[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAAq0SUkdvA7SKsM_a5z+_Dx8wAEdyw2KKgE=VnyQ7Q-ocZHq4g@mail.gmail.com>
Date: Fri, 1 Aug 2025 10:05:50 -0300
From: Wander Lairson Costa <wander@...hat.com>
To: Peter Zijlstra <peterz@...radead.org>
Cc: Ingo Molnar <mingo@...hat.com>, Juri Lelli <juri.lelli@...hat.com>,
Vincent Guittot <vincent.guittot@...aro.org>, Dietmar Eggemann <dietmar.eggemann@....com>,
Steven Rostedt <rostedt@...dmis.org>, Ben Segall <bsegall@...gle.com>, Mel Gorman <mgorman@...e.de>,
Valentin Schneider <vschneid@...hat.com>, Masami Hiramatsu <mhiramat@...nel.org>,
Mathieu Desnoyers <mathieu.desnoyers@...icios.com>, David Woodhouse <dwmw@...zon.co.uk>,
Thomas Gleixner <tglx@...utronix.de>, Boqun Feng <boqun.feng@...il.com>,
open list <linux-kernel@...r.kernel.org>,
"open list:TRACING" <linux-trace-kernel@...r.kernel.org>,
Arnaldo Carvalho de Melo <acme@...nel.org>, Clark Williams <williams@...hat.com>,
Gabriele Monaco <gmonaco@...hat.com>
Subject: Re: [PATCH v3 2/2] tracing/preemptirq: Optimize preempt_disable/enable()
tracepoint overhead
On Tue, Jul 8, 2025 at 3:47 PM Peter Zijlstra <peterz@...radead.org> wrote:
>
Sorry for the late reply. A mix of PTO + higher priority backlog items.
> On Tue, Jul 08, 2025 at 10:09:45AM -0300, Wander Lairson Costa wrote:
> > On Mon, Jul 07, 2025 at 01:26:22PM +0200, Peter Zijlstra wrote:
> > > On Fri, Jul 04, 2025 at 02:07:43PM -0300, Wander Lairson Costa wrote:
> > > > +#if defined(CONFIG_DEBUG_PREEMPT) || defined(CONFIG_TRACE_PREEMPT_TOGGLE)
> > > > +#define preempt_count_dec_and_test() \
> > > > + ({ preempt_count_sub(1); should_resched(0); })
> > > > +#endif
> > >
> > > Also this is terrible. Surely you can do better.
> > >
> >
> > Thank you for pointing this out. I'm not sure I've fully understood the
> > concern here. My understanding was that this logic was pre-existing and
> > my patch only reorganized it.
> >
> > I'm clearly missing something. Could you please elaborate a bit on the
> > issue you've spotted?
>
> The normal (!DEBUG) case uses __preempt_count_dec_and_test(), which is
> significantly better.
>
Maybe I am missing something, but my understanding is that this behavior didn't
change. When DEBUG_PREEMPT and TRACE_PREEMPT_TOGGLE are not defined,
__preempt_count_dec_and_test() is used.
Powered by blists - more mailing lists