From owner-freebsd-hackers Wed Jan 4 21:42:15 1995 Return-Path: hackers-owner Received: (from root@localhost) by freefall.cdrom.com (8.6.9/8.6.6) id VAA25619 for hackers-outgoing; Wed, 4 Jan 1995 21:42:15 -0800 Received: from Root.COM (implode.Root.COM [198.145.90.1]) by freefall.cdrom.com (8.6.9/8.6.6) with ESMTP id VAA25609 for ; Wed, 4 Jan 1995 21:42:11 -0800 Received: from corbin.Root.COM (corbin.Root.COM [198.145.90.18]) by Root.COM (8.6.8/8.6.5) with ESMTP id VAA19004; Wed, 4 Jan 1995 21:41:39 -0800 Received: from localhost (localhost [127.0.0.1]) by corbin.Root.COM (8.6.9/8.6.5) with SMTP id VAA00421; Wed, 4 Jan 1995 21:41:38 -0800 Message-Id: <199501050541.VAA00421@corbin.Root.COM> X-Authentication-Warning: corbin.Root.COM: Host localhost didn't use HELO protocol To: jim.bryant@whytel.com cc: hackers@freebsd.org Subject: Re: HELP!@# In-reply-to: Your message of "Wed, 04 Jan 95 20:47:16 CST." <9501042047.0T7BF00@whytel.com> From: David Greenman Reply-To: davidg@Root.COM Date: Wed, 04 Jan 1995 21:41:37 -0800 Sender: hackers-owner@freebsd.org Precedence: bulk > The speed of the OS seems to be a helluva lot slower than 1.1.5.1, under > 1.1.5.1 the nasa/ames pitest program runs in about 350 seconds, but under > 2.* it runs in 840+ seconds. I'll upload pitest.tar.gz to: > wcarchive.cdrom.com:/pub/incoming/FreeBSD > -- > config and dmesg follow: You have absolutely the *weirdest* config file I've ever seen. *Why* do you have all of those options? FASTLINKS? pseudo-device *pager? ...and a whole slew of stuff that you either don't need or don't want. Your performance problems are in part related to this - 4MB machines simply don't have enough memory to support this. > maxusers 32 Why? You need at least 16MB of memory to support that many users effectively. > maxfdescs 2048 Why? - deprecated a long time ago. > timezone 6 dst 5 Why? - deprecated a long time ago. > options BOUNCE_BUFFERS Why? Bounce buffer are only needed on systems with >16MB of memory and SCSI. > pseudo-device pty 32 Why? > options "NCONS=8" This is fine if you plan to use them all, but it requires a lot of memory. On a 4MB machine, *4* is more reasonable. > options SYSVSEM > options SYSVMSG Why? Do you actually have an application that needs Semaphore and SYSV messaging? This code in 2.0 is partly broken anyway. > options SYSVSHM > options SHMMAXPGS=64 Why? You can't possibly be running X in 4MB...do you have an application that needs SYSV shared memory support? > pseudo-device sl2 > pseudo-device ppp2 Why? Are you actually doing *both* SLIP and PPP at the same time on 2 serial ports at the same time? > options QUOTA Why? Do you actually need filesystem quota support? > options FASTLINKS Why? Deprecated a long time ago. > options "UPDATE_INTERVAL=1" Why? This actually will severely lower system performance. > pseudo-device devpager > pseudo-device swappager > pseudo-device vnodepager Why? Deprecated a long time ago. > pseudo-device gzip Okay if you plan to use it, but not normally necessary. > options "USER_LDT" Only needed for WINE support. > real memory = 4063232 (992 pages) > avail memory = 3002368 (733 pages) This is you problem. Basically, you're trying to generate a kernel that requires much more memory to work efficiently than you have. The reason that your program runs so slowly is that 1) The minimum 2.0 kernel is much larger than than the 1.1.5 kernel, and 2) you have a fat bloated kernel with options that most people will never use. I suggest that you start with the GENERIC kernel config file, remove device support that you don't need, and build a small custom kernel. -DG