Re: g_malloc overhead

["Emmanuel Rodriguez" <emmanuel rodriguez gmail com>]

On Tue, Jan 20, 2009 at 12:48 PM, Larry Reaves <larry yrral net> wrote:
> On Tue, 2009-01-20 at 09:01 +0100, MartÃn Vales wrote:
> > BJÃrn Lindqvist escribiÃ:
> > > Actually, a custom allocator could be useful even in the general case.
> > > Malloc is a system call and has quite bad performance on certain
> > > platforms (windows in particular i think). Something like the gslice
> > > allocator could
> > > Probably improve performance a bit.
> > gslice i believe use malloc internally. I believe you always need
> > malloc/new-(C/C++) because you depend on ms Windows API.
> >
> > I am not sure if you can build your own malloc because you depend on the
> > operating system.
> Sure, you must malloc to get new memory, but you can malloc bigger than
> what you need and hand out the extra memory later at a much lower cost.
I recall reading somewhere that mmap can be used to build custom
memory allocators. If that's true than one can bypass malloc. I think
that you can request memory through mmap by using MAP_ANONYMOUS.

Emmanuel Rodriguez