Content-Type: multipart/mixed; boundary="===============8908514102076463225==" MIME-Version: 1.0 From: Mel Gorman To: lkp@lists.01.org Subject: Re: [mm] 3484b2de949: -56.2% vm-scalability.throughput, +9.3% turbostat.Pkg_W Date: Tue, 19 Aug 2014 15:29:25 +0100 Message-ID: <20140819142925.GB10146@suse.de> In-Reply-To: <20140819043252.GA8798@localhost> List-Id: --===============8908514102076463225== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable On Tue, Aug 19, 2014 at 12:32:52PM +0800, Fengguang Wu wrote: > Hi Mel, > = > We noticed the below vm-scalability performance/power regressions on > commit 3484b2de9499df23c4604a513b36f96326ae81ad ("mm: rearrange zone > fields into read-only, page alloc, statistics and page reclaim lines"). > = > 24b7e5819ad5cbe 3484b2de9499df23c4604a513 testbox/testcase/testparams > --------------- ------------------------- --------------------------- > %stddev %change %stddev > \ | / > 9.95 =C2=B1 2% +69.1% 16.83 =C2=B1 5% brickland3/vm-scala= bility/300s-lru-file-mmap-read > 2.32 =C2=B1 6% +229.4% 7.63 =C2=B1 5% brickland3/vm-scala= bility/300s-lru-file-readonce > 12.27 =C2=B1 3% +99.4% 24.46 =C2=B1 5% TOTAL vm-scalabilit= y.stddev > = > 24b7e5819ad5cbe 3484b2de9499df23c4604a513 = > --------------- ------------------------- = > 13882598 =C2=B1 0% -35.8% 8915310 =C2=B1 1% brickland3/vm-scala= bility/300s-lru-file-mmap-read > 36379953 =C2=B1 1% -64.0% 13093373 =C2=B1 0% brickland3/vm-scala= bility/300s-lru-file-readonce > 50262551 =C2=B1 0% -56.2% 22008683 =C2=B1 0% TOTAL vm-scalabilit= y.throughput > = What units are these? It's completely unclear what is good and bad from the figures. 300s-lru-file-mmap-read appears multiple times in this report, each with different numbers beside them but little clue as to what they mean or what I'm meant to be looking for :( This is the same patch that was reported as having a performance gain in another set of tests from lkp so am a little confused. More importantly, as this patch is primary abougt cache misses it should be very unlikely that it makes a noticable difference to IO as the relative cost of a cache miss is so low. Similarly any difference it makes to reclaim activity is likely to be a coincidence or due to test variance. -- = Mel Gorman SUSE Labs --===============8908514102076463225==--