linux-kernel - Re: [RFC PATCH 0/6] augmented rbtree changes

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20120724014611.GA6974@google.com>
Date:	Mon, 23 Jul 2012 18:46:11 -0700
From:	Michel Lespinasse <walken@...gle.com>
To:	riel@...hat.com, peterz@...radead.org, daniel.santos@...ox.com,
	aarcange@...hat.com, dwmw2@...radead.org, akpm@...ux-foundation.org
Cc:	linux-mm@...ck.org, linux-kernel@...r.kernel.org
Subject: Re: [RFC PATCH 0/6] augmented rbtree changes

On Fri, Jul 20, 2012 at 05:31:01AM -0700, Michel Lespinasse wrote:
> Patch 5 speeds up the augmented rbtree erase. Here again we use a tree
> rotation callback during rebalancing; however we also have to propagate
> the augmented node information above nodes being erased and/or stitched,
> and I haven't found a nice enough way to do that. So for now I am proposing
> the simple-stupid way of propagating all the way to the root. More on
> this later.

So, I looked at it again and finally figured out a decent way to avoid
unnecessary propagation here. Going to resend patches 5/6 as replies to
their original postings.

> - The prio tree of all VMAs mapping a given file (struct address_space)
> could be switched to an augmented rbtree based interval tree (thus removing
> the prio tree library in favor of augmented rbtrees)

I actually have a prototype for that already. The augmented rbtree based
implementation is slightly faster than prio tree on insert/erase, and
considerably faster on lookups. However, this is with a synthetic test
exercising prio and rbtrees directly, not with a realistic workload going
through the MM layers. Do we know of situations where prio tree performance
is currently a concern ?

> As they stand, patches 3-6 don't seem to make a difference for basic rbtree
> support, and they improve my augmented rbtree insertion/erase benchmark
> by a factor of ~2.1 to ~2.3 depending on test machines.

After rewriting patches 5-6 as discussed above, augmented rbtrees are now
~2.5 - ~2.7 times faster than before this patch series.

-- 
Michel "Walken" Lespinasse
A program is never fully debugged until the last user dies.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/