From owner-cvs-all Mon Jun 24 19:43:37 2002 Delivered-To: cvs-all@freebsd.org Received: from cs.rice.edu (cs.rice.edu [128.42.1.30]) by hub.freebsd.org (Postfix) with ESMTP id 8AA2437B40C; Mon, 24 Jun 2002 19:43:14 -0700 (PDT) Received: from localhost (localhost [127.0.0.1]) by cs.rice.edu (Postfix) with ESMTP id 1236B4A9E2; Mon, 24 Jun 2002 21:43:14 -0500 (CDT) Received: by cs.rice.edu (Postfix, from userid 19572) id 8F23D4A9BB; Mon, 24 Jun 2002 21:43:13 -0500 (CDT) Date: Mon, 24 Jun 2002 21:43:13 -0500 From: Alan Cox To: Peter Wemm Cc: Ian Dowse , cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/powerpc/include types.h src/sys/arm/include types.h src/sys/i386/include types.h Message-ID: <20020625024313.GI11391@cs.rice.edu> References: <200206232157.g5NLvKF31769@freefall.freebsd.org> <20020624022343.3FDDB380E@overcee.wemm.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20020624022343.3FDDB380E@overcee.wemm.org> User-Agent: Mutt/1.3.28i X-Virus-Scanned: by AMaViS snapshot-20010714 Sender: owner-cvs-all@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG On Sun, Jun 23, 2002 at 07:23:43PM -0700, Peter Wemm wrote: > Ian Dowse wrote: > > iedowse 2002/06/23 14:57:19 PDT > > > > Modified files: > > sys/powerpc/include types.h > > sys/arm/include types.h > > sys/i386/include types.h > > Log: > > Make vm_pindex_t 64-bit on all platforms. This is necessary to avoid > > overflows with the large file sizes that UFS2 permits. > > Ouch, is this really necessary? The whole index::offset thing was > explicitly to avoid using 64 bit arithmetic here due to the performance > hit. Dyson posted some numbers somewhere in the distant past, they can > probably be found on the archives. > ... Since Ian is following up on some of the other issues related to this change, I ran a test using a script that performs three rounds of "make clean", "make depend", and "make" on a kernel source tree. On each round, the script times the final "make". The only variable in this test was the size of vm_pindex_t. In short, there was virtually no difference: /tmp/vm_pindex-32: 1135.28 real 1031.08 user 110.99 sys /tmp/vm_pindex-32: 1134.02 real 1029.93 user 111.84 sys /tmp/vm_pindex-32: 1137.11 real 1030.32 user 111.08 sys /tmp/vm_pindex-64: 1134.62 real 1031.13 user 110.43 sys /tmp/vm_pindex-64: 1133.16 real 1030.69 user 110.30 sys /tmp/vm_pindex-64: 1132.50 real 1029.74 user 111.01 sys Do you have any idea what sorts of tests John was running? Alan To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message