Date: Sat, 18 Jun 2005 00:13:52 +0100 From: Peter Edwards <peadar.edwards@gmail.com> To: Poul-Henning Kamp <phk@phk.freebsd.dk> Cc: Peter Edwards <peadar@freebsd.org>, current@freebsd.org Subject: Re: Towards a working "wine". [long] Message-ID: <34cb7c8405061716139488699@mail.gmail.com> In-Reply-To: <58747.1119044361@critter.freebsd.dk> References: <20050617180232.GA25818@freefall.freebsd.org> <58747.1119044361@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On 6/17/05, Poul-Henning Kamp <phk@phk.freebsd.dk> wrote: > phkmalloc works just fine with a fragmented heap, but allocates too > much memory for the page-map if all the memory is too far away > from "_end". >=20 > The correct (and portable) fix is to give phkmalloc a treee-structure > instead of a linear array to manage the page table. >=20 Sorry, I didn't mean to imply that it was just never going to work, just the structure of the mapping from addresses to the page table: the god-awful hack was the quickest way to test that things would work better in wine. I'll get some decent patches done RSN, but I'd imagine there'll be a minor amount of overhead involved. (eg, ptr2index would never be as simple as it is now) The change would also affect the behaviour of applications in terms of how MAXDSIZ impacted them (it'd be pretty much meaningless for a non-brk-based malloc, but that's not neccessarily a bad thing)
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?34cb7c8405061716139488699>