RE: [MiNT] memory fragmentation

[Jo-Even.Skarstein@gjensidige.no]

> -----Original Message-----
> From:	Guido Flohr [SMTP:gufl0000@stud.uni-sb.de]
> Sent:	Wednesday, August 18, 1999 5:09 PM
> To:	mint@stud.uni-sb.de
> Subject:	Re: [MiNT] memory fragmentation
> 
> Anyway, Thomas has already implemented it and it is worth a try.  Let's
> have a look at the result.  There are two aspect's of MiNT that Tannenbaum
> 
Yes, only (controlled) experiments can answer this...

> does not talk about:
> 
> - speed: of course finding the "right" block will be slower with best-fit.
> 
I don't think this is an issue at all under MiNT. Even if we have 200
fragments, you'll never notice this speed-wise.

> - best-fit will produce smaller holes.  Uhm, yes, that's the intention of
> it.  OK, I know what you/Tannenbaum mean.  But with little memory I am
> more interested in one large memory block left free than in the number or
> 
That's a good point, but I'm not sure how valid it is if you have a lot of
memory.

> size of several tiny wasted blocks.  Furthermore, most Unix software
> tends to malloc a lot of small and smallest blocks.  I think that these
> tiny holes will not be wasted.
> 
Like Thomas Binder already pointed out, most of these mallocs from their own
heap which might be quite large. But as I said, only experiments can tell us
whether this is a good idea or not. Surely it must have been tested in
MagiC?

Jo Even Skarstein