, and, rather the opposite, are
quite good.  You might want to re-examine your position on the subject.  I
don't know anything about IDL, but, if what you say about RSI is true, then
they might want to re-think their position as well.


	- pz.

John Pezaris