ward@cfa.harvard.EDU (Steve Ward) (11/11/88)
Subject: Re: 32K "do-it-yourself" kits... Newsgroups: comp.sys.nsc.32k Keywords: The Ultimate Scrooge... Summary: why no MMU for SOME situations homogenous software environs, passive backplane References: <2613@sultra.UUCP> <2629@sultra.UUCP> ... MMU: The point about MMU's being great to protect the system from the errant process is true, true, true. Of course, one also gets virtual memory out of it, too. This can be bad for some things. I do realtime work and predictable, fast context switches are a must. I measure these in low microseconds. Virtual memory inflicts non-deterministic latencies and if process parts are swapping, quite often many milliseconds of overhead. Sometimes one can just turn off the MMU. I am not familiar with the 32K family to know what is possible and with what effect. No MMU IN EFFECT would be just as good as no MMU physically. Maybe one design will work for both user camps. I prefer a single homogenous software environment using only the selected 32K cpu. Still, the IBM PC bus and I/O harware can be utilized by designing the CPU board to function with an IBM bus passive backplane (you get them for $75 - $99), acting as the IBM PC bus master. This gives you one software environment, but uses all that PC hardware. Just pop out your PC mom board and pop in a passive backplane. The first place to start is with a real design and I have only heard D. Rand offer one, so far. I guess we'll see what pops up in the next few days. For the record, nobody responded to my posted message offering some CAE/CAD service toward a board project - I called for direct email to me on it. Two messages have been posted here that touch on my message. I am not sure there is much interest in making a board, though there may be interest in using a board, providing SOMEBODY ELSE makes it, of course. Also, the real doers may already have made something as there have been cryptic references to some secret project. And, of course, a few days isn't enough time for all interest to be expressed. Wait and see, I guess. Steve W. ward@cfa.harvard.edu
dtynan@sultra.UUCP (Der Tynan) (11/12/88)
In article <1268@cfa.cfa.harvard.EDU>, ward@cfa.harvard.EDU (Steve Ward) writes: > > I prefer a single homogenous software environment using only the > selected 32K cpu. Still, the IBM PC bus and I/O harware can be > utilized by designing the CPU board to function with an IBM bus > passive backplane (you get them for $75 - $99), acting as the IBM > PC bus master. This gives you one software environment, but uses > all that PC hardware. Just pop out your PC mom board and pop in > a passive backplane. This isn't a bad idea either. The data I have received so far (Thanks to Steve Ligett) says that a bare PC board would cost ~$200 (including setup) if we could get 20 people interested. I still can't decide between a "plug-in" card, and a complete motherboard replacement. I guess I'm currently favoring the motherboard idea. Maybe some sort of vote is in order? A straight "co-processor" arrangement is a simpler design (read cheaper), because such nasties as DMA, etc, are already provided by the host 8088. > The first place to start is with a real design and I have only > heard D. Rand offer one, so far. Yeah, Dave how about some pricing (preliminary or otherwise) info? > For the record, nobody responded to > my posted message offering some CAE/CAD service toward a board > project - I called for direct email to me on it. Two messages > have been posted here that touch on my message. I am not sure > there is much interest in making a board, though there may be > interest in using a board, providing SOMEBODY ELSE makes it, > Steve W. ward@cfa.harvard.edu I guess you didn't get my mail. Hence I'm posting this. If you like, you can pretend I e-mailed it to you :-) I don't mind doing the design. Provided I get help/support. On top of that, the one thing I hate worse than wirewrapping, is slaving over schematics. If someone else wants the task of schematic generation, I'll happily send him/her some drawings on tissue paper :-) - Der -- dtynan@sultra.UUCP (Dermot Tynan @ Tynan Computers) {mips,pyramid}!sultra!dtynan --- God invented alcohol to keep the Irish from taking over the planet ---
gbs@stratus.UUCP (George B. Smith) (11/12/88)
In article <1268@cfa.cfa.harvard.EDU> ward@cfa.harvard.EDU (Steve Ward) writes: >Also, the real doers may already have made something >as there have been cryptic references to some secret project. > >Steve W. ward@cfa.harvard.edu The cryptic reference was to a board being done by the cg16 app group, also known as the National Imaging Group. It has already been announced to the Field Application Engineers, aka FAEs, so I don't know why Dave Rand is being so shy about it. George B. Smith disclaimer: I thought that the cg16 group Stratus Computer, Inc. would be anxious to talk about this gbs@stratus.stratus.com project, but I guess I was wrong.
dlr@daver.UUCP (Dave Rand) (11/13/88)
In article <2644@sultra.UUCP> dtynan@sultra.UUCP (Der Tynan) writes: >This isn't a bad idea either. The data I have received so far (Thanks to >Steve Ligett) says that a bare PC board would cost ~$200 (including setup) >if we could get 20 people interested. I still can't decide between a >"plug-in" card, and a complete motherboard replacement. I guess I'm >currently favoring the motherboard idea. I don't think you can blow a full PC-Motherboard board for $200. A PC Plug-in would be possible to do for that price, or even less. >> The first place to start is with a real design and I have only >> heard D. Rand offer one, so far. > >Yeah, Dave how about some pricing (preliminary or otherwise) info? The PD-32 design is free. It has been published in Micro-C, and reprints are available from them. The software is not free :-( You can get the hardware together for about $200, without PCB. The expensive bit is the RAMS. You can run in 1 Meg, and that would cost (roughly) $400. Another Meg - another $400. Fully assembled and tested boards were available from Definicon, but I'm not sure if they are any longer. The Zaiaz offer was probably one of the best deals to come along in a long time for 32k - how about checking them out again? A few last questions - do people _really_ want a full Unix environment? How much can you afford for this? Do you just want a C compiler, assembler, debugger, etc. that runs on (and for) a 32k processor? How much can you afford for this? (These are not idle questions...) Disclaimer: These are all personal opinions. -- Dave Rand {pyramid|hoptoad|sun|vsi1}!daver!dlr
adh@anumb.UUCP (a.d.hay) (11/15/88)
In article <2644@sultra.UUCP> dtynan@sultra.UUCP (Der Tynan) writes: -->In article <1268@cfa.cfa.harvard.EDU>, ward@cfa.harvard.EDU (Steve Ward) writes: -->> [] -->> --> I still can't decide between a -->"plug-in" card, and a complete motherboard replacement. I guess I'm -->currently favoring the motherboard idea. Maybe some sort of vote is in -->order? i vote for the motherboard plan. -->> For the record, nobody responded to -->> my posted message offering some CAE/CAD service toward a board -->> project - I called for direct email to me on it. --> -->> Steve W. ward@cfa.harvard.edu --> -->I guess you didn't get my mail. Hence I'm posting this. If you like, -->you can pretend I e-mailed it to you :-) -->-- --> dtynan@sultra.UUCP (Dermot Tynan @ Tynan Computers) likewise, i tried email and it bounced. -- Andrew Hay +------------------------------------------------------+ Holistic Specialist | I will design a computer for you, so powerful that | AT&T-BL Ward Hill MA | organic life will form part of its operational matrix| mvuxq.att.com!adh +------------------------------------------------------+
rfg@nsc.nsc.com (Ron Guilmette) (11/17/88)
In article <7465@daver.UUCP> dlr@daver.UUCP (Dave Rand) writes: >A few last questions - do people _really_ want a full Unix environment? >How much can you afford for this? Do you just want a C compiler, assembler, >debugger, etc. that runs on (and for) a 32k processor? How much can >you afford for this? (These are not idle questions...) Dear Dave, Some of these are idle questions! I can give you a FREE (well, copylefted) C compiler for the 32000's TODAY! It's called GNU C. I also have GNU C++, and the GNU assembler (GAS) has already been rehosted & retargeted for the 32000. GAS needs more work though. Any volunteers? I'll gladly send you all the new GAS code I have and let anybody have a whack at it. I would be working on it myself, but I don't have the time. If you want to help test & fix GAS/32000, I suggest that you have access to a Sequent-Balance, Encore-Multimax, or some other BSD (i.e. a.out) based 32000 system for you testing base. P.S. Also available ios the GNU loader GLD. Also FREE (and copylefted). many other binary utilities are also available (such as size, strip, etc., and many general utilities such as gmake, gnews, gdiff, gawk, ggrep). Oh yes! And GNU Emacs of course. -- Ron Guilmette National SemiConductor, 1135 Kern Ave. M/S 7C-266; Sunnyvale, CA 94086 Internet: rfg@nsc.nsc.com or amdahl!nsc!rfg@ames.arc.nasa.gov Uucp: ...{pyramid,sun,amdahl,apple}!nsc!rfg
glennw@nsc.nsc.com (Glenn Weinberg) (11/18/88)
In article <7856@nsc.nsc.com> rfg@nsc.nsc.com.UUCP (Ron Guilmette) writes: >If you want to help test & fix GAS/32000, I suggest that you have access to >a Sequent-Balance, Encore-Multimax, or some other BSD (i.e. a.out) based >32000 system for you testing base. I'd just like to point out that both the BSD and System V versions of UMAX (the Multimax Un*x port) use COFF, not a.out. This was done to allow the same binaries (e.g., utilities) to run on both systems. As I recall, the only reason COFF was chosen over a.out was that the tools Encore decided to use already generated COFF. -- Glenn Weinberg Email: glennw@nsc.nsc.com National Semiconductor Corporation Phone: (408) 721-8102 (My opinions are strictly my own, but you can borrow them if you want.)