From owner-freebsd-smp Mon Jan 7 4: 6:46 2002 Delivered-To: freebsd-smp@freebsd.org Received: from svr-ganmtc-appserv-mgmt.ncf.coxexpress.com (svr-ganmtc-appserv-mgmt.ncf.coxexpress.com [24.136.46.5]) by hub.freebsd.org (Postfix) with ESMTP id E2EAA37B420; Mon, 7 Jan 2002 04:06:23 -0800 (PST) Received: from darkstar.doublethink.cx (cpe-oca-24-136-59-202-cmcpe.ncf.coxexpress.com [24.136.59.202]) by svr-ganmtc-appserv-mgmt.ncf.coxexpress.com (8.11.4/8.11.4) with ESMTP id g07C6Ml29455; Mon, 7 Jan 2002 07:06:23 -0500 Received: by darkstar.doublethink.cx (Postfix, from userid 1000) id 105F52B99; Mon, 7 Jan 2002 07:06:22 -0500 (EST) Date: Mon, 7 Jan 2002 07:06:22 -0500 From: Chris Faulhaber To: peter@FreeBSD.org, scottl@FreeBSD.org Cc: smp@FreeBSD.org Subject: P5 vs. SMP, part 2 Message-ID: <20020107120621.GA23585@darkstar.doublethink.cx> Mime-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="sm4nu43k4a2Rpi4c" Content-Disposition: inline User-Agent: Mutt/1.3.24i Sender: owner-freebsd-smp@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.org --sm4nu43k4a2Rpi4c Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable I finally got around to narrowing down the commit (see below) that slowed by dual P5 box. From my original message to -smp: On Tue, 25 Sep 2001, Chris Faulhaber wrote: > I just updated my Dual Pentium (P54C) box from RELENG_4_3 to > RELENG_4 and have found that the box is significantly slower > when running an SMP kernel versus UP. Under 4.3, an SMP -j4 > buildworld took just over 4.5 hours. It now takes upwards > of 20 hours w/SMP and 6 (no -j) UP. When logging in and > running basic commands the system seems quite sluggish. Backing out this commit (with a slight conflict in /sys/sys/kernel.h) allows my box to build an up-to-date -stable world is less than 5 hours (instead of 20) again. Any chance someone with i386/smp-foo can look at this? Original message at: http://docs.freebsd.org/cgi/getmsg.cgi?fetch=3D33169+0+archive/2001/freebsd= -smp/20010930.freebsd-smp --=20 Chris D. Faulhaber - jedgar@fxp.org - jedgar@FreeBSD.org -------------------------------------------------------- FreeBSD: The Power To Serve - http://www.FreeBSD.org From: Scott Long To: cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: cvs commit: src/sys/sys kernel.h src/sys/kern init_main.c kern_descrip.c kern_kthread.c src/sys/i386/include md_var.h cpu.h src/sys/i386/i386 vm_machdep.c mp_machdep.c machdep.c locore.s Date: Fri, 15 Jun 2001 02:38:07 -0700 (PDT) scottl 2001/06/15 02:38:07 PDT Modified files: (Branch: RELENG_4) sys/sys kernel.h=20 sys/kern init_main.c kern_descrip.c kern_kthread.c=20 sys/i386/include md_var.h cpu.h=20 sys/i386/i386 vm_machdep.c mp_machdep.c machdep.c=20 locore.s=20 Log: MFC Peter Wemm's init and kthread enhancements from 08/11/2000 of -curren= t. This allows kthreads to be used by device drivers long before root is mou= nted and init is officially started. This also brings RAIDframe one step clos= er. =20 Reviewed by: peter =20 Revision Changes Path 1.63.2.1 +8 -7 src/sys/sys/kernel.h 1.134.2.6 +62 -100 src/sys/kern/init_main.c 1.81.2.8 +13 -7 src/sys/kern/kern_descrip.c 1.5.2.2 +2 -2 src/sys/kern/kern_kthread.c 1.35.2.1 +2 -1 src/sys/i386/include/md_var.h 1.43.2.2 +1 -2 src/sys/i386/include/cpu.h 1.132.2.3 +9 -8 src/sys/i386/i386/vm_machdep.c 1.115.2.8 +3 -12 src/sys/i386/i386/mp_machdep.c 1.385.2.12 +21 -1 src/sys/i386/i386/machdep.c 1.132.2.4 +1 -46 src/sys/i386/i386/locore.s --sm4nu43k4a2Rpi4c Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.6 (FreeBSD) Comment: FreeBSD: The Power To Serve iEYEARECAAYFAjw5jz0ACgkQObaG4P6BelDIMgCfdO1QI9waj3Qk5ePWGxC4BVJh FZkAoKKuFPQvLRe8Im/TVFGwIeNxJoFq =uSnL -----END PGP SIGNATURE----- --sm4nu43k4a2Rpi4c-- To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-smp" in the body of the message