Re: dealloc vs dealloc(n)

["Mark S. Waterbury" <mark.s.waterbury@xxxxxxx>]

Hi, Simon:

You are correct -- I described an overly simplistic approach. (It was 
late in the evening.)

For cases where a long-running job needs to use dynamic storage, I have 
found that using "pools" of like storage is useful. The idea is to 
allocate all objects of a given "kind" (and lifetime) from the same 
"pool" -- this way, when you are done with that "kind" of thing, you 
just delete the entire pool. This simplified approach provides many of 
the benefits of "generational garbage collection" as seen in Lisp, 
Smalltalk, Java, etc., without all the overhead of true garbage 
collection. 

It is fairly easy to do this using ILE heap support by creating a new 
ILE heap space for each unique "pool" (via the CRTHS MI instruction) 
then allocating from that specific heap, as needed (via the ALCHSS MI 
instruction). Then, when done with that "pool" of objects, issue the 
DESHS MI instruction for that  particular heap.

Cheers,

Mark S. Waterbury

> Simon Coulter wrote:
> Might be fine for a program that always runs in its own job, starts,  
> does some work, and ends, but for anything that may run frequently in  
> an interactive job or a long-running batch server type job this is  
> just asking for trouble. Memory leaks abound in code with plenty of  
> storage available and eventually they cause problems.
>
> I'm fairly sure you know that but there are others who might consider  
> this lazy way an acceptable one.
>
> Pointers are like farm gates: if you open (allocate) it you should  
> close (deallocate) it. If it's already open leave it open. With  
> pointers however, there are some sloppy functions that return an  
> allocated pointer and expect the caller to deallocate so it's more of  
> a guideline than a rule but the point (hardeharhar) is obvious.
>