From mboxrd@z Thu Jan 1 00:00:00 1970 Content-Type: multipart/mixed; boundary="===============4511649497829526496==" MIME-Version: 1.0 From: Ye Xiaolong To: lkp@lists.01.org Subject: Re: [lkp-robot] [mm, vmscan] 5e56dfbd83: fsmark.files_per_sec -11.1% regression Date: Tue, 21 Feb 2017 09:10:09 +0800 Message-ID: <20170221011009.GB10024@yexl-desktop> In-Reply-To: <20170220101942.GE2431@dhcp22.suse.cz> List-Id: --===============4511649497829526496== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable On 02/20, Michal Hocko wrote: >On Mon 20-02-17 09:43:29, Ye Xiaolong wrote: >> On 02/19, Michal Hocko wrote: >> >On Mon 13-02-17 14:46:37, Ye Xiaolong wrote: >> >> On 02/07, Michal Hocko wrote: >> >> >On Tue 07-02-17 10:22:13, Ye Xiaolong wrote: >> >> > >> >> [snip] >> >> = >> >> >Could you retest with a single NUMA node? I am not familiar with the >> >> >benchmark enough to judge it was set up properly for a NUMA machine. >> >> = >> >> Sorry for the late, I'm not quite familiar with NUMA, could you tell = me how >> >> to configure a machine to use a single NUMA node? through cmdline? >> > >> >I am not aware of any parameter which would disble numa via command >> >line. You can do that when starting qemu, which I assume you are using. >> >Just configure a single memory node (-numa parameter). >> = >> Actually we were doing the test on a physical machine, not a guest. >> Any kernel config will do? > >You can disbale CONFIG_NUMA. But a much better approach would be >limitting the test to a single numa node via numactl. Thanks, numactl is supported in 0day system, I'll try it. > >-- = >Michal Hocko >SUSE Labs --===============4511649497829526496==--