Date: Fri, 19 Jan 2007 10:02:19 -0800 From: Paul Allen <nospam@ugcs.caltech.edu> To: Jason Evans <jasone@freebsd.org> Cc: freebsd-current@freebsd.org, Stefan Ehmann <shoesoft@gmx.net> Subject: Re: very high memory usage in -current Message-ID: <20070119180219.GD8574@heave.ugcs.caltech.edu> In-Reply-To: <45B00BF5.6030200@FreeBSD.org> References: <200701172045.35137.shoesoft@gmx.net> <45B00BF5.6030200@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
>From Jason Evans <jasone@freebsd.org>, Thu, Jan 18, 2007 at 04:08:21PM -0800:
> Personally, in the absence of a dynamic boundary between the data
> segment and the heap, I would be quite happy to completely disable
> sbrk() support in jemalloc, and let those who really need that last 512
> MB of address space adjust resource limits for their applications as
> necessary. In practice, I expect this would cause people far less
> trouble than does the current state of affairs.
Well it might be reasonable to use a malloc flag. Nonetheless, it should
be possible for you to MADV_FREE brk memory without moving the brk point.
While munmap has some advantages in better coexisting with direct use of
mmap by a program, MADV_FREE has the advantage that virtual pages can be
immediately reused without incurring a syscall cost. It isn't obvious
a priori what the right balance of these operations is.
Paul
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070119180219.GD8574>