jensen@utgpu.UUCP (04/30/87)
Perhaps I have missed the discussion but I'm hearing a number of rumours about bugs in the Intel 80386 when running under Unix/Xenix. Apparently the chips can be returned for new ones even. Has anyone ever run SCO Xenix on the 386? -- Mike Jensen
rab@rabtrs (Rich Bilancia) (05/03/87)
> Perhaps I have missed the discussion but I'm hearing a number of rumours about > bugs in the Intel 80386 when running under Unix/Xenix. Apparently the chips > can be returned for new ones even. Has anyone ever run SCO Xenix on the 386? > -- Mike Jensen Not only have I gotten SCO XENIX to run on a Compaq Deskpro/386 flawlessly, I've also installed an Intel Inboard/386 processor in an AT compatible and have that working with SCO XENIX. I also have written an article for the August issue of UNIX/World taking a first look at XENIX System V/386 in both of the above hardware environments. ^^^ Keep an eye open for it. --- "Adversity is the crucible from which greatness is forged." - Ken Baker ____ ____ ___ / _ \ / _ \ / _ \ Richard A. Bilancia / / \_/ / / \_/ _ \ \\/ Computer Guidance & Support / / / / _ | | \ \ P. O. Box 620127 \ \ _ \ \ | \ | + | \ \ Littleton, Colorado 80162 \ \_/ \ \ \_| \ |_| /\_| \ voice: (303) 973-4035 \____/ \____/ \____/ hao!{udenva|isis}!bilanc!rab Multi-User Accounting Solutions attmail!bilanc!rab
tanj@ucscc.UCSC.EDU (Michael Berman) (05/04/87)
> From: Michael Jensen <jensen@gpu.utcs.toronto.EDU> > Subject: Unix/Xenix on the 386 > Date: 30 Apr 87 18:45:07 GMT > To: info-unix@brl-sem.arpa > > Perhaps I have missed the discussion but I'm hearing a number of rumours about > bugs in the Intel 80386 when running under Unix/Xenix. Apparently the chips > can be returned for new ones even. Has anyone ever run SCO Xenix on the 386? > -- Mike Jensen > ----------------------------- Certain Intel 386 chips come up with the wrong answer when doing a 32 bit add. There is a fairly simple program available from Intel which tests for this bug. The current release of SCO Xenix is not effected by this bug because it is running 286 code--not 386. The new 386 OS release will/has this test program incorporated in it as part of the install procedure. Since the addition bug will adversely effect the OS, if the test program detects an error in the CPU the install will fail and notify the user that they should replace their CPU. Michael Berman
bobr@zeus.TEK.COM (Robert Reed) (05/04/87)
Certain Intel 386 chips come up with the wrong answer when doing a 32
bit add.
Funny, what I read is that in protected mode (and only protected mode),
there were 16 x 10^18th (exponent from memory and probably approximate)
pairs of operands to the multiply instruction which will generate the wrong
results. Perhaps there is more than one bug left in the 386.
--
Robert Reed, Tektronix CAE Systems Division, bobr@zeus.TEKtony@killer.UUCP (Tony Holden) (05/08/87)
in article <146@rabtrs>, rab@rabtrs (Rich Bilancia) says: > > I also have written an article for the August issue of UNIX/World taking a > first look at XENIX System V/386 in both of the above hardware environments. > ^^^ > Keep an eye open for it. Gee Rich, not that I won't look for and read your article, but how about a hint. Do we have to wait till August to at least find out if it's any good or not? Just a one liner might do. Tony Holden ihnp4!killer!tony
neighorn@qiclab.UUCP (Steven C. Neighorn) (05/09/87)
In article <1987Apr30.144507.18240@gpu.utcs.toronto.edu> jensen@gpu.utcs.toronto.edu (Michael Jensen) writes: >Perhaps I have missed the discussion but I'm hearing a number of rumours about >bugs in the Intel 80386 when running under Unix/Xenix. Apparently the chips >can be returned for new ones even. Has anyone ever run SCO Xenix on the 386? >-- Mike Jensen Intel has admitted to a problem with incorrect results on some multiply combinations in native 32-bit mode. These problems will not surface running msdos or '286 unix/xenix on your 80386, but they *could* appear when running an 80386 operating system. Yes, Xenix *and* unix are running on the 80386. Microport Systems Inc of Scotts Valley Ca is taking orders for a full 80386 version of SV3 that is to be delivered in July. Rumor has it that they are currently getting SVID ok from AT&T. The price is supposed to be around $800 for the software development kit, the text preparation kit, and the runtime system. The recent large posting of Dhrystone benchmarks listed several entries for Xenix 386 (also SV3) running on Multibus equipment. While I haven't seen any non-beta/end-user sites running either of these systems, I HOPE it is only a matter of time until anyone who has a 80386 can start actually using the other 50% of the silicon that until now has gone to waste emulating lesser chips. -- Steven C. Neighorn tektronix!{psu-cs,reed}!qiclab!neighorn Portland Public Schools "Where we train young Star Fighters to defend the (503) 249-2000 ext 337 frontier against Xur and the Ko-dan Armada"
rab@rabtrs (Rich Bilancia) (05/13/87)
> in article <146@rabtrs>, rab@rabtrs (Rich Bilancia) says: > > > > I also have written an article for the August issue of UNIX/World taking a > > first look at XENIX System V/386 in both of the above hardware environments. > > ^^^ > > Keep an eye open for it. > > > Gee Rich, not that I won't look for and read your article, but how about a > hint. Do we have to wait till August to at least find out if it's any good > or not? Just a one liner might do. > > Tony Holden > ihnp4!killer!tony OK, here's some of the article Either UNIX systems really are getting easier to install or I'm just getting used to the level of difficulty. But, I really think that it's the former because XENIX 386 is easier to install than any other version of XENIX that I've used. The default installation is with a single hard disk partition on the primary hard disk and a reasonable default swap space of 2.5MB. The system seems very fast while reading diskettes, formatting the hard drive, booting the system, and screen refreshing in general. Those impressions were not incorrect as you'll see below. One of the significant limitations of XENIX on the other Intel microprocessors is the way that memory is allocated in 64K segments, requiring the implementation of memory models that are selected by the applications programmer at compilation time. The four 80286 memory models are called small, medium, large and huge. The 80386 microprocessor also allocates its virtual memory in segments, but the segments in the 386 world are 4 gigabytes large. This limitation should hardly ever, if ever at all, be a problem for application developers. In fact, the elimination of this memory model (if you'll excuse me) nonsense will likely attract many MC68000 enthusiasts to the roles of XENIX 386 developers. Of course I had to verify the promise of upwards XENIX software portability, so I chose to check two of my favorite XENIX System V/286 tools: the Korn Shell from Aspen Technology (Parsipanny, New Jersey) and the Accell Integrated Development System from Unify Corporation (Sacramento, California). True to the XENIX promise of binary code compatibility, both products performed flawlessly in the XENIX 386 environment. The only problem that I experienced with XENIX 386 was a bug in the Gamma Release kernel that was documented and corrected too late for inclusion in the Gamma Release 3.05. The one other thing that did surprise me with many of the utilities that are distributed with XENIX 386 is that they are not recompiled 386 code, but rather the same executables that are distributed with XENIX System V/286.
korn@altger.UUCP (05/25/87)
altos is shipping xenix-V on their machines 'model 2000'.
it uses an intel 80386 ( 16.7Mhz, 32kb cache mem )
questions? details? feel free to mail me!
----------------------------------------------------------------------------
Hans Korneder UUCP: somwhere!mcvax!unido!altger!korn
SNAILMAIL: Gebsattelstr. 32, D-8000 Muenchen 90
VOICE: +49 89 4488373
----snip----snip----snip----snip----snip----snip----snip----snip----snip----
+-------------------------------------+
The opinions above are my own. | Advertising space for rent ! |
They can be yours too !! | Send requests to the adress above ! |
Send $57.95 to the adress above. +-------------------------------------+
(Offer void where prohibited by law or good taste)rab@rabtrs (Rich Bilancia) (05/29/87)
> altos is shipping xenix-V on their machines 'model 2000'. > it uses an intel 80386 ( 16.7Mhz, 32kb cache mem ) > > questions? details? feel free to mail me! > > ---------------------------------------------------------------------------- > Hans Korneder UUCP: somwhere!mcvax!unido!altger!korn Yes, but if Altos XENIX 386 is not binary code campatible with all other versions of XENIX 386 (just as Altos XENIX 286 IS NOT binary code compatible with all other XENIX 286 products, including Tandy, IBM, and SCO which are binary code compatible with each other) why would you want to by it? --- "Adversity is the crucible from which greatness is forged." - Ken Baker ____ ____ ___ / _ \ / _ \ / _ \ Richard A. Bilancia / / \_/ / / \_/ _ \ \\/ Computer Guidance & Support / / / / _ | | \ \ P. O. Box 620127 \ \ _ \ \ | \ | + | \ \ Littleton, Colorado 80162 \ \_/ \ \ \_| \ |_| /\_| \ voice: (303) 973-4035 \____/ \____/ \____/ ihnp4!sys1!bilanc!rab Multi-User Accounting Solutions attmail!bilancs, un
dc@sdd.UUCP (Daniel Corbett) (05/30/87)
Great news... I have a couple of questions.... for Altos. Is this System V available for your 286-based machines? Is there still a 192K byte limit on total code space per executable still on the 286 machines? When will it be available. We are developing a very powerfull and sophisticated accounting package utilizing some innovative database techniques, and it will make it much easier to port if the code space limitation is increased! Also, is sdb (or other source level debugger) available on either the future 286 release of Xenix 5 or the 386 release of Xenix 5? -dc
ron@vsedev.VSE.COM (Ron Flax) (06/02/87)
In article <153@rabtrs> rab@rabtrs (Rich Bilancia) writes: >> altos is shipping xenix-V on their machines 'model 2000'. >> it uses an intel 80386 ( 16.7Mhz, 32kb cache mem ) >> >> questions? details? feel free to mail me! >> >> ---------------------------------------------------------------------------- >> Hans Korneder UUCP: somwhere!mcvax!unido!altger!korn > >Yes, but if Altos XENIX 386 is not binary code campatible with all other >versions of XENIX 386 (just as Altos XENIX 286 IS NOT binary code compatible >with all other XENIX 286 products, including Tandy, IBM, and SCO which are >binary code compatible with each other) why would you want to by it? Actually Rich you are quite wrong here, Altos Xenix 286 *is* binary code compatible with SCO, Tandy, and IBM Xenix. At one time I had an Altos 2086 (Xenix 80286 box) in my shop, along with an AT running SCO Xenix V and there where many programs, including The USENET News distribution, that simply would not compile properly under SCO's version of the compiler. However, I was able to port the code to the 2086 and then move the binaries to the AT and they ran just fine! I did this with many programs, in fact I even went as far as to take Altos distributed software and move it over to the AT just for the hell of it, and guess what? Yep it worked too. I have also had occasion to use IBM and Tandy Xenix as well, and they use the same, non-industry-standard-intel, x.out format executables as SCO. -- ron@vsedev.VSE.COM (Ron Flax) UUCP: ..!seismo!vsedev!ron INET: vsedev.VSE.COM!ron@seismo.CSS.GOV
rab@rabtrs.UUCP (06/08/87)
> In article <153@rabtrs> rab@rabtrs (Rich Bilancia) writes: > >> altos is shipping xenix-V on their machines 'model 2000'. > >> it uses an intel 80386 ( 16.7Mhz, 32kb cache mem ) > >> > >> questions? details? feel free to mail me! > >> > >> ---------------------------------------------------------------------------- > >> Hans Korneder UUCP: somwhere!mcvax!unido!altger!korn > > > >Yes, but if Altos XENIX 386 is not binary code campatible with all other > >versions of XENIX 386 (just as Altos XENIX 286 IS NOT binary code compatible > >with all other XENIX 286 products, including Tandy, IBM, and SCO which are > >binary code compatible with each other) why would you want to by it? > > Actually Rich you are quite wrong here, Altos Xenix 286 *is* binary code > compatible with SCO, Tandy, and IBM Xenix. At one time I had an Altos > 2086 (Xenix 80286 box) in my shop, along with an AT running SCO Xenix V > and there where many programs, including The USENET News distribution, > that simply would not compile properly under SCO's version of the > compiler. However, I was able to port the code to the 2086 and then > move the binaries to the AT and they ran just fine! > > I did this with many programs, in fact I even went as far as to take > Altos distributed software and move it over to the AT just for the > hell of it, and guess what? Yep it worked too. > > I have also had occasion to use IBM and Tandy Xenix as well, and they > use the same, non-industry-standard-intel, x.out format executables as > SCO. Unfortunately it doesn't work the other way! I have been unable to get any binary that runs on IBM, SCO, or TANDY versions of XENIX to run on either a 2086 or a 3086. Since most of the XENIX software available today was ported to either IBM or SCO XENIX, Altos purchasers lose a great deal (as far as software availability is concerned) because the compatibility only goes one way. An interesting side-light is that while Tandy XENIX will run code compiled on SCO and IBM XENIX, not all code compiled on a Tandy will run on SCO or IBM. I guess this all means we still live in an imperfect world! Rich --- "Adversity is the crucible from which greatness is forged." - Ken Baker ____ ____ ___ / _ \ / _ \ / _ \ Richard A. Bilancia / / \_/ / / \_/ _ \ \\/ Computer Guidance & Support / / / / _ | | \ \ P. O. Box 620127 \ \ _ \ \ | \ | + | \ \ Littleton, Colorado 80162 \ \_/ \ \ \_| \ |_| /\_| \ voice: (303) 973-4035 \____/ \____/ \____/ ihnp4!sys1!bilanc!rab Multi-User Accounting Solutions attmail!bilanc!rab