From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ophir Munk Subject: Re: [PATCH v2 4/7] net/mlx4: merge Tx path functions Date: Tue, 24 Oct 2017 20:36:52 +0000 Message-ID: References: <1508752838-30408-1-git-send-email-ophirmu@mellanox.com> <1508768520-4810-1-git-send-email-ophirmu@mellanox.com> <1508768520-4810-5-git-send-email-ophirmu@mellanox.com> <20171024135149.fyg4nzcbygo2amtz@laranjeiro-vm> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Cc: Adrien Mazarguil , "dev@dpdk.org" , Thomas Monjalon , Olga Shern , Matan Azrad To: =?iso-8859-1?Q?N=E9lio_Laranjeiro?= Return-path: Received: from EUR03-AM5-obe.outbound.protection.outlook.com (mail-eopbgr30078.outbound.protection.outlook.com [40.107.3.78]) by dpdk.org (Postfix) with ESMTP id D280E1B7F7 for ; Tue, 24 Oct 2017 22:36:54 +0200 (CEST) In-Reply-To: <20171024135149.fyg4nzcbygo2amtz@laranjeiro-vm> Content-Language: en-US List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" Hi, On Tuesday, October 24, 2017 4:52 PM, N=E9lio Laranjeiro wrote: >=20 > On Mon, Oct 23, 2017 at 02:21:57PM +0000, Ophir Munk wrote: > > From: Matan Azrad > > > > Merge tx_burst and mlx4_post_send functions to prevent double asking > > about WQ remain space. > > > > This should improve performance. > > > > Signed-off-by: Matan Azrad > > --- > > drivers/net/mlx4/mlx4_rxtx.c | 353 > > +++++++++++++++++++++---------------------- > > 1 file changed, 170 insertions(+), 183 deletions(-) >=20 > What are the real expectation you have on the remaining patches of the > series? >=20 > According to the comment of this commit log "This should improve > performance" there are too many barriers at each packet/segment level to > improve something. >=20 > The point is, mlx4_burst_tx() should write all the WQE without any barrie= r as > it is processing a burst of packets (whereas Verbs functions which may on= ly > process a single packet). =20 > The lonely barrier which should be present is the > one to ensure that all the host memory is flushed before triggering the T= x > doorbell. >=20 There is a known ConnectX-3 HW limitation: the first 4 bytes of every TXWBB= (64 bytes chunks) should be=20 written in a reversed order (from last TXWBB to first TXWBB).=20 The last 60 bytes of any TXWBB can be written in any order (before writing = the first 4 bytes). Is your last statement (using lonely barrier) is in accordance with this li= mitation? Please explain. > There is also too many cases handled which are useless in bursts situatio= n, > this function needs to be re-written to its minimal use case i.e. proces= sing a > valid burst of packets/segments and triggering at the end of the burst th= e Tx > doorbell. >=20 > Regards, >=20 > -- > N=E9lio Laranjeiro > 6WIND