From owner-freebsd-hardware Thu Apr 20 12:17:43 1995 Return-Path: hardware-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id MAA06830 for hardware-outgoing; Thu, 20 Apr 1995 12:17:43 -0700 Received: from gndrsh.aac.dev.com (gndrsh.aac.dev.com [198.145.92.241]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id MAA06823 for ; Thu, 20 Apr 1995 12:17:38 -0700 Received: (from rgrimes@localhost) by gndrsh.aac.dev.com (8.6.11/8.6.9) id MAA06479; Thu, 20 Apr 1995 12:14:43 -0700 From: "Rodney W. Grimes" Message-Id: <199504201914.MAA06479@gndrsh.aac.dev.com> Subject: Re: PCI plug-n-play on Intel Premiere Baby II? To: ljo@po.CWRU.Edu Date: Thu, 20 Apr 1995 12:14:42 -0700 (PDT) Cc: fenner@parc.xerox.com, freebsd-hardware@FreeBSD.org In-Reply-To: <199504201743.MAA00831@amcell2.accumed.com> from "L Jonas Olsson" at Apr 20, 95 12:43:25 pm X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 3105 Sender: hardware-owner@FreeBSD.org Precedence: bulk > > The Intel Premiere II (aka Plato) uses shared ISA interrupts if you > don't mark enough ISA interrupts as available (vs "used by ISA card") > in the plug and play subportion of the BIOS setup. > > I've used three PCI cards all at different IRQs on this board. > (video, frame grabber, and NCR810). That does indeed work, if the cards are of different class (I have forgotten a lot of the details about this card since I haven't touched one in almost a year). The real big problem comes in when you stick 2 of the same ``class'' cards in the machine, they will end up on the same IRQ. The ``class'' I am refuring to is one of storage (disk controller), network, memory, graphics, etc. This created a severve problem when I first updated wcarchive to a PCI board (it was infact a Plato), in that the 2 Bt946's ended up sharing the same interrupt. At that time FreeBSD did not have support for this :-(. I never did get it working with dual BT946's, and ended up using AHA1542CF's to get the damn thing back on line. From a recent converstation with David Greenman he has not been able to make it work with more than 2 PCI scsi controllers. I have since done a lot of testing here on 2 Neptune based boards and found they work okay with 2 controllers, but add that third one and your disks shall be scrambled like eggs :-(. > I'm more worried about the things Rod have said about multiple bus > masters. I plan to use this board with NCR810 and an Imaging > Technology bus-mastering PCI frame grabber. You should be fine there, that is only 2 masters. Don't try to add a third, it will cause you grief. To the point that it is almost impossible to go multiuser without a panic, and if you do get past that point expect some scambled disks as soon as you hit 3 masters at once. I am currently testing the ASUS PCI/I-P54TP4 board with 4 NCR810 controllers and I can *NOT* make the thing crash. Tonights test while I sleep will be 4 concurrent make worlds running on 4 disks in chroot trees. (I'm going to have to pull memory from other boxes to expand this thing and build 3 new chroot test disks so It might not get started tonight, but that is the target). I will be writting up a complete ``Accurate Automation Motherboard Evaluation Report'' on this board once I am finished with all of my testing. But so far it is a ``just works'' kind of card. I did have a reboot problem, but that was fixed by my posted patch to -current. > > PS I've succesfully used the BitFlow Raptor PCI frame grabber under > FreeBSD. This was all in user mode using /dev/io to read some PCI > BIOS variables (memory address and IRQ line) and /dev/mem to mmap > the registers and memory (8MB region). This is shared memory access > and the read speed is ~15.5MB/s on 90MHz Plato. Was that from a ``C'' program using bcopy, or hand optimized assembler? I have done concurrent disk I/O on PCI now in excess of 12MB/sec using 4 controllers :-). -- Rod Grimes rgrimes@gndrsh.aac.dev.com Accurate Automation Company Custom computers for FreeBSD