Date: Mon, 12 Oct 1998 14:16:31 +0200
From: Hans-Bernhard Broeker <broeker AT physik DOT rwth-aachen DOT de>
Message-Id: <199810121216.OAA13359@acp3bf.physik.rwth-aachen.de>
To: djgpp AT delorie DOT com
Cc: heckler AT mat DOT ufrgs DOT br
Subject: Re: Memory allocation
Newsgroups: comp.os.msdos.djgpp
Organization: RWTH Aachen, III. physikalisches Institut B
X-Newsreader: TIN [version 1.2 PL2]
Reply-To: djgpp AT delorie DOT com

In article <361E6945 DOT 5FCA5FAC AT tecnotron DOT com DOT br> you wrote:
> Hi there!  the following program allocate lots of memory
> and free it.  Then, try to allocate again, but with a diferent
> block size.  It dosn't work ! -> first part runs OK, but the
> second gets a "virtual memory exceed" error !

The phenomenon you're observed is called 'heap fragmentation'. By allocating
all the available memory in a certain block size, you limit yourself to
that particular block size for the rest of your program's lifetime.

[...]

> I need to allocate lots of memory, in blocks with diferent sizes,
> and at diferent times.  under some situations, the operations of
> "new" and "delete" cause the DPMI server to block the interrupts
> for a long time (more than 5ms), and it create problems for a
> hardware I drive on the program.

I'm not sure such long delays should be happening if you're using a
proper DPMI server (i.e. *not* Winblows 9x) and have disabled
swapping.

> Cloud: it seems for me that I need to reuse the previously
> allocated blocks in the same block-size I did first, otherwise
> I will receive a "virtual memory exceed in new" ...

> ps.: I really don't want to "reinvent the weel", i.e., made
> my own memory managing.

You may not want to, but your set of requirements blocks all other
possibilities, I suspect.

--
Hans-Bernhard Broeker (broeker AT physik DOT rwth-aachen DOT de)
Even if all the snow were burnt, ashes would remain.