DTACK GROUNDED, The Journal of Simple 68000/16081 Systems
Issue # 30 - April 1984 - Copyright Digital Acoustics, Inc
As HALGOL development rapidly progresses, we are coming closer and closer to having a standard computer language, which naturally must be handled in a standard way. So release 3 of HALGOL will be distributed on copy-protected diskettes. This is being done to benefit the user, who will take satisfaction in knowing that any and every other HALGOL user paid just as much as he or she did. What could be more fair?
We are going to provide a version of HALGOL which supports the Nat Semi 16081 math processor. Judging by the benchmark data provided by Chet Sensenig later in this issue, that will provide far greater performance than is needed by any one user. This version of HALGOL will accordingly be a multi-user language system which will not run unless five users are logged onto the system (thus assuring that no one user improperly tries to hog too much performance).
For those persons who insist on using HALGOL with math chip support in a single-user environment, a special version of HALGOL, written in C, will be provided so that the user will not be overwhelmed by too much performance or take unfair advantage of those folks not fortunate enough to own a math chip accessory board.
We will need a common operating system for those versions of HALGOL and so we have decided to adopt UNIX, which is going to become the de facto standard mass-market 68000 operating system REAL SOON NOW.
Digital Acoustics is opening a new Customer Service Department. If you have a particular Apple accessory which you want to be supported by HALGOL, just write down your requirements and mail them to that department. Your request will be implemented immediately, once the department is staffed. This department will also design and build, for free, any custom interfaces you might need,
And pigs will sprout wings.
Digital Acoustics has ordered a Tandy Model 2000, and expects to have taken delivery by the time you are reading this. There are a variety of reasons why we went with the Model 2000 rather than the PC:
Are we saying the 2000 is better than an Apple II? We certainly hope it is! For starters, the price tag is over $4700 including sales tax! And it is a newer-vintage model, no?
Math chips p26b. Dsex catalog p27b. Price increase p23b. Mail call p10 & p19. Benchmarks p10b. BIGFORTH p14. Improved Schmidt DTACK/PASCAL p11b. Improved Soule monitor p14. Chinese Chessboard p22b. Saybrook report p18b. MAC vs MACK vs MAQVE p11b. Mack pricing p4. Memory limit p4B. Parts shortage p23. A.T.&T. onslaught p28. Disk stuff p20b. OSI interface p25b. UNIX: mass-mkt p14b, high-end p24b. Horror stories p17b. Busted BASIC p2. Doubling forever? p24b. British p5. Other editor? p27. Secret standards? p19b. S.A.T. = PASCAL p6. HALGOL: structured p7b. Data types p8b. Strings p21. Redlands p28 (barely).
Most of you are aware that Microsoft's Bill Gates has announced that Microsoft is developing an unbusted BASIC for Mackintosh, one that will be (is predicted to be) ten times faster than Microsoft's BASIC for the IBM PC (which is written largely in 8080 assembly). But the Microsoft BASIC that you can buy NOW is the carefully busted 68000 BASIC that Microsoft wrote to protect Big Blue back when they were still in bed together. John M. has just sent us the figures for Terry Peterson's benchmark: 621 seconds! Remember, the Apple II runs the benchmark in 488 seconds under Appiesoft. And the 'accuracy' is only 4.19E-11 - ten times less accurate than the HALGOL results, and over 33 times slower!
Did you get that? THIRTY-THREE TIMES SLOWER? Look, folks: it is absolutely impossible to make a 68000 run that slowly unless one does so on purpose. Those of you who have asserted that Microsoft did NOT deliberately bust its non-IBM BASICS, the ones written after IBM's entry into the personal computer marketplace, on purpose haven't a leg to stand on...
However, there is another firm which is busting its BASIC (and other languages) by accident - they don't know what they are doing! We refer to a firm known as Digital Research, which obviously needs to do some digital research. The following quotation is from 23 Feb '84 Electronics magazine, p100:
"The key to Digital Research's technical strategy is to achieve portability of languages and operating systems by writing them all in C. The language compiler itself, not the application program, is then portable from microprocessor to microprocessor, and applications can be ported from one system to another just by being recompiled.
"'We have completely abandoned assembly language,' notes DRI systems-software director Beebe. 'It makes you write every new generation of products from scratch.' Although assembly code tends [TENDS? YE GODS! - FNE] to be tighter and faster than high-level languages, cheap memory and high-performance processors makes these advantages less important.'
Isn't this where we came in a couple of years back? Where have those DRI guys BEEN lately? Are they blind and deaf or just extraordinarily dumb? How many BASICs written in C do they think they are going to sell now that Microsoft has decided not to protect IBM's interests? How many spreadsheets written in C are going to sell now that Mitch Kapor has become wealthy proving that businessmen like software that runs fast?
Many of you may not know that DRI has been trying to sell an 8080 BASIC written in C in recent months. Yes, that BASIC is just as slooow as that implies, and Microsoft's competing 8080 BASIC is written in 8080 assembly, you will recall. So how are sales of that DRI busted-BASIC doing? Well, you hadn't heard of it until we told you about it, right?
Perhaps DRI now plans to write their C in PASCAL to gain the approval of the computer science theocracy, then write their PASCAL in COBOL to please the data processing managers, and finally write their COBOL in ADA to meet mil specs. After all, "...cheap memory and high-performance makes (efficiency) less important." Mama mia!
A T & T, which is no longer restricted by law, is planning to enter the computer marketplace, including the personal computer marketplace. Considering the resources which A T & T commands, we are surprised how little one reads about that elsewhere in the personal computer press. Let us explain why you have not read much about that here, either:
A T & T developed a lot of software (C and UNIX) while its profits were protected from competitive pressures - it was a utility whose profitability was assured by a federal regulatory agency, remember. Having been born in a non-competitive environment, C and UNIX were then nurtured in yet another non-competitive environment - the universities. It should therefore come as no surprise that C and UNIX comprise an absolutely perfect software system for non-competitive environments.
At UCI, which happens to be located near us, 20 seconds is the usual and accustomed response time for a student sitting at a terminal ("on a good day" our other HALGOL programmer, a recent UCI computer science graduate, tells us). In such an environment, UNIX is a perfectly normal operating system. We can't see a 20 second response time being tolerated in a competitive environment.
A Motorola systems type recently pointed out to us the significant difference between Motorola's VersaDOS and UNIX: if three programmers are using the same compiler, only one copy of the compiler is in the system (not true of UNIX). If these three programmers have each requested enough memory that none is left, a fourth programmer attempting to log onto the system is told in effect, 'the system is fully loaded - go away!" UNIX just does more thrashing back and forth to disk and the response time gets longer and longer. VersaDOS does almost no thrashing while UNIX is ALWAYS thrashing back and forth -everything in UNIX is a file, the files talk to other files, and they are all on the disk!
We don't like multi-user systems (to put it mildly), but it seems to us that VersaDOS is designed for a competitive environment. UNIX isn't, due to the environment where it was born and the other environment where it was nurtured.
The A T & T folks likely see themselves as a fierce, powerful giant about to be unleashed to scourge the landscape. We see A T & T as a clumsy and unworldly oaf who is likely to be unmanned if he tries to venture out from beneath Ma Bell's skirt. Trying to tackle the MASS, competitive, marketplace with C and UNIX - and that is A T & T's plan - is like Czech cavalry trying to counter German tanks and Stukas with swords and lances.
(Did you know that Commodore is now making 400,000 C-64s a month? And not keeping up with demand? That if the installed base of C-64s was over two million as of 1 DEC '83 then the installed base of C-64s will be OVER FOUR MILLION by 1 MAY '84? Did you know that almost all of the software sold to support the C-64 is written in untransportable assembly? With that installed base, do you think anybody CARES?)
Even the Apple Computer Company has finally figured out that you win with assembly and lose - and die - with high-level. (Place fresh flowers occasionally on LISA I's grave.) When an unbusted 68000 BASIC arrives for Mackintosh to go with that assembly-based operating system, Mackintosh is going to become VERY popular, as we have been predicting for some time now.
And DRI has just "completely abandoned assembly language." Gee, what a shrewd move on DRI's part!
The current MULTIPLAN and BASIC which are available from Microsoft for Mackintosh were both written a while back for the 68000 at a time when Microsoft was protecting its IBM connection. Therefore, both are lovingly busted in C and hence horribly slow. Terry Peterson's benchmark, the one which runs in 18.6 sec in HALGOL oozes in 621 sec in Mackintosh busted-BASIC, 33.4 times slower! We have just spoken with two folks who got Macks in the past week; one has had his Mack for a whole 3 hours. The following is a composite of those two conversations:
"Boy, do I need a second floppy drive! When you boot up Multiplan it will ask you for the other disk, then ask for the first again, then the other - you have to do 7 or B disk swaps before you are up and running. You only have 40K of data available after the program is loaded, about the same as in a 64K Apple II. I already knew it was written in PASCAL or C because I recognized that characteristic 'flicker' as it
(Multiplan) wrote to the screen. You get the same thing with PASCAL using the Apple II or the SAYBROOK board! But that graphics program is not half shabby, and the mouse really works out pretty well. Sure wish the CRT was a little bigger. BOY, DO I NEED A SECOND FLOPPY DISK DRIVE..."
It is important to keep track of the various Mackintosh BASICs. We understand the only one you can buy is Microsoft's busted-BASIC, written in C. Apple has a BASIC which they are not choosing to make available, it seems. That is almost certainly an offshoot of that very early LISA busted-BASIC, the one carefully busted in PASCAL. We just heard that Apple refers to this as a "learner's BASIC". "Learner's BASICs" don't have to run fast, right? The third entrant is Microsoft's announced but unavailable UNbusted-BASIC. Lots of folks are confusing that one with the busted-BASIC, the excruciatingly slow BASIC that you can actually buy today.
(Proving that nothing is as simple as it seems, we have heard from two sources that the very fast 68000 BASIC, the one which compiled each line as it was entered and which was attributed by InfoWorld to Microsoft, is actually being developed inside Apple Computer. On the other hand it does appear that Microsoft IS working on an un-busted 68000 BASIC to replace the busted-BASIC it now sells for Mack. Here is how we put this together: both Apple and Microsoft are busily developing an unbusted 68000 BASIC for Mackintosh. Both of them are 'compiling' each line as it is entered.)
(We know another company which is developing a 68000 BASIC like that [guess who]. We wonder who will be done first?)
This thing is almost like a soap opera. Having kicked Microsoft out of bed, IBM is showing signs of climbing into bed with Digital Research. Is it a coincidence that Microsoft deliberately busted its non-IBM software in C (much earlier, in PASCAL) while it was in bed with IBM? Is it a coincidence that they announced a forthcoming UNbusted 68000 BASIC about two microseconds after IBM kicked them out of bed (by supporting a UNIX clone that was not XENIX)? Is it a coincidence that Digital Research is announcing adherence to other languages busted in C just as it is climbing into bed with IBM? Maybe DRI isn't goofing by accident at all!
The funny thing is, DRI is simultaneously trying to climb in bed with IBM and A T & T! Talk about strange bedfellows! (No, we are not making this up, and we are not describing the plot of an episode of Dynasty.)
Yep! The PCjr has arrived on dealer's shelves - which, according to the WSJ, is where they are still sitting. Tsk. It really upsets us to see IBM not doing well in the marketplace. In today's L. A. Times newspaper, the $6995 list IBM PC/XT is on sale for $4995. A few months back, the IT was heavily backlogged. IBM has cut back its orders for hard disks by a factor of at least two, and we understand they have stood up Tandon, which was building a plant just to supply IBM. Gosh, we are really saddened to see IBM's sales fall off like that. Let's see now: IBM is going to sell 2.5 million personal computers in 1984, right? Would Big Blue lie to us?
Will someone please explain Mackintosh pricing to us? Apple is apparently trying to sell a $1000 Mackintosh to all the nation's college students and all of the nation's computer salespersons. They appear to be offering a half-price Mack ($1250) to software or hardware developers. And they will be selling Macks to dealers for about $1600-$1800 for resale to the public at $2495 list. We confess that none of this makes sense to us.
It appears to us that the raw manufacturing cost of Mack is just under $400, so Apple isn't going to lose money selling Mack for $1000 to those college students. But they are going to try and hit your FNE, who is an EX-college student, for 2 1/2 times that amount? Your FNE is dumb but he is not THAT dumb! For the record, our project engineer is still taking night classes and is going to try to get one of those $1000 Macks personally or through a buddy. We plan to apply for the $1250 developer's special (we happen to be legitimate in this case - remember HALGOL?) and we spoke yesterday to Bob Tripp of Micro magazine who is ALSO trying to buy the developer's special.
In fact, anybody with any sense at all is going to try to get a Mack at one of those two lowball prices! In the meantime, the computer retailers are doubtless charmed by that 2 1/2 times ratio in prices, and by having to themselves pay nearly double what a college student end-user pays. We are not entirely certain what Apple's motivation for creating this situation was, but it was most certainly not to keep the retailers happy.
People have already observed that a computer which can easily be carried can easily be carried by someone else. If a student's $1000 Mack is stolen, does the student pay $1000 for a replacement or $2495? Think about the implications of THAT - both ways - for a while.
Oh, yes: that $400 raw manufacturing cost: $200 max to Sony for the drive, $70 tops for the monitor and what's left costs less than double the raw cost of the C-64 and hence less than $126. Glad you asked.
Sure you pay more than those prices yourself. But you buy onesie-twosies, and Apple has quite literally placed an order for more than one million drives with Sony (more likely TWO million) and they have ordered one million of those monitors.
Timex has just decided to drop personal computers. Do you realize how few entrants are left in the low-end personal computer market? With Commodore cranking out 400,000 C-64s per month and not meeting demand, do you realize that the remaining entrants are nearly irrelevant?
If Trameil had been given just one more year, there would have been NOBODY left in the low-end marketplace. Even with Trameil on the outside looking in - for now - his legacy, the C-64, dominates. (Are you going to write software for an installed base of 4 million C-64s or 50,000 Tomys?)
Trameil is now converting the FIRST third of his Commodore holdings - a third is about $70 million - into cash. We strongly believe he will convert another third or two in the not-distant future. What is kindly Uncle Jack going to do with all that money? Buy a new toaster? A sailboat? Is Chuck Peddle's personal traveling engineering team (e.g. Bill Seiler) going to do some more traveling soon? Will Chuck himself be included IN or included OUT? Stay tuned for future developments...
Several folks have 'busted their pick' on a simple algorithm for detecting the upper limit of memory. They have all found that they can write a word to a non-extant address (that is, an address where there is no memory) and yet do a comparison and discover that the test word is there! The reason is very simple: there are two undocumented bytes of DRAM included with the DTACK boards, even the static RAM versions!
If you write a word to a dummy location, that location will be off-board via the expansion bus. ESPECIALLY if you have a cable on that bus, there is capacitance associated with the expansion bus. When you write, the data word is stored in this capacitance. If you immediately read that word back, it is still there, stored in the capacitance! Therefore, there appears to be memory at a location where there is no memory in the usual sense.
The fix is to do a long-word store and read, where the second of the two words is the complement of the first (e.g. $AAAA5555). The second word will be the one stored in the capacitance, and an immediate read-back will read the last word twice, and your memory-test algorithm will now correctly test the memory.
The European Theatre DSEx Herr Direktor asks whether we send info packs overseas these days. The answer is mostly no - it costs us a tad over two bucks to do so, and there is no way we can recover that money. But we are not totally consistent; a request from a university will usually be honored, for instance.
According to an item in the latest Peelings II magazine, an East German firm has developed electron beam lithography equipment ten times better than exists in the west. We wonder how long the Pentagon brass are going to claim that it would be a breach of national security for us to ship Japanese DRAM, TTL and 12.5 MHz microprocessors to those commie West Germans? Or those pinko socialist Limeys? Or...
We are looking forward to being enlightened by the erudite editors of Britain's Micro Computer Printout magazine regarding Sir Clive's new QL machine, which we understand uses a 68008 CPU.
We recall the 8088, which has an 8-bit data bus, 8-bit instructions and some 16-bit registers. MCP rated this machine at 16-bits.
The 68008 has an 8-bit data bus, 20-bit addressing, and 32-bit registers. We are eagerly awaiting MCP's decision on the 68008. Will MCP join Apple Computer's leader John Scully in rating the 6800X as a 32-bit machine? If not, just what size will MCP select and on what basis will it make that selection?
It seems the British press is dumping on Sir Clive's QL 68008 machine because its tape 'waferettes' are not suited to the storage of business records. We believe that a point is being missed:
In the late 1960s, computers used little magnetic doughnuts, collectively called 'core', as RAM. This 'core' did not lose its stored programs and data when the power was shut down. Accordingly, those computers used for a dedicated purpose had to have their programs reloaded only rarely. So IBM developed a very low-cost program loader. We'll tell you about that loader, but promise not to laugh!
In the first place, the magnetic oxide was deposited NOT on a proper hard surface but on a cheap flexible plastic called Mylar. Next, the disk was ridiculously small - only 8 inches in diameter! Because the device would be used only rarely, long life was not a design goal and so the head was designed - you promised not to laugh, now - to actually stay in contact with the media! You can imagine how that design was regarded back in those days, when 'HEAD CRASH' was the foulest phrase heard in computer environments.
Well, that design seems to have proliferated and become more reliable, no? So what sort of storage device does the Sinclair QL use? An absolutely perfect program loader for rock-shooting games that use 32-bit registers! We understand the data transfer rate approaches 20K bytes per second, over an order of magnitude faster than Apple's DOS 3.3 and incomparably faster than the disk drive Commodore sells for use with the C-64.
Sir Clive is making a mistake introducing the QL in the British marketplace. The QL is in fact ideally suited for the American home computer marketplace! There are lots of folks in the U.S. with $500 discretionary funds. Apple and Mackintosh would be the best salesmen for the QL, and Bob Tripp of Micro magazine would be its second-best salesman. Bob is tiring of 8-bit machines and is no longer Focusing on the 6809. The QL would fit perfectly into the pages of Micro.
Too, the QL is a heck of a good machine for somebody to learn about 16ea 32-bit registers, and cheap enough to try as a lark. We honestly believe that sales of the QL in this country could approach the level of C-64 sales, and certainly exceed the combined sales of the Atari computers AND the TRS-Color machines - if it is shipped into the U.S. market SOON! Even this little dinky newsletter would push the QL, and maybe support it with HALGOL. We can see the sales pitch now:
"Buy a QL and we will sell you a HALGOL waferette and manual for $39.95. When you learn what the 68000 architecture and linear addressing means and also need higher performance, come to us. We will sell you a real 16-bit machine that runs at 12.5 MHz, has hardware math chip support and so will run all of your HALGOL software over ten times faster, in up to a megabyte of RAM."
We emphasize again that the QL has an absolutely perfect cheap, fast program loader to get the largish HALGOL object code up and running. Think about it. And pass this on to Sir Clive? We assume that British subjects have no objection to Sir Clive importing substantial amounts of U.S. dollars!
Somebody sent us some (doubtless illegal) photocopies of a PASCAL-related publication. Included was an interview with Dr. Kenneth Bowles, who is now an ADA person. Bowles was discussing a problem with teaching PASCAL. At UCSD, the sacred Mecca of PASCAL (U.S. division) it seems that 50% of all students failed to successfully complete the first semester of an introductory programming course if that course is PASCAL-based. They have also (re)discovered that practically everybody can pass an introductory programming course in BASIC. We suspect that the reaction of the computer science profs is to find a way to weed out the intellectually inferior students before they are permitted to study an (ahem!) PROPER computing language.
Our project engineer, the one who designed the Stuffer board and much of the Grande and the VDHR board set, happened to overhear us discussing this matter. He said, "Yeah! I dropped out of an introductory course in PASCAL after 3 1/2 weeks. It bored me!" At the time, he had about 4 years' experience as a 6502 assembly language programmer!
It happens that your FNE thinks (right or wrong) that he is an above-average programmer. If we were forced to take a class in PASCAL, we would flunk it for the same reason that we would flunk a course in selling life insurance. Our project engineer now programs in BASIC, 6502 assembly and 68000 assembly. This guy is intellectually inferior?
Dartmouth University learned, a long time ago, that the average engineering student (NOT computer science student) could learn BASIC in 30 minutes. Slow learners took two hours. Sure, BASIC is a terrible language to write a 5000 line program in and an IMPOSSIBLE language for a project requiring 30 man-years - but so what? BASIC is a language for SOLVING PROBLEMS. If we have an engineering problem to solve we can often have a solution in five minutes using a medium-simple BASIC program. Sure the program will not be structured "properly". If we get the correct answer five minutes after the problem comes up while a PASCAL programmer devises a "correctly" structured program in 30 minutes and gets all the semicolons in place in 90 minutes and the answer in 2 hours, who is intellectually inferior?
We believe that there are a hell of a lot more folks looking for an answer to a problem than there are folks working as part of a 50-person team coding PASCAL or ADA for the Pentagon. This newsletter is directed primarily to persons interested in personal computers and PASCAL is DYING among that group. PASCAL (in personal computerdom) is retreating to the bastions of the universities from which it originated. WE ARE NOT MAKING THIS UP: We have been contacted in the past
30 days by THREE university profs, two from California, who have asserted that they have statistical analysis programs which must be written in PASCAL so that they can be exchanged with other universities!
Fine. We believe in everyone having the freedom to choose their own brand of poison. We even like FORTH persons, although we would lay hands on FORTH only briefly, as an intellectual challenge. But there is a problem.
The problem is that the academics behind the S.A.T. (Scholastic Aptitude Test) have decided that prospective freshmen (freshpersons?) will be tested for programming ability using PASCAL EXCLUSIVELY! This apparently means that the academics want to weed out the 50% who flunk the introductory programming course (when taught in PASCAL) at UCSD, before they arrive on campus! If we were just starting our college education, we - and our project engineer - would surely be among those rejected as intellectually inferior.
It worries us that the S.A.T. academics are locking PASCAL into the college curriculum, or at least the college prep curriculum, at precisely the time that PASCAL is DYING in the personal computer community. The important characteristic of the personal computing community is that the persons who program get to choose their own programming languages instead of having a boss force them to use a particular language! So what language is winning in the personal computer marketplace? (That is called a 'rhetorical' question, folks.)
If you are among those extremely vocal persons who support (worship?) PASCAL, you are confusing mouthpower with sales. Plot the reported sales of SofTech for the past six quarters and take particular notice of the two most recent quarters. If that does not tell you something, then you would not have understood us when we reported that one of two LISA production plants had been closed completely and the other cut back to less than half size last year, either.
No amount of mouthpower can compensate for a lack of sales. There is no place for PASCAL in today's personal computer marketplace. (If you wish to controvert that statement, please do so with comparative sales figures, NOT mouthpower! Please note that universities are isolated from the marketplace - university professors can be as inefficient as they please.) PASCAL is still being used by 30-man programming teams for the Pentagon and a few other employers. I guess that must be what the S.A.T. folks and the computer science folks are training all those PASCAL programmers for. Come to think of it, maybe it's a good thing the S.A.T. folks are going to exclude future FNEs from that sort of education...
If we can agree that BASIC is unsuitable for projects requiring 30 man-years' programming, and that PASCAL is unsuitable for problems which can be solved in five minutes using BASIC, we might then ask, "where is the middle ground?"
We answer: the middle ground is a mildly structured language. Is there such a language? Yes. Its name is COMAL, and it should be familiar to the few PET types amongst our readers because there is a public domain version for the PET which comes by way of Denmark. (There are also commercial versions.)
There are also at least two supporting organizations for COMAL in this country. We should also tell you that there appears to be some disagreement among those supporters which is partly financial and partly a good old political power struggle. The power struggle is sort of like DSEx setting up shop and voting your FNE O - U - T out except that the COMAL folks apparently take such things seriously! Len Lindsey, who founded the precursor of COMPUTE! magazine, started a COMAL newsletter and sort of turned it over to somebody else and has now started a NEW newsletter, COMAL TODAY. Vol 1 #1 consists of a lot of old reprints which is good if you are interested, because it will bring you up to date from the beginning.
It even contains the obligatory anti-BASIC diatribe by the inventor of COMAL (one of those ignorant un-American foreigners). The inventor seems to think that teaching somebody how to program BASIC is a criminal act. Geez. Don't these one-way types understand that if a programmer knows more than one way to write a program that he will invariably use the most efficient method to solve a particular problem? And that since there are more simple problems than vastly complex problems, he will invariably use BASIC if he has two brain cells to rub together? (Oh Great Spirit: please send your FNE a non-sexist substitute for "he"!)
If reading about programming concepts is your thing, by all means subscribe by sending $14.95 to:
COMAL User's Group U.S.A.
5501 Groveland Terrace
Madison WI 53716
Our prejudice against PASCAL is NOT due to its structured form, but due to the fact that PASCAL is incomplete ("the thalidomide baby of programming languages" - Kernighan) and, in its most common implementation, unbearably slow and burdened by an atrocious operating system. One author (not us) described that operating system as like "kicking a dead whale along the beach"!
and if you had not figured that out already, shame on you! HALGOL is in fact a technique and (we hope) an invisible operating system. The technique is the use of the power of the most modern microprocessor to obscure a compiled language from the programmer. What the programmer sees is a nice, interactive BASIC in a nice, friendly, almost modeless environment complete with screen editor. What he gets when he keys RUN(RETURN) is, instantly, a blindingly-fast compiled 68000 program.
Does the programmer HAVE to see BASIC? NO? He can be made to see COMAL, or FORTRAN, or even (heaven forfend) COBOL! Why have we chosen to make the programmer see BASIC? Because BASIC can be learned and used by the largest percentage of those who can program at all, and because, being an engineer, we often have engineering problems to solve which have less than massive complexity. (And, with a nod to Mammon, because we want to make the largest number of DTACK board purchasers happy.)
Anybody who understands the 68000 knows that it has a lot more oomph than one person traditionally needs. There are differing ways of expending that extra oomph. Carl Helmers said it was O.K. to write your operating system in p-code, so the LISA folks did and LISA is dead. Others said that since the 68000 had too much oomph for one person, let us chain multiple users to that one CPU. Hence Fortune, the Tandy 16 and Alpha Micro. Fortune is having bad fortune (please excuse the tacky pun) while the Tandy 16 is on oxygen in the intensive care ward. Interestingly, Alpha Micro is prospering.
Our approach is to use that extra oomph to obscure the real program from the computer operator. It would be almost trivially simple to make the operator see, for example, COMAL rather than BASIC.
You already know the HALGOL (BASIC) conditional operator IF:
100 IF VARN = X THEN (do this) ELSE (do that)
Here is the corresponding COMAL statement:
IF VARN = X THEN (do something) (maybe something more ELSE (do that) (maybe more that) END IF
While we might like to see the COMAL form in a long program, especially one involving a dozen or so programmers, we can completely solve a short engineering problem in the time it takes to get those damn indentations right.
The point is, it is just as easy to program (obscure) one form or the other! EXACTLY as easy (using the HALGOL technique)!
for structured programming languages. We happen not to have such a need ourselves. Also, it just so happens that we have customers. Why should we horse around with a language like PASCAL which, the evidence shows, half of our customers (and all of our newsletter editors) cannot or will not learn?
Still, the technique of HALGOL clearly can be applied to COMAL (moderately structured HALGOL), MODULA II (structured HALGOL) and ADA (massively structured HALGOL).
Oh, yes: the operating system. One moment while we put on our 'elitist' hat (seems tight - hmmm). Now we can explain to you, an intelligent person like us, that we are attempting to obscure the operating system from the user just as we are trying to obscure the inner workings of the language from the user. (But not from YOU, dear reader - there's nobody here but us elitists!) As we know, a CP/M user or a UNIX user or a SofTech p-system user is INTENSELY aware of the operating system. Our fondest hope is that the casual HALGOL programmer, when asked what operating system is used, will scratch his (her?) head and may, "Gee! I don't think HALGOL has an operating system!"
We just read the other day that there is a controversy in progress, we think centered around MIT, over modal versus modeless programming environments. We made the decision to make the HALGOL environment as modeless as possible about 18 months before we became aware that a controversy existed. You Apple types: the 'monitor' is a mode, entered by CALL -151 and exited by 3D0G. You Pet (8032) types: your 'monitor' is also a mode and so is the INSERT mode of the screen editor. The HALGOL monitor is not a mode, it is a natural part of the environment - and INSERT is a keystroke, not a mode to enter and exit.
It does appear that runtime is one mode and edittime is another - so we cannot claim that HALGOL is modeless. Suggestions, anyone?
Mixed-mode calculations will be forbidden in HALGOL for good reasons, plural. One, we want HALGOL to run as fast as possible with acceptable errors. Consider the following (illegal) mixed-mode calculation:
100 LET VARN = X%*Y%
If we perform the indicated multiplication using integer arithmetic and convert the result to floating point before storing the result in VARN, we might get overflow which would NOT occur if the two integers were first converted to floating point and the multiplication performed in floating point. But if we did that ALL the time (and Microsoft 6502 BASIC does) we would be slower than possible. It's best to let the programmer make the decision.
Another reason, as we will see later on, is to put pressure on lazy and/or inexperienced programmers to use integers for string and array indices.
Integers? Oh, yes: integers. Integers are 16-bit two's complement values which can take any value from -32768 to +32767. Integer arithmetic is very fast, especially using a 68000. Integer arithmetic is well suited for graphics and as index values (indices) (e.g. VARN[HEIGHT%,WIDTH%] or MID$[A$,I%, J%] because of their speed compared to floating point values. Consider the following two equations:
Each equation contains the constant 2. If both formulas occur in a HALGOL program, as might be the case, those are DIFFERENT constants! The first "2" occurs in a data type real (real = floating point) equation, so it has the 8-byte F.P. value $1001 8000 0000 0000 and the name "2" in the F.P. constant name table. The second "2" is an index and therefore defaults to an integer constant whose value is $0002 and whose name, "2", is in the integer constant name table.
The constant "2" is carried in two locations because we do not want to perform any unnecessary number conversions at run time.
Type 0: real variable (floating point) Type 1: real constant (floating point) Type 2: real array, dimension 1 (F.P.) Type 3: real array, dimension 2 (F.P.) Type 4: real array, dimension > 2 (F.P.) Type 5: integer variable (16 bit)
Type 6: integer constant " Type 7: integer array, dimension 1 (16 bit) Type 8: integer array, dimension 2 " Type 9: integer array, dimension > 2 Type 10: string variable Type 11: string variable array, dimension 1 Type 12: string variable array, dimension 2 Type 13: string variable array, dimension > 2
You may have noticed the absence of a string constant in the above list. HALGOL will not have a separate table for string constants, but will instead carry them around in the body of the program as 'literals'. Note that three separate types of arrays are available for data types real, integer and string. Slightly faster operation is obtained by knowing in advance whether the index (dimension 1) or indices (dimension 2) are either integer variables or integer constants, and compiling separate action addresses for run-time code which knows in advance whether a given offset is data type real variable, real constant, integer variable or integer constant. When the number of dimensions increase, this is impractical and the number and types of the indices must be stored with the action address and evaluated at run time. Yes, this IS a form of interpretation, something we would prefer to avoid. Total compilation might be extended to arrays of dimension 3 - we will wait and see.
This provides yet a third reason for not permitting array or MID$ indices to be data type real - it would force sixteen action addresses for an array of dimension 2 (if each index could take on one of four data types), and 64 action addresses if we decide to fully compile arrays of dimension 3. You can forget the possibility of an index or indices of data type real right now, in case you were planning to write us a letter pointing out how absolutely and utterly essential that is. (We know how the innards of HALGOL work and you probably don't - yet!) Besides, you CAN us data type real for an index. You just have to explicitly convert it to an integer value just before using it as an index.
Incidentally, any HALGOL function which requires multiple action addresses to implicitly specify the data type of the offset value ALSO requires multiple listing addresses - but that is simpler; each list link is located two bytes behind each of those action addresses. And the code to which the link points can simply load a number (MOVEQ #X,Dn) which then can be easily interpreted so that the list code is not excessively long.
We use 'infamous' in place of stronger anglo-saxon expressions which might unnecessarily offend. While
many of you may be thinking by now that we are worrying unnecessarily about the speed with which data is moved about, considering the relative slowness of, say, floating point computations, please remember that HALGOL is being planned from day one to include hardware floating point chip support! It comes as a considerable shock to most folks to discover that once one has hardware math support, data movement becomes the bottleneck!
To drive the point home, remember the Weitek two-chip set which can perform single-imprecision floating point at 1/4 microsecond per operation (see WONDERFUL WEITEK, issue 26 p.14). O.K., chum, now move the data to that chip set and back and see if your math operation doesn't take somewhat longer than 1/4 microsecond... NOW do you see why we are concentrating so hard on reducing #&%$@! data movement overhead? (Most folks think of hardware math chip support as a distinct afterthought when designing their programming languages. Not us! Floating point chip support is very definitely the wave of the [very near] future!)
That noise you hear in the background comes from molars being ground together by Motorola executives who are infuriated at the thought of someone promoting the use of a non-Motorola math chip (the Nat Semi 16081). In the meantime, their 68881 is being continually delayed (production of the mask has not yet begun) because Motorola is concentrating on the 68020. And they have a limited number of skilled workers (surprise?)'.
If the Nat Semi 16081 reaches production status this spring, it will establish a new world record for the time between mask start and production availability of about 3 1/4 years. (Partially functioning 16081s existed in June '82!) The 8087 took over 3 1/2 years! And there is no assurance that the 16081 WILL become a production part this spring - past events discourage us from being optimistic. It therefore seems reasonable to suggest that three years will elapse from mask start to production status for the 68881, which is a far more complex chip than the 68881. Since we are well into 1984, we add three years and find ourselves in, um, 1987? Yep, 1987. If you plan to use a math chip with the 68000 before 1987, you better get friendly with the 16081 fast - unless you have sufficient resources, like IBM, that you can diddle the 8087's microcode. Motorola's executives will HATE you for using the 16081, of course.
Well, it is not a first for Digital Acoustics to buck the tide and do something which does not meet the approval of Motorola's executives.
Let us look at this in a positive manner. If folks in Micro land think the 68000 doesn't have math chip support, we won't have trouble buying 12.5MHz 68000s!
"Let us travel back in time when FNE raged [huh?] across a whole gamut of topics and sacred cows. Newsletter #3 included some timings for simple BASIC operations on the 8032, back in late 1981. Using the same test program, effectively identical, the PCjr produced these timings [tabulated below]:
"Have we progressed in 2 1/2 years? Slightly. LOG seems to be much faster. Multiply followed by divide runs slower. The interpreter is significantly different. Keywords in imbedded form are disregarded. Keywords must, in fact, be separated by blanks or operators to be recognized. FORGET is a valid variable name. FORI= will not work - it must be FOR I=. The line search problem is eliminated at a cost of 2.5K of RAM for label/address table contents...
"FC wishes that FNE would stop talking to computer store dealers, whose primary interest is the most expensive (highest commission) micros or to Applkle owners whose software and time investment makes a changeover impractical." Nils D., Wethersfield CT
Nils has been crusading for some time for us to make a 68000 attached processor, with 128K RAM, 16 ea I/O ports each with its own 64K DRAM, and expansion capability - for 87 cents. He also despises the Apple, probably because it costs over 87 cents as well. Sigh. FC, did you play too many downs without a helmet when you were in school? What you want is, literally, impossible. Ask us for a perpetual-motion machine or an engine which violates Newton's third law and we might be able to come through; but the 68000 board you want at the price you are willing to pay? NEVER!
Whether you like the fruitarian machine or not, it is a fact that all those empty plug-in slots attracted a large proportion of the hackers and our 68000 boards are directed largely towards hackers, not drooling nitwits using electronic door-stops to shoot rocks. If there are five or fifty non-droolers amongst the owners of those electronic door-stops, that is not a big enough market even for a small company like us to go after.
Now for your comments about PCjr's BASIC: except for the one sentence about something running slower, we would have sworn that you were describing HALGOL. Lots of languages today require that spaces surround keywords; it happens to be necessary if FORGET is to be a variable name. We are more pleased than displeased at the similarity because it indicates to us that we are not deviating too far from the mainstream in our quest for a combination of speed and a friendly interactive environment.
Your reference to newsletter #3 revives painful memories. You see, almost anyone can write ONE newsletter. Our first one was a modification of a letter which we wrote and mailed to every small computer publication we could think of; so #2 was our first real newsletter. We didn't really read #3 until after it had been printed and mailed; when we DID read it we were appalled! We very nearly decided that we were not up to the task of writing a newsletter. Until we received your letter, we had not peeked at #3 since back in those days.
With some trepidation (a bow to Ivan S.) we here repeat the benchmarks and timings from page 2 of issue #3 along with some recent additions, including yours:
130 B=SQR(3):FORI=1TO5000:A=B+I:NEXTI 140 B=SQR(3):FORI=1TO5000:A=B*I:NEXTI 150 B=SQR(3):FORI=1TO5000:A=B/I:NEXTI 160 FORI=1T05000:A=LOG(I):NEXTI 170 B=SQR(3):C=1234:FORI=1TO5000:A=B*C/I+I:NEXTI 180 FORI=0TO5STEP.001:A=SIN(I):NEXTI
It is interesting to note that no spaces are needed to surround the keywords in those benchmarks, which were written for Microsoft's 6502 BASIC. Here is the current HALGOL version of line 130:
130 LET B=SQR(3) 131 FOR I = 1 TO 5000 132 LET A=B+I 133 NEXT I 134 PRINT "DONE" TEST RESULTS: 130 140 150 160 170 180 LISA (a) 45 46 52 254 85 222 LISA (b) 25 24 30 143 47 109 MACK (c) 13 20 35 245 49 149 MACK (d) 12 11 17 121 25 101 MACK (e) 19 18 23 131 40 108 APPLE II 18.9 22.8 27.8 114.2 43.9 133.3 IBM PCjr 20.9 22.2 27.4 58.2 100.8 132.0 IBM PC (f) 23.7 25.9 109.0 50.5 131.9 103.6 IBM PC (g) 4.9 6.0 7.1 23.0 13.2 19.7 IBM PC (h) 6.7 6.6 6.9 7.0 13.2 9.5 HALGOL (i) .8 .95 2.23 8.8 2.9 7.1 68000/16081 (j) .31 .32 .35 1.98 .56 2.15 (a) LISA BASIC REL 2 (b) LISA BASIC REL 3 (c) MICROSOFT BASIC for MACK (d) MAC BASIC (e) MAC PASCAL (f) PC BASIC (interpreted) (g) PC BASIC, COMPILED
(h) PC BASIC, COMPILED with 8087 (i) HALGOL v1.02, no wait state (j) Chet Sensenig compiled BASIC
This started as a simple 5-liner, one line per machine. We received lots of inputs, which means folks are paying attention and we guess that's good. We tossed out everything that was not double precision floating point so that we are, to some degree, comparing Apricots and Apricots. LISA BASIC REL 3 is not available now but may be released later. Only the first of the three MACK benchmarks uses software which is currently available. We believe that MAC BASIC and MAC PASCAL may be languages which are currently under development in the Apple Computer Co. HALGOL v1.02 is 68000 software using the DTACK GROUNDED board. The Grande should be about 18% slower. A version of Sensenig BASIC which supports the 16081 math chip is in the public domain and is available from DSEx.
130 140 150 160 170 180 LISA (a) .4 .5 .5 .4 .5 .6 LISA (b) .8 1.0 .9 .8 .9 1.2 MACK (c) 1.5 1.1 .8 .5 .9 .9 MACK (d) 1.6 2.1 1.6 .9 1.8 1.3 MACK (e) 1.0 1.3 1.2 .9 1.1 1.2 APPLE II 1.0 1.0 1.0 1.0 1.0 1.0 IBM PCjr .9 1.0 1.0 2.0 .4 1.0 IBM PC (f) .8 .9 .3 2.3 .3 1.3 IBM PC (g) 3.9 3.8 3.9 5.0 3.3 6.8 IBM PC (h) 2.8 3.4 4.0 16.3 3.3 14.0 HALGOL (i) 23.6 24.0 12.5 13.0 15.1 18.8 68000/16081 62 71 79 58 78 62
We have been telling you that HALGOL was gonna be fast, and that it would be faster yet with a 16081. The results above, which show HALGOL whipping IBM compiled BASIC using the 8087 5 out of 6 trys, and the 68000 (compiled BASIC) with the 16081 running off and hiding from all other contenders seems to prove that our predictions have been accurate.
About half the letters we are now receiving are suggesting that Mackintosh would make a dandy terminal/host for the Grande (we add: a Grande with a math chip peripheral). Here is an example:
"The idea seems irresistible, especially when you could name the system the MacinTAC... (MacGrande and Great Mac came in second).
"P.S. Is it true that you are going to come out with a portable version of your newsletter?" Pete S., Palos Verdes Estates CA
Pete, haven't you noticed that the misspelling of McIntosh which we have adopted for this newsletter is NOT the same misspelling that Apple Computer has selected? Make that 'MackinTACK', please. Incidentally, you will be pleased to learn that we were stuffing issue #29, all 34 pages of it, into envelopes as your letter arrived. More on MackinTACK elsewhere this issue (maybe).
Lots of folks obviously are not planning to sell many computers or magazines in France. "Maque", pronounced "mack", is the French word for pimp. Since the French have long had a relaxed attitude towards sex, pimps (who prey on honest working girls of the evening) are an even lower form of life over there than in this country. "maque" is to France what "bloody" is to polite English society and what the "f" word used to be in this country.
"I would like to announce a new version of 'Inter68'. The Apple Pascal compatible P-code interpreter. Here are the main differences between version 1.0 and version 1.1:
"The package comes with a modified PASCAL compiler that can generate code to access the Inter68 transcendentals: the compiler corrects several bugs in the Apple PASCAL compiler and provides new options to handle all byte sex questions in a uniform and consistent manner.
"The price for Inter68 v1.1 is $90. For all those who bought version 1.0 I will sell the new version for $60.
An der Junkersmuhle 33/35
"Incidentally, why did you implement the slow 'shift and maybe subtract' algorithm in your double precision floating point division? Especially after using a faster algorithm in your Microsoft compatible package? I think I am going to rewrite your divide routine using the hardware divide. Are you interested in such a routine?" Ulrich S.
Ulrich, you almost make Apple PASCAL look good - and coming from your FNE, that is a compliment. We implemented the 'shift and maybe subtract' algorithm because we thought it would be FASTER than the earlier approach. If you examine that algorithm closely, you should see what we mean. We would prefer not to see proprietary code since we program the 68000 ourselves and make most of the code generally available to DTACK customers. But, you will be happy to note, we DID replace LSL.L #1,Dn with ADD.L Dn,Dn at your suggestion as the code runs two clocks faster per instruction. (And we replaced ROLX.L #1, Dn with ADDX.L Dn,Dn.)
If you are using our transcendentals (with our permission, we assure our other readers) we will send you the 'bug fixes' for SQR and PRINT and CPFPOS - FNE
"I want to make two comments about your refusal of P-code in general and P-code FORTRAN in particular.
"1st: P-code is the simplest way to transfer a programming environment from one computer to another. I am sure that neither PASCAL nor FORTRAN would be available on the DTACK board if Ulrich Schmidt had not done this excellent job in transferring the Apple PASCAL system to your board [agreed - FNE].
"2nd: To run FORTRAN on the DTACK board you do not need to know anything about the (language) PASCAL, but only about the UCSD Operating System. Although this itself is written in PASCAL, the normal user will not recognize this and I do not know about any FORTRAN compiler you can use without some knowledge of about the operating system it runs on. I also think that this system is quite easy to learn and understand (in contrast to CP/M for example).
"I personally do not program in FORTRAN but I wanted to enable everybody who wants to do so using a fast microprocessor and a nice operating system.
"One last note: to transfer programs/data from CP/M to UCSD (and vice versa) I can provide two programs called UCSD and CP/M. These might be of interest to anyone who wishes to leave the user-unfriendly CP/M system and to use UCSD PASCAL on the DTACK board instead." Peter S. Neuss W. Germany
Peter, we bow with respect to the magnificent work that is generally coming out of W. Germany. We acknowledge
that this work is beneficial to the DTACK community and hence to Digital Acoustics. Still, this is a newsletter; we are its editor, and we are not yet ready to turn to prostitution. We honestly cannot see many FORTRAN fans jumping on a P-code FORTRAN which runs on an operating system which has been described elsewhere as "like kicking a dead whale along the beach." However, we have given your system as much publicity as if we were PASCAL fans. If we prove wrong, and you are inundated with orders from FORTRAN fans, write and let us know. We do acknowledge mistakes - FNE
[Hey! This IS an American publication - isn't it?]
Thomas Wieland, the European Theatre DSEx Herr Direktor, would like us to remind you (again) that he did NOT write that letter about PASCAL on the front page of issue #28. He also writes:
"I would like to know how far you want to go on developing HALGOL before you send your last free update. (Will you then) stop working on HALGOL or will you go on and charge for updates? What made you reverse your free update policy? How much do you plan to charge for updates and will there be any at all?"
Thomas, we have not reversed a policy because we never HAD a policy. We sent early purchasers of our boards free updates because we think early purchasers should be rewarded, not penalized. (Owners of S/N #1, both Pet and Apple variety, got free 12.5MHz upgrades 18 months ago using 68000 chips that cost us $332 each. This is the first time we have mentioned that.) We are going to stop sending free updates for the simple reason that we can't afford them. It costs us $10 to mail a floppy disk. One general update is equal to the profit on 40-50 Grandes. You see the arithmetic?
We hope to continue implementing HALGOL. After having the equivalent of a standard BASIC (closer to Wang's BASIC 2 than Microsoft BASIC) we plan to support the 16081 and then to provide high level support for linear algebra and matrix arithmetic. We hope. Our problem at the moment is not SELLING systems, but getting parts to BUILD the systems that have been ordered. This is becoming a serious problem.
In an earlier newsletter, we optimistically told you we were going to double in size in fiscal 1984. WE WILL - IF we can get the parts! - FNE
"I do remember 'wall clock time', especially after reading the (very dubious) times a certain FACOM M150 claimed I used its CPU. Five seconds of CPU time didn't impress me when I knew damn well I'd waited all afternoon for a printed output.
"Still interested in the 16081, although it is really overkill for the sorts of things I need. If HALGOL ends up supporting it, I would almost certainly buy, but I don't need it urgently (hell, all I really want is the fastest, largest personal computer in Australia - that isn't too much to ask, is it?)." Eric L, Faulconbridge Australia
That's what we like about Australians, Eric - they're modest and unassuming, just like your FNE. Shame on you for your apparent collusion with a miscreant in obtaining a Grande, on account of we most certainly did NOT ship a Grande to Faulconbridge. In fact, we have never shipped ANY DTACK board to Faulconbridge, which is slightly larger than three dingos and a billabong - so how come you guys have TWO DTACK boards? Maybe you guys ought to continue to be careful not to mention which serial number you have in your correspondence?
You see, there are some brain-damaged types who think Japanese DRAMs and 12.5MHz 68000s - made in Japan - should not be permitted to pass beyond the borders of the U. S. of A. or CANADA. We are reminded of our Pete Seeger recording of his GAZETTE (Folkways Records FN 2501, copyright 1958). Here are selections from the song, "TO BE OR NOT TO BE":
"I'm gonna preach you a sermon 'bout Old Man Atom,
Now, I don't mean the Adam in the Bible datum,
No, I don't mean the Adam that Mother Eve mated,
I mean the thing that science liberated.
You know, Einstein said he was scared;
And he's scared - boy, I'm scared!
(two stanzas later)
"'Course, the cartel crowd up and put on a show,
They're gonna turn back the clock on the UNO.
Grab a corner on atoms and maybe extinguish
Every damn atom can't speak English.
American for American atoms;
Step right up folks,
Let's atomize world peace!
"Ah, but the atom's international, in spite of hysteria,
Flourishes in Utah, also Siberia;
The atom don't care about politics,
Or who got what into whichever fix.
All he wants to do is sit around
And have his nucleus
Bombarded by neutrons.
(one stanza later)
"Yes, the answer to it all ain't military datum,
Like 'Who gets thar fustest with the mostest atoms;'
But the people of the world must decide their fate;
We gotta stick together -- or disintegrate!
We hold these truths to be self-evident:
All men could be cremated equal!"
(End of song by Pete Seeger, a folk singer.)
On page 58 of the Feb '84 issue of IEEE SPECTRUM magazine is an article entitled "Technology Transfer".
On page 61 is a very UN-humorous (to your FNE) multi-color cartoon showing the "U.S. INDUSTRY STREAM" dammed up, with 10% spillage. It also shows the "TECHNICAL BYPASS RIVER", containing the other 90% of the 'water' flowing into the "RED BEAR WATERING HOLE". All that refers to U.S. technology. Will someone kindly tell us what U.S. technology is in a Japanese 64K DRAM? But under current U.S. law, your FNE can get his fanny tossed in jail for shipping Japanese DRAM out of the country. Or a Japanese 12.5MHz 68000. No, (sigh) we are NOT kidding!
Down with foreign-born microprocessors (or DRAMs)!
Let us suppose, for the sake of argument, that Digital Acoustics and Analytical Engines (the SAYBROOK under-the-hood board) were direct competitors, which we aren't. Suppose that A.E., which is located in Texas, was only permitted to sell within Texas while Digital Acoustics could sell anywhere in the U.S. and CANADA. Who would win? (That's a rhetorical question, folks!)
Consider this: a company located in Japan or West Germany is permitted to sell to an audience of about THREE BILLION PEOPLE! Digital Acoustics is permitted to sell only to the U.S. and CANADA, about a tenth of Japan's marketplace! Have you read lately that this country has a SERIOUS trade deficit? Have you ever wondered WHY? Your FNE can TELL you why - just come by and read our correspondence file of foreign folks who would like to buy some Japanese DRAMs and maybe Japanese 12.5MHz 68000s from us!
(We will step down off our soap-box and rest for a moment.)
"(Here's $15 to renew) ...I am expecting to see (in issue #29) you eat at least half a crow vis a vis the Apple Mac.
"Because of your newsletter, I immediately bought stock in Apple as soon as the facts about the Mac's price, assembly-language monitor - OS - graphics, etc and automated factory became available. If I get rich, I won't share with you. If I drop a bundle on this, expect a visit from my associates." Wayne W. Eugene OR
Drop by our office sometime and we will show you our other letters congratulating us for getting the price absolutely dead on, Wayne. With that OS written in assembly and an automated factory Apple has two thirds of what we predicted was needed to be wildly successful. The third is an included BASIC written entirely in assembly, and they don't have that yet. We never expected Apple to do something really stupid like not allowing for a second internal drive, but history shows that a personal computer with only one serious defect can be successful. If you drop a bundle, we will be able to afford larger and meaner associates than you. Did we ever tell you that your FNE is really not a nice person?
"I sent a copy of SSMON/RC V1.2 with a manual to you right at Christmas time. I still hope you'll point that out in the newsletter since it has some things several people have asked for in the past. As I mentioned at the time the main features are that register contents are controlled on entry to programs and the inclusion of ANASOURCE which will will produce ASSEM68K textfiles from 68K code in the DTACK board. If the people send their old disk back I'll update it for $3, otherwise it's $10 for a new disk and manual."
969 Via Del Monte
Palos Verdes Estates
We caught that substitution of 'loose' for 'lose' earlier in your letter, Pete. You never let a guy off the hook, do you? Watch out or we will reveal that we had to correct three misspelled words and a grammatical error in the paragraph above - FNE
The following is a press release:
"Lone Star Computer is pleased to announce BIGFORTH, a 32-bit implementation of fig-FORTH for the 68000. BIGFORTH 1.1, the version that runs on the DTACK Grande, is now available. BIGFORTH is modeled after Bruce Walker's excellent 16-bit FORTH for the Apple/DTACK system and most FORTH screens will load with little or no modification. It features 32-bit single precision arithmetic, 64-bit double precision arithmetic, virtually unlimited dictionary space, standard fig editor, optional assembler, and some nice operating characteristics which become nicer as memory size increases. One can, for example, almost eliminate disk access times. BIGFORTH also supports 80-col. card/dual-CRT operation. The (growing) manual demonstrates all of these and other features via a walk-through presentation.
"BIGFORTH 1.1 for the DTACK Grande is $55; the source code diskette is $35; the assembler diskette is $25. Send check or money order to:"
Lone Star Computer
P.O. Box 5044
Bryan TX 77805
"I hadn't heard of your graphics card before. Sounds great, but my resolution is only 640 X 480 on my RGB. You don't suppose that since 'all active hardware designs (have been brought) to an apparently successful conclusion' you could put together a moderate res 7220 for the rest of us? Sigh! I thought not. Michael E. Renton WA
Michael, you will find elsewhere in this issue a real horror story about the 7220. Also, we have been unable to locate a reasonably priced non-interlaced 640 X 480 RGB. If your monitor meets that description, would you tell us more about it?
We just got a renewal from a guy who works for one of those outfits which make very good, and very expensive, UNIX boxes. Instead of printing his letter we are printing our (lightly edited) reply. For one thing, he admits that his company's UNIX manual is not 1,000 pages long but 3,000! And he admits they do not like to publicize that fact. Also, we are getting tired of that "we" used by royalty and by editors:
"Welcome back! I thought you were mad at me. When you get issue #29 you may recognize yourself on page 23, column 2. The system illustrated on the front cover uses a math processor that is almost as fast as [company deleted], and a whole bunch cheaper!
"FNE originally meant FAITHFUL Newsletter Editor. If you don't believe it, take a look at the number and size of the newsletters I have turned out! I composed the subscription renewal letter on an Eagle II and turned it over to my secretary to re-type on letterhead. Then I didn't notice that 'Eloi was misspelled (you are the second person to call that to my attention).
"You should like UNIX - UNIX is an absolutely perfect tool for a highly motivated, full-time computer professional who has the cash (or whose boss has the cash) to buy a really good UNIX machine. For $37,000 I certainly hope that your system has good response! I happen to get (I believe) equally fast response from the Eagle II I am keying on at the moment - and you can buy an Eagle II for under $2000. Where is a UNIX box that will give good response and cost under $2000?
"At $37,000, the UNIX box on your desk, which we both agree is very good, is in no way targeted at the mass computer marketplace. When I write that UNIX ain't gonna make it, I am referring to the mass computer marketplace. Have you noticed all the IBM PC owners discarding PC-DOS and rushing to buy PC/IX? Neither have I. And although you and I appear to have similar sentiments re one-man-one-CPU I just read after breakfast a prediction by one Scanlon (I'll send you a copy) which claims just the opposite trend will soon occur.
"A mass-market personal computer is going to be a cheap computer, or at least a relatively inexpensive computer. It is not clear to me that virtual memory can be properly implemented on a mass-market computer.
"I have a neat little file system I have been building since 1961. Right now I am in the process of filling a fourth four-drawer cabinet. Last year when our project engineer was still in school, he needed some information on Hall effect devices. So I reached into the top drawer of file #1 and pulled out the Hall effect folder and handed it to him. Nearly all of the articles were from 1964, because that's when interest in Hall effect devices peaked.
"This relates to your assertion that "people in the computer world are notoriously antagonistic to taking measures of performance." Voila! Enclosed is a copy of a 1972 DATAMATION article on "The Profit in Bad Software." The GAO is trying to get the Air Force to sell off its CRAY-1 in New Mexico. It seems the beast is only being utilized 55%. We now segue to Minnesota, where CRAY is busy reporting that it is going to adopt UNIX! I wonder why..." Felgercarb E., Santa Ana CA
heavily influence one's opinions. We all know about the blind men and the elephant. This newsletter is written from the point of view of folks who own, or want to own, a high performance computer for their own use. Folks who will likely pay for that computer out of their own pocket. Most of these folks are employed in some field other than the computer industry and hence are not full-time computer professionals (DSEx founder Jeff H. is a pediatrician, for instance). Only intelligent, highly motivated full-time computer professionals are ever going to be able to master UNIX, whether the manual is 1,000 pages long or 3,000 pages.
Let us now examine UNIX from the point of view of the guy the letter above was addressed to: he has a $37,000 UNIX box sitting on his desk and he does not share CPU cycles with anyone. He is, obviously, a full-time computer professional. The $37,000 came out of his employer's pocket, not his own. He is obviously
intelligent because corporations do not place single-user $37,000 UNIX boxes on the desks of dummies. You can see how this guy would like UNIX.
There is a newsletter about the world of full-time computer professionals working with very good, very expensive (from our point of view) UNIX boxes. The newsletter is titled "Unique" and is edited (written) by David Fiedler, who also serves as editor of UNIX REVIEW. From the point of view of "Unique", UNIX is a very good operating system indeed! And $37,000 is really a moderate price for your employer to pay for a very good UNIX box...
Look, Gucci sells $3,000 purses on Rodeo Drive in Beverly Hills. Gucci does not have any shops in the Bronx or in East St. Louis or in Arab, Alabama.
Let us leave Rodeo Drive and return to Santa Ana where folks are not full-time computer professionals and they buy their own computers. Do you realize that there is a virtually unanimous consensus on Rodeo Drive that they are going to come down to Santa Ana and sell us some $3,000 Gucci purses? That about 100 companies have invested big, big bucks in (mostly 680X0-based) UNIX systems which are all going to be highly successful in the mass marketplace? That UNIX is going to become the de facto standard 68000 operating system?
How can they possibly believe that? Because of their point of view! That full-time computer professional looks at the $37,000 UNIX box on his desk and the 3,000 page manual sitting beside it and thinks, "Yea, this is great! Everybody ought to have a system like this!"
The other day we called Jeff H, the DSEx founder, at about 7:15 PM Texas time. His wife gave us his work number. We figured he was taking care of some paperwork. Uh, uh. We couldn't talk to him because he was with a patient! He returned the call after 7:30 PM and was going to see another patient after our brief chat. The guys on Rodeo Drive look at Jeff's pediatrician income and his interest in computers and say, "we're gonna sell that guy a UNIX box!" The truth is, Jeff can afford a UNIX box and he is probably interested, having read a lot about UNIX.
But as soon as the Jeff H's of this world learn about that 1,000 page manual and how long it takes to master UNIX, they lose interest REAL fast.
Here is the point which the folks on Rodeo Drive can't seem to figure out: the reason UNIX will NEVER succeed in the mass marketplace has nothing to do with the price tag. It has EVERYTHING to do with the time and dedication required to master the system. In the mass computer marketplace folks are pediatricians or engineers or economists or dentists or math profs.
The reason the folks whooping it up on the permanently stalled UNIX bandwagon cannot see this is that they have a different point of view (they have hold of a different part of the elephant). Every single person on that bandwagon is a dedicated full-time computer professional who thinks UNIX is WONDERFUL - as it really IS, for them!
Some of you have perhaps wondered how your FNE forms opinions on various subjects. Well, we try hard to ignore promotional puffery and we watch the marketplace VERY carefully. Examples: about 15 months ago we saw a classified ad in the Sunday LA Times by a Victor dealer who was offering three Victor 9000s for sale for $2400. Since the list price at the time was $3995 that was obviously a dealer unloading at his cost. So we watched closely as Victor became a separate entity and then dove and crashed.
While a lot of folks were still whooping it up over the $10,000 LISA I, we paid close attention to the letters we were getting talking about LISA's word processor not being able to keep up with a touch typist. WAIT A MINUTE! we thought to ourselves. OUR CBM 8032, with a lousy 1MHz 6502, has no trouble at all keeping up BUT LISA WITH A 68000 CAN'T? SOMETHING IS ROTTEN IN DENMARK! And then we read about one of two LISA production plants being closed down and the other cut back 55%. Look, we run a (very small) production plant and we understand that VERY clearly. And then came the price cut and the unbundling of the software...
But everybody can see those things, you say? Yes, if they will look. But some folks obviously don't look. InfoWorld's Kathy Chin praised LISA I's software as one of the three products of the year when that software was the reason LISA I in fact died a month after Kathy made that award! The LISA folks are now frantically writing assembly-language software for LISA II, some of which is being stolen from Mackintosh.
Are some signals emanating from the UNIX camp? Boy, are there! One cannot open an industry publication these days without reading of more implementations of UNIX on 680X0 machines, 16000 machines, 286 machines, the PC... All of this naturally means that the companies already in the marketplace are doing well, right? Let us examine how the two most visible mass-market UNIX outfits are doing.
One is the Tandy 16. According to Tandy's released sales figures, the 16 sold better in the year before XENIX arrived (a year in which XENIX was eagerly anticipated) than in the year AFTER XENIX arrived! The model 16 is now in the intensive care ward and is breathing raggedly, about the way LISA I was last fall.
The most prominent is the Fortune 68000/UNIX system, which you will recall was the smash-hit of COMDEX a little over three years ago. They finally got their production up in late '82, went public WITH THE TENTH LARGEST STOCK PLACEMENT ($110 million) IN AMERICAN BUSINESS HISTORY in spring of '83. Nearly two-thirds of that $110 million is still sitting in various bank accounts drawing interest so Fortune has no cash problems at all, and after three years' design time and over a year of mass production they should be doing as well now as they ever will.
So how are they doing? They have had MASSIVE losses in the past two quarters on sales conspicuously lower than when they went public. You want more? In this past week, an ad appeared twice in the Wall Street Journal. An outfit was selling Fortune computers for exactly HALF the list price! What does that mean? YOU figure it out!
Let us add one more datum: there is an outfit named Wicat which makes good $10,000 68000 machines. For the first year or year and a half after they got production going, they did not offer UNIX. Roughly a year ago, they DID offer UNIX. We have a research project for you: Examine Wicat's profitability for the four quarters BEFORE they had UNIX and then examine their more recent post-UNIX profitability.
It is apparent that UNIX, in the mass marketplace, does better in the anticipation than after the fact. We just can't wait for Analytical Engines (the SAYBROOK under-the-hood 68000 folks) to actually follow through on their promise to get UNIX up and running - on 143K floppy disks!
XENIX, a UNIX work-alike, has been around for some time now. Name ONE product using XENIX as its primary operating system which is growing and profitable. We'll wait. Hum diddle de dum dum. Cough! You've thought of one? No? Well, we'll wait a little while longer...
Now, if it's product announcements that get your undivided attention and your blood racing hotly, there are lots of new UNIX product announcements appearing weekly. With each one, a new roar of excitement and approval goes up from all the folks sitting on that stalled UNIX bandwagon.
In the meantime, your FNE is getting closer to the status of a full-time computer professional, a status that many of our readers (who are NOT a typical cross-section of the mass market) attained long ago. When Digital Acoustics gets a bit more flush we may try to convince our employer that we are sufficiently
intelligent and motivated to rate a $37,000 UNIX box on our desk that we don't have to share with anybody!
Naturally, we will have to give up writing this newsletter so we will have time to learn that 3,000 page manual (we will want the BEST of those good, expensive UNIX boxes).
UNIX may cause us to stop writing this newsletter! Because we have told you all we know of the situation with respect to the mass market, and because other industry pundits are catching on (Dvorak, Pournelle and even Jeffries, a strong UNIX supporter, have noticed that UNIX boxes don't seem to sell in the mass market), we are going to teach each of you how to be your own FNE in re mass-market UNIX.
If you will follow these rules, you should, in a very few months, come to the same conclusions we did a long while back with regard to UNIX's suitability in the mass computer marketplace. And now, we lay this subject to rest (honest... we think).
This is a bona fide horror story, one of two such in this issue. Jack is vice-president of the Computer Systems Division of A T & T Technology Systems - which used to be known as Bell Labs. In the 8 Mar '84 issue of Electronic Design, on p.52, Jack asserts that:
"As 32-bit microprocessors begin to pave the way toward multitasking systems that serve several users, people will be forced to ask whether they need a personal computer on their desk. The answer will be a resounding 'no'!"
We simply don't have the heart to report Scanlon's views further. Needless to say, his company is busy preparing shackles and chains to tie you and some other worthless peons to a multi-user system. Using what operating system? Well, Jack IS from A T & T...
is about NEC's 7220. This is also a story about how it is better to be lucky than good. The 7220 was designed to make possible the inexpensive design and construction of displays which are quite good but not state-of-the-art. 640 X 480 would be a typical 7220 display, 800 X 600 would represent the best one could do with a premium high-speed 7220. The 7220 is designed to not only draw pixels to memory but also to sequence the memory to the display.
800 X 600 was not good enough for the VDHR so we used not one but two memories, each dual-ported. At any one time, one of the memories is being displayed using 'F' and 'LS' TTL logic. The 7220 can write into (or read from) the other. This means, among other things, that the 7220 clock is completely independent of the display clock. We use the 7220-1 and a 5MHz clock for a faster writing rate, and a 67.888 MHz pixel clock for a 1024 X 796 display. We can, of course, swap these two memories back and forth while the horizontal scan is off the screen, so no flicker can be seen. Ned W. asked us at the time if that was an original design and we replied that, although we came up with the idea ourselves, the design approach was so obvious that we assumed other folks were doing the same thing. That was 9 months back and we have not seen that approach publicized elsewhere.
Because of that design approach, we just happened to get real lucky. You see, we don't use any of the 7220 stuff that sequences the screen display. That is a very good thing, because it happens that the 7220 is MASSIVELY bug-ridden when it comes to driving things to the display. We have a 6-page NEC product bulletin sitting beside us (PB# 29) which itemizes 13 bugs. Only one of those could possibly affect the VDHR, and
it turns out that there are better ways of accomplishing what that one bug (#8) affects. So we got lucky.
But are we P.O.'d! (As Bob P. would point out, that could stand for Post Office but in this case it does not.) We began the VDHR design early in '83. We only received this bug list a couple of weeks ago. THE PRODUCT BULLETIN IS DATED AUG '82!
We have heard reports of late (or no) delivery by, for instance, Number Nine, which makes an under-the-hood conventional 7220 display driver. Now we can understand and even sympathize.
Why did NEC finally reveal that bug list to us? They are beginning to sample the 7220A, which naturally has those bugs fixed. It also has a higher clock frequency up to 8MHz. 7.2 MHz is as fast as we can go while meeting the 275 nsec cycle time spec on the 150 nsec DRAM we use. But since our display clock and our 7220(A) clock is independent, we will be able to provide a higher-performance (by 44%) display by simply changing clocks and switching over to the new 7220A (and a few timing adjustments). It's nice to be lucky!
are what you can get as a sample today - if you are with the right company. The following quotation is from Electronics magazine, 8 Mar '84, p.97:
"A 16032 with a 6MHz clock rate is in production and samples with a 10MHz chip are available. The 32032 has a 4-MHz clock rate; a 6MHz is due in 60 days; a 10-MHz in the third quarter."
High performance microprocessors, like software, are always delivered on schedule. On the same page, Electronics asserts that 68020 samples are scheduled for this summer, while we have told you that WE have a 50-50 chance of getting a 68020 sample by 1 Jan '85. Why the contradiction? There is no contradiction! Electronics is describing the time frame in which General Dynamics can get one sample, Honeywell one sample. IBM two samples (one to turn over to Intel). Two months later Tandy can get one, three months later Charles River, and six months later Digital Acoustics gets what the littlest pig got.
We have consistently been trying to impress upon you how long it takes for things to happen in the state-of-the-art microprocessor game. If you learn, you will not charge into our offices in 1982 and demand to see our prototype 68020 board, and you will glance askance at Electronic's assertion that the 32032 is going to go from 4MHz to 10MHz in six months...
Yes, NOT InfoWorld watch. Those guys at InfoWorld keep stepping on our toes lately. Twelve days ago, on a Saturday morning, we wrote the bit which appears elsewhere in this issue under "COMETH THE ONSLAUGHT (YAWN)". The next morning, the Sunday L.A. Times printed in the business section a column which was strikingly similar. Now InfoWorld, the one with the FBI agent in a trenchcoat on the cover, prints an editorial on page 5 that is even more similar to what we wrote.
John Dvorak's column in the back of that same issue asserts that there is not a heck of a lot of pride and honor in making the umpteenth exact copy of the PC. Now, it is true that we have never written that in these pages but it is most certainly what we have been thinking for a long time. We are going to see our attorney about this blatant copying of our material. WATCH IT, InfoWorld!
The following quotation is from David Bunnel's column in April PC World:
"A few months ago, I declared 1984 'the year of the mouse,' by which I meant the year of personal computer pointing devices. On second thought I see that 1984 is the year of the PC-standard personal computer."
David is the publisher of both PC World and MacWorld. It is truly a pleasure to see somone like David stick firmly to his principles, whatever they might be at a given time. We sure hope that John Dvorak understands that there is a great deal more pride and honor in making a 'PC-standard personal computer' than there is in just making an exact copy... isn't there?
The folks at Analytical Engines have been busy, it seems. They are still promising UNIX in the spring, this time the spring of '84. They have even had a price decrease - the 8MHz base system now costs just $895. (We wonder where they get their TTL?) The SAYBROOK is now the SAYBROOK II. It might be that they switched from their former Multiwire construction to standard 4-layer PC boards, which should be more economical.
They even have a 14MHz version, which is interesting considering the fastest TESTED 68000 you can buy is 12.5MHz. (It's fairly easy to select a 12.5MHz which will MOSTLY work at 14MHz.) Here is a quotation from a flyer dated 1/84: "The 68000 math chip has not yet been introduced by Motorola; therefore, there are no
immediate plans to incorporate a math chip into SAYBROOK II." Boy, those A.E. folks stay right on top of things, don't they?
Is the SAYBROOK II fast at 14MHz? Is sure the heck is! The Terry Peterson benchmark that runs in 18.6 seconds using HALGOL/DTACK runs in 77 seconds on the 14MHz SAYBROOK II! For PASCAL, that is really moving out! If you are interested in a fast under-the-hood board, their phone number in Texas is (512) 346-8430. (If you call them, ask them why they don't place copyright notices on their product literature? We're curious.)
(following are the concluding paragraphs of a two-page howl of complaint about deficiencies in HALGOL, written 5 weeks AFTER the writer received HALGOL alpha release #1 and 1 week BEFORE he received alpha release #2:)
"Page 24 [nl #29] is a real cop-out. There are n software guys out there that support m hardware devices without having those m hardware devices. You put the micro revolution into the MINI league - CAVEAT EMPTOR if you don't got what I got, then I will just be too lazy to follow the standards and you will do without.
"Except as a rock shooting toy, you have made certain that HALGOL will be an artifact and unuseful. Take a vote - find out how many plain janes there are out there with full-house Apples. In fact how many DTACK owners have an Epson, much less a plain jane interface? Sincerely (and that is appropriate)" Bob P., ST. Louis MO
(We think Bob is slightly angry.) "Caveat emptor" is Latin for "let the buyer beware." We have here a classic case of a buyer who is unhappy with the quality or suitability of the goods which he has purchased. Bob, we will apply the classic remedy: here is a complete refund ($0.00) of your purchase price.
We would never want to force you or anyone else to use a programming language that you did not like, just as we hope you would not want to force US to use PASCAL or COBOL! However, we feel that the product we are developing, and the path we are following to develop that product, is not totally devoid of merit. May we have equal time? Thank you.
HALGOL is being developed despite the fact that there is no possible financial return to Digital Acoustics in the short or medium term - and remember, we have to pay our bills in the short term. HALGOL is being developed by folks who know a lot about the Wang 2200 and the 68000, a little about the CBM 8032, and very little about the Apple. You write, "...too lazy to follow the standards..." WHAT standards?
How do those n software guys pay their bills? What's that? Oh, you say they charge $195 to $995 for their programming languages, which are provided on locked diskettes? We can see that you would want to hold HALGOL, provided for $0.00 on an unlocked diskette, to the same "standard" of performance.
We don't suppose that it would do any good to point out that HALGOL is not yet completed - little things like strings and arrays are missing, not to mention a DOS. Instead of continuing to forge ahead toward a reasonably complete language, it makes a lot of sense to stop and work on the 10,000 different add-ons which have been sold for use with the Apple II?
Bob, we sure apologize for that little hint we left on page 24, the page which you say is a real cop-out. Shucks, anybody who followed that hint would have to actually do nasty old (ugh) WORK - and that would interfere with one's inherent right to squat and howl until somebody ELSE does the work, right?
Isn't that little headline above silly? Who ever heard of standards that are secret? Well, Bob P. has a supporter in the Apple community (pseudo-) named Jim who asserts that there are indeed standards for control-key usage. Fine, we said. Where can we get a copy? "Oh, these standards are not written down anywhere. You are supposed to absorb them by osmosis over a number of years of programming!" Jim replied.
But Jim, we said, we have NOT spent a number of years programming the Apple. In fact, although we have six or seven Apples at Digital Acoustics (we have lost the exact count) we don't do any Apple programming on them, just 68000 related programming. How can you criticize us for not using Apple standards if those standards are not available to us?
"Why, if you are going to offer software into the Apple marketplace you will simply have to do your homework!" Jim exclaimed, "Since the standards aren't written down everyone has to go through this 'trial by fire'!" We replied, we think the system stinks! We think somebody ought to PUBLISH those quote standards unquote. As it is, we can get criticized - we ARE being criticized - for not knowing something which there is no way we could have known! Uh, did you get that?
"Your logic is a bit turgid, but I think I got the picture" Jim replied. Hmmm... Jim, maybe you could give us just a teensie bit of help here? we enquired. (Stay tuned.)
(We are reminded of the recent Pentagon proposal to make the list of devices which could not be exported
secret. "If you dare export one of the devices on this list we are going to throw you in jail!" the brass hat thunders. What are the devices we are not supposed to export, sir? we meekly reply. "Oh, I can't tell you that!" responds the brass hat, "The list is secret!"
The one by Microsoft. We just bought the Eagle II version, $350 list discounted down to $295. Why? About the same reason we bought (ordered) the Tandy 2000. It will be refreshing - we are told - to find a Microsoft BASIC which permits the use of MID$ on the left hand side of the equation. Also, we want to see how the disk file handling under that BASIC interacts with CP/M. (As we want to see how GW BASIC on the 2000 interacts with MS-DOS 2.0.)
One of the nice things about working in this business is you never have to worry about how to fill your spare time.
There is this here pop song which opens with some teeny-bloppers singing in chorus "Wolf Creek Pass up on the Great Divide - trucking on down... the other side!" The "Great Divide" refers to the very top of the Rocky Mountains, where two raindrops which fall closely together might flow in opposite directions - one east, one west (that's the "Divide" part, and if you know all about this we apologize for explaining).
Its seems these two country type gentlemen are delivering a truck-load of chickens to a township on the other side of the Great Divide. Now, it is one heck of a steep, winding road going down the other side of the Pass so it's a good thing they have good brakes - OOPS, THERE GO THE BRAKES! Well, a skilled trucker can always use the gear shift for braking purposes - OOPS, THE GEAR SHIFT LEVER JUST FELL OFF! Better settle back, folks, it's gonna be an interesting ride!
Commodore is traveling down the other side of its Great Divide (before/after Trameil's egress). It lost the brakes a while back when new management (from the garment district?) misread the entire purpose and marketing plan of the new 264/364 line. Well, the gear shift lever just broke off. Commodore is actually, get this now, going to manufacture the hundred-and-leventy'-seventh PC clone. Not only that, but they are not going to design their own clone - they have instead, according to the current Electrical Engineering Times, elected to build (under license) the Hyperion portable PC clone! Better settle back, folks, it's gonna be an interesting ride!
Now you, too, can suffer from a fashionable ailment without being athletic at all. Just buy a Mackintosh. In some circles, "Mackintosh wrist" (from swapping microdisks) is beginning to rival "tennis elbow". If you do not understand that remark, you obviously have not yet met someone who actually owns a Mackintosh!
We just got ProDOS, but not the documentation needed to access it at the machine language level. Yes, we know BYTE magazine said that was available but BYTE operates on a six-month lead time. Perhaps it is available (for now) only to licensed (by Apple) software vendors! When we spoke with Don Worth a month back he had just finished Chapter 6 of "Beneath ProDOS", which we suspect is going to sell pretty well.
One good thing about ProDOS is that it supports hard disks. We are getting pretty doggone tired of that 140K Disk II, let us tell you. On the other hand, backing up a hard disk on that 140K Disk II is not appealing at all.
That brings us to the subject of removable hard disks. It seems that everybody (every company) in the world is EVALUATING hard disks but hardly any have ADOPTED one of those hard disks. Even more disheartening is that one continually reads in the trade journals that all is not well with the removable hard disks. Which is a rotten shame, because a two-drive system with removable media sure as heck whips the backup problem.
There is a company named IOMEGA which makes a removable-media 5 1/4 inch form-factor 10 megabyte drive that uses, get this, a FLOPPY disk cartridge (single surface). It's actually FASTER than a lot of Winchesters, being capable of backing up 10 megabytes in 30 seconds. That's right, backing up. You see, IOMEGA's ads (in Mini-Micro Systems, for instance) clearly assert that the purpose of the drive is BACKING UP HARD DISKS! In other words, the IOMEGA drive is intended for secondary storage. Naturally, an outfit in Georgia is selling that drive for the Apple II for use as PRIMARY storage!
In the meantime, floppy disks are advancing (everywhere but at the Apple Computer Co) into the megabyte range, with Amlyn producing a 3.3 megabyte 5 1/4 inch floppy drive which is second-sourced by Kodak, the camera/film folks. Why, oh why can't we get any of the above for our Apple? (We wonder how loud the screams would be if we went ahead and tied that IOMEGA drive directly into the 68000 via an interface board - one that had a socket for a math chip and maybe an RS-422 port...)
the Walrus said, to speak of many strings. Strings? Well, (sigh) we guess it IS time to speak of HALGOL strings. Based on the correspondence we have had up to flows most of you are not going to like the way we are going to do strings, for now. On the other hand, 18 months ago most of you simply could not understand why we did not use the super-intelligent 6502 simultaneously with the 68000 so that TWO wonderful things could happen at once. We are very pleased that we don't hear much about such simultaniety (NOT concurrency) these days. We strongly suspect that once you are exposed to strings which work correctly, you will come over to our side - again!
On our side, unformatted string arrays are possible. That means executable object code can be manipulated as strings - something which is totally impossible using dynamic string arrays. An unformatted string array is one in which the byte value $0D (or $8D) has no particular significance and is only one of the 256 possible byte values. Unformatted string arrays look the same in memory as on disk. A FORMATTED string array (done correctly) is one in which the value $0D (or $8D) terminates a particular string when stored on disk. When stored in memory, each element of the array is stored in a fixed-length string which is padded out with spaces.
Why store formatted strings arrays in unformatted arrays when in memory? Because otherwise we have dynamic strings, that's why!
Since our strings are fixed in length, we need to have some control over their length. One simply DIMensions the string length as needed just as numeric arrays are DIMed. Just like numeric arrays, there is a default string length, in this case 16 bytes. If we want a string 40 characters long, named "ADDRESS$" we DIM ADDRESS$40. Nothing to it.
String arrays must be dimensioned as well. If we have an array dimensioned DIM A$[15,255] or A$[15,255]16 we have a nice 16 X 256 X 16 alpha array. That's 64K, or about 6% of a full-gallon Grande. To find A$[7,141] we multiply 16 X 141, multiply that result times 7, and then add the base address of array $A and VOILA, we have found our string. That is not only fast, it is trivially simple to program.
If we want one-dimensional string array of 100 elements, each six bytes long, DIM B$6. Of what use is such a string array? One could keep up to 100 addresses to any place in a sixteen megabyte memory space in such an array. You see, with fixed strings
one is not restricted to the manipulation of alphanumeric characters...
are being adopted so that we do not have a conflict between the mathematical operator LOG( and a numeric array named LOG. We do not mind at all that PASCAL uses square brackets (for the same reason?). We are absolutely shameless when it comes to stealing good programming ideas.
(The following strings are assumed not to have been dimensioned, and so all have the default length of 16 bytes.)
10 LET A$="ABCDEFGH" 20 LET B$="1234567678" 30 LET MID$(B$,3,2)="YZ" 40 PRINT "B$ = ";B$ B$ = 12YZ5678
Now note the distinction in these two operations:
50 LET MID$(A$,1,3)=B$ 60 PRINT "A$ = ";A$ A$ = 123DEFGH 70 LET A$="K" 80 PRINT "A$ = ";A$ A$ = K
If "MID$" is used on the left-hand side of the equation, then only the designated characters are changed. If "MID$" is not used, as in line 70 above, all of the characters are affected. In the case of line 70, the first character becomes "K" and the remaining 15 characters are padded out with spaces.
Now follow these two program lines:
90 LET X = LEN(A$) 100 PRINT "X = ";X X = 1
This example shows that the LEN function is performed by fetching the dimension of the string, and then subtracting one for each right-most space, terminating on the first non-space character.
A pseudo-algorithm for calculating LEN follows:
500 "LEN" 510 LET LEN=DIM(A$) 520 "LENCOUNT" 530 IF MID$(A$,LEN,1)<> " " GOTO "LENDONE" ELSE LEN=LEN-1 540 IF LEN > 0 GOTO "LENCOUNT" 550 "LENDONE" 560 RETURN
But since we do not approve of the practice of writing a high-level language in another high-level language, we will write the LEN function in 68000 assembly.
You would think we would have better sense than this, but we really cannot see any big problems (lots of little detail, yes) in implementing strings. Maybe that's why Tiny BASICs (remember them?) generally had strings but not floating point.
On the other hand, the last time we made such an incautious assertion was when we told Le Roy offhand that programming a floating point add would be a great deal simpler than a multiply! Ho ho ho indeed.
The second HALGOL alpha release is now being evaluated by a couple dozen folks. On the last page is a sample program written in HALGOL v1.02, which is Terry Peterson's benchmark program from issue #25, p. 12, with a few added bells and whistles to demonstrate features. Please note that all those primitives such as LOAD and STORE and CHS (change sign) appear to be missing. (That last sentence is literally correct: the primitives APPEAR to be missing.)
In fact, the primitives are still there and are still doing all the work - it's just that we have kept our promise and have used the 68000 to obscure from the programmer the details of how HALGOL really works.
Actually, that program on the last page looks a lot like Applesoft. So what is the easiest way to distinguish between Applesoft and HALGOL? With a stopwatch! Like 18.6 vs 488 seconds!
If you are impatient and would like a copy of the v1.02 demo disk (and the 17 page 'manual') send $10 (U.S & CANADA), $13 elsewhere. If you would like the demo disk AND THE SOURCE CODE, all crammed onto three DYSAN disks complete with instructions for un-cramming, send $30 (U.S. & CANADA, $35 elsewhere. Remember, there will be a third release which will obsolete v1.02 in another six or eight weeks.
There is this apocryphal story about a chinese peasant who did something to greatly please his emperor. When the emperor asked the peasant to choose his own reward, the peasant asked for a chessboard. On the first square he wanted one grain of rice. On the second two, and on the third four... it took the emperor a while to catch on to what was going on, but when he did the peasant lost his head. Literally. The last square would have to be covered with two to the 63rd power grains of rice, and there ain't that much rice.
That is a very old story which all of you have heard already. So why did we repeat it here? We dunno. Naturally it has nothing to do with you and us.
has been doubling every year for a number of years. The personal computer market is sixteen times larger today than it was four years ago, and it is one-sixteenth the size it will be four years from now. The dollar volume of personal computers is rapidly approaching that of mainframe and minicomputers combined. The consumption of TTL and memory by the personal computer industry is proportionally greater than the dollar volume. (A TTL gate costs a lot more as part of a mini than as part of a VIC-20.)
Delivery of TTL parts is now running about 4 - 8 months behind and lead times are lengthening. Prices are doing exactly what you might think as a result. Most of the major semi houses are making ENORMOUS profits. (T.I.'s semi manufacturing is in fact yielding enormous profits, but is trying to fill the huge hole caused by the 99/4A.)
Where are all these parts going? Take apart a Commodore 64 and look at it. Commodore is shipping 400,000 64s a month and can't meet the demand. Apple is shipping 100,000 IIes a month, IBM at least that many PCs and maybe soon more than that in Peanuts. Coleco is planning on selling a LOT of Adams.
Memory stuff: unless you know something we don't, the average IBM PC has about 400K of RAM once it has been in the field four months (RAM disks are VERY popular). We don't know what the proportion is of 64K vs. 128K IIes but we'll guess 3 - 1. Here's what we get:
Computer number/month megabytes/month CBM 64 400,000 25,000 Apple IIe 100,000 8,000 IBM PC 100,000 40,000 ------ 73,000
According to our figures, those three personal computers alone are eating DRAM at the rate of nearly ONE MILLION MEGABYTES per year! Burroughs is a big company in data processing; DATAMATION ranks it number three in the U.S. behind DEC and you-know-who. Burroughs' sales in 1982 barely exceeded $4 billion. Apple and Commodore are already selling at a $1 billion-plus rate, and Tandy is right behind, IBM's PC division is certainly selling at a $1 billion-plus rate - in fact, the 2.5 million personal computers they have predicted this year amount to over $4 Billion, the way we figure it.
Back to Commodore. They are shipping 300,000 megabytes a year to shoot rocks with. At Burroughs' prices, that's about $4 billion worth of DRAM. Since a majority of Burroughs' sales dollars are for CPUs, disks, tapes, printers, software, etc. and yet their TOTAL sales barely exceed $4 billion, it is pretty obvious that Commodore is consuming 4 times as many DRAMS as Burroughs - and Burroughs is #3 in data processing (U.S.). And Commodore is growing at a yearly compounded rate of 2.3 - 1, or was until kindly Uncle Jack left. Burroughs isn't.
Now you know why there is beginning to be a shortage of TTL and memory of all kinds. Think about this: in three years, the personal computer industry will be consuming eight times the TTL and 32 times the megabytes that it is consuming today. (32 times the megabytes means 8 times the number of 256K DRAMS.)
How does this affect you? DRAM pricing is about the same right now as it was two years ago. While RAM pricing has in fact halved every 18 months historically, there have been hiccups in the curve from time to time. Right now is the start of one gigantic hiccup that, in our opinion, is going to last for at least three years.
When you double the number of rice grains on every successive square you can go from smiling, happy emperor to beheaded peasant in only a few squares. When a marketplace doubles every year something goes crunch, eventually. Something just did.
The semiconductor manufacturing industry is highly capital-intensive. If we are going to have a growth factor of eight in the personal computer industry in the next three years (and we will) then we are going to need two to three times the total integrated circuit manufacturing capability, worldwide, three years from now than we have now. Because the industry is so capital-intensive, that means lots of borrowing.
Because we (the United States) are spending $200 billion per year more than we have got, most of that borrowing is going to be done in Japan and most of that
additional manufacturing capacity is going to built in Japan. What's that? You don't like what we are writing? We have news for you: we like it even less than you do.
Are we leading up to an announcement of a price increase? Yes. You never expected to see a DTACK price increase? Neither did we. But we priced the Grande last summer when DRAM and TTL were both cheap and available. Now, DRAM is not so cheap and TTL is not even available (for new buyers). We just paid $2.25 each for some 74LS32s on the gray market - and the 74LS32 is normally a 28 cent part!
Last summer the outfit that makes our crystals (Murata - Erie) was reasonably happy to make us three sample 67.888 MHz crystals for our prototype VDHR boards and quite happy to sell us 100 pieces at that same frequency for production. Last month we went back for 150 more pieces and they turned down the order! Now, if we would like to buy 5000 pieces... What we wound up doing was ordering 250 pieces each of three frequencies (750 pieces total), the price was jacked up 28%, and they magnanimously condescended to accept our order. This, boys and girls, is the reality of today's marketplace.
You may have noticed that we are not running ads in the personal computer - or any other - magazines these days. In recent months, that was because we figured it would be stupid to buck the feverish expectations over Mackintosh. We had planned to place new ads in February, but we didn't. We are deathly afraid that we will just get lots MORE folks placing orders that we can't ship because of the shortage of TTL.
In fact, the mythical case, which is becoming a mythical case/power supply and growing wider accordingly, might be artificially delayed even more if the parts shortage worsens - and we are afraid that is going to happen. We know that a bunch of you readers are planning to buy a Grande when the case/power supply arrives. If we do not have the ability to fill those orders, we would rather you be disgusted with us not producing the case than infuriated by us not shipping you a 68000 board that you have ordered. If this situation makes YOU unhappy, dear reader, imagine how WE feel!
Oh, yes, that. The Grande's price has been increased by a flat $100 across the board, effective 1 March '84. That means a 128K Grande is now $895 and a full gallon is $2095.
The Stuffer board is nothing but some epoxy-glass and a whole bunch of LSTTL or ALSTTL. We very nearly decided
to discontinue the Stuffer for the duration of the TTL shortage but decided that might deal severe injury to a few folk's plans. So we decided instead to plant a great big "KEEP OFF" sign by raising the price $150. Now it is $245 without documentation and $260 with documentation. You think that's too much? Good! That's precisely what you are supposed to think - we don't want to sell Stuffers right now.
Imagine not being able to sell a full-gallon Grande for lack of TTL that got shipped in a Stuffer and you will see the point.
A flat $100 increase is not bad considering market conditions. (And thank heaven we do not use any Intel LSI parts because some companies that do are going out of business for lack of parts. We are not kidding even a little bit!) In any event, we have proven in the past that we are willing to drop our prices when our costs decrease.
Some folks see an easing of the situation this fall. We see personal computers doubling every year and we see that they are already a VERY significant factor in THIS year's shortage - and we get worried. DRAM prices are now the same as two years ago (and $.80 higher than last summer when we priced the Grande). This is unprecedented, given the historic price trend. It means that the folks who have the capacity to crank out DRAM are making high profits - and guess which country has the most capacity?
The following is NOT a prediction, it is a hopeful scenario: although personal computers are doubling in sales, many are being purchased by businessmen and there are MANY more persons than businessmen. So, the last businessman buys his PC and sales of the PC and its clones plummet. PCjr sales continue to be minimal because of the busted keyboard and because a PCjr expanded to be useful costs about the same as a real PC. IBM's prediction that it will sell 2.5 million personal computers in 1984 falls short by 1.5 million and IBM cancels many of the contracts it now has for delivery of TTL - and those parts find their way to the market instead (and to Digital Acoustic's shelves!).
Mackintosh becomes enormously popular with a real unbusted 68000 BASIC and begins to eat into PC sales - and Mack uses a lot less TTL per computer!
The (LS)TTL makers switch to (ALS)TTL this fall and get lots more parts per wafer because the ALS parts are physically smaller than LS parts. (This one is likely to happen.)
Japan, Inc. expands their capacity and resumes shipping to the U.S. (which at the moment they aren't, TTL-wise). That would be good news for Digital Acoustics, if not for Motorola and T.I.
Some folks are beginning to assert that the yearly doubling of the personal computer market is going to taper off soon. For the record, we think that the three-piece computer market will taper off before the personal computer market - there are a lot fewer businessmen than there are situations into which a personal computer can fit. Further, there will be considerable upgrading in the home market as folks learn enough about their electronic doorstops to know that they can do a lot more with a real computer.
We think the doubling-per-year will continue for two years, at least, after the personal business computer market flattens. There are some signals, faint ones as yet, emanating from the IBM PC marketplace which suggest that market is approaching saturation. It is going to be REAL interesting this fall to see if Big Blue is on target to triple its sales in 1984 over 1983 - and it was an IBM honcho, you will recall, who made that (very uncharacteristic) assertion before a bunch of financial analysts. Remember: THIS FALL!
Having disposed of personal-computer UNIX, it is time to look at UNIX in its natural habitat - the $20,000 and up systems, either minicomputers or super-micros. You will recall the predictions in 1982 that vast numbers of UNIX boxes were due to be sold in 1983. We called those predictions ridiculous at the time, leaving unspoken the thought that they were also irresponsible. Well, it is now 1984 and we can look back and compare those predictions with the facts.
But we don't have to; three other folks have suddenly done so for us! Brian Boyle in the new publication UNIX/WORLD (not affiliated with either InfoWorld or ComputerWorld) points out that one forecaster published, in a respectable journal, a prediction that 600,000 (six hundred thousand) UNIX boxes would be sold in 1983! (We missed that one but caught the prediction of $3 billion in XENIX licenses.) In fact, Boyle suggests, about 25,000 UNIX boxes were sold in 1983. The UNIX market is a good, solid "B" market with "A+" expectations, he continues. But EVERYBODY; the companies, investors, distributors, all had geared their plans for "A+" results. When what they got was "B", company ledgers bled red ink all over, executives got fired (he specifically mentions Gary Friedman of Fortune), and stockholders filed suit over 'misleading forecasts' (Wicat, Fortune and others),
Geoff Lewis in, of all places, Mini-Micro Systems magazine (Mar '84 p.101) has a six-page article entitled, "Whatever happened to the UNIX supermicro boom?" According to Geoff's figures, there were 23,000 micros and 5,500 minicomputers running - not sales, running - in 1983, That suggests that even Brian Boyle's figures might be slightly optimistic.
In Electronic Engineering Times, 12 Mar '84 p.80 there is an article on Supermicros by Omri Serlin. Serlin devotes 40 column inches to debunking UNIX, saying pretty much the same things we have been saying here all along. This makes three articles which have suddenly appeared in the mainstream (non-personal) computer journals which assert that UNIX ain't wearing any clothes.
We are so polite that we are not even going to tell you that we told you so, beginning with issue #10 (pp2,3) in June of 1982 and frequently enough since then that we have bored some (all?) of you. But if you had paid attention you would not have bought Fortune Systems when it opened at 22 last spring and be holding it today at 5 7/8. The reason Fortune opened at 22 was those "A+" expectations.
It's a little late, but if UNIX actually sold 18,000 to 25,000 [high-end] boxes in 1983, how many are going to be sold in 1984? Twice that number seems a reasonable or perhaps even mildly optimistic guess. Now, count all the folks making UNIX boxes (count FAST, more are appearing daily). Divide the number of boxes that are going to be sold by the number of vendors and tell us how many of those vendors are going to be profitable. [This is no fun at all. What's fun, as Brian Boyle suggests, is for a wild-eyed foaming-at-the-mouth UNIX booster to jump up on the soapbox and boldly predict that "one million - no, make that TWO MILLION UNIX BOXES ARE GOING TO BE SOLD IN 1984!]
Since the depressing facts about the actual sales of UNIX are becoming apparent and since it does not seem possible to sweep those facts under the rug, the wild-eyed UNIX boosters are looking to IBM as a savior. After all, IBM is omnipotent [right?] and has laid plans to sit everybody in those corporations at one of the terminals of a new multi-user system (Popcorn, iAPX 286-based) running IBM's variant of UNIX and plugged into an IBM mainframe?
Sigh. We have news: IBM is NOT omnipotent, as anybody who has observed that company for more than the last five minutes can attest. It makes fairly frequent small mistakes and even (rare) massive blunders of
company-threatening magnitude. If you do not believe us, look up what happened to the line of mainframes which IBM introduced just after the 360 series and just before the 370 series. Hint: by comparison, LISA I was wildly successful!
If IBM really thinks it is going to chain comptrollers and sales managers etc. to a UNIX terminal then that will be one of their frequent small mistakes. For starters, a lot of those folks have already had a taste of freedom with a genuine personal computer with genuine instant response and the freedom to select ones' own software. Then there is that huge manual - but isn't this where we came in?
Our Belgium DTACK enthusiast is almost ready to start selling his DTACK/Ohio Scientific interface, including both hardware and software. Besides the hardware, you get: 1) The (Microsoft compatible) floating point code for the 68000, 2) The Utilities code modified for the OSI, 3) A BASIC program to load the 68000 code and utilities into the proper memory location based on memory size, 4) The Hand Assembler's Helper modified for the O.S.I., 5) The monitor program modified to run on the O.S.I., 6) A few demonstration programs.
The price for the interface board with software will be $148 with shipping and customs (if any) collect at the customers' end. Order from:
David M. Livesay
Ave de la Resistance No.
B-4920 Embourg, Belgium
David wishes to emphasize that his first name is David, not Eric. Right, Eric, er, David? Incidentally, we hope for you that there are more O.S.I./DTACK fans out there than we suspect. David is in Europe supporting a new military vehicle that uses - he swears this is the truth - the Intel 4040! - FNE
Two months after introducing Mackintosh, Apple obviously is not receiving significant guantities of Sony microfloppies, which means only token numbers of Macks are being shipped, and ALSO means that your chances of buying that desperately-needed second disk drive are slim and none (and slim just ducked out the door). Say, didn't we read last Dec and early Jan that Apple had learned its lesson via LISA I and that Mack would be ready to ship in quantity before being introduced? On the other hand, we read somewhere else that Apple did not get around to placing the production order with Sony until last Nov. Now, let's see: where did we read that? (Actually, in Electronic News.)
As predicted, you either love Mack or hate it. Right now there are more folks who love Mack than there are Macks for sale. But we certainly hope that switch on the bottom left of the chassis permits overriding that damn DOS which insists on booting any disk you insert, whether you want it to or not. And we hope it is possible to have a nice, simple environment like Applesoft or, ah, HALGOL.
Sometimes the worst bugs are revealed by a computer's most ardent admirers: Jeff H. reports that when one has one's attention focused on the CRT and is busily pushing the rodent, it is easy to knock over one's beer.
Not only that, but too much time spent with MackDraw (as we think it's called) might cause you to miss Bowling for Dollars and the latest episode of the supersonic helicopter on TV. You would not want to be left out of the mainstream of American life, would you? (The average American TV set is now on more than seven hours a day!) Yes, we said supersonic helicopter. No, we are not kidding. Yes, we know the rotors wouldn't stay on for more than 10 milliseconds.
Executive V.P. for marketing and sales Floyd Kvamme, who was hired by Markulla B.S. (before Scully), has now departed and his duties have been assumed, for now, by Scully. We now have an opening for another Scully man when a suitably loyal yes-man can be located. In fact, Scully has now made a clean sweep of the important positions, perhaps excepting two guys named Steve and Steve.
Floyd was originally hired in over the head of John Couch, who is now operating an Apple retail outlet in Riverside, CA (we think).
Question: when the only Steve who is in management (one is a techie and hence not a threat to Scully) resigns his Apple position to run for the U.S. Senate, will it be his decision or Scully's?
You can't have a palace revolution if all the princelings are loyal to the king.
In today's (17 Mar) L.A. Times sports section, there are lots and lots of ads for discounted IBM PCs. For instance, Intec Computer Center in Beverly Hills is advertising a 64K PC with 2 ea DS/DD disk drives, color/graphic card and Amdek monitor for $2649. As recently as last month, the PC was the only IBM model that was scarce, with ample supplies of the XT and the PCjr gathering dust on dealers' shelves. It almost
makes us regret that we predicted an excess of PCs for THIS FALL a couple of pages back.
Perhaps we should take our own advice and remember that things happen at the chip-makers' level very slowly (e.g. 68020) but very quickly in the (retail) marketplace. Do you realize the fire-fight in the home computer market lasted only 15 months from start to end? A permanent non-shortage of the IBM PC will signal the beginning of a fire-fight among the PC clones for survival.
Before we mailed the last newsletter, we received an issue of IEEE MICRO magazine with an article on the 68881. We had intended to provide a summary of that article, but the article was itself a summary! So if you want to know more about the 68881, you will have to find that issue. The article has the best summary of the IEEE (draft) floating point standard that we have yet seen - that standard is COMPLICATED! Also, it is apparently not making much progress toward adoption, because of the usual personal petty politics and backstabbing. (Your FNE is not above cutting somebody up, but the wounds will invariably be found on the FRONT of the torso, not in the back.)
Besides, we could get a lot more excited about the 68881 if we thought the mask layout had begun. Since that mask is going to be laid out by the same folks who are currently finishing up the 68020 mask, it almost certainly has NOT begun. Do not hold your breath until the 68881 arrives.
But we will tell you that the 68881 has a 32-bit data bus width to work perfectly with the 68020...
We are hearing reports that some Motorola folks are not inalterably opposed to the use of the Nat Semi math chip with the 68000. We don't know yet whether this represents an official policy change or some freelancing by a few of the line troops - so we will not be more specific.
There is a very minor bug in the current mask version of the 16081. If a double-to-single precision floating point conversion is done with a zero argument, the underflow flag will be (incorrectly) set. Also, Nat Semi has only recently begun testing whether the Negate Long instruction works correctly after a certain series of operations. If you already bought your 16081, it is possible, though not certain, that you might have a problem with Negate Long.
We have unofficially enquired as to when the 16081 might reach production status. The reply was, it is a marketing decision. In other words, that underflow flag bug isn't that big a problem. (Underflow just means that the result went to zero, after all - and the argument was already zero.) The 16081 has a heck of a lot fewer bugs than, say, the NEC 7220 which has been in production for about 18 months.
We are no longer wondering why the Supermicro guys aren't picking up on the 68000/16081 combo - you (and we) must remember that that is a multiuser/multitasking environment. Now imagine switching tasks in the middle of a floating point operation. No joy. Of course (stated brightly) us single-user single-taskers don't have to worry about context switching, hmm?
Have you noticed that the higher the performance, the more difficult it is to change tasks/contexts? Did we ever tell you that the T.I. programmable controllers can switch contexts VERY quickly? What we are trying to say is, one should REJOICE rather than complain that the 68000/16081 change contexts slowly...
Some folks have questioned our figures on C-64 shipments, asserting that SURELY Commodore is not shipping that many a month in the U.S. Of course they aren't! Commodore is an international company, remember? They're shipping 200,000 a month in the U.S. If you still have some doubts, look at their dollar volume in the most recent quarter ($425 million) and remember that represents WHOLESALE prices, not retail. That comes, if you are slow with figures, to $354.17 per C-64, which is about right at wholesale prices when peripherals are figured in.
Commodore has announced that sales have not slackened at all in the post-Christmas season, which is reasonable considering the stores were sold out at Christmas. In fact, C-64s aren't sitting on K-MART shelves for very long right now. If you want to see some nice, stable, dusty computers sitting on shelves look up the PCjr...
Without mentioning names, there is this outfit in Texas which has not only ripped off the name HALGOL for a quote newsletter unquote but which has an editor who has an EN-TIRE 8 pages - almost - in print vs. our 673 (but who's counting?). Do you know how that #%&*#!$ signed a note to us? "THE OTHER NEWSLETTER EDITOR!" Geez, talk about self-aggrandizement! We, on the other hand, are noted for our limitless modesty... Naturally, we would not wish to cause this person embarrassment by revealing his identity.
(The only worse acronym we have ever heard of was the year the Federal Underwriter's Conference was held in Wichita.) Jeff Hull has sent the latest list of DSEx offerings: "Doc. 1" is the Sensenig documentation modified for use with conventional upper/lower case (conventional Sensenig documentation, as in Disk 3, is in all capital letters, which gets to be a drag after a while).
"All disks are $3 each postpaid or duplicated onto your disks free if you send return postage: One disk 54 cents, two or three disks 71 cents, four or five disks $1.25, six disks $1.42. Document price as listed, postpaid. MAKE CHECKS PAYABLE TO THE 'DTACK Software Exchange."
DISK 1 - DAS, DTACK linked Applesoft by Pete Soule
Disk 2 - Sensenig BASIC, sample programs, intro and index. For Grande/GROUNDED boards, 32K min (60K recommended)
Disk 3 - Sensenig BASIC, main manual
Disk 4 - Pascal Stuff
Disk 5 - Assorted assembly stuff (not in official distribution, order at own risk)
Disk 6 - Sensenig BASIC for the 16081 math chip, sample programs and manual addition
Disk 7 - FigFORTH for static DTACK boards
Doc. 1 - Printout of Sensenig BASIC documentation. $6 postpaid.
DTACK Software Exchange
4326 Congressional Dr.
Corpus Christi TX 78413
PERMISSION IS HEREBY granted to anyone whomever to make unlimited copies of any part or whole of this newsletter provided a copy of THIS page, with its accompanying subscription information, is included.
THE FOLLOWING TRADEMARKS ARE ACKNOWLEDGED: Apple; II, II+, IIe, soft: ProDOS, LISA, Mackintosh?: Apple Computer Co. Anybody else need a 176th million ack, have your legal beagles send us the usual threat.
SUBSCRIPTIONS: Beginning with issue #19, subscriptions are $15 for 10 issues in the U.S. and Canada (U.S. funds), or $25 for 10 issues elsewhere. Make the check payable to DTACK GROUNDED. The address is:
1415 E. McFadden, Ste. F
SANTA ANA CA 92705
100 HOME 110 PRINT "WELCOME TO UN-PRIMITIVE HALGOL!" 120 PRINT 130 PRINT "TERRY PETERSON'S BENCHMARK IN ISSUE #25" 140 PRINT "USED 2500 LOOPS." 150 PRINT 160 "START" 170 INPUT "INPUT THE NUMBER OF LOOPS ",N 160 IF N=0 GOTO "ZERERROR" 190 IF N<0 GOTO "NEGERROR" 200 LET F=FRAC(N) 210 IF F>0 GOTO "FRACERROR" 220 LET ESTTIME=N*.00748 230 HOME 240 PRINT " ESTIMATED TIME IN SEC: ",ESTTIME 250 VTAB 9 260 HTAB 16 270 PRINT "CHARGE!" 280 FOR I=1 TO N STEP 1 290 LET B=TAN(ARCTAN(EXP(LOG(SQR(I*I)))))/I-1 300 LET Z=Z+B*B 310 NEXT I 320 LET RMSERROR=SQR(Z/N) 330 VTAB 14 340 PRINT "THE RMS ERROR IS",RMSERROR 350 END 360 REM TRAP'S FOR ERRONEOUS LOOP COUNT INPUTS 370 "ZERERROR" 380 HOME 390 VTAB 10 400 PRINT "JUST EXACTLY WHOM DO WE THINK WE ARE" 410 PRINT "KIDDING, NOW? ZERO LOOPS? REALLY??" 420 VTAB 16 430 GOTO "START" 440 "NEGERROR" 450 HOME 460 VTAB 10 470 PRINT "SINCE WHEN ARE NEGATIVE LOOP COUNTS" 480 PRINT "PERMITTED, PAL? YOU'D BETTER SHAPE" 490 PRINT "UP OP SHIP OUT!" 500 VTAB 16 510 GOTO "START" 520 "FRACERROR" 530 HOME 540 VTAB 10 550 PRINT "WHAT?? A FRACTIONAL LOOP COUNT? LOOK," 560 PRINT "ONCE MORE AND WE'LL SEND OUR TWO ATTACK" 570 PRINT "TEDDY BEARS, DTACK AND GROUNDED, TO PAY" 580 PRINT "YOU A CALL!" 590 VTAB 18 600 GOTO "START"