From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Burakov, Anatoly" Subject: Re: [RFC v2 00/23] Dynamic memory allocation for DPDK Date: Wed, 14 Feb 2018 09:32:12 +0000 Message-ID: References: <1513892309.2658.80.camel@intel.com> <1514308764.2658.93.camel@intel.com> <20180202192832.GA42096@yongseok-MBP.local> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit Cc: "Walker, Benjamin" , "dev@dpdk.org" , Thomas Monjalon , "andras.kovacs@ericsson.com" , "Wiles, Keith" , "Richardson, Bruce" , =?UTF-8?Q?N=c3=a9lio_Laranjeiro?= , Shahaf Shuler , "Xueming(Steven) Li" To: Yongseok Koh Return-path: Received: from mga07.intel.com (mga07.intel.com [134.134.136.100]) by dpdk.org (Postfix) with ESMTP id 9CA871B2F7 for ; Wed, 14 Feb 2018 10:32:18 +0100 (CET) In-Reply-To: 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" On 14-Feb-18 2:01 AM, Yongseok Koh wrote: > >> On Feb 5, 2018, at 2:03 AM, Burakov, Anatoly wrote: >> >> Thanks for your feedback, good to hear we're on the right track. I already have a prototype implementation of this working, due for v1 submission :) > > Anatoly, > > One more suggestion. Currently, when populating mempool, there's a chance to > have multiple chunks if system memory is highly fragmented. However, with your > new design, it is unlikely to happen unless the system is really low on memory. > Allocation will be dynamic and page by page. With your v2, you seemed to make > minimal changes on mempool. If allocation fails, it will still try to gather > fragments from malloc_heap until it acquires enough objects and the resultant > mempool will have multiple chunks. But like I mentioned, it is very unlikely and > this will only happen when the system is short of memory. Is my understanding > correct? > > If so, how about making a change to drop the case where mempool has multiple > chunks? > > Thanks > Yongseok > Hi Yongseok, I would still like to keep it, as it may impact low memory cases such as containers. -- Thanks, Anatoly