From mboxrd@z Thu Jan  1 00:00:00 1970
From: Stephen Hemminger <stephen@networkplumber.org>
Subject: Re: [PATCH 1/3] timer: add per-installer pending lists
 for each lcore
Date: Tue, 29 Aug 2017 08:11:12 -0700
Message-ID: <20170829081112.57b693a3@xeon-e3>
References: <1503499644-29432-1-git-send-email-erik.g.carrillo@intel.com>
 <1503499644-29432-2-git-send-email-erik.g.carrillo@intel.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Cc: rsanford@akamai.com, dev@dpdk.org
To: Gabriel Carrillo <erik.g.carrillo@intel.com>
Return-path: <dev-bounces@dpdk.org>
Received: from mail-pg0-f50.google.com (mail-pg0-f50.google.com [74.125.83.50])
 by dpdk.org (Postfix) with ESMTP id B3DE32BAC
 for <dev@dpdk.org>; Tue, 29 Aug 2017 17:11:15 +0200 (CEST)
Received: by mail-pg0-f50.google.com with SMTP id y15so11693823pgc.1
 for <dev@dpdk.org>; Tue, 29 Aug 2017 08:11:15 -0700 (PDT)
In-Reply-To: <1503499644-29432-2-git-send-email-erik.g.carrillo@intel.com>
List-Id: DPDK patches and discussions <dev.dpdk.org>
List-Unsubscribe: <http://dpdk.org/ml/options/dev>,
 <mailto:dev-request@dpdk.org?subject=unsubscribe>
List-Archive: <http://dpdk.org/ml/archives/dev/>
List-Post: <mailto:dev@dpdk.org>
List-Help: <mailto:dev-request@dpdk.org?subject=help>
List-Subscribe: <http://dpdk.org/ml/listinfo/dev>,
 <mailto:dev-request@dpdk.org?subject=subscribe>
Errors-To: dev-bounces@dpdk.org
Sender: "dev" <dev-bounces@dpdk.org>

On Wed, 23 Aug 2017 09:47:22 -0500
Gabriel Carrillo <erik.g.carrillo@intel.com> wrote:

>  	__TIMER_STAT_ADD(manage, 1);
> -	/* optimize for the case where per-cpu list is empty */
> -	if (priv_timer[lcore_id].pending_head.sl_next[0] == NULL)
> -		return;
> -	cur_time = rte_get_timer_cycles();
> +	for (i = 0, installer_lcore = enabled_lcores[i]; i < n_enabled_lcores;
> +	     installer_lcore = enabled_lcores[++i]) {
> +		list = &priv_tim->pending_lists[installer_lcore];
> +
> +		/* optimize for the case where list is empty */
> +		if (list->head.sl_next[0] == NULL)
> +			continue;
> +		cur_time = rte_get_timer_cycles();

This code is critical. We expect applications using timers to call timer_manage
very often and the case of no timers present, and no timer due must consume as
few cycles as possible

This change will add significant performance delays to these applications.