From mboxrd@z Thu Jan  1 00:00:00 1970
From: Stephen Hemminger <stephen@networkplumber.org>
Subject: Re: [PATCH v8 2/3] lib/gro: add TCP/IPv4 GRO support
Date: Thu, 29 Jun 2017 10:51:50 -0700
Message-ID: <20170629105150.74b8bf88@xeon-e3>
References: <1498459430-116048-1-git-send-email-jiayu.hu@intel.com>
 <1498733940-117800-1-git-send-email-jiayu.hu@intel.com>
 <1498733940-117800-3-git-send-email-jiayu.hu@intel.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Cc: dev@dpdk.org, konstantin.ananyev@intel.com, jianfeng.tan@intel.com,
 yliu@fridaylinux.org, jingjing.wu@intel.com, tiwei.bie@intel.com,
 lei.a.yao@intel.com
To: Jiayu Hu <jiayu.hu@intel.com>
Return-path: <dev-bounces@dpdk.org>
Received: from mail-pg0-f53.google.com (mail-pg0-f53.google.com [74.125.83.53])
 by dpdk.org (Postfix) with ESMTP id DDEA658EC
 for <dev@dpdk.org>; Thu, 29 Jun 2017 19:51:53 +0200 (CEST)
Received: by mail-pg0-f53.google.com with SMTP id f127so51365724pgc.0
 for <dev@dpdk.org>; Thu, 29 Jun 2017 10:51:53 -0700 (PDT)
In-Reply-To: <1498733940-117800-3-git-send-email-jiayu.hu@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 Thu, 29 Jun 2017 18:58:59 +0800
Jiayu Hu <jiayu.hu@intel.com> wrote:

> +	/* allocate a reassembly table for TCP/IPv4 GRO */
> +	uint32_t tcp_item_num = RTE_MIN(item_num,
> +			RTE_GRO_MAX_BURST_ITEM_NUM);
> +	struct gro_tcp_tbl tcp_tbl;
> +	struct gro_tcp_key tcp_keys[tcp_item_num];
> +	struct gro_tcp_item tcp_items[tcp_item_num];

Variable size structures on stack were not supported by
CLANG, last time I checked. Why not always max size?

> +
> +	struct rte_mbuf *unprocess_pkts[nb_pkts];
> +	uint16_t unprocess_num = 0;
> +	int32_t ret;
> +	uint64_t current_time;
> +
> +	if ((param->desired_gro_types & RTE_GRO_TCP_IPV4) == 0)
> +		return nb_pkts;
> +
> +	memset(tcp_keys, 0, sizeof(struct gro_tcp_key) *
> +			tcp_item_num);
> +	memset(tcp_items, 0, sizeof(struct gro_tcp_item) *
> +			tcp_item_num);

Variable size memset's are slow. They generate 'rep; stoz' last
I checked in GCC and because of rep instruction it kills multi-execution
pipeline.