From mboxrd@z Thu Jan  1 00:00:00 1970
From: =?iso-8859-1?Q?N=E9lio?= Laranjeiro <nelio.laranjeiro@6wind.com>
Subject: Re: [PATCH 2/5] net/mlx5: free buffers in bulk on Tx
	completion
Date: Fri, 30 Jun 2017 14:30:47 +0200
Message-ID: <20170630123047.GO18305@autoinstall.dev.6wind.com>
References: <20170628230403.10142-1-yskoh@mellanox.com>
 <20170628230403.10142-3-yskoh@mellanox.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
Cc: ferruh.yigit@intel.com, dev@dpdk.org, adrien.mazarguil@6wind.com
To: Yongseok Koh <yskoh@mellanox.com>
Return-path: <dev-bounces@dpdk.org>
Received: from mail-wr0-f181.google.com (mail-wr0-f181.google.com
 [209.85.128.181]) by dpdk.org (Postfix) with ESMTP id 096155323
 for <dev@dpdk.org>; Fri, 30 Jun 2017 14:30:56 +0200 (CEST)
Received: by mail-wr0-f181.google.com with SMTP id k67so203003322wrc.2
 for <dev@dpdk.org>; Fri, 30 Jun 2017 05:30:56 -0700 (PDT)
Content-Disposition: inline
In-Reply-To: <20170628230403.10142-3-yskoh@mellanox.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, Jun 28, 2017 at 04:04:00PM -0700, Yongseok Koh wrote:
> When processing Tx completion, it is more efficient to free buffers in bulk
> using rte_mempool_put_bulk() if buffers are from a same mempool.
> 
> Signed-off-by: Yongseok Koh <yskoh@mellanox.com>
> ---
>  drivers/net/mlx5/mlx5_rxtx.c | 36 +++++++++++++++++++++++++++---------
>  1 file changed, 27 insertions(+), 9 deletions(-)
> 
> diff --git a/drivers/net/mlx5/mlx5_rxtx.c b/drivers/net/mlx5/mlx5_rxtx.c
> index 43db06ad8..d81d630f7 100644
> --- a/drivers/net/mlx5/mlx5_rxtx.c
> +++ b/drivers/net/mlx5/mlx5_rxtx.c
> @@ -264,6 +264,8 @@ txq_complete(struct txq *txq)
>  	uint16_t cq_ci = txq->cq_ci;
>  	volatile struct mlx5_cqe *cqe = NULL;
>  	volatile struct mlx5_wqe_ctrl *ctrl;
> +	struct rte_mbuf *m, *free[elts_n];
> +	unsigned int blk_n = 0;
>  
>  	do {
>  		volatile struct mlx5_cqe *tmp;
> @@ -296,21 +298,37 @@ txq_complete(struct txq *txq)
>  	assert((elts_tail & elts_m) < (1 << txq->wqe_n));
>  	/* Free buffers. */
>  	while (elts_free != elts_tail) {
> -		struct rte_mbuf *elt = (*txq->elts)[elts_free & elts_m];
> -		struct rte_mbuf *elt_next =
> -			(*txq->elts)[(elts_free + 1) & elts_m];
> -
> +		m = rte_pktmbuf_prefree_seg((*txq->elts)[elts_free++ & elts_m]);
> +		if (likely(m != NULL)) {
> +			if (blk_n) {
> +				if (likely(m->pool == free[0]->pool)) {
> +					free[blk_n++] = m;
> +				} else {
> +					rte_mempool_put_bulk(
> +							free[0]->pool,
> +							(void *)free,
> +							blk_n);

The indentation is strange here, free[0] should be on the same line as
rte_mempool_put_bulk.

> +					free[0] = m;
> +					blk_n = 1;
> +				}
> +			} else {
> +				free[0] = m;
> +				blk_n = 1;
> +			}
> +		}

This loop could be smaller, blk_n can only be equal to 0 in the first
iteration, otherwise is >= 1.
The first if statement can be merged with the second one: 

	if (likely(m != NULL)) {
		if (likely(blk_n && m->pool == free[0]->pool)) {
		...
		} else {
		...
		}

Thanks,

-- 
Nélio Laranjeiro
6WIND