Path: utzoo!attcan!uunet!seismo!ukma!tut.cis.ohio-state.edu!brutus.cs.uiuc.edu!apple!bionet!ames!decwrl!megatest!djones
From: djones@megatest.UUCP (Dave Jones)
Newsgroups: comp.sw.components
Subject: Re:  Re: Real-time Garbage Collection
Message-ID: <8222@goofy.megatest.UUCP>
Date: 21 Sep 89 21:16:34 GMT
Organization: Megatest Corporation, San Jose, Ca
Lines: 22


A very wise man once said, "Floating point representation is a mistake.
If you don't know enough about the problem to scale the quantities, you
don't know enough about the problem."

May he rest in peace. He was martyred by an angry hoard of Fortran
application programmers.

If he had lived, today he might say, "Garbage collection is a mistake.
If you don't know enough about the problem to free the unreferenced
memory, you don't know enough about the problem."

If I had lived, I probably would agree with him.

I have some library packages, oops -- I meant to say "ADT's" --
which help.  I can declare Stacks, which can be marked.  Later you can
deallocate all the memory above the mark in one very fast pop. That
covers many many situations. I also have Heaps which contain blocks
all of one size. Not only is it quicker than your typical malloc(),
sometimes by a factor of several hundred, but if you are through with
all the records in a Heap, just destroy the entire Heap. Both of these,
like all my librar.. er ADT's, grow in size indefinitely, as required.