From owner-freebsd-current Sat Aug 24 10:48:51 1996 Return-Path: owner-current Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id KAA19927 for current-outgoing; Sat, 24 Aug 1996 10:48:51 -0700 (PDT) Received: from rover.village.org (rover.village.org [204.144.255.49]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id KAA19918 for ; Sat, 24 Aug 1996 10:48:47 -0700 (PDT) Received: from rover.village.org (localhost [127.0.0.1]) by rover.village.org (8.7.5/8.6.6) with ESMTP id LAA02804 for ; Sat, 24 Aug 1996 11:48:45 -0600 (MDT) Message-Id: <199608241748.LAA02804@rover.village.org> To: current@freebsd.org Subject: Speedingup the "worldstone" Date: Sat, 24 Aug 1996 11:48:45 -0600 From: Warner Losh Sender: owner-current@freebsd.org X-Loop: FreeBSD.org Precedence: bulk Greetings, is there a FAQ on speeding up the worldstone benchmark? I have an old 486 DX2-66 that I'd like to get the most out of when it comes to building the world, since it takes just about exactly 9 hours (for cvs-cur CTM #2389)[*] to do a make world. Are there some easy things that I can do? I have my source on a jaz drive (10ms seek times for reads) and /usr/obj on a second drive (8.5ms Quantum) driven by an UltraStor 34F controller. The second drive is where /usr is mounted from also. Would I gain by a third drive that had /usr/obj on it? I have a 500M hard disk that I could use for this if it would be a big (>20%) win. I already have 32M of memory, so I'm not memory limited. Are there some options that would allow me to take advantage of this memory? If I run top while doing a make world, I see that gcc only uses about 20M or so of memory. Does anybody have a list of the typical bottle necks in this process? I know I can solve this problem by getting a good P6/200 with 64M of memory and a good Adaptech Ultra Wide PCI card and a fast wide scsi disk. But I don't have the $7500 (or even $5000) to do that just now :-). Since I only do a clean make world about once a month or so, this isn't a huge deal. [*] 9:01:32 was the time for this build. Warner P.S. Is there some way I can win if I know the compiler hasn't changed? make depend all install seems faster, but still takes at least 6 hr or so (it has been a while since I've done this since the builds take long enough that overnight is the only way I can do them, I opt for safety). P.P.S. If someone comes up with a clean way to include the CTM level in the uname string, would there generally be support for including it in the standard sources? I have an idea I'd like to explore, but don't want to waste my time if no one is keen on this idea....