Project14 | You Decide the Next Monthly Project Competition!

Another potential idea for a Project14 is a project that has custom assembly language code in it.

Or another one could be a project that has an FPGA in it.

I like the FPGA idea, but I wonder who still does Assembly? Optimizers have gotten too good these days, I only use Assembler if there just isn't a good C compiler.

I still add in small sections of pure assembly into C code when I want to ensure strict timing requirements are met and I want to know exactly what is happening. Apart from that I moved across a few years ago from assembly to C and was so glad I did as it is great to be able to use all the structures and constructs of C.

Rod

Totally agree. The only time I revert to Assembly is with a new processor that requires a start up operation that needs to finish before other devices come on line.

Jack

Rod with the compilers today you really almost don't need to use assembly at all. the biggest problem I have is sub-microsecond timing as the built-in library time() only returns things in the microsecond range.  At that point, I found that it is easier to build a TTL based oscillator that is synced to time() which in turn gets set two ways one is via the keyboard/bios when you build the pc or you can be clever and set up which I call tic, toc from Simple Network Time Protocol (SNTP) servers which get their time from major time servers (like Bolder Co.) and then you, slave them and well life is good. In fact tic, toc becomes your network defacto time servers.

For instance, if I were in an aircraft, I could use tic, toc fed from two separate GPS receivers with data outputs, the antenna would have to be about 12 feet apart though.

This all can be done in C.  

phoenixcomm  wrote:

 

Rod with the compilers today you really almost don't need to use assembly at all. .

..

This all can be done in C.  

I suppose the same argument could be made of C.  With the advancements of the modern compilers, there really is no reason to write code in scary C instead of C++.  However, this will be environmentally as well as company dependent.  I know of a particular Music Instrument company that requires its Firmware Engineers to know and code in Assembly for DSP processing.

The correct tool is the one available and the one that does the job best.

Abandoning C for C++ is a mixed bag in my opinion. The temptations of C++ to abstract the operations and operators can cause a program or function to lose its self documenting ability. This is correctable with strong commenting and well thought out and implemented documentation (something I know everyone here follows).

Not sure about the example described by phoenixcomm , The way it is written uses post decrement on the a variable. C compilers should perform correctly as written, but that means the example would run 2,3,1,4,5. The 68xxx has a pre-decrement addressing mode, which is there so stack operations are more automated (pre-decrement, post-increment).

I did a significant amount of work with 68k series processors over the years, and agree they are significantly better than 808x series. I think the reason Intel was selected for IBM PC had something to do with marketing, not "who was first". Particularly since architecturally, the 68k was very much like the 360 mainframes. "How do you sell a $$MM$$ mainframe, if VM runs on a desktop?"

Jack.

fat fingers.. sorry it should be:  V++ = --a * b++ 

1. decrement a
2. multiply a times b
3. store solution in V
4. increment both V and b
5. fins.

Also, IBM, in fact, was and built a 68000pc  but when IBM asked Motorola for a pre-allotment, Moto said no! so IBM went to Moto little cousin Intel who said hell yes.  having DOS was an accident. When I believe Digital Research CPM blew them off. The rest is history. LOL  

IBM's decision to use the Intel 8088 for their PC seems to be mainly their familiarity with the 8086 and the fact that the Motorola 68000 was still a few months away from having all support chips in production.

IBM's decision to use the Intel 8088 for their PC seems to be mainly their familiarity with the 8086 and the fact that the Motorola 68000 was still a few months away from having all support chips in production.

Douglas, I do believe you are mistaken. IBM had actually built the prototype around the 68010. But when IBM asked for an "allotment" Motorola turned them down.  They rebuilt with the 8088 the rest is history!

BZZZZZZZZ, sorry, wrong answer!!!

I asked a director of marketing at IBM back around 1985(ish), why IBM went with the Intel instead of Motorola. I pointed out, the timing of the availability was only a week or two, and the architecture of the 68000 was much better and cleaner than the 8088.  The response, because this person was a close friend was:

Officially, the release date was the overall deciding factor, and the availability of the new DOS from the newly formed Microsoft.

Un-officially, the architecture and addressing of the 68000 was very close to the 360 mainframe, and Motorola was in the works of developing a memory management unit (MMU). Because, it would then be a very simple process to port VM to the desktop, the possibility of destroying the mainframe market would possibly kill the company.

BBBBZZZZZ Dito

IBM mainframe 360/370 was never dyadic ie you can do this in one operation: A- = A+ + -B  or ie add and store but there are post and pre-increment and decrement.  While the general architecture was similar the 68000 did not have the speed that a  360/50 had. And while your marketing guy was not an engineer in Florida and they did in fact have a 68010 unit built. And about the DOS not true at all. There were two companies on the west coast that had OS one was C/PM could be ported to 68k codebase the other was DOS and NOT Micro$oft.

IBM when up to see one of them but I believe they were surfing at the time and blew the meeting. one bright soul at IBM said MS might be able to do it in Phoenix, MS lied of course and said no problem. then they bought the company with DOS. which they could not get ported, One of my mentors and dear friends went to Phoenix and got their code working.

phoenixcomm  wrote:

 

... MS lied of course and said no problem. then they bought the company with DOS. which they could not get ported, One of my mentors and dear friends went to Phoenix and got their code working.

It's not lying when you can do it. They delivered, so that settles the fact that they weren't lying.

Jan Cumps wrote:

........ It's not lying when you can do it. They delivered, so that settles the fact that they weren't lying.

At the point, they said Yes it was what we called vaporware! (Lied)  They first had to buy the dam west coast company that had DOS, then port it. which they had major problems with and thats when Paul Zilber was brought in The first release never really never worked right. CP/M was better at the time.

I could easily be misinformed. I haven't talked to any of the decision makers - just regurgitating a consensus of all the stories I've heard on the subject.

The failure of Digital Research to work out an agreement with IBM has to be one of the all time biggest missed opportunities in the industry.

dougw

Don't worry, I don't know where the misinformation about 68000 is coming from I was slumming at Garret Airresearch Special Test Dept in LA. My boss put on a special project to pick the next CPU for our custom test which at the time we using 40/40 cards, I specked out 68010 back then! It was clean silicon which intel has never done. To this date, it holds the record as the most widely embedded CPU family in the world.  The other thing that I sould talk about is that the 680x0 family used memory addressing for all I/O while IBM used Channels

In reality,  what IBM scared them was a little company in Oceanport NJ or Interdata the 7/16, 7/32 and 8/32 were minicomputers that could replace Big Blue's Boxes, and the instructions sets were almost identical. In fact, 8/32 #1 when to Murry Hill Bell Labs, for the first non-DEC port of UNIX!

du time frame about 1970 just before Airforce