malloc changes

[Kevin L. Kane]

Looking into this now, hopefully ill have a updated patch soonish.

-Kevin

On 2/3/06, David Rhodus <drhodus at xxxxxxxxxxx> wrote:
> On 2/3/06, Matthew Dillon <dillon at xxxxxxxxxxxxxxxxxxxx> wrote:
> >
> > :
> > :FYI,
> > :
> > :FreeBSD just imported a new malloc to replace the aged
> > :phkmalloc.
> > :
> > :There was much back and forth with the author and David Xu
> > :so he might have an opinion as to whether the changes in the
> > :FreeBSD malloc are worth looking into.
> > :
> > :The new malloc is not derived from the phkmalloc so likely
> > :these changes would not apply.  Many of the features are
> > :similar, except the new jemalloc was engineered with true
> > :concurrent threaded applications in mind.
> >
> >    I've been following that thread in the FreeBSD groups.
> >    We will definitely *NOT* going to be importing FreeBSD's
> >    new malloc.
> >
> >    My feeling is that if a threaded program allocates so much
> >    data that a fast MP-safe malloc is required, then we should
> >    simply port the slab allocator from the DragonFly kernel to
> >    userland and use that.
> >
> >    For now its not on the table.  The current proposed malloc
> >    changes for DragonFly IS on the table and I do want to see
> >    it committed, but I'd appreciate it if some GUI users could
> >    test it out with e.g. mozilla, X, and so forth, first.
> >
> >                                         -Matt
>
> I haven't been able to get a pkgsrc bulk build to complete using the
> malloc patch.
>
> -DR
>
>