[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20170726071609.GN20323@X58A-UD3R>
Date: Wed, 26 Jul 2017 16:16:09 +0900
From: Byungchul Park <byungchul.park@....com>
To: Peter Zijlstra <peterz@...radead.org>
Cc: mingo@...nel.org, tglx@...utronix.de, walken@...gle.com,
boqun.feng@...il.com, kirill@...temov.name,
linux-kernel@...r.kernel.org, linux-mm@...ck.org,
akpm@...ux-foundation.org, willy@...radead.org, npiggin@...il.com,
kernel-team@....com
Subject: Re: [PATCH v7 08/16] lockdep: Avoid adding redundant direct links of
crosslocks
On Tue, Jul 25, 2017 at 05:41:36PM +0200, Peter Zijlstra wrote:
> On Wed, May 24, 2017 at 05:59:41PM +0900, Byungchul Park wrote:
> > We can skip adding a dependency 'AX -> B', in case that we ensure 'AX ->
> > the previous of B in hlocks' to be created, where AX is a crosslock and
> > B is a typical lock. Remember that two adjacent locks in hlocks generate
> > a dependency like 'prev -> next', that is, 'the previous of B in hlocks
> > -> B' in this case.
> >
> > For example:
> >
> > in hlocks[]
> > ------------
> > ^ A (gen_id: 4) --+
> > | | previous gen_id
> > | B (gen_id: 3) <-+
> > | C (gen_id: 3)
> > | D (gen_id: 2)
> > oldest | E (gen_id: 1)
> >
> > in xhlocks[]
> > ------------
> > ^ A (gen_id: 4, prev_gen_id: 3(B's gen id))
> > | B (gen_id: 3, prev_gen_id: 3(C's gen id))
> > | C (gen_id: 3, prev_gen_id: 2(D's gen id))
> > | D (gen_id: 2, prev_gen_id: 1(E's gen id))
> > oldest | E (gen_id: 1, prev_gen_id: NA)
> >
> > On commit for a crosslock AX(gen_id = 3), it's engough to add 'AX -> C',
> > but adding 'AX -> B' and 'AX -> A' is unnecessary since 'AX -> C', 'C ->
> > B' and 'B -> A' cover them, which are guaranteed to be generated.
> >
> > This patch intoduces a variable, prev_gen_id, to avoid adding this kind
> > of redundant dependencies. In other words, the previous in hlocks will
> > anyway handle it if the previous's gen_id >= the crosslock's gen_id.
> >
>
> Didn't we talk about an alternative to this?
Yes, we did. You said the optimazation was unnecessary, and I was not
sure if it's true, so added it at this time.
But *I will exclude this from next spin*.
>
> /me goes dig
>
> https://lkml.kernel.org/r/20170303091338.GH6536@twins.programming.kicks-ass.net
>
> There and replies.
>
> So how much does this save vs avoiding redundant links?
No different on my qemu machine. The answer was:
https://lkml.org/lkml/2017/3/14/103
Powered by blists - more mailing lists