From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Monjalon Subject: Re: xstats performance Date: Wed, 29 Jun 2016 18:40:08 +0200 Message-ID: <8870166.MVVViF2IlQ@xps13> References: <5773EB79.1090509@6wind.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7Bit Cc: dev@dpdk.org, Olivier MATZ To: Remy Horton Return-path: Received: from mail-wm0-f42.google.com (mail-wm0-f42.google.com [74.125.82.42]) by dpdk.org (Postfix) with ESMTP id 1ADBD2B84 for ; Wed, 29 Jun 2016 18:40:14 +0200 (CEST) Received: by mail-wm0-f42.google.com with SMTP id f126so189018614wma.1 for ; Wed, 29 Jun 2016 09:40:14 -0700 (PDT) In-Reply-To: List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" 2016-06-29 17:03, Remy Horton: > On 29/06/2016 16:38, Olivier MATZ wrote: > > And assume that the id field in rte_eth_xstat corresponds to > > the index in the rte_eth_xstat_name table? > > It was an assumption I wanted to avoid setting in stone, even though at > the moment it is actually true in implementation. > > The mappings are driver-specific but I wanted to leave open, amoung > other things, the possibility of string-id mappings being standardised > across all drivers. For that to work drivers would have to be able to > use a set of ids that may contain gaps. I don't think it is possible to standardize stats ids, for two reasons: - it is hard to maintain and avoid conflicts between drivers - the drivers would have to lookup the names which degrades performance I think the idea of Olivier would improve the performance of stats retrieval, which was the idea of this rework :) Unfortunately we need someone available to fix it quickly for RC2.