From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from smtp.kernel.org (aws-us-west-2-korg-mail-1.web.codeaurora.org [10.30.226.201]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 190E21B86C7; Mon, 9 Sep 2024 12:57:23 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=10.30.226.201 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1725886646; cv=none; b=d2BzVc+1f+dMchy70n4Pt5ZdiWuFvPyurkIrNAJ/WmJ7BlHwDMb+lJ1SKWlMF3JOEwxJTaV7beYhoHimgluM+A3x6Y41ONeFHcCjjc5EPHy9XcnRTv+UyEld0HzTKMD2/xv+0AcQHG86KYeIhfe51LYrqN+ITZ1S48C1C3rdKs0= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1725886646; c=relaxed/simple; bh=gMDLQsyGr+aVnTlYtwC5Ee0IsYtzSPk+Rgyj1y9L18s=; h=Message-ID:Subject:From:To:Cc:Date:In-Reply-To:References: Content-Type:MIME-Version; b=OhpxtWU+6ZOqY2+KKFOkQrP+ezMVDt/FoauhCSsLHz9czPeS7N7cg/UqsN3ua8NIsLdS49yWTy0emvZFksucYzhfkMrxWs2KkSowg4ZrO5VVfKPZg+YTxUeOcYOmrnVYbsuVWwlXwV/oHZmpjdnKBsxwwuYnE88K9R65x8KWWNA= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=kernel.org header.i=@kernel.org header.b=Pravk/Cl; arc=none smtp.client-ip=10.30.226.201 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=kernel.org header.i=@kernel.org header.b="Pravk/Cl" Received: by smtp.kernel.org (Postfix) with ESMTPSA id 7A7A5C4CEC5; Mon, 9 Sep 2024 12:57:22 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=k20201202; t=1725886643; bh=gMDLQsyGr+aVnTlYtwC5Ee0IsYtzSPk+Rgyj1y9L18s=; h=Subject:From:To:Cc:Date:In-Reply-To:References:From; b=Pravk/ClXhDgDogjGSMfxro9ygP1cUyKbepMErRbFI7Su+nRCYfyH8F21Pqq9YMPC zzhXdPU213sKtAONZ00TBZg+QCnVxbgnpBwtsLkbRnxITgb+wwfAXsTr3Hl/2hgiGC Pw8y4CCsHqUgo3MTXGwmAlQoa66k2oGAcusl0ETnjkzLTsoTSWmXGhueO/j24IWzFo 3juEpHL/0x+ZObLWZ2o1za4DbDax5zjpijsu50/C2N9XB5XGsetB9DOMAurt1MG5U0 SUDSptAmOvzK/zJWzGV3ezI2E961ZL48o/KXoIjvIyRYDzRDXyPY/0XIvV/8NyRIhI nJMvSkiF8qZHA== Message-ID: Subject: Re: [brauner-vfs:vfs.mgtime] [fs] a037d5e7f8: will-it-scale.per_thread_ops -5.5% regression From: Jeff Layton To: kernel test robot Cc: oe-lkp@lists.linux.dev, lkp@intel.com, Christian Brauner , Christian Brauner , "Darrick J. Wong" , Josef Bacik , Jan Kara , linux-fsdevel@vger.kernel.org, ying.huang@intel.com, feng.tang@intel.com, fengwei.yin@intel.com Date: Mon, 09 Sep 2024 08:57:21 -0400 In-Reply-To: <202409091303.31b2b713-oliver.sang@intel.com> References: <202409091303.31b2b713-oliver.sang@intel.com> Autocrypt: addr=jlayton@kernel.org; prefer-encrypt=mutual; keydata=mQINBE6V0TwBEADXhJg7s8wFDwBMEvn0qyhAnzFLTOCHooMZyx7XO7dAiIhDSi7G1NPxw n8jdFUQMCR/GlpozMFlSFiZXiObE7sef9rTtM68ukUyZM4pJ9l0KjQNgDJ6Fr342Htkjxu/kFV1Wv egyjnSsFt7EGoDjdKqr1TS9syJYFjagYtvWk/UfHlW09X+jOh4vYtfX7iYSx/NfqV3W1D7EDi0PqV T2h6v8i8YqsATFPwO4nuiTmL6I40ZofxVd+9wdRI4Db8yUNA4ZSP2nqLcLtFjClYRBoJvRWvsv4lm 0OX6MYPtv76hka8lW4mnRmZqqx3UtfHX/hF/zH24Gj7A6sYKYLCU3YrI2Ogiu7/ksKcl7goQjpvtV YrOOI5VGLHge0awt7bhMCTM9KAfPc+xL/ZxAMVWd3NCk5SamL2cE99UWgtvNOIYU8m6EjTLhsj8sn VluJH0/RcxEeFbnSaswVChNSGa7mXJrTR22lRL6ZPjdMgS2Km90haWPRc8Wolcz07Y2se0xpGVLEQ cDEsvv5IMmeMe1/qLZ6NaVkNuL3WOXvxaVT9USW1+/SGipO2IpKJjeDZfehlB/kpfF24+RrK+seQf CBYyUE8QJpvTZyfUHNYldXlrjO6n5MdOempLqWpfOmcGkwnyNRBR46g/jf8KnPRwXs509yAqDB6sE LZH+yWr9LQZEwARAQABtCVKZWZmIExheXRvbiA8amxheXRvbkBwb29jaGllcmVkcy5uZXQ+iQI7BB MBAgAlAhsDBgsJCAcDAgYVCAIJCgsEFgIDAQIeAQIXgAUCTpXWPAIZAQAKCRAADmhBGVaCFc65D/4 gBLNMHopQYgG/9RIM3kgFCCQV0pLv0hcg1cjr+bPI5f1PzJoOVi9s0wBDHwp8+vtHgYhM54yt43uI 7Htij0RHFL5eFqoVT4TSfAg2qlvNemJEOY0e4daljjmZM7UtmpGs9NN0r9r50W82eb5Kw5bc/r0km R/arUS2st+ecRsCnwAOj6HiURwIgfDMHGPtSkoPpu3DDp/cjcYUg3HaOJuTjtGHFH963B+f+hyQ2B rQZBBE76ErgTDJ2Db9Ey0kw7VEZ4I2nnVUY9B5dE2pJFVO5HJBMp30fUGKvwaKqYCU2iAKxdmJXRI ONb7dSde8LqZahuunPDMZyMA5+mkQl7kpIpR6kVDIiqmxzRuPeiMP7O2FCUlS2DnJnRVrHmCljLkZ Wf7ZUA22wJpepBligemtSRSbqCyZ3B48zJ8g5B8xLEntPo/NknSJaYRvfEQqGxgk5kkNWMIMDkfQO lDSXZvoxqU9wFH/9jTv1/6p8dHeGM0BsbBLMqQaqnWiVt5mG92E1zkOW69LnoozE6Le+12DsNW7Rj iR5K+27MObjXEYIW7FIvNN/TQ6U1EOsdxwB8o//Yfc3p2QqPr5uS93SDDan5ehH59BnHpguTc27Xi QQZ9EGiieCUx6Zh2ze3X2UW9YNzE15uKwkkuEIj60NvQRmEDfweYfOfPVOueC+iFifbQgSmVmZiBM YXl0b24gPGpsYXl0b25AcmVkaGF0LmNvbT6JAjgEEwECACIFAk6V0q0CGwMGCwkIBwMCBhUIAgkKC wQWAgMBAh4BAheAAAoJEAAOaEEZVoIViKUQALpvsacTMWWOd7SlPFzIYy2/fjvKlfB/Xs4YdNcf9q LqF+lk2RBUHdR/dGwZpvw/OLmnZ8TryDo2zXVJNWEEUFNc7wQpl3i78r6UU/GUY/RQmOgPhs3epQC 3PMJj4xFx+VuVcf/MXgDDdBUHaCTT793hyBeDbQuciARDJAW24Q1RCmjcwWIV/pgrlFa4lAXsmhoa c8UPc82Ijrs6ivlTweFf16VBc4nSLX5FB3ls7S5noRhm5/Zsd4PGPgIHgCZcPgkAnU1S/A/rSqf3F LpU+CbVBDvlVAnOq9gfNF+QiTlOHdZVIe4gEYAU3CUjbleywQqV02BKxPVM0C5/oVjMVx3bri75n1 TkBYGmqAXy9usCkHIsG5CBHmphv9MHmqMZQVsxvCzfnI5IO1+7MoloeeW/lxuyd0pU88dZsV/riHw 87i2GJUJtVlMl5IGBNFpqoNUoqmvRfEMeXhy/kUX4Xc03I1coZIgmwLmCSXwx9MaCPFzV/dOOrju2 xjO+2sYyB5BNtxRqUEyXglpujFZqJxxau7E0eXoYgoY9gtFGsspzFkVNntamVXEWVVgzJJr/EWW0y +jNd54MfPRqH+eCGuqlnNLktSAVz1MvVRY1dxUltSlDZT7P2bUoMorIPu8p7ZCg9dyX1+9T6Muc5d Hxf/BBP/ir+3e8JTFQBFOiLNdFtB9KZWZmIExheXRvbiA8amxheXRvbkBzYW1iYS5vcmc+iQI4BBM BAgAiBQJOldK9AhsDBgsJCAcDAgYVCAIJCgsEFgIDAQIeAQIXgAAKCRAADmhBGVaCFWgWD/0ZRi4h N9FK2BdQs9RwNnFZUr7JidAWfCrs37XrA/56olQl3ojn0fQtrP4DbTmCuh0SfMijB24psy1GnkPep naQ6VRf7Dxg/Y8muZELSOtsv2CKt3/02J1BBitrkkqmHyni5fLLYYg6fub0T/8Kwo1qGPdu1hx2BQ RERYtQ/S5d/T0cACdlzi6w8rs5f09hU9Tu4qV1JLKmBTgUWKN969HPRkxiojLQziHVyM/weR5Reu6 FZVNuVBGqBD+sfk/c98VJHjsQhYJijcsmgMb1NohAzwrBKcSGKOWJToGEO/1RkIN8tqGnYNp2G+aR 685D0chgTl1WzPRM6mFG1+n2b2RR95DxumKVpwBwdLPoCkI24JkeDJ7lXSe3uFWISstFGt0HL8Eew P8RuGC8s5h7Ct91HMNQTbjgA+Vi1foWUVXpEintAKgoywaIDlJfTZIl6Ew8ETN/7DLy8bXYgq0Xzh aKg3CnOUuGQV5/nl4OAX/3jocT5Cz/OtAiNYj5mLPeL5z2ZszjoCAH6caqsF2oLyAnLqRgDgR+wTQ T6gMhr2IRsl+cp8gPHBwQ4uZMb+X00c/Amm9VfviT+BI7B66cnC7Zv6Gvmtu2rEjWDGWPqUgccB7h dMKnKDthkA227/82tYoFiFMb/NwtgGrn5n2vwJyKN6SEoygGrNt0SI84y6hEVbQlSmVmZiBMYXl0b 24gPGpsYXl0b25AcHJpbWFyeWRhdGEuY29tPokCOQQTAQIAIwUCU4xmKQIbAwcLCQgHAwIBBhUIAg kKCwQWAgMBAh4BAheAAAoJEAAOaEEZVoIV1H0P/j4OUTwFd7BBbpoSp695qb6HqCzWMuExsp8nZjr uymMaeZbGr3OWMNEXRI1FWNHMtcMHWLP/RaDqCJil28proO+PQ/yPhsr2QqJcW4nr91tBrv/MqItu AXLYlsgXqp4BxLP67bzRJ1Bd2x0bWXurpEXY//VBOLnODqThGEcL7jouwjmnRh9FTKZfBDpFRaEfD FOXIfAkMKBa/c9TQwRpx2DPsl3eFWVCNuNGKeGsirLqCxUg5kWTxEorROppz9oU4HPicL6rRH22Ce 6nOAON2vHvhkUuO3GbffhrcsPD4DaYup4ic+DxWm+DaSSRJ+e1yJvwi6NmQ9P9UAuLG93S2MdNNbo sZ9P8k2mTOVKMc+GooI9Ve/vH8unwitwo7ORMVXhJeU6Q0X7zf3SjwDq2lBhn1DSuTsn2DbsNTiDv qrAaCvbsTsw+SZRwF85eG67eAwouYk+dnKmp1q57LDKMyzysij2oDKbcBlwB/TeX16p8+LxECv51a sjS9TInnipssssUDrHIvoTTXWcz7Y5wIngxDFwT8rPY3EggzLGfK5Zx2Q5S/N0FfmADmKknG/D8qG IcJE574D956tiUDKN4I+/g125ORR1v7bP+OIaayAvq17RP+qcAqkxc0x8iCYVCYDouDyNvWPGRhbL UO7mlBpjW9jK9e2fvZY9iw3QzIPGKtClKZWZmIExheXRvbiA8amVmZi5sYXl0b25AcHJpbWFyeWRh dGEuY29tPokCOQQTAQIAIwUCU4xmUAIbAwcLCQgHAwIBBhUIAgkKCwQWAgMBAh4BAheAAAoJEAAOa EEZVoIVzJoQALFCS6n/FHQS+hIzHIb56JbokhK0AFqoLVzLKzrnaeXhE5isWcVg0eoV2oTScIwUSU apy94if69tnUo4Q7YNt8/6yFM6hwZAxFjOXR0ciGE3Q+Z1zi49Ox51yjGMQGxlakV9ep4sV/d5a50 M+LFTmYSAFp6HY23JN9PkjVJC4PUv5DYRbOZ6Y1+TfXKBAewMVqtwT1Y+LPlfmI8dbbbuUX/kKZ5d dhV2736fgyfpslvJKYl0YifUOVy4D1G/oSycyHkJG78OvX4JKcf2kKzVvg7/Rnv+AueCfFQ6nGwPn 0P91I7TEOC4XfZ6a1K3uTp4fPPs1Wn75X7K8lzJP/p8lme40uqwAyBjk+IA5VGd+CVRiyJTpGZwA0 jwSYLyXboX+Dqm9pSYzmC9+/AE7lIgpWj+3iNisp1SWtHc4pdtQ5EU2SEz8yKvDbD0lNDbv4ljI7e flPsvN6vOrxz24mCliEco5DwhpaaSnzWnbAPXhQDWb/lUgs/JNk8dtwmvWnqCwRqElMLVisAbJmC0 BhZ/Ab4sph3EaiZfdXKhiQqSGdK4La3OTJOJYZphPdGgnkvDV9Pl1QZ0ijXQrVIy3zd6VCNaKYq7B AKidn5g/2Q8oio9Tf4XfdZ9dtwcB+bwDJFgvvDYaZ5bI3ln4V3EyW5i2NfXazz/GA/I/ZtbsigCFc 8ftCBKZWZmIExheXRvbiA8amxheXRvbkBrZXJuZWwub3JnPokCOAQTAQIAIgUCWe8u6AIbAwYLCQg HAwIGFQgCCQoLBBYCAwECHgECF4AACgkQAA5oQRlWghUuCg/+Lb/xGxZD2Q1oJVAE37uW308UpVSD 2tAMJUvFTdDbfe3zKlPDTuVsyNsALBGclPLagJ5ZTP+Vp2irAN9uwBuacBOTtmOdz4ZN2tdvNgozz uxp4CHBDVzAslUi2idy+xpsp47DWPxYFIRP3M8QG/aNW052LaPc0cedYxp8+9eiVUNpxF4SiU4i9J DfX/sn9XcfoVZIxMpCRE750zvJvcCUz9HojsrMQ1NFc7MFT1z3MOW2/RlzPcog7xvR5ENPH19ojRD CHqumUHRry+RF0lH00clzX/W8OrQJZtoBPXv9ahka/Vp7kEulcBJr1cH5Wz/WprhsIM7U9pse1f1g Yy9YbXtWctUz8uvDR7shsQxAhX3qO7DilMtuGo1v97I/Kx4gXQ52syh/w6EBny71CZrOgD6kJwPVV AaM1LRC28muq91WCFhs/nzHozpbzcheyGtMUI2Ao4K6mnY+3zIuXPygZMFr9KXE6fF7HzKxKuZMJO aEZCiDOq0anx6FmOzs5E6Jqdpo/mtI8beK+BE7Va6ni7YrQlnT0i3vaTVMTiCThbqsB20VrbMjlhp f8lfK1XVNbRq/R7GZ9zHESlsa35ha60yd/j3pu5hT2xyy8krV8vGhHvnJ1XRMJBAB/UYb6FyC7S+m QZIQXVeAA+smfTT0tDrisj1U5x6ZB9b3nBg65kc= Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable User-Agent: Evolution 3.52.4 (3.52.4-1.fc40app2) Precedence: bulk X-Mailing-List: linux-fsdevel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 On Mon, 2024-09-09 at 16:24 +0800, kernel test robot wrote: >=20 > Hello, >=20 > kernel test robot noticed a -5.5% regression of will-it-scale.per_thread_= ops on: >=20 >=20 > commit: a037d5e7f81bae8ff69eb670b2ec3f25ad4d2cc2 ("fs: add infrastructure= for multigrain timestamps") > https://git.kernel.org/cgit/linux/kernel/git/vfs/vfs.git vfs.mgtime >=20 > testcase: will-it-scale > test machine: 64 threads 2 sockets Intel(R) Xeon(R) Gold 6346 CPU @ 3.10G= Hz (Ice Lake) with 256G memory > parameters: >=20 > nr_task: 100% > mode: thread > test: pipe1 > cpufreq_governor: performance >=20 >=20 > In addition to that, the commit also has significant impact on the follow= ing tests: >=20 > +------------------+-----------------------------------------------------= -----------------------------------------------+ > > testcase: change | will-it-scale: will-it-scale.per_thread_ops -2.4% re= gression | > > test machine | 64 threads 2 sockets Intel(R) Xeon(R) Gold 6346 CPU = @ 3.10GHz (Ice Lake) with 256G memory | > > test parameters | cpufreq_governor=3Dperformance = | > > | mode=3Dthread = | > > | nr_task=3D100% = | > > | test=3Dwriteseek1 = | > +------------------+-----------------------------------------------------= -----------------------------------------------+ > > testcase: change | will-it-scale: will-it-scale.per_thread_ops -5.5% re= gression | > > test machine | 224 threads 4 sockets Intel(R) Xeon(R) Platinum 8380= H CPU @ 2.90GHz (Cooper Lake) with 192G memory | > > test parameters | cpufreq_governor=3Dperformance = | > > | mode=3Dthread = | > > | nr_task=3D100% = | > > | test=3Dpipe1 = | > +------------------+-----------------------------------------------------= -----------------------------------------------+ >=20 >=20 > If you fix the issue in a separate patch/commit (i.e. not just a new vers= ion of > the same patch/commit), kindly add following tags > > Reported-by: kernel test robot > > Closes: https://lore.kernel.org/oe-lkp/202409091303.31b2b713-oliver.san= g@intel.com >=20 >=20 It's not too surprising that some of these microbenchmarks might regress, given that we're adding some atomic ops in the write codepath. That said, if this the commit it landed on, then this is before any filesystems have enabled multigrain timestamps. Most of the new stuff we're adding here is dead code at this point. The main difference is that we're fetching the ctime_floor value when calling current_time (and doing the ktime_t comparison). static ktime_t coarse_ctime(ktime_t floor) { ktime_t coarse =3D ktime_get_coarse(); /* If coarse time is already newer, return that */ if (!ktime_after(floor, coarse)) return ktime_get_coarse_real(); return ktime_mono_to_real(floor); } I forget who it was that suggested changing it, but originally this patch just did this when the coarse time was after the floor: return ktime_mono_to_real(coarse); I wonder if that might shave off a few cycles? The other possibility is the is_mgtime check. That has to do some pointer chasing to get at the fs_flags field. If that's too costly, we could look at adding a flag to the inode that mirrors that value. I'll see if I can reproduce this. > Details are as below: > -------------------------------------------------------------------------= -------------------------> >=20 >=20 > The kernel config and materials to reproduce are available at: > https://download.01.org/0day-ci/archive/20240909/202409091303.31b2b713-ol= iver.sang@intel.com >=20 > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > compiler/cpufreq_governor/kconfig/mode/nr_task/rootfs/tbox_group/test/tes= tcase: > gcc-12/performance/x86_64-rhel-8.3/thread/100%/debian-12-x86_64-2024020= 6.cgz/lkp-icl-2sp7/pipe1/will-it-scale >=20 > commit:=20 > v6.11-rc1 > a037d5e7f8 ("fs: add infrastructure for multigrain timestamps") >=20 > v6.11-rc1 a037d5e7f81bae8ff69eb670b2e=20 > ---------------- ---------------------------=20 > %stddev %change %stddev > \ | \ =20 > 98752 +1.3% 100032 proc-vmstat.nr_active_anon > 103972 +1.2% 105220 proc-vmstat.nr_shmem > 98752 +1.3% 100032 proc-vmstat.nr_zone_active_= anon > 84330110 -5.5% 79683588 will-it-scale.64.threads > 1317657 -5.5% 1245055 will-it-scale.per_thread_op= s > 84330110 -5.5% 79683588 will-it-scale.workload > 4.678e+10 +1.2% 4.733e+10 perf-stat.i.branch-instruct= ions > 0.03 =C2=B1 7% -0.0 0.02 =C2=B1 6% perf-stat.i.branc= h-miss-rate% > 12781080 =C2=B1 7% -19.2% 10321748 =C2=B1 6% perf-stat.i.branc= h-misses > 1.01 -2.3% 0.99 perf-stat.i.cpi > 1.946e+11 +2.4% 1.993e+11 perf-stat.i.instructions > 0.99 +2.4% 1.01 perf-stat.i.ipc > 0.03 =C2=B1 7% -0.0 0.02 =C2=B1 6% perf-stat.overall= .branch-miss-rate% > 1.01 -2.4% 0.99 perf-stat.overall.cpi > 0.99 +2.4% 1.01 perf-stat.overall.ipc > 695016 +8.4% 753440 perf-stat.overall.path-leng= th > 4.661e+10 +1.2% 4.717e+10 perf-stat.ps.branch-instruc= tions > 12713767 =C2=B1 7% -19.3% 10263468 =C2=B1 6% perf-stat.ps.bran= ch-misses > 1.939e+11 +2.4% 1.986e+11 perf-stat.ps.instructions > 5.861e+13 +2.4% 6.004e+13 perf-stat.total.instruction= s > 7.03 -0.4 6.60 perf-profile.calltrace.cycl= es-pp.clear_bhb_loop.write > 6.89 -0.4 6.48 perf-profile.calltrace.cycl= es-pp.clear_bhb_loop.read > 6.72 -0.4 6.35 perf-profile.calltrace.cycl= es-pp.copy_page_from_iter.pipe_write.vfs_write.ksys_write.do_syscall_64 > 5.69 -0.3 5.37 perf-profile.calltrace.cycl= es-pp.entry_SYSCALL_64.read > 5.46 -0.3 5.17 perf-profile.calltrace.cycl= es-pp._copy_from_iter.copy_page_from_iter.pipe_write.vfs_write.ksys_write > 5.68 -0.3 5.40 perf-profile.calltrace.cycl= es-pp.entry_SYSCALL_64.write > 5.30 -0.3 5.03 perf-profile.calltrace.cycl= es-pp.copy_page_to_iter.pipe_read.vfs_read.ksys_read.do_syscall_64 > 49.56 -0.2 49.34 perf-profile.calltrace.cycl= es-pp.read > 4.39 -0.2 4.17 perf-profile.calltrace.cycl= es-pp._copy_to_iter.copy_page_to_iter.pipe_read.vfs_read.ksys_read > 0.56 -0.1 0.43 =C2=B1 50% perf-profile.calltrace= .cycles-pp.aa_file_perm.apparmor_file_permission.security_file_permission.r= w_verify_area.vfs_write > 2.66 -0.1 2.52 perf-profile.calltrace.cycl= es-pp.__wake_up_sync_key.pipe_write.vfs_write.ksys_write.do_syscall_64 > 1.71 -0.1 1.61 perf-profile.calltrace.cycl= es-pp._raw_spin_lock_irqsave.__wake_up_sync_key.pipe_write.vfs_write.ksys_w= rite > 1.81 -0.1 1.72 perf-profile.calltrace.cycl= es-pp.mutex_lock.pipe_read.vfs_read.ksys_read.do_syscall_64 > 1.77 -0.1 1.69 perf-profile.calltrace.cycl= es-pp.mutex_lock.pipe_write.vfs_write.ksys_write.do_syscall_64 > 1.04 -0.1 0.97 perf-profile.calltrace.cycl= es-pp.fput.ksys_read.do_syscall_64.entry_SYSCALL_64_after_hwframe.read > 0.86 -0.1 0.80 perf-profile.calltrace.cycl= es-pp.testcase > 1.11 -0.1 1.05 perf-profile.calltrace.cycl= es-pp.mutex_unlock.pipe_read.vfs_read.ksys_read.do_syscall_64 > 1.12 -0.1 1.06 perf-profile.calltrace.cycl= es-pp.mutex_unlock.pipe_write.vfs_write.ksys_write.do_syscall_64 > 0.88 -0.0 0.84 perf-profile.calltrace.cycl= es-pp.syscall_return_via_sysret.read > 1.04 -0.0 1.00 perf-profile.calltrace.cycl= es-pp.fput.ksys_write.do_syscall_64.entry_SYSCALL_64_after_hwframe.write > 0.72 -0.0 0.68 perf-profile.calltrace.cycl= es-pp.x64_sys_call.do_syscall_64.entry_SYSCALL_64_after_hwframe.read > 0.65 -0.0 0.61 perf-profile.calltrace.cycl= es-pp.x64_sys_call.do_syscall_64.entry_SYSCALL_64_after_hwframe.write > 0.55 -0.0 0.52 =C2=B1 2% perf-profile.calltrace= .cycles-pp.aa_file_perm.apparmor_file_permission.security_file_permission.r= w_verify_area.vfs_read > 0.58 -0.0 0.55 =C2=B1 2% perf-profile.calltrace= .cycles-pp.__cond_resched.mutex_lock.pipe_write.vfs_write.ksys_write > 53.26 +0.1 53.34 perf-profile.calltrace.cycl= es-pp.write > 0.00 +0.6 0.56 =C2=B1 5% perf-profile.calltrace= .cycles-pp.ktime_get_coarse_ts64.coarse_ctime.current_time.atime_needs_upda= te.touch_atime > 0.00 +0.6 0.64 perf-profile.calltrace.cycl= es-pp.timestamp_truncate.current_time.atime_needs_update.touch_atime.pipe_r= ead > 0.00 +0.6 0.64 perf-profile.calltrace.cycl= es-pp.timestamp_truncate.current_time.inode_needs_update_time.file_update_t= ime.pipe_write > 32.72 +0.8 33.51 perf-profile.calltrace.cycl= es-pp.entry_SYSCALL_64_after_hwframe.read > 31.75 +0.8 32.58 perf-profile.calltrace.cycl= es-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.read > 36.32 +1.0 37.27 perf-profile.calltrace.cycl= es-pp.entry_SYSCALL_64_after_hwframe.write > 28.50 +1.0 29.49 perf-profile.calltrace.cycl= es-pp.ksys_read.do_syscall_64.entry_SYSCALL_64_after_hwframe.read > 35.33 +1.0 36.34 perf-profile.calltrace.cycl= es-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.write > 0.00 +1.1 1.08 =C2=B1 2% perf-profile.calltrace= .cycles-pp.coarse_ctime.current_time.inode_needs_update_time.file_update_ti= me.pipe_write > 32.09 +1.1 33.20 perf-profile.calltrace.cycl= es-pp.ksys_write.do_syscall_64.entry_SYSCALL_64_after_hwframe.write > 0.00 +1.2 1.23 perf-profile.calltrace.cycl= es-pp.coarse_ctime.current_time.atime_needs_update.touch_atime.pipe_read > 23.02 +1.3 24.30 perf-profile.calltrace.cycl= es-pp.vfs_read.ksys_read.do_syscall_64.entry_SYSCALL_64_after_hwframe.read > 26.62 +1.4 27.97 perf-profile.calltrace.cycl= es-pp.vfs_write.ksys_write.do_syscall_64.entry_SYSCALL_64_after_hwframe.wri= te > 19.03 +1.6 20.60 perf-profile.calltrace.cycl= es-pp.pipe_write.vfs_write.ksys_write.do_syscall_64.entry_SYSCALL_64_after_= hwframe > 15.11 +1.6 16.71 perf-profile.calltrace.cycl= es-pp.pipe_read.vfs_read.ksys_read.do_syscall_64.entry_SYSCALL_64_after_hwf= rame > 1.48 +2.1 3.56 perf-profile.calltrace.cycl= es-pp.current_time.atime_needs_update.touch_atime.pipe_read.vfs_read > 3.51 +2.2 5.72 perf-profile.calltrace.cycl= es-pp.touch_atime.pipe_read.vfs_read.ksys_read.do_syscall_64 > 3.05 +2.2 5.28 perf-profile.calltrace.cycl= es-pp.atime_needs_update.touch_atime.pipe_read.vfs_read.ksys_read > 1.81 =C2=B1 2% +2.4 4.20 perf-profile.calltrace= .cycles-pp.inode_needs_update_time.file_update_time.pipe_write.vfs_write.ks= ys_write > 2.23 =C2=B1 2% +2.5 4.71 perf-profile.calltrace= .cycles-pp.file_update_time.pipe_write.vfs_write.ksys_write.do_syscall_64 > 0.00 +3.3 3.26 perf-profile.calltrace.cycl= es-pp.current_time.inode_needs_update_time.file_update_time.pipe_write.vfs_= write > 14.06 -0.8 13.22 perf-profile.children.cycle= s-pp.clear_bhb_loop > 7.01 -0.4 6.62 perf-profile.children.cycle= s-pp.copy_page_from_iter > 6.63 -0.4 6.28 perf-profile.children.cycle= s-pp.entry_SYSCALL_64 > 5.61 -0.3 5.32 perf-profile.children.cycle= s-pp._copy_from_iter > 5.46 -0.3 5.18 perf-profile.children.cycle= s-pp.copy_page_to_iter > 5.00 -0.3 4.73 perf-profile.children.cycle= s-pp.entry_SYSRETQ_unsafe_stack > 50.00 -0.2 49.78 perf-profile.children.cycle= s-pp.read > 4.46 -0.2 4.24 perf-profile.children.cycle= s-pp._copy_to_iter > 3.88 -0.2 3.69 perf-profile.children.cycle= s-pp.mutex_lock > 2.87 -0.1 2.73 perf-profile.children.cycle= s-pp.__wake_up_sync_key > 2.38 -0.1 2.24 perf-profile.children.cycle= s-pp.mutex_unlock > 2.23 -0.1 2.10 perf-profile.children.cycle= s-pp.fput > 1.81 -0.1 1.71 perf-profile.children.cycle= s-pp._raw_spin_lock_irqsave > 1.53 -0.1 1.44 perf-profile.children.cycle= s-pp.x64_sys_call > 1.10 -0.1 1.01 perf-profile.children.cycle= s-pp.testcase > 1.26 -0.1 1.19 =C2=B1 2% perf-profile.children.= cycles-pp.aa_file_perm > 1.48 -0.1 1.41 perf-profile.children.cycle= s-pp.__cond_resched > 2.14 -0.1 2.09 perf-profile.children.cycle= s-pp.syscall_exit_to_user_mode > 0.35 -0.0 0.32 =C2=B1 2% perf-profile.children.= cycles-pp.entry_SYSCALL_64_safe_stack > 0.57 -0.0 0.54 perf-profile.children.cycle= s-pp.__wake_up_common > 0.51 -0.0 0.48 perf-profile.children.cycle= s-pp.kill_fasync > 0.26 -0.0 0.23 =C2=B1 2% perf-profile.children.= cycles-pp.__x64_sys_read > 0.49 -0.0 0.47 perf-profile.children.cycle= s-pp._raw_spin_unlock_irqrestore > 0.25 -0.0 0.23 perf-profile.children.cycle= s-pp.__x64_sys_write > 0.30 -0.0 0.28 perf-profile.children.cycle= s-pp.make_vfsuid > 0.23 -0.0 0.22 =C2=B1 2% perf-profile.children.= cycles-pp.write@plt > 0.00 +0.7 0.65 perf-profile.children.cycle= s-pp.set_normalized_timespec64 > 0.71 +0.7 1.43 perf-profile.children.cycle= s-pp.timestamp_truncate > 0.00 +0.9 0.93 perf-profile.children.cycle= s-pp.ns_to_timespec64 > 28.79 +1.0 29.76 perf-profile.children.cycle= s-pp.ksys_read > 0.00 +1.0 1.04 perf-profile.children.cycle= s-pp.ktime_get_coarse_with_offset > 32.43 +1.1 33.54 perf-profile.children.cycle= s-pp.ksys_write > 0.00 +1.2 1.15 =C2=B1 3% perf-profile.children.= cycles-pp.ktime_get_coarse_ts64 > 23.24 +1.3 24.52 perf-profile.children.cycle= s-pp.vfs_read > 26.88 +1.3 28.22 perf-profile.children.cycle= s-pp.vfs_write > 19.69 +1.5 21.23 perf-profile.children.cycle= s-pp.pipe_write > 15.78 +1.6 17.35 perf-profile.children.cycle= s-pp.pipe_read > 69.41 +1.7 71.14 perf-profile.children.cycle= s-pp.entry_SYSCALL_64_after_hwframe > 67.61 +1.8 69.42 perf-profile.children.cycle= s-pp.do_syscall_64 > 3.65 +2.2 5.86 perf-profile.children.cycle= s-pp.touch_atime > 3.34 +2.2 5.56 perf-profile.children.cycle= s-pp.atime_needs_update > 2.08 =C2=B1 2% +2.3 4.35 perf-profile.children.= cycles-pp.inode_needs_update_time > 2.45 =C2=B1 2% +2.4 4.85 perf-profile.children.= cycles-pp.file_update_time > 0.00 +2.7 2.72 perf-profile.children.cycle= s-pp.coarse_ctime > 1.63 +5.9 7.55 perf-profile.children.cycle= s-pp.current_time > 13.92 -0.8 13.07 perf-profile.self.cycles-pp= .clear_bhb_loop > 1.04 =C2=B1 3% -0.3 0.73 perf-profile.self.cycl= es-pp.inode_needs_update_time > 5.38 -0.3 5.09 perf-profile.self.cycles-pp= ._copy_from_iter > 4.83 -0.3 4.56 perf-profile.self.cycles-pp= .entry_SYSRETQ_unsafe_stack > 4.55 -0.2 4.33 perf-profile.self.cycles-pp= .vfs_read > 4.37 -0.2 4.15 perf-profile.self.cycles-pp= ._copy_to_iter > 3.34 -0.2 3.14 perf-profile.self.cycles-pp= .write > 3.21 -0.2 3.02 perf-profile.self.cycles-pp= .pipe_read > 3.34 -0.2 3.15 perf-profile.self.cycles-pp= .read > 2.06 -0.1 1.93 perf-profile.self.cycles-pp= .fput > 4.43 -0.1 4.30 perf-profile.self.cycles-pp= .vfs_write > 2.21 -0.1 2.09 perf-profile.self.cycles-pp= .mutex_unlock > 2.54 -0.1 2.42 perf-profile.self.cycles-pp= .do_syscall_64 > 2.37 -0.1 2.27 perf-profile.self.cycles-pp= .mutex_lock > 1.88 -0.1 1.78 perf-profile.self.cycles-pp= .entry_SYSCALL_64 > 1.73 -0.1 1.64 perf-profile.self.cycles-pp= ._raw_spin_lock_irqsave > 1.46 -0.1 1.37 perf-profile.self.cycles-pp= .copy_page_from_iter > 1.79 -0.1 1.71 perf-profile.self.cycles-pp= .entry_SYSCALL_64_after_hwframe > 1.36 -0.1 1.28 perf-profile.self.cycles-pp= .x64_sys_call > 1.33 -0.1 1.26 perf-profile.self.cycles-pp= .security_file_permission > 0.87 -0.1 0.80 perf-profile.self.cycles-pp= .testcase > 1.05 -0.1 0.99 perf-profile.self.cycles-pp= .rw_verify_area > 0.99 -0.1 0.93 perf-profile.self.cycles-pp= .copy_page_to_iter > 1.05 -0.1 0.99 perf-profile.self.cycles-pp= .aa_file_perm > 1.27 -0.1 1.21 perf-profile.self.cycles-pp= .atime_needs_update > 0.83 -0.1 0.78 perf-profile.self.cycles-pp= .__cond_resched > 1.38 -0.0 1.34 perf-profile.self.cycles-pp= .syscall_exit_to_user_mode > 0.34 =C2=B1 2% -0.0 0.31 =C2=B1 2% perf-profile.self= .cycles-pp.entry_SYSCALL_64_safe_stack > 0.50 -0.0 0.48 perf-profile.self.cycles-pp= .__wake_up_sync_key > 0.18 -0.0 0.16 =C2=B1 2% perf-profile.self.cycl= es-pp.__x64_sys_read > 0.17 -0.0 0.16 perf-profile.self.cycles-pp= .__x64_sys_write > 0.43 +0.1 0.49 perf-profile.self.cycles-pp= .file_update_time > 0.00 +0.5 0.51 perf-profile.self.cycles-pp= .set_normalized_timespec64 > 0.58 +0.6 1.15 perf-profile.self.cycles-pp= .timestamp_truncate > 0.00 +0.8 0.78 perf-profile.self.cycles-pp= .ns_to_timespec64 > 0.00 +0.9 0.88 =C2=B1 4% perf-profile.self.cycl= es-pp.ktime_get_coarse_ts64 > 0.00 +0.9 0.89 =C2=B1 2% perf-profile.self.cycl= es-pp.ktime_get_coarse_with_offset > 1.07 +0.9 1.98 =C2=B1 2% perf-profile.self.cycl= es-pp.current_time > 0.00 +1.2 1.17 perf-profile.self.cycles-pp= .coarse_ctime >=20 >=20 > *************************************************************************= ************************** > lkp-icl-2sp7: 64 threads 2 sockets Intel(R) Xeon(R) Gold 6346 CPU @ 3.10G= Hz (Ice Lake) with 256G memory > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > compiler/cpufreq_governor/kconfig/mode/nr_task/rootfs/tbox_group/test/tes= tcase: > gcc-12/performance/x86_64-rhel-8.3/thread/100%/debian-12-x86_64-2024020= 6.cgz/lkp-icl-2sp7/writeseek1/will-it-scale >=20 > commit:=20 > v6.11-rc1 > a037d5e7f8 ("fs: add infrastructure for multigrain timestamps") >=20 > v6.11-rc1 a037d5e7f81bae8ff69eb670b2e=20 > ---------------- ---------------------------=20 > %stddev %change %stddev > \ | \ =20 > 79.00 =C2=B1 9% +38.8% 109.67 =C2=B1 20% perf-c2c.HITM.rem= ote > 75816166 -2.4% 73999365 will-it-scale.64.threads > 1184627 -2.4% 1156239 will-it-scale.per_thread_op= s > 75816166 -2.4% 73999365 will-it-scale.workload > 1.29 =C2=B1 15% +32.0% 1.70 =C2=B1 15% perf-sched.sch_de= lay.avg.ms.__cond_resched.vfs_write.ksys_write.do_syscall_64.entry_SYSCALL_= 64_after_hwframe > 5.46 =C2=B1221% -99.7% 0.02 =C2=B1 15% perf-sched.sch_de= lay.max.ms.schedule_timeout.rcu_gp_fqs_loop.rcu_gp_kthread.kthread > 2.57 =C2=B1 15% +32.0% 3.40 =C2=B1 15% perf-sched.wait_a= nd_delay.avg.ms.__cond_resched.vfs_write.ksys_write.do_syscall_64.entry_SYS= CALL_64_after_hwframe > 846.14 =C2=B1 37% -64.5% 300.69 =C2=B1117% perf-sched.wait_a= nd_delay.avg.ms.devkmsg_read.vfs_read.ksys_read.do_syscall_64 > 1.29 =C2=B1 15% +32.0% 1.70 =C2=B1 15% perf-sched.wait_t= ime.avg.ms.__cond_resched.vfs_write.ksys_write.do_syscall_64.entry_SYSCALL_= 64_after_hwframe > 839.24 =C2=B1 37% -64.4% 298.49 =C2=B1116% perf-sched.wait_t= ime.avg.ms.devkmsg_read.vfs_read.ksys_read.do_syscall_64 > 8283953 +18.7% 9832101 =C2=B1 6% perf-stat.i.branch-mis= ses > 1.13 -1.7% 1.11 perf-stat.i.cpi > 1.746e+11 +1.7% 1.775e+11 perf-stat.i.instructions > 0.89 +1.7% 0.90 perf-stat.i.ipc > 0.02 +0.0 0.02 =C2=B1 6% perf-stat.overall.bran= ch-miss-rate% > 1.13 -1.6% 1.11 perf-stat.overall.cpi > 0.89 +1.6% 0.90 perf-stat.overall.ipc > 696240 +4.1% 725044 perf-stat.overall.path-leng= th > 8247837 +18.6% 9784317 =C2=B1 6% perf-stat.ps.branch-mi= sses > 1.74e+11 +1.7% 1.769e+11 perf-stat.ps.instructions > 5.279e+13 +1.6% 5.365e+13 perf-stat.total.instruction= s > 26.93 -0.8 26.10 perf-profile.calltrace.cycl= es-pp.llseek > 31.20 -0.6 30.58 perf-profile.calltrace.cycl= es-pp.generic_perform_write.shmem_file_write_iter.vfs_write.ksys_write.do_s= yscall_64 > 12.61 -0.5 12.12 =C2=B1 2% perf-profile.calltrace= .cycles-pp.entry_SYSCALL_64_after_hwframe.llseek > 11.70 =C2=B1 2% -0.5 11.23 =C2=B1 2% perf-profile.call= trace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.llseek > 8.91 =C2=B1 2% -0.4 8.50 =C2=B1 3% perf-profile.call= trace.cycles-pp.ksys_lseek.do_syscall_64.entry_SYSCALL_64_after_hwframe.lls= eek > 13.72 -0.3 13.38 perf-profile.calltrace.cycl= es-pp.copy_page_from_iter_atomic.generic_perform_write.shmem_file_write_ite= r.vfs_write.ksys_write > 6.36 -0.2 6.16 perf-profile.calltrace.cycl= es-pp.clear_bhb_loop.llseek > 6.60 -0.2 6.42 perf-profile.calltrace.cycl= es-pp.shmem_write_begin.generic_perform_write.shmem_file_write_iter.vfs_wri= te.ksys_write > 5.69 -0.2 5.52 perf-profile.calltrace.cycl= es-pp.shmem_get_folio_gfp.shmem_write_begin.generic_perform_write.shmem_fil= e_write_iter.vfs_write > 5.85 -0.1 5.72 perf-profile.calltrace.cycl= es-pp.clear_bhb_loop.write > 5.05 -0.1 4.93 perf-profile.calltrace.cycl= es-pp.entry_SYSCALL_64.write > 4.86 -0.1 4.75 perf-profile.calltrace.cycl= es-pp.entry_SYSCALL_64.llseek > 3.23 -0.1 3.14 perf-profile.calltrace.cycl= es-pp.shmem_write_end.generic_perform_write.shmem_file_write_iter.vfs_write= .ksys_write > 2.23 -0.1 2.17 perf-profile.calltrace.cycl= es-pp.filemap_get_entry.shmem_get_folio_gfp.shmem_write_begin.generic_perfo= rm_write.shmem_file_write_iter > 1.55 -0.0 1.51 perf-profile.calltrace.cycl= es-pp.down_write.shmem_file_write_iter.vfs_write.ksys_write.do_syscall_64 > 1.52 -0.0 1.49 perf-profile.calltrace.cycl= es-pp.mutex_lock.__fdget_pos.ksys_write.do_syscall_64.entry_SYSCALL_64_afte= r_hwframe > 0.98 -0.0 0.95 perf-profile.calltrace.cycl= es-pp.fput.ksys_lseek.do_syscall_64.entry_SYSCALL_64_after_hwframe.llseek > 1.56 -0.0 1.52 perf-profile.calltrace.cycl= es-pp.mutex_lock.__fdget_pos.ksys_lseek.do_syscall_64.entry_SYSCALL_64_afte= r_hwframe > 0.95 -0.0 0.92 perf-profile.calltrace.cycl= es-pp.folio_unlock.shmem_write_end.generic_perform_write.shmem_file_write_i= ter.vfs_write > 1.06 -0.0 1.03 perf-profile.calltrace.cycl= es-pp.mutex_unlock.ksys_lseek.do_syscall_64.entry_SYSCALL_64_after_hwframe.= llseek > 0.84 -0.0 0.82 perf-profile.calltrace.cycl= es-pp.folio_mark_accessed.shmem_get_folio_gfp.shmem_write_begin.generic_per= form_write.shmem_file_write_iter > 0.64 -0.0 0.62 perf-profile.calltrace.cycl= es-pp.folio_mark_dirty.shmem_write_end.generic_perform_write.shmem_file_wri= te_iter.vfs_write > 0.76 -0.0 0.74 perf-profile.calltrace.cycl= es-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwfram= e.llseek > 0.00 +0.6 0.57 perf-profile.calltrace.cycl= es-pp.timestamp_truncate.current_time.inode_needs_update_time.file_update_t= ime.shmem_file_write_iter > 75.78 +0.8 76.57 perf-profile.calltrace.cycl= es-pp.write > 61.25 +1.0 62.22 perf-profile.calltrace.cycl= es-pp.entry_SYSCALL_64_after_hwframe.write > 60.37 +1.0 61.35 perf-profile.calltrace.cycl= es-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.write > 57.56 +1.0 58.61 perf-profile.calltrace.cycl= es-pp.ksys_write.do_syscall_64.entry_SYSCALL_64_after_hwframe.write > 0.00 +1.1 1.11 =C2=B1 3% perf-profile.calltrace= .cycles-pp.coarse_ctime.current_time.inode_needs_update_time.file_update_ti= me.shmem_file_write_iter > 49.83 +1.2 51.07 perf-profile.calltrace.cycl= es-pp.vfs_write.ksys_write.do_syscall_64.entry_SYSCALL_64_after_hwframe.wri= te > 39.42 +1.4 40.86 perf-profile.calltrace.cycl= es-pp.shmem_file_write_iter.vfs_write.ksys_write.do_syscall_64.entry_SYSCAL= L_64_after_hwframe > 1.95 +2.1 4.04 perf-profile.calltrace.cycl= es-pp.inode_needs_update_time.file_update_time.shmem_file_write_iter.vfs_wr= ite.ksys_write > 2.33 +2.2 4.54 perf-profile.calltrace.cycl= es-pp.file_update_time.shmem_file_write_iter.vfs_write.ksys_write.do_syscal= l_64 > 0.00 +3.1 3.14 perf-profile.calltrace.cycl= es-pp.current_time.inode_needs_update_time.file_update_time.shmem_file_writ= e_iter.vfs_write > 27.11 -0.9 26.24 perf-profile.children.cycle= s-pp.llseek > 31.79 -0.6 31.15 perf-profile.children.cycle= s-pp.generic_perform_write > 9.30 =C2=B1 2% -0.4 8.87 =C2=B1 2% perf-profile.chil= dren.cycles-pp.ksys_lseek > 13.81 -0.3 13.47 perf-profile.children.cycle= s-pp.copy_page_from_iter_atomic > 12.33 -0.3 11.99 perf-profile.children.cycle= s-pp.clear_bhb_loop > 6.76 -0.2 6.56 perf-profile.children.cycle= s-pp.shmem_write_begin > 5.95 -0.2 5.77 perf-profile.children.cycle= s-pp.shmem_get_folio_gfp > 5.77 -0.1 5.64 perf-profile.children.cycle= s-pp.entry_SYSCALL_64 > 3.43 -0.1 3.34 perf-profile.children.cycle= s-pp.shmem_write_end > 3.51 -0.1 3.42 perf-profile.children.cycle= s-pp.__cond_resched > 3.34 -0.1 3.26 perf-profile.children.cycle= s-pp.mutex_lock > 2.37 -0.1 2.30 perf-profile.children.cycle= s-pp.filemap_get_entry > 2.01 -0.1 1.96 perf-profile.children.cycle= s-pp.fput > 1.57 -0.1 1.52 perf-profile.children.cycle= s-pp.rcu_all_qs > 1.68 -0.0 1.63 perf-profile.children.cycle= s-pp.down_write > 2.15 -0.0 2.10 perf-profile.children.cycle= s-pp.mutex_unlock > 0.11 =C2=B1 4% -0.0 0.07 =C2=B1 11% perf-profile.chil= dren.cycles-pp.ktime_get_update_offsets_now > 1.76 -0.0 1.72 perf-profile.children.cycle= s-pp.syscall_exit_to_user_mode > 1.02 -0.0 0.98 perf-profile.children.cycle= s-pp.folio_unlock > 0.89 -0.0 0.87 perf-profile.children.cycle= s-pp.folio_mark_accessed > 0.55 -0.0 0.52 perf-profile.children.cycle= s-pp.shmem_file_llseek > 0.77 -0.0 0.75 perf-profile.children.cycle= s-pp.folio_mark_dirty > 0.26 -0.0 0.25 perf-profile.children.cycle= s-pp.__f_unlock_pos > 0.22 +0.0 0.24 perf-profile.children.cycle= s-pp.inode_to_bdi > 0.00 +0.3 0.30 perf-profile.children.cycle= s-pp.set_normalized_timespec64 > 0.00 +0.4 0.42 =C2=B1 3% perf-profile.children.= cycles-pp.ns_to_timespec64 > 74.20 +0.5 74.66 perf-profile.children.cycle= s-pp.entry_SYSCALL_64_after_hwframe > 0.00 +0.5 0.47 perf-profile.children.cycle= s-pp.ktime_get_coarse_with_offset > 72.54 +0.5 73.04 perf-profile.children.cycle= s-pp.do_syscall_64 > 0.00 +0.6 0.56 =C2=B1 6% perf-profile.children.= cycles-pp.ktime_get_coarse_ts64 > 76.24 +0.8 77.00 perf-profile.children.cycle= s-pp.write > 57.97 +1.0 59.02 perf-profile.children.cycle= s-pp.ksys_write > 50.19 +1.2 51.42 perf-profile.children.cycle= s-pp.vfs_write > 0.00 +1.3 1.29 =C2=B1 2% perf-profile.children.= cycles-pp.coarse_ctime > 39.96 +1.4 41.40 perf-profile.children.cycle= s-pp.shmem_file_write_iter > 2.18 +2.0 4.17 perf-profile.children.cycle= s-pp.inode_needs_update_time > 2.52 +2.1 4.67 perf-profile.children.cycle= s-pp.file_update_time > 0.00 +3.5 3.46 perf-profile.children.cycle= s-pp.current_time > 1.13 -0.4 0.72 perf-profile.self.cycles-pp= .inode_needs_update_time > 13.62 -0.3 13.28 perf-profile.self.cycles-pp= .copy_page_from_iter_atomic > 12.19 -0.3 11.86 perf-profile.self.cycles-pp= .clear_bhb_loop > 2.31 -0.1 2.24 perf-profile.self.cycles-pp= .llseek > 4.25 -0.1 4.19 perf-profile.self.cycles-pp= .entry_SYSRETQ_unsafe_stack > 2.18 -0.1 2.12 perf-profile.self.cycles-pp= .shmem_get_folio_gfp > 2.18 -0.1 2.12 perf-profile.self.cycles-pp= .do_syscall_64 > 1.72 -0.1 1.67 perf-profile.self.cycles-pp= .filemap_get_entry > 1.88 -0.0 1.83 perf-profile.self.cycles-pp= .fput > 2.14 -0.0 2.09 perf-profile.self.cycles-pp= .mutex_lock > 2.01 -0.0 1.96 perf-profile.self.cycles-pp= .mutex_unlock > 1.66 -0.0 1.61 perf-profile.self.cycles-pp= .entry_SYSCALL_64_after_hwframe > 1.18 -0.0 1.14 perf-profile.self.cycles-pp= .rcu_all_qs > 0.54 =C2=B1 2% -0.0 0.50 perf-profile.self.cycl= es-pp.timestamp_truncate > 1.94 -0.0 1.90 perf-profile.self.cycles-pp= .__cond_resched > 1.09 -0.0 1.05 perf-profile.self.cycles-pp= .down_write > 1.56 -0.0 1.52 perf-profile.self.cycles-pp= .shmem_write_end > 0.10 =C2=B1 4% -0.0 0.07 =C2=B1 11% perf-profile.self= .cycles-pp.ktime_get_update_offsets_now > 1.62 -0.0 1.59 perf-profile.self.cycles-pp= .entry_SYSCALL_64 > 2.68 -0.0 2.65 perf-profile.self.cycles-pp= .__fsnotify_parent > 1.13 -0.0 1.10 perf-profile.self.cycles-pp= .syscall_exit_to_user_mode > 0.95 -0.0 0.92 perf-profile.self.cycles-pp= .folio_unlock > 0.83 -0.0 0.81 perf-profile.self.cycles-pp= .folio_mark_accessed > 0.42 -0.0 0.41 perf-profile.self.cycles-pp= .shmem_file_llseek > 0.49 -0.0 0.47 perf-profile.self.cycles-pp= .balance_dirty_pages_ratelimited_flags > 0.34 -0.0 0.33 perf-profile.self.cycles-pp= .folio_mapping > 0.15 =C2=B1 3% +0.0 0.17 =C2=B1 2% perf-profile.self= .cycles-pp.inode_to_bdi > 0.39 +0.1 0.48 perf-profile.self.cycles-pp= .file_update_time > 0.00 +0.2 0.24 =C2=B1 3% perf-profile.self.cycl= es-pp.set_normalized_timespec64 > 0.00 +0.4 0.35 perf-profile.self.cycles-pp= .ns_to_timespec64 > 0.00 +0.4 0.41 =C2=B1 2% perf-profile.self.cycl= es-pp.ktime_get_coarse_with_offset > 0.00 +0.4 0.44 =C2=B1 8% perf-profile.self.cycl= es-pp.ktime_get_coarse_ts64 > 0.00 +0.6 0.55 perf-profile.self.cycles-pp= .coarse_ctime > 0.00 +0.9 0.89 perf-profile.self.cycles-pp= .current_time >=20 >=20 >=20 > *************************************************************************= ************************** > lkp-cpl-4sp2: 224 threads 4 sockets Intel(R) Xeon(R) Platinum 8380H CPU @= 2.90GHz (Cooper Lake) with 192G memory > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > compiler/cpufreq_governor/kconfig/mode/nr_task/rootfs/tbox_group/test/tes= tcase: > gcc-12/performance/x86_64-rhel-8.3/thread/100%/debian-12-x86_64-2024020= 6.cgz/lkp-cpl-4sp2/pipe1/will-it-scale >=20 > commit:=20 > v6.11-rc1 > a037d5e7f8 ("fs: add infrastructure for multigrain timestamps") >=20 > v6.11-rc1 a037d5e7f81bae8ff69eb670b2e=20 > ---------------- ---------------------------=20 > %stddev %change %stddev > \ | \ =20 > 101460 =C2=B1 14% -40.5% 60365 =C2=B1 54% numa-numastat.nod= e2.other_node > 101460 =C2=B1 14% -40.5% 60365 =C2=B1 54% numa-vmstat.node2= .numa_other > 214956 +1.3% 217796 proc-vmstat.nr_shmem > 2.864e+08 -5.5% 2.706e+08 will-it-scale.224.threads > 1278568 -5.5% 1207975 will-it-scale.per_thread_op= s > 2.864e+08 -5.5% 2.706e+08 will-it-scale.workload > 0.29 =C2=B1220% +486.6% 1.70 =C2=B1 71% perf-sched.sch_de= lay.avg.ms.schedule_hrtimeout_range_clock.ep_poll.do_epoll_wait.__x64_sys_e= poll_wait > 0.14 =C2=B1104% +238.3% 0.49 =C2=B1 26% perf-sched.sch_de= lay.avg.ms.wait_for_partner.fifo_open.do_dentry_open.vfs_open > 4.33 =C2=B1196% +426.6% 22.79 =C2=B1 55% perf-sched.sch_de= lay.max.ms.devkmsg_read.vfs_read.ksys_read.do_syscall_64 > 15.03 =C2=B1101% +153.2% 38.05 =C2=B1 19% perf-sched.total_= sch_delay.max.ms > 0.07 =C2=B1141% +201.5% 0.20 perf-sched.wait_and_de= lay.avg.ms.schedule_hrtimeout_range_clock.usleep_range_state.ipmi_thread.kt= hread > 0.70 =C2=B1100% +137.6% 1.66 =C2=B1 20% perf-sched.wait_a= nd_delay.avg.ms.schedule_timeout.__wait_for_common.wait_for_completion_stat= e.kernel_clone > 35.83 =C2=B1143% +476.3% 206.50 =C2=B1 56% perf-sched.wait_a= nd_delay.count.schedule_hrtimeout_range_clock.usleep_range_state.ipmi_threa= d.kthread > 1.40 =C2=B1100% +116.9% 3.03 =C2=B1 6% perf-sched.wait_t= ime.avg.ms.rcu_gp_kthread.kthread.ret_from_fork.ret_from_fork_asm > 2.43 =C2=B1143% +472.3% 13.90 =C2=B1 56% perf-sched.wait_t= ime.avg.ms.schedule_hrtimeout_range_clock.do_select.core_sys_select.do_psel= ect.constprop > 0.64 =C2=B1100% +137.0% 1.51 =C2=B1 20% perf-sched.wait_t= ime.avg.ms.schedule_timeout.__wait_for_common.wait_for_completion_state.ker= nel_clone > 0.14 =C2=B1104% +229.2% 0.47 =C2=B1 25% perf-sched.wait_t= ime.avg.ms.wait_for_partner.fifo_open.do_dentry_open.vfs_open > 2.33 =C2=B1101% +114.3% 4.99 perf-sched.wait_time.m= ax.ms.rcu_gp_kthread.kthread.ret_from_fork.ret_from_fork_asm > 4.45 =C2=B1143% +600.8% 31.18 =C2=B1 64% perf-sched.wait_t= ime.max.ms.schedule_hrtimeout_range_clock.do_select.core_sys_select.do_psel= ect.constprop > 1.612e+11 +1.6% 1.638e+11 perf-stat.i.branch-instruct= ions > 0.42 =C2=B1 2% -0.1 0.37 perf-stat.i.branch-mis= s-rate% > 6.629e+08 -9.0% 6.034e+08 perf-stat.i.branch-misses > 13.77 =C2=B1 3% +1.3 15.04 =C2=B1 6% perf-stat.i.cache= -miss-rate% > 1.12 -2.6% 1.09 perf-stat.i.cpi > 6.635e+11 +2.9% 6.827e+11 perf-stat.i.instructions > 0.89 +2.7% 0.91 perf-stat.i.ipc > 0.41 -0.0 0.37 perf-stat.overall.branch-mi= ss-rate% > 1.12 -2.6% 1.09 perf-stat.overall.cpi > 0.89 +2.6% 0.91 perf-stat.overall.ipc > 702093 +8.3% 760166 perf-stat.overall.path-leng= th > 1.604e+11 +1.7% 1.631e+11 perf-stat.ps.branch-instruc= tions > 6.591e+08 -8.9% 6.005e+08 perf-stat.ps.branch-misses > 6.6e+11 +2.9% 6.795e+11 perf-stat.ps.instructions > 2.011e+14 +2.3% 2.057e+14 perf-stat.total.instruction= s > 5.47 -0.4 5.04 perf-profile.calltrace.cycl= es-pp.copy_page_from_iter.pipe_write.vfs_write.ksys_write.do_syscall_64 > 6.78 -0.4 6.37 perf-profile.calltrace.cycl= es-pp.clear_bhb_loop.write > 6.84 -0.3 6.51 perf-profile.calltrace.cycl= es-pp.clear_bhb_loop.read > 2.72 =C2=B1 2% -0.3 2.41 =C2=B1 8% perf-profile.call= trace.cycles-pp.apparmor_file_permission.security_file_permission.rw_verify= _area.vfs_read.ksys_read > 5.38 -0.3 5.08 perf-profile.calltrace.cycl= es-pp.entry_SYSCALL_64.read > 5.36 -0.3 5.08 perf-profile.calltrace.cycl= es-pp.entry_SYSCALL_64.write > 3.98 -0.3 3.70 perf-profile.calltrace.cycl= es-pp._copy_from_iter.copy_page_from_iter.pipe_write.vfs_write.ksys_write > 4.16 -0.2 3.96 perf-profile.calltrace.cycl= es-pp.copy_page_to_iter.pipe_read.vfs_read.ksys_read.do_syscall_64 > 52.78 -0.2 52.59 perf-profile.calltrace.cycl= es-pp.write > 2.09 -0.1 1.95 perf-profile.calltrace.cycl= es-pp.__wake_up_sync_key.pipe_write.vfs_write.ksys_write.do_syscall_64 > 1.04 -0.1 0.90 perf-profile.calltrace.cycl= es-pp.mutex_unlock.pipe_write.vfs_write.ksys_write.do_syscall_64 > 3.17 -0.1 3.06 perf-profile.calltrace.cycl= es-pp._copy_to_iter.copy_page_to_iter.pipe_read.vfs_read.ksys_read > 1.56 -0.1 1.46 perf-profile.calltrace.cycl= es-pp._raw_spin_lock_irqsave.__wake_up_sync_key.pipe_write.vfs_write.ksys_w= rite > 0.99 -0.1 0.90 perf-profile.calltrace.cycl= es-pp.x64_sys_call.do_syscall_64.entry_SYSCALL_64_after_hwframe.read > 1.67 -0.1 1.59 perf-profile.calltrace.cycl= es-pp.mutex_lock.pipe_write.vfs_write.ksys_write.do_syscall_64 > 0.64 -0.1 0.56 perf-profile.calltrace.cycl= es-pp.testcase > 0.96 -0.1 0.89 perf-profile.calltrace.cycl= es-pp.x64_sys_call.do_syscall_64.entry_SYSCALL_64_after_hwframe.write > 1.24 -0.1 1.17 perf-profile.calltrace.cycl= es-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwfram= e.read > 1.24 -0.1 1.17 perf-profile.calltrace.cycl= es-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwfram= e.write > 0.98 -0.1 0.92 perf-profile.calltrace.cycl= es-pp.fput.ksys_read.do_syscall_64.entry_SYSCALL_64_after_hwframe.read > 1.15 -0.1 1.10 perf-profile.calltrace.cycl= es-pp.fput.ksys_write.do_syscall_64.entry_SYSCALL_64_after_hwframe.write > 0.94 -0.1 0.88 perf-profile.calltrace.cycl= es-pp.mutex_unlock.pipe_read.vfs_read.ksys_read.do_syscall_64 > 50.06 +0.1 50.13 perf-profile.calltrace.cycl= es-pp.read > 1.64 +0.1 1.75 perf-profile.calltrace.cycl= es-pp.mutex_lock.pipe_read.vfs_read.ksys_read.do_syscall_64 > 0.59 =C2=B1 6% +0.3 0.86 =C2=B1 2% perf-profile.call= trace.cycles-pp.anon_pipe_buf_release.pipe_read.vfs_read.ksys_read.do_sysca= ll_64 > 0.25 =C2=B1100% +0.4 0.68 =C2=B1 2% perf-profile.call= trace.cycles-pp.__cond_resched.mutex_lock.pipe_read.vfs_read.ksys_read > 0.00 +0.5 0.54 =C2=B1 2% perf-profile.calltrace= .cycles-pp.timestamp_truncate.current_time.inode_needs_update_time.file_upd= ate_time.pipe_write > 0.00 +0.6 0.58 =C2=B1 3% perf-profile.calltrace= .cycles-pp.ktime_get_coarse_ts64.coarse_ctime.current_time.atime_needs_upda= te.touch_atime > 35.68 +0.8 36.45 perf-profile.calltrace.cycl= es-pp.entry_SYSCALL_64_after_hwframe.write > 34.78 +0.8 35.61 perf-profile.calltrace.cycl= es-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.write > 0.00 +1.0 1.03 =C2=B1 14% perf-profile.calltrace= .cycles-pp.coarse_ctime.current_time.inode_needs_update_time.file_update_ti= me.pipe_write > 32.94 +1.0 33.97 perf-profile.calltrace.cycl= es-pp.entry_SYSCALL_64_after_hwframe.read > 32.05 +1.1 33.13 perf-profile.calltrace.cycl= es-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.read > 30.48 +1.1 31.57 perf-profile.calltrace.cycl= es-pp.ksys_write.do_syscall_64.entry_SYSCALL_64_after_hwframe.write > 25.24 +1.2 26.49 perf-profile.calltrace.cycl= es-pp.vfs_write.ksys_write.do_syscall_64.entry_SYSCALL_64_after_hwframe.wri= te > 27.80 +1.4 29.19 perf-profile.calltrace.cycl= es-pp.ksys_read.do_syscall_64.entry_SYSCALL_64_after_hwframe.read > 0.00 +1.4 1.42 =C2=B1 4% perf-profile.calltrace= .cycles-pp.coarse_ctime.current_time.atime_needs_update.touch_atime.pipe_re= ad > 22.79 +1.5 24.30 perf-profile.calltrace.cycl= es-pp.vfs_read.ksys_read.do_syscall_64.entry_SYSCALL_64_after_hwframe.read > 17.46 +1.6 19.04 perf-profile.calltrace.cycl= es-pp.pipe_write.vfs_write.ksys_write.do_syscall_64.entry_SYSCALL_64_after_= hwframe > 14.85 +2.2 17.02 perf-profile.calltrace.cycl= es-pp.pipe_read.vfs_read.ksys_read.do_syscall_64.entry_SYSCALL_64_after_hwf= rame > 4.02 +2.3 6.29 perf-profile.calltrace.cycl= es-pp.touch_atime.pipe_read.vfs_read.ksys_read.do_syscall_64 > 1.72 +2.3 4.02 perf-profile.calltrace.cycl= es-pp.current_time.atime_needs_update.touch_atime.pipe_read.vfs_read > 3.63 +2.3 5.94 perf-profile.calltrace.cycl= es-pp.atime_needs_update.touch_atime.pipe_read.vfs_read.ksys_read > 1.78 =C2=B1 8% +2.3 4.12 =C2=B1 4% perf-profile.call= trace.cycles-pp.inode_needs_update_time.file_update_time.pipe_write.vfs_wri= te.ksys_write > 2.06 =C2=B1 8% +2.3 4.41 =C2=B1 3% perf-profile.call= trace.cycles-pp.file_update_time.pipe_write.vfs_write.ksys_write.do_syscall= _64 > 0.00 +3.3 3.34 =C2=B1 5% perf-profile.calltrace= .cycles-pp.current_time.inode_needs_update_time.file_update_time.pipe_write= .vfs_write > 13.71 -0.7 12.97 perf-profile.children.cycle= s-pp.clear_bhb_loop > 5.60 -0.4 5.15 perf-profile.children.cycle= s-pp.copy_page_from_iter > 6.87 -0.4 6.48 perf-profile.children.cycle= s-pp.entry_SYSCALL_64 > 4.28 -0.3 3.98 perf-profile.children.cycle= s-pp._copy_from_iter > 4.06 -0.2 3.83 perf-profile.children.cycle= s-pp.entry_SYSRETQ_unsafe_stack > 4.23 -0.2 4.03 perf-profile.children.cycle= s-pp.copy_page_to_iter > 2.06 -0.2 1.86 perf-profile.children.cycle= s-pp.mutex_unlock > 52.98 -0.2 52.80 perf-profile.children.cycle= s-pp.write > 2.14 -0.2 1.96 perf-profile.children.cycle= s-pp.x64_sys_call > 2.18 -0.2 2.03 perf-profile.children.cycle= s-pp.__wake_up_sync_key > 2.60 -0.1 2.45 perf-profile.children.cycle= s-pp.syscall_exit_to_user_mode > 3.43 -0.1 3.31 perf-profile.children.cycle= s-pp._copy_to_iter > 2.14 -0.1 2.02 perf-profile.children.cycle= s-pp.fput > 1.58 -0.1 1.48 perf-profile.children.cycle= s-pp._raw_spin_lock_irqsave > 0.74 =C2=B1 2% -0.1 0.65 perf-profile.children.= cycles-pp.testcase > 0.77 -0.0 0.74 perf-profile.children.cycle= s-pp.entry_SYSCALL_64_safe_stack > 0.36 -0.0 0.34 perf-profile.children.cycle= s-pp.__x64_sys_read > 0.39 -0.0 0.37 perf-profile.children.cycle= s-pp.__x64_sys_write > 0.14 =C2=B1 3% -0.0 0.12 =C2=B1 6% perf-profile.chil= dren.cycles-pp.make_vfsuid > 0.26 -0.0 0.24 =C2=B1 2% perf-profile.children.= cycles-pp.__wake_up_common > 0.30 -0.0 0.28 perf-profile.children.cycle= s-pp._raw_spin_unlock_irqrestore > 0.42 =C2=B1 3% +0.1 0.48 =C2=B1 7% perf-profile.chil= dren.cycles-pp.rcu_all_qs > 50.24 +0.1 50.31 perf-profile.children.cycle= s-pp.read > 1.36 =C2=B1 3% +0.1 1.44 =C2=B1 2% perf-profile.chil= dren.cycles-pp.rep_movs_alternative > 1.27 +0.1 1.41 perf-profile.children.cycle= s-pp.__cond_resched > 0.60 =C2=B1 6% +0.3 0.86 =C2=B1 2% perf-profile.chil= dren.cycles-pp.anon_pipe_buf_release > 0.00 +0.5 0.54 =C2=B1 8% perf-profile.children.= cycles-pp.set_normalized_timespec64 > 0.46 =C2=B1 10% +0.7 1.11 =C2=B1 3% perf-profile.chil= dren.cycles-pp.timestamp_truncate > 0.00 +0.9 0.94 =C2=B1 2% perf-profile.children.= cycles-pp.ktime_get_coarse_with_offset > 0.00 +1.0 0.98 =C2=B1 7% perf-profile.children.= cycles-pp.ktime_get_coarse_ts64 > 0.00 +1.0 1.04 perf-profile.children.cycle= s-pp.ns_to_timespec64 > 30.73 +1.1 31.81 perf-profile.children.cycle= s-pp.ksys_write > 25.42 +1.2 26.66 perf-profile.children.cycle= s-pp.vfs_write > 27.96 +1.4 29.32 perf-profile.children.cycle= s-pp.ksys_read > 22.87 +1.5 24.37 perf-profile.children.cycle= s-pp.vfs_read > 17.63 +1.6 19.19 perf-profile.children.cycle= s-pp.pipe_write > 68.90 +1.8 70.69 perf-profile.children.cycle= s-pp.entry_SYSCALL_64_after_hwframe > 67.06 +1.9 68.95 perf-profile.children.cycle= s-pp.do_syscall_64 > 15.15 +2.1 17.30 perf-profile.children.cycle= s-pp.pipe_read > 4.10 +2.3 6.37 perf-profile.children.cycle= s-pp.touch_atime > 3.73 +2.3 6.02 perf-profile.children.cycle= s-pp.atime_needs_update > 1.90 =C2=B1 8% +2.3 4.20 =C2=B1 4% perf-profile.chil= dren.cycles-pp.inode_needs_update_time > 2.13 =C2=B1 7% +2.3 4.48 =C2=B1 3% perf-profile.chil= dren.cycles-pp.file_update_time > 0.00 +2.5 2.52 =C2=B1 5% perf-profile.children.= cycles-pp.coarse_ctime > 1.76 +6.0 7.77 =C2=B1 2% perf-profile.children.= cycles-pp.current_time > 13.63 -0.7 12.89 perf-profile.self.cycles-pp= .clear_bhb_loop > 1.07 =C2=B1 5% -0.4 0.67 perf-profile.self.cycl= es-pp.inode_needs_update_time > 3.81 -0.3 3.46 perf-profile.self.cycles-pp= ._copy_from_iter > 4.10 =C2=B1 2% -0.3 3.80 =C2=B1 2% perf-profile.self= .cycles-pp.vfs_read > 3.58 -0.2 3.34 perf-profile.self.cycles-pp= .pipe_read > 3.35 -0.2 3.11 perf-profile.self.cycles-pp= .read > 3.92 -0.2 3.70 perf-profile.self.cycles-pp= .entry_SYSRETQ_unsafe_stack > 2.90 -0.2 2.70 perf-profile.self.cycles-pp= .do_syscall_64 > 1.97 -0.2 1.78 perf-profile.self.cycles-pp= .mutex_unlock > 3.00 -0.2 2.81 perf-profile.self.cycles-pp= .entry_SYSCALL_64 > 2.90 -0.2 2.74 perf-profile.self.cycles-pp= ._copy_to_iter > 1.71 -0.2 1.54 perf-profile.self.cycles-pp= .atime_needs_update > 2.01 -0.2 1.85 perf-profile.self.cycles-pp= .x64_sys_call > 3.39 -0.1 3.25 perf-profile.self.cycles-pp= .write > 1.32 =C2=B1 2% -0.1 1.18 perf-profile.self.cycl= es-pp.copy_page_from_iter > 1.87 -0.1 1.76 perf-profile.self.cycles-pp= .entry_SYSCALL_64_after_hwframe > 2.15 -0.1 2.05 perf-profile.self.cycles-pp= .mutex_lock > 1.66 -0.1 1.56 perf-profile.self.cycles-pp= .syscall_exit_to_user_mode > 1.89 -0.1 1.79 perf-profile.self.cycles-pp= .fput > 0.61 -0.1 0.51 perf-profile.self.cycles-pp= .testcase > 1.52 -0.1 1.43 perf-profile.self.cycles-pp= ._raw_spin_lock_irqsave > 0.81 =C2=B1 4% -0.1 0.72 =C2=B1 2% perf-profile.self= .cycles-pp.copy_page_to_iter > 1.16 -0.1 1.11 perf-profile.self.cycles-pp= .ksys_write > 1.06 -0.0 1.02 perf-profile.self.cycles-pp= .ksys_read > 0.36 -0.0 0.32 perf-profile.self.cycles-pp= .__wake_up_sync_key > 0.37 -0.0 0.34 =C2=B1 2% perf-profile.self.cycl= es-pp.touch_atime > 0.76 -0.0 0.73 perf-profile.self.cycles-pp= .entry_SYSCALL_64_safe_stack > 0.26 -0.0 0.24 perf-profile.self.cycles-pp= .__wake_up_common > 0.27 -0.0 0.26 perf-profile.self.cycles-pp= .__x64_sys_read > 0.27 -0.0 0.26 perf-profile.self.cycles-pp= ._raw_spin_unlock_irqrestore > 0.12 =C2=B1 3% -0.0 0.11 =C2=B1 3% perf-profile.self= .cycles-pp.make_vfsuid > 0.29 -0.0 0.28 perf-profile.self.cycles-pp= .__x64_sys_write > 0.26 +0.0 0.30 perf-profile.self.cycles-pp= .file_update_time > 0.30 =C2=B1 4% +0.1 0.37 =C2=B1 9% perf-profile.self= .cycles-pp.rcu_all_qs > 0.85 +0.1 0.92 perf-profile.self.cycles-pp= .__cond_resched > 0.89 =C2=B1 5% +0.1 0.99 =C2=B1 4% perf-profile.self= .cycles-pp.rep_movs_alternative > 0.56 =C2=B1 6% +0.3 0.82 =C2=B1 2% perf-profile.self= .cycles-pp.anon_pipe_buf_release > 0.00 +0.5 0.52 =C2=B1 7% perf-profile.self.cycl= es-pp.set_normalized_timespec64 > 0.43 =C2=B1 10% +0.6 1.03 =C2=B1 3% perf-profile.self= .cycles-pp.timestamp_truncate > 0.00 +0.9 0.86 perf-profile.self.cycles-pp= .ns_to_timespec64 > 0.00 +0.9 0.90 =C2=B1 2% perf-profile.self.cycl= es-pp.ktime_get_coarse_with_offset > 0.00 +0.9 0.94 =C2=B1 8% perf-profile.self.cycl= es-pp.ktime_get_coarse_ts64 > 0.00 +1.0 1.02 =C2=B1 4% perf-profile.self.cycl= es-pp.coarse_ctime > 1.06 =C2=B1 4% +1.3 2.38 =C2=B1 3% perf-profile.self= .cycles-pp.current_time >=20 >=20 >=20 >=20 >=20 > Disclaimer: > Results have been estimated based on internal Intel analysis and are prov= ided > for informational purposes only. Any difference in system hardware or sof= tware > design or configuration may affect actual performance. >=20 >=20 --=20 Jeff Layton