INFO-PC@USC-ISIB (02/05/83)
From: Dick Gillmann <INFO-PC@USC-ISIB>
Info-PC Digest Saturday, 5 February 1983 Volume 2 : Issue 7
Today's Topics:
C86 C Compiler
Lattice C Portability to CP/M-86
Sritek (2 msgs)
Hardware References Wanted
IBM PC II (2 msgs)
TI PC Clone
RF Modulators
----------------------------------------------------------------------
Date: 4 Feb 83 05:54:47 PST (Fri)
From: npois!inuxd!stevens@Berkeley
Subject: C86 C Compiler
To: INFO-PC@ISIB
A while back, I asked for peoples' opinions on the C86 C compiler from
Computer Innovations. Here are the replies I received.
========================================
I have the C86 compiler for my IBM PC.
General comments: it's a good implementation of the "standard"
(i.e. white book) C language.
It's a four pass process, each pass must be run by separate
invocation by the user (or set up a command file; I don't know
why they didn't supply one).
========================================
It's good and worth the $250 to get it (the price will be ~$400
soon) but the update will only be $20, so it's cheaper to get
it now.
========================================
I have been using the Computer Innovations C compiler on the
IBM for about 8 months. I have a product on the market that
was written in it and asm86. The compiler will support ALL of
C, but George (the author) comes from the 8080 CP/M world, so
his generated code gets a little sloppy and his handling of the
heap is slow. Overall, his performance is probably average and
his code size is improving. He has promised an optimizer
version by the end of January. Computer Innovations biggest
plus is support. George is a really nice person, from England
by his accent, and is 90% available by phone. He and I have
had several debugging sessions by phone, where he has lead me
through the compiled code to find either his problems or mine.
This is worth more, to me anyways, than front end performance.
The only other C compiler I have any experience with is Telcon
in San Jose, Calif. BAD NEWS. He requires the assembler while
George generates linkable, with his linker, object. This is
not so bad, but the compiler builds an execution shell around
your program to allow command line parsing and redirected io.
George handles it the way you would want, IO is redirected at
the invocation. Also, you will need a hard disk with Telcon,
Computer Innovations will compile using 2 floppies, as he has a
3 pass, separated sections. This allows you to compile pass 1
on one disk, pass 2 with another etc. I have compiled 1700
line files on 2 160k floppies.
========================================
I purchased the subject compiler for use with CP/M-86. My
experience has been mostly good.
It is a bit cumbersome to use, requiring each pass to be
invoked as a separate command (a la Whitesmith's). To do a
compile and link, you'd do:
CC1 foo.c
CC2 foo
CC3 foo
CL foo clib
The generated code is a bit large, but that will improve when
their new code generator is out in a couple of months. To that
extent, Computer Innovations' update policy is reasonable: $20
for an upgrade, or free if you find a bug serious enough to
require a new release. I found a couple of bugs and they made
good with no hesitation.
K & R is implemented fairly faithfully. No extensions like
enums or structure assignments, however. Included in the
libraries are a fair number of V7 functions with a reasonably
complete rendition of stdio. And when I complained to the
company about the lack of a setbuf(), I was advised that if it
existed under Unix(TM) and I wanted it, they'd include it in
the next distribution.
There is an 8087 library that gets included automagically if
you are using that FP co-processor. There are enough low-level
routines to do some decent system programming, too.
I suspected that the front end was produced by YACC; this was
confirmed by George Eberhard of Computer Innovations. By the
way, George is a good guy to deal with. He is technically
competent and gets to the heart of problems quickly. When my
first problem arose, I called him on a Friday and again on
Monday. He actually apologized on Monday for not getting back
to me on Friday! (Unheard of in the micro-software business).
The problem was smoked out that Monday afternoon.
I have heard from folks who use C86 under the IBM PC that the
DOS implementation is isomorphic to the CP/M one with a couple
of differences in utilities and, of course, a compatible set of
library routines.
========================================
I have been using the Computer Innovations C86 compiler for
about 8 months, on and off, on an 8086 system with 2 8"
floppies and 128 kB RAM. Note that this is a minimum for C86
(PC version or 8086). I like C86 much better than the other two
non-Unix C compilers that I have used (Whitesmith's and
DECUS). (Correction on line 2 - the MEMORY is the minimum, but
you'd want 2 disks too.) The C86 author (George Eberhardt) is
trying to make C86 as standard as possible, and he is very
helpful if you call him with questions or problems. The first
versions had some bugs, but V1.3 is pretty good. I haven't
exercised the floating point, but everything else works very
well. Unlike many C's for micros, C86 has just about all the
data structures (even bit fields) and control statements. It
will compile several hundred lines of code in about 1.5 minutes
on these slow floppies, and shouldn't be too much worse on a PC
with the 5" drives. (Note - compile and LINK!). The run-time
library is fairly complete.
========================================
-- Scott Stevens
-- inuxd!stevens
-- npois!inuxd!stevens@berkeley
------------------------------
Date: Saturday, 5 Feb 1983 09:39-PST
To: info-pc at USC-ISIB
Subject: Re: Lattice C Portability to CP/M-86
From: jim at RAND-UNIX
The question was: How portable is the run-time produced by Lattice C?
Will it run under CP/M-86 without much modification?
Depending on the application, Lattice C (as any language!) makes heavy
use of MS-DOS' BDOS for communication with the terminal, the file
system, and the operating system. All such code is necessarily
non-portable, and would have to be translated into the (admittedly
similar) system calls for CP/M-86. If all such communication were
done through BIOS interrupts (in the PC ROM) and if the user is
sophisticated enough to dummy up the executable file to run under
CP/M-86 instead of MS-DOS, I suppose it could be done.
I'd rather wait for Lattice or somebody else to come out with a native
CP/M-86 version, though, or a cross-compiler. For one thing, reliance
on BIOS interrupts will lose you compatibility with the
"PC-compatibles", who, I suspect, will not want to challenge IBM's
legal staff by duplicating their ROM.
Why go to CP/M-86 anyway? I suspect that MS-DOS will stay the
standard, since it costs 1/4 what CP/M-86 does. Also they keep
blithering about upward compatibility from MS-DOS to XENIX...
Jim Gillogly
------------------------------
Date: 4 Feb 1983 8:51:53 EST (Friday)
From: Martin Schoffstall <schoff at BBN-UNIX>
Subject: Sritek
To: info-pc at isib
I have a few questions about the discussion of Sritek? Will they
allow someone to buy their boards and software to update their already
owned PC?
The second question has to do with memory. Given the three boards
as diagrammed does that mean that the 68000 will only be allowed to
address 256K words of memory? I guess if I had a 68000 I would
really want to address at least a Meg. Maybe I have to wait for
the 256K chips if I wanted to go along with this IBM upgrade.
schoff at bbn-unix
------------------------------
Date: 3 Feb 83 11:06:35-PST (Thu)
To: info-pc@isib
From: npois!cbosgd!djb@Berkeley.arpa
Subject: Sritek price list
Thanks to John Leon of Bell Labs Denver, here is the pricing info for
the various Sritek boards and XENIX packages. Prices are list, with
no mention of available academic or quantity discounts.
PC VersaCard........................... $995
Ram Module............................. $595
Z80 MicroCard.......................... $195
8086 MicroCard......................... $595
68000 MicroCard........................ $795
XENIX operating system and utilities... $495
Software Development System............ $495
Text processing package ............... $295
Multi-user package..................... $495
I don't know what's included in the XENIX packages so it isn't clear
which ones you need. I'd assume at least the OS and SDS packages,
making the total price about $3400 (VersaCard + Ram Module + 68000
MicroCard + XENIX OS + XENIX SDS) plus the cost of the IBM PC.
Sritek's quoted price of "under $6000" seems pretty reasonable.
John saw the system at COMDEX and had the opportunity to play with it.
He was impressed with the system speed (using a 10MHz 68000) and with
the people he talked to.
Thanks, John for passing along the info...
David Bryant Bell Labs Columbus, OH (614) 860-4516
(cbosg!djb)
------------------------------
Date: 4 Feb 83 8:34:11-PST (Fri)
To: info-pc@isib
From: ihnss!ihldt!ihnp4!ihuxx!marks (Mark Wm. Beckner)@Berkeley.arpa
Subject: Hardware References Wanted
The recent discussion concerning the Sritek line of products for the
PC has re-kindled my interest in the PC. So I thought I'd try to dig
up some more information - particularly, hardware architecture.
The problem is I can't seem to find any. I dug out my January '82
issue of BYTE, but the article on the PC is not as in depth as I would
like to see.
Does anyone have a good reference on the PC hardware architecture that
would detail the hardware to the point of providing schematics? I'm
looking for something in the public domain since I can always go out
and purchase the manuals.
Mark Beckner - Bell Labs
Naperville, Illinois
..ihuxx!marks / 312-979-2297
------------------------------
Date: 4-Feb-83 06:31:30-EST (Fri)
From: cbosgd!djb@Berkeley (David J. Bryant)
Subject: IBM PC II
To: brackenridge@usc-isib
Remailed-To: Info-pc
The latest rumor is that in March IBM will officially announce the IBM PC II,
which will be a new "edition" of the PC. The PC II will use the Intel
80188 CPU - a new, faster version of the 8088 which also has more of the
support hardware built in. This will reduce the chip count on the system
board by as much as 20 IC's, making it less expensive to produce than the
PC, and resulting in a reduced price. Some sources expect that the new
system box price will drop to under $1000. Other less substantial rumors
are that there will be more expansion slots, that a floppy controller will
be built onto the system board, and that 4164 dynamic rams (64Kx1) will be
used instead of 4116s (16Kx1) to give more memory "room" on the system board.
This is all supposed to happen at some conference (Computer Faire). At the
same time IBM is supposed to announce the Peanut, which is a low priced
unit aimed to compete with the Commodore 64, Atari 400/800, etc.
David Bryant Bell Labs Columbus, OH (614) 860-4516
(cbosg!djb)
------------------------------
Date: 4 Feb 1983 1158-PST
Subject: Re: IBM PC II
From: Billy <BRACKENRIDGE@USC-ISIB>
To: cbosgd!djb@UCB-C70 (David J. Bryant)
cc: info-pc
Exactly that course has seemed the rational path to me. For one thing
the Peanut sure explains why the PC has that absurd cassette
interface! IBM seems to be making so much money with the PC that I
wonder that the marketing types won't slow the PC-II for at least
another year. Apple could have released the Apple-IIe a lot sooner
than it did but held back as they were still making money on the
Apple-II.
If you refer to last December's Scientific American article on
personal computers you will find that the hardware costs are an
insignificant portion of the final PC cost to the consumer. I suspect
production costs for the IBM system unit run under $200. If I can buy
an 8088 for $15 at Radio Shack, I hesitate to guess what IBM is paying
for the 20,000 they use a month. I can't believe that it is worth
slowing production to introduce a new design that might save at most
$10 a unit in parts cost.
A much better justification for a new design is in cutting out the
after market for memory, etc. Here a good marketing man has a dilemma.
On the one hand this aftermarket helps sell machines. Have you priced
a full page in PC magazine? Hundreds of small companies are willing to
risk borrowed money to convince folks an IBM-PC with brand X memory is
the answer to all their problems. These companies are willing to work
on slim profit margins in effect subsidizing IBM.
I have found our IBM salesmen is again willing to sell us 16K systems.
These systems had been scarce, but suddenly have reappeared again.
They know we are using other companies after market devices and this
saves us about $1500 per system. There is no competition for
cost/performance when I configure a system using aftermarket memory,
I/O ports, and disks. It just knocks DEC and their bundled system out
of the market. The IBM salesmen know this and they know I know it.
I think IBM will hold off announcement of a newer PC until a few
months before the Macintosh is due to arrive. Who knows what system
they will be able to put together by that time. I too believe it will
have all the features you mentioned in your speculation and who knows
about things like mice and a better display for VisiOn support. VisiOn
is a trademark of Visicorp. IBM is, well you all know what IBM is...
------------------------------
Date: 4 Feb 83 14:01:38-PST (Fri)
To: info-pc@isib
From: harpo!eagle!mhuxt!evans (crandall)@Berkeley.arpa
Subject: TI PC Clone
Has anyone seen enough of the new TI PC look-alike to pass judgment
on it? Is it really compatible, how is the keyboard, etc.
Steve Crandall
mhuxt!evans
------------------------------
Date: Saturday, 5 Feb 1983 09:29-PST
To: info-pc at USC-ISIB
Subject: RF Modulators
From: jim at RAND-UNIX
I got a SUP 'R' MOD II modulator to hook my PC to the TV; it cost
about $40 at The Computer Store in Santa Monica. The Computerland was
trying to sell me something different and more expensive. The SUP'R'MOD II
worked fine, but an ordinary TV is no good for 80 columns, so I
almost immediately bought a cheapo Apple III monitor for $61, which
worked fine until I could afford my Sony Profeel (the latter is hard
to interface, so don't get one until you know how you're going to put
it on). With the Apple III (or equivalent) monitor you don't need an
RF modulator anyway, so unless you'd rather have color than 80 columns
you may as well go to that to begin with.
------------------------------
End of Info-PC Digest
******************************
-------