Build a Computer From Nand Logic To Tetris

[Rich Hart]

I just saw a great video from TED about an online course that teaches you how to build a computer from nand logic to an operating system that can play tetris.  It sounds, incredibly challenging but well worth trying, especially for anyone that is looking to learn about computers.  

Check out the video: Shimon Schocken: The self-organizing computer course

Also, the website is at this link: From NAND to TetrisBuilding a Modern Computer From First Principles

I have to say it wasn't until I was required to build a simple program in assembly code, when I finally started to wrap my head around how computers worked.  I think that starting all the way back from the logic which computers are build on would be even more insightful.

Anyway, thought you'd find it interesting.

Rich

[Sean McBeth]

Wow, that is really cool. Hmmm, considering doing this. Aaaah, but there are so many other projects!

Rich

--
To post to this group, send email to hive76-d...@googlegroups.com
To unsubscribe send email to hive76-discuss...@googlegroups.com
For more awesome goto http://groups.google.com/group/hive76-discussion?hl=en

[Rich Hart]

I know.  This might have to take a back seat until the summer when I'm a little more free.  But I REALLY want to do this. 

[Sean McBeth]

Actually, I suspect there is a lot of tedious work to this. Maybe it would be something to do as a class/group project at the space.

[Rich Hart]

Agreed.  I think that having people at Hive to lend a hand with it would make it a much more manageable project. 

[andrew sooy]

i would be totally game for this as a group project

[Dave]

I agree with the video, I think self-directed learning is essential (at least to me). I think they should get people into assembler earlier than they usually do, it really makes it more clear what's going on under the hood so you know what you're really asking your computer to do when you get to the higher level languages.

When you guys talk about doing this as a group project, are you referring to basically a study group to go through the course? It sounded to me like people were actually talking about building a physical computer from discrete ICs ('thousands' of them, to quote the video), whereas the course seems from my quick look to be more about doing it all in simulation the way you would before synthesizing the hardware for an FPGA or something. Both are very cool, I was just curious what people meant here.

[pezman]

I can't think of anything crazier than trying to build a computer from discrete chips.  Actually, I lie -- gathering sand, purifying it, making silicon dioxide ingots, doping the ingots to make chips and !then! making a home-made computer from the home-made chips would be a tad crazier.

Making a core from an FPGA, though, seems pretty cool.   I would be all over fpgas,. except that last time I checked, they were a little pricey compared to MCUs

I think that building and/or learning each of the pieces that comprise a CPU is instructive -- i.e. build logic gates from discrete components, build adders and multipliers from discrete logic, learning how microcoding works etc. etc. (all of which I have done) --  but doing all of that in an effort to build a working machine that will kind of suck by any objective measure seems like a complete waste to me -- you will spend far more time building (and fixing) the machine than learning about concepts. 

But that's just my $0.02 and I generally like exploits and re-purposing much more than outright building stuff anyway.

[Louis Gerbarg]

On Thu, Oct 4, 2012 at 11:22 PM, pezman <mikeh...@gmail.com> wrote:
> I can't think of anything crazier than trying to build a computer from
> discrete chips. Actually, I lie -- gathering sand, purifying it, making
> silicon dioxide ingots, doping the ingots to make chips and !then! making a
> home-made computer from the home-made chips would be a tad crazier.

When I was 16 I designed and built a RISC processor from TTL parts
(7400 series SSI). I wonder if I still have the designs around. Back
then then introductory text books didn't spend any time discussing
VHDL or Verilog except to say they existed, and spent some time
discussing the characteristics of CMOS, BJT, and ECL(!) transistors.

In my opinion the most interesting part of the processor, especially
if you want to understand performance of modern systems, tends to be
the pipeline and the rename interface. Once you implement a scoreboard
or rename registers you'll either go "cool" or "wtf!?!" Probably both
;-)

Louis

[Dave]

Louis: that's awesome! Sounds like the legendary Homebrew Computer Club. You've got more patience than me.

Pez: I completely agree. I just wanted to understand what people are saying they want to do as a group project. If they just want to do a study group, not unlike the one for the AI class that was posted about a while ago, that's cool. If they want something more out of it, whether it is a physical working computer from discrete ICs, or just a soft core on an FPGA, it seems like this course may be the long way to go about that. There are tons of open source soft cores out there now, and does anybody even know what HDL this course is using? They keep referring to it as "the" or "a" Hardware Description Language, but never seem to name it. Can it be directly synthesized to hardware with existing tools the way Verilog or VHDL can? They link to some guy's project in China, an "implementation" of the course work on an FPGA, which to me suggests he ported the code to a real HDL first (otherwise what would be notable about his project?). They don't link to his code though.

If you ask me, if you're going to get into learning HDLs, CPUs are about the most boring thing you can do with them. You're taking a system capable of massive, true parallelism, and configuring it do one thing at a time that can be done more cheaply with existing hardware. Where these things really shine is when you just have gobs and gobs of data to process, or you need a lot of things happening at exactly the same time. We think nothing of implementing fairly complex state machines processing pipelines on our favorite microcontrollers (PICs, AVRs, the venerable MSP430 [hallowed be its name]) or even DSPs, but for the right type of problem you can completely blow their capabilities out of the water with even a cheap FPGA.

I think doing something with FPGAs at the space would be really cool. Wish I had more time. I'm hoping to do my next phased array project on an inexpensive FPGA. Dirt-cheap sidescan sonar would be cool as hell to make from scratch.

-Dave

[andrew sooy]

My cousin gave me his class work and all slides for FPGA development i have not had a chance to go through them yet. But i did manage to borrow an altera dev board cyclone 2 and a production board with a FPGA and a DSP on it if we wanted to use those 2 for like a group class on FPGA'S the class my cousin gave me does go over making you own assembler language i think. we could use those to make the soft core computer. or i don't even mind making it out of nands i think doing it soft core then building it and porting it over would be a much more involved and rewarding learning experience.

--

[pezman]

Well, I'd dearly love to fiddle with FPGAs -- my only reservation is price.  However, I did look at Xylinx parts on Avnet, and some of the smaller chips seemed pretty darn reasonable.

I tend to think of small MCUs as "virtual hardware" anyway, so in many cases for a purely digital design, FPGAs are probably the better "virtual hardware" mousetrap.

For mixed signal designs, I presume that MCUs are probably the way to go.  For what it's worth, the "MS" in "MSP" stands for mixed signal.

[Rich Hart]

Actually, getting back to Sean's suggestion.  I think this would make an excellent class.  Maybe we could offer it once a week at some point.  It'd definitely volunteer to manage it.

[andrew sooy]

to teach it we would have to learn it first but if we did it as "class"/study group and work together on it that would work maybe 1-2 times a month to start then go into it more to work around schedules and other projects.

--

[Rich Hart]

Sorry, that's what I meant.  

[Jack Zylkin]

When i was a kid I made a robot out of a coffee can, some tin foil, and some pipe cleaners.  Just sayin.

--
Jack Zylkin
usbtypewriter.com
Philadelphia, PA

[andrew sooy]

http://www.nand2tetris.org/

http://www1.idc.ac.il/tecs/course/talk.pdf

here is some of the coarse work to take a look at there is a $15 kindle book on this.

[andrew11235]

It looks that it is all done on the computer virtually with their software.

http://hackaday.com/2012/10/11/programming-tetris-by-first-building-a-logic-gate-then-a-computer-then/

http://www.nand2tetris.org/course.php

http://www.nand2tetris.org/software.php

But once the concepts are down we could do the same thing with FPGA's and verilog because they just use a HDL (Hardware Description Language) and go through this course learn the basics as a study group and then learn verilog and FPGA's, play with those for a bit and then we can use this and the stuff we do with the FPGA's and create our own class for it from nand to tetris to FPGA and beyond.

On Thursday, October 4, 2012 3:55:23 PM UTC-4, Rich Hart wrote:

[Dave]

That is sooooo much work, what you just described.

[andrew11235]

Never said it would be over night.

I would tho love to/will do this and want to learn FPGA's.

On Thursday, October 4, 2012 3:55:23 PM UTC-4, Rich Hart wrote: