Mail Archives: djgpp/1998/02/05/05:40:52
On Wed, 4 Feb 1998 GAMMELJL AT SLU DOT EDU wrote:
> // For zsize3=2^n, the actual available memory drops to 1/2
> //of the available memory reported by go32-v2.
> // For zsize3=(2^n) - 1, the actual available memory is close
> //to the available memory reported by go32-v2.
This is a known misfeature of the current `malloc': it adds a 4-byte
overhead to the requested size and then rounds up the result to the
nearest integral power of 2. The next DJGPP release will have a
different implementation of `malloc'. You can find several
experimental versions on DJ Delorie's site,
http://www.delorie.com/djgpp/. Or you can use some alternate
implementation of `malloc' that is available from the net. Check the
DJGPP mail archives for some pointers.
- Raw text -