Date: Fri, 22 May 2015 15:42:42 +0200
From: Peter Zijlstra <peterz@infradead.org>
To: mingo@kernel.org
Cc: vincent.weaver@maine.edu, eranian@google.com, jolsa@redhat.com,
        kan.liang@intel.com, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 02/11] perf/x86: Improve HT workaround GP counter
 constraint
Message-ID: <20150522134242.GH3644@twins.programming.kicks-ass.net>
References: <20150522132905.416122812@infradead.org>
 <20150522133135.447912500@infradead.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20150522133135.447912500@infradead.org>
User-Agent: Mutt/1.5.21 (2012-12-30)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1359
Lines: 46

On Fri, May 22, 2015 at 03:29:07PM +0200, Peter Zijlstra wrote:
> @@ -696,11 +699,16 @@ static bool __perf_sched_find_counter(st
>  				goto done;
>  		}
>  	}
> +
>  	/* Grab the first unused counter starting with idx */
>  	idx = sched->state.counter;
>  	for_each_set_bit_from(idx, c->idxmsk, INTEL_PMC_IDX_FIXED) {
> +		if (!__test_and_set_bit(idx, sched->state.used)) {
> +			if (sched->state.nr_gp++ >= sched->max_gp)
> +				return false;

Note the placement inside the success path of the GP allocation, instead
of the attempt a GP place we had before.

> +
>  			goto done;
> +		}
>  	}
>  
>  	return false;


> @@ -2000,6 +1999,11 @@ intel_get_excl_constraints(struct cpu_hw
>  	 * across HT threads
>  	 */
>  	is_excl = c->flags & PERF_X86_EVENT_EXCL;
> +	if (is_excl && !(event->hw.flags & PERF_X86_EVENT_EXCL_ACCT)) {
> +		event->hw.flags |= PERF_X86_EVENT_EXCL_ACCT;
> +		if (!cpuc->n_excl++)
> +			WRITE_ONCE(excl_cntrs->has_exclusive[tid], 1);
> +	}
>  
>  	/*
>  	 * xl = state of current HT

And that is what keeps repeated get_event_constraints() calls from ever
increasing our n_excl count.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/