xfs
[Top] [All Lists]

Re: [Patch] Cacheline align xlog_t

To: Andi Kleen <andi@xxxxxxxxxxxxxx>
Subject: Re: [Patch] Cacheline align xlog_t
From: Timothy Shimmin <tes@xxxxxxx>
Date: Fri, 04 Apr 2008 12:02:06 +1100
Cc: David Chinner <dgc@xxxxxxx>, Lachlan McIlroy <lachlan@xxxxxxx>, xfs-dev <xfs-dev@xxxxxxx>, xfs-oss <xfs@xxxxxxxxxxx>
In-reply-to: <20080403064608.GS29105@one.firstfloor.org>
References: <20080401231552.GV103491721@sgi.com> <47F3293C.6090708@sgi.com> <20080402054403.GF103491721@sgi.com> <87myocek4o.fsf@basil.nowhere.org> <20080402222347.GK103491721@sgi.com> <20080403064608.GS29105@one.firstfloor.org>
Sender: xfs-bounce@xxxxxxxxxxx
User-agent: Thunderbird 2.0.0.12 (Macintosh/20080213)
Andi Kleen wrote:
On Thu, Apr 03, 2008 at 08:23:47AM +1000, David Chinner wrote:
For the dynamic allocation you would rather need to make sure it
starts at a cache line boundary explicitely because the allocator doesn't
know the alignment of the target type, otherwise your careful
padding might be useless.

There isn't one, right?

You can always align yourself with kmalloc (or any other arbitary size allocator) with the standard technique:
get L1_CACHE_BYTES-1 or possibly better cache_line_size() - 1 bytes
more and then align the pointer manually with ALIGN. Only tricky part
is that you have to undo the alignment before freeing.


Yeah, how do you know how much was offset for the alignment
to go back for the freeing. I guess you could also keep the original pointer
or the offset handy (a pain). Or can the allocator help out?

--Tim



<Prev in Thread] Current Thread [Next in Thread>