View unanswered posts | View active topics It is currently Thu Mar 28, 2024 7:30 pm



Reply to topic  [ 49 posts ]  Go to page Previous  1, 2, 3, 4  Next
 Introducing the 65m32 

65m32: Stupid or neat?
Stupid 0%  0%  [ 0 ]
Neat 86%  86%  [ 6 ]
Undecided 0%  0%  [ 0 ]
65m32? 14%  14%  [ 1 ]
Total votes : 7

 Introducing the 65m32 
Author Message

Joined: Sat Feb 02, 2013 9:40 am
Posts: 2095
Location: Canada
Do you build from the bottom up or top down ?

Vivado has a simulator built into the IDE so it’s possible to simulate the HDL code destined for synthesis directly. One doesn’t have to write one’s own cycle accurate simulator when the HDL code can be simulated directly. One can do things like dump the type of the bus cycle to the screen as literal text. I haven’t written a cycle accurate simulation yet, haven’t needed one. But I’m small time. I’ve been able to find all sorts of bugs using the toolset simulator. I have however written a couple of software ISA emulators. They are a very useful tool. It gets to be too cumbersome to use the simulator for debugging when millions of cycles are happening.

_________________
Robert Finch http://www.finitron.ca


Tue Jan 10, 2017 6:18 am
Profile WWW

Joined: Tue Dec 31, 2013 2:01 am
Posts: 116
Location: Sacramento, CA, United States
robfinch wrote:
Do you build from the bottom up or top down ?

Are you asking me, Rob? I hesitate to call what I'm doing "building". I have a goal but no solid plan, so I'm just following my instincts to see how far I can get before I run out of time, energy or sanity, and have to give up or ask for professional help.

Mike B.


Tue Jan 10, 2017 6:31 am
Profile

Joined: Wed Jan 09, 2013 6:54 pm
Posts: 1780
robfinch wrote:
Do you build from the bottom up or top down ?

In industry, in my experience, definitely from the top down. That is, a CPU must be architected before it can be designed before it can be implemented.

That said, there have always been small specialist teams who build useful low level things: logic cell libraries, pad libraries, memories, PLLs. There will be low level expertise applied to the design of adders, multipliers and so on.

I do see that for some combination of sufficiently simple design and sufficiently clever designer it's possible to start with an idea and proceed to HDL, which can then be simulated as a way to explore the validity of the idea. I don't think the approach scales up well to complex designs and teams of people - it might be possible, but it has hazards.

The following is only tangentially relevant, but probably worth writing about somewhere...

As an anecdote, the T9000 project was a disaster. It was broadly architected, and then the design was handled by half a dozen teams of half a dozen people each. All the teams proceeded bottom up, and were not especially experienced. The management was inexperienced too. Each of the component parts had a unique design style and many cell libraries were used. Communication between the parts was one failure point. Timing closure within some parts was near impossible - scratch that, timing analysis was near impossible - and timing between parts was another failure point. We went through at least 8 revisions. The initial vision might have been 50MHz, the target was reduced to 30MHz and I think more, and when the thing eventually almost worked it was 10MHz or less, and several years late. It was much larger than it needed to be for the price point, and at various points was too large for the package and too large for the photoreduction optics.

Now, that project's failure is not necessarily applicable to all bottom-up design efforts! Many different things went wrong, and the tooling at the time was primitive and the computers (four microVaxes, a hundred or so transputers) rather limited too. Inexperience played a big part, and a lack of analysis and corrective action didn't help.


Tue Jan 10, 2017 8:37 am
Profile

Joined: Tue Dec 31, 2013 2:01 am
Posts: 116
Location: Sacramento, CA, United States
barrym95838 wrote:
robfinch wrote:
Do you build from the bottom up or top down ?

Are you asking me, Rob? I hesitate to call what I'm doing "building". I have a goal but no solid plan, so I'm just following my instincts to see how far I can get before I run out of time, energy or sanity, and have to give up or ask for professional help.

Mike B.

My progress this year has been very spotty, due to a thousand distractions, but I have reached an important decision point. After coding many examples and analyzing the results, I have arrived at an inexorable conclusion: 32-bit words aren't quite wide enough for what I'm trying to do. I think I could make it work, but it just would feel a bit too constrained, and I wouldn't be 100% satisfied with the eventual results. Starting now, I'm going to put 65m32 development on hiatus and recycle almost everything I have done so far (many hundreds of hours of thinking and typing, spread over many years) into development of the 65m36. 36-bit word-addressed machines aren't exactly "main-stream", but neither am I, so I am cautiously optimistic that it could be a match made in heaven. Whether or not anyone else in the universe will be the least bit interested is open for discussion.

Mike B.


Tue Nov 21, 2017 8:28 am
Profile

Joined: Wed Jan 09, 2013 6:54 pm
Posts: 1780
It'll be interesting to see what happens with that - it's a word width with a history, of course:
Quote:
Computers with 36-bit words included the MIT Lincoln Laboratory TX-2, the IBM 701/704/709/7090/7094, the UNIVAC 1103/1103A/1105, the UNIVAC 1100/2200, the General Electric GE-600/Honeywell 6000, the Digital Equipment Corporation PDP-6/PDP-10 (as used in the DECsystem-10/DECSYSTEM-20), and the Symbolics 3600 series.
Smaller machines like the PDP-1/PDP-9/PDP-15 used 18-bit words, so a double word was 36 bits.


Tue Nov 21, 2017 9:05 am
Profile

Joined: Sat Feb 02, 2013 9:40 am
Posts: 2095
Location: Canada
Itanium has 41 bit instructions packed into 128 bits. I think it's not a bad idea to expand on the number of bits. 32 bit instructions are definitely cramped. I've worked on a couple of machine with 40 bit instructions as a result. 36 bits would probably work well with 9 bit bytes, putting a use to the parity bit available with some rams. 48 bit machines with 12 bit bytes also have some appeal. 10 bit byte machines could make use of six extra bits in a 16 bit word in order to do single bit error correction on the byte (SBDD).

_________________
Robert Finch http://www.finitron.ca


Wed Nov 22, 2017 1:38 pm
Profile WWW

Joined: Tue Dec 31, 2013 2:01 am
Posts: 116
Location: Sacramento, CA, United States
Yikes! It's 2020, and I'm still floundering amid a huge dose of Real Life™. I have been prompted by Dieter to attempt to provide weekly updates, so here's the first (and hopefully not last) one for the new decade.

The 65m36 is the 36-bit version, and is going to be a direct upgrade to the 65m32a, which is still a 32-bit accumulator-based design like the 65m32, but with some instruction word modifications.

    The 65m32a instruction word is now 8:4:4:2:14 -- operation:conditional:operand register:address mode:inherent numeric.
    The 65m36 instruction word is 8:4:4:2:18 -- more symmetrical, but otherwise similar.
    There are now 256 operations and 16 registers -- a, b, c, d, e, f, i, j, k, s, t, u, w, x, y, z.
    It is still "one-address", more or less (four bits of the 8-bit operation code have a tendency to specify a register as well).
    There is still the "magic" inherent numeric value of %1000...000 to trigger a full-width numeric load from the following instruction word.
    Mnemonics are all still three letters, but that space is getting a bit crowded ... I still haven't stumbled into any "bad" three-letter words, at least not in English.
    Several of the registers are special-purpose (dependent on instruction), and several are also dual-purpose (dependent on operating mode).

I can't spill everything else right now, but I just wanted get a public record of a tiny snapshot of progress, in the hope of continuing from here with small weekly updates. I fear that this may be my only realistic chance to progress ... baby steps.

Mike B.


Fri Feb 07, 2020 4:38 am
Profile

Joined: Wed Jan 09, 2013 6:54 pm
Posts: 1780
Thanks for the update! If weekly is too challenging, monthly should be good. It gives you a chance to realise you need to do something in order to have something to report!


Fri Feb 07, 2020 9:18 am
Profile
User avatar

Joined: Tue Jan 15, 2013 5:43 am
Posts: 189
Go, Mike, go! :)

_________________
http://LaughtonElectronics.com


Fri Feb 07, 2020 1:29 pm
Profile WWW

Joined: Mon Oct 07, 2019 2:41 am
Posts: 585
I vote for the 65m40. 10 bit byte. I have not looked the 65xxx at all by the 650x cpu is 8 bit data only. 2 more bits will buy you byte,short,unsigned short and long. 3 byte branches would give you ample code space. (19 bits). The stack/pc would be extended to 24 bits. +-9 bits give a good offset
for the larger data types. Ben.


Sat Feb 08, 2020 9:50 pm
Profile

Joined: Tue Dec 31, 2013 2:01 am
Posts: 116
Location: Sacramento, CA, United States
I have stated before that I think 8-bit bytes are too limiting for my taste, but it looks like we're stuck with them for now ... except for those like me who don't mind hopping the curb and driving on the sidewalk from time to time! The only nod to 8-bit bytes in my 65mxx is the "byt" instruction, which is like "bit" but only tests bits 0-7.

Mike B.


Sun Feb 09, 2020 9:00 am
Profile

Joined: Tue Dec 31, 2013 2:01 am
Posts: 116
Location: Sacramento, CA, United States
Work progresses in fits and starts ... I'm still filling in my op-code matrix, which is double the width of the original due to the eight new registers. I'm running into aesthetic issues with my three-letter mnemonics, due to the limitations of a "basic" 26-letter alphabet and my intense desire to keep a "6xxx feel" to them (which to me means that the third letter has a strong tendency to refer to one of the sixteen registers). I know, I almost certainly shouldn't get bogged down on such cosmetic trivia, but I have pledged to myself that I won't settle for anything less than my very best effort (regardless of how much time that requires), so I'll keep at it ...

Mike B.


Fri Feb 14, 2020 5:36 am
Profile

Joined: Tue Dec 31, 2013 2:01 am
Posts: 116
Location: Sacramento, CA, United States
My concentration and sleep patterns have been disrupted for the last week by awful shoulder pain, but I wanted to share a nice .png Jeff made for me last week. Thanks, Dr. J!

Attachment:
65m36 format.png
65m36 format.png [ 5.59 KiB | Viewed 5985 times ]


Mike B.


Mon Feb 24, 2020 6:23 am
Profile

Joined: Wed Jan 09, 2013 6:54 pm
Posts: 1780
A picture is worth 1k words!


Mon Feb 24, 2020 9:17 am
Profile

Joined: Tue Dec 31, 2013 2:01 am
Posts: 116
Location: Sacramento, CA, United States
BigEd wrote:
A picture is worth 1k words!

A generous estimate, Ed! Thank you.

I still haven't been able to populate my op-code matrix to my satisfaction, but I have settled on register names zyxwusqkjiabcdef (at least for now). My operand format utilizes the operand register (r), address mode (m) and numeric (n) fields like so:

Attachment:
65m36 operand.PNG
65m36 operand.PNG [ 11.24 KiB | Viewed 5848 times ]


My shoulder pain isn't interfering with my sleep anymore, so I should be able to concentrate a little better this coming week!

Mike B.


Mon Mar 02, 2020 7:14 am
Profile
Display posts from previous:  Sort by  
Reply to topic   [ 49 posts ]  Go to page Previous  1, 2, 3, 4  Next

Who is online

Users browsing this forum: No registered users and 7 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
cron
Powered by phpBB® Forum Software © phpBB Group
Designed by ST Software