Message-ID: <20090430063306.GA27431@Krystal>
Date:	Thu, 30 Apr 2009 02:33:06 -0400
From:	Mathieu Desnoyers <compudj@...stal.dyndns.org>
To:	Ingo Molnar <mingo@...e.hu>
Cc:	Nick Piggin <nickpiggin@...oo.com.au>,
	Peter Zijlstra <a.p.zijlstra@...llo.nl>,
	Yuriy Lalym <ylalym@...il.com>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	ltt-dev@...ts.casi.polymtl.ca,
	Andrew Morton <akpm@...ux-foundation.org>, thomas.pi@...or.dea,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	Christoph Lameter <cl@...ux.com>
Subject: Re: [ltt-dev] [PATCH] Fix dirty page accounting in
	redirty_page_for_writepage()

* Ingo Molnar (mingo@...e.hu) wrote:
> 
> * Mathieu Desnoyers <mathieu.desnoyers@...ymtl.ca> wrote:
> 
> > And thanks for the review! This excercise only convinced me that 
> > the kernel memory accounting works as expected. All this gave me 
> > the chance to have a good look at the memory accounting code. We 
> > could probably benefit of Christoph Lameter's cpu ops (using 
> > segment registers to address per-cpu variables with atomic 
> > inc/dec) in there. Or at least removing interrupt disabling by 
> > using preempt disable and local_t variables for the per-cpu 
> > counters could bring some benefit.
> 
> Note, optimized per cpu ops are already implemented upstream, by 
> Tejun Heo's percpu patches in .30:
> 
>  #define percpu_read(var)	percpu_from_op("mov", per_cpu__##var)
>  #define percpu_write(var, val)	percpu_to_op("mov", per_cpu__##var, val)
>  #define percpu_add(var, val)	percpu_to_op("add", per_cpu__##var, val)
>  #define percpu_sub(var, val)	percpu_to_op("sub", per_cpu__##var, val)
>  #define percpu_and(var, val)	percpu_to_op("and", per_cpu__##var, val)
>  #define percpu_or(var, val)	percpu_to_op("or", per_cpu__##var, val)
>  #define percpu_xor(var, val)	percpu_to_op("xor", per_cpu__##var, val)
> 
> See:
> 
>   6dbde35: percpu: add optimized generic percpu accessors
> 
> From the changelog:
> 
>     [...]
>     The advantage is that for example to read a local percpu variable,
>     instead of this sequence:
>     
>      return __get_cpu_var(var);
>     
>      ffffffff8102ca2b:  48 8b 14 fd 80 09 74    mov    -0x7e8bf680(,%rdi,8),%rdx
>      ffffffff8102ca32:  81
>      ffffffff8102ca33:  48 c7 c0 d8 59 00 00    mov    $0x59d8,%rax
>      ffffffff8102ca3a:  48 8b 04 10             mov    (%rax,%rdx,1),%rax
>     
>     We can get a single instruction by using the optimized variants:
>     
>      return percpu_read(var);
>     
>      ffffffff8102ca3f:  65 48 8b 05 91 8f fd    mov    %gs:0x7efd8f91(%rip),%rax
>     [...]
> 
> So if you want to make use of it, percpu_add()/percpu_sub() would be 
> the place to start.
> 

Great !

I see however that it's only guaranteed to be atomic wrt preemption.
What would be even better would be to have the atomic ops wrt local irqs
(as local.h does) available in this percpu flavor. By doing this, we
could have interrupt and nmi-safe per-cpu counters, without even the
need to disable preemption.

In terms of counters, except maybe for tri-values for some
architectures, I don't see how we could manage synchronization in a
better way.

Mathieu

> 	Ingo
> 
> _______________________________________________
> ltt-dev mailing list
> ltt-dev@...ts.casi.polymtl.ca
> http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
> 

-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68
--
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/

Powered by blists - more mailing lists