From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tom Barbette Subject: Re: [PATCH v4] mlx5: Support for rte_eth_rx_queue_count Date: Mon, 5 Nov 2018 09:01:03 +0000 Message-ID: <1541408462940.66979@kth.se> References: <1540653055-67051-1-git-send-email-barbette@kth.se>, <1540976475938.69727@kth.se>, Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Cc: Yongseok Koh To: Shahaf Shuler , "dev@dpdk.org" , Ferruh Yigit , Thomas Monjalon , Andrew Rybchenko , "olivier.matz@6wind.com" Return-path: Received: from smtp-3.sys.kth.se (smtp-3.sys.kth.se [130.237.48.192]) by dpdk.org (Postfix) with ESMTP id 6844558F6 for ; Mon, 5 Nov 2018 10:01:09 +0100 (CET) In-Reply-To: Content-Language: fr-FR List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" =0A= > It will tell you if in a given offset on the rxq you have a packet ready.= I think it will fit your needs, see below.=0A= So we just loose in precision here. We're looking at ML techniques that wil= l play better with a numerical value and benefit from more entropy than arb= itrarily poking of some thresholds.=0A= If one had to remain, I'd say rx_descriptor_* should be removed as it can b= e deduced from queue_count. Moreover, in the MLX5 case, calling the functio= n 3 times to get "half busy", "partially busy" or "idle" will lead to scann= ing the queue 3 times...=0A= =0A= > But for the latency it is better to work w/ NIC host coherent clock + tim= estamps (like you implemented on a different patch).=0A= We're looking at both as you noticed. But timestamping is much more costly.= Hence latency is more for monitoring, reporting to the operator, while que= ue statistics are used for scheduling. It's harder to understand a latency = value as a higher latency may only mean that a bunch of packets were more c= omplex to handle. But we want to look at the relation between the two.=0A= =0A= I think this patch should go in (with maybe a solution for vectorized?). An= d when the removal of queue_count will be discussed I'll follow. But I'm su= re other people will jump in the discussion at that time.=0A= =0A= Thanks for your time anyway,=0A= Tom=