http://guineh.blogspot.com/search/label/cpu%20build

I'll try to keep things current. Still designing the thing.

-- 
~Mike

Post a reply to this message

On 5/17/2010 11:56 AM, Mike Raiford wrote:
>
> http://guineh.blogspot.com/search/label/cpu%20build
>
> I'll try to keep things current. Still designing the thing.
>

Looks like fun.  I've been playing around with Verilog code lately, and 
am going to be purchasing an FPGA starter kit soon.  A basic CPU is one 
of the things that I thought would be fun to do in the FPGA, although, I 
imagine it's been done to death.


Kyle

Post a reply to this message

Mike Raiford wrote:

> I'll try to keep things current. Still designing the thing.

Oh, but that's the hardest part. Once you have the finished design, 
merely putting a few wires together is *easy* by comparison! ;-)

(Of course, _this_ is the stage where you discover that real electronic 
components don't work like in the simulator, and so your design doesn't 
work right anyway...)

I must say, this is like the coolest thing *ever*! :-D Almost makes me 
want to try again. But I don't have any decent diagramming or simulation 
tools, and I don't think my GUI skills are up to the task of creating some.

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

On 5/17/2010 1:59 PM, Orchid XP v8 wrote:
> Mike Raiford wrote:
>
>> I'll try to keep things current. Still designing the thing.
>
> Oh, but that's the hardest part. Once you have the finished design,
> merely putting a few wires together is *easy* by comparison! ;-)
>
> (Of course, _this_ is the stage where you discover that real electronic
> components don't work like in the simulator, and so your design doesn't
> work right anyway...)
>
> I must say, this is like the coolest thing *ever*! :-D Almost makes me
> want to try again. But I don't have any decent diagramming or simulation
> tools, and I don't think my GUI skills are up to the task of creating some.
>

Here's the sim I'm using:

http://sourceforge.net/projects/circuit/

It's actually a pretty decent app. written in Java, seems up to handling 
some pretty complex things. It has a few UI quirks, and has a major 
drawback in that you cannot add bi-directional pins to the circuits, 
which is a bit annoying when you're designing something that has a data 
bus that is mean to be read from and written to.

-- 
~Mike

Post a reply to this message

On 5/17/2010 1:18 PM, Kyle wrote:

> Looks like fun. I've been playing around with Verilog code lately, and
> am going to be purchasing an FPGA starter kit soon. A basic CPU is one
> of the things that I thought would be fun to do in the FPGA, although, I
> imagine it's been done to death.

Yup. but isn't that the point of an FPGA? ;)

I mean... basically to build a custom, application specific, chunk of 
logic. If it's a customized CPU (er, say a DSP...), then all the more fun.

It's tempting to get something like an FPGA test board and poke 
something into it, though.... but I feel this will take some of the "I 
built it myself" satisfaction from the project.

.... I still have to figure out how to pare this sucker down so I can do 
the big clacking pile of relays version of it ;)

-- 
~Mike

Post a reply to this message

>> Looks like fun. I've been playing around with Verilog code lately, and
>> am going to be purchasing an FPGA starter kit soon. A basic CPU is one
>> of the things that I thought would be fun to do in the FPGA, although, I
>> imagine it's been done to death.
> 
> Yup. but isn't that the point of an FPGA? ;)
> 
> I mean... basically to build a custom, application specific, chunk of 
> logic. If it's a customized CPU (er, say a DSP...), then all the more fun.
> 
> It's tempting to get something like an FPGA test board and poke 
> something into it, though.... but I feel this will take some of the "I 
> built it myself" satisfaction from the project.

 From time to time, people try to build ray tracers using FPGA. And, 
occasionally, they make one that out-performs a software ray tracer 
running on a quad-core Xeon or something. So don't go thinking that just 
because it's FPGA it can't be fast; special-purpose machines can and 
sometimes do out-perform general-purpose machines. ;-)

> .... I still have to figure out how to pare this sucker down so I can do 
> the big clacking pile of relays version of it ;)

Uh, yeah. Good luck with that.

You know that the first computers filled entire buildings, right? ;-)

Post a reply to this message

>> I must say, this is like the coolest thing *ever*! :-D Almost makes me
>> want to try again. But I don't have any decent diagramming or simulation
>> tools, and I don't think my GUI skills are up to the task of creating 
>> some.
>>
> 
> Here's the sim I'm using:
> 
> http://sourceforge.net/projects/circuit/
> 
> It's actually a pretty decent app. written in Java, seems up to handling 
> some pretty complex things. It has a few UI quirks, and has a major 
> drawback in that you cannot add bi-directional pins to the circuits, 
> which is a bit annoying when you're designing something that has a data 
> bus that is mean to be read from and written to.

Well, so far I've tried two systems:

1. KLogic. Unfortunately, this crashes constantly, and sometimes gives 
you THE WRONG ANSWER. As in, I've drawn circuits which are simple enough 
that I damned-well *know* the answer being displayed is incorrect.

2. The freebie version of the Xilinx simulation/synthesis tool. Here you 
have to type in VHDL as input, and (IIRC), you don't get any kind of 
visual representation of the circuit. And once you've designed the 
circuit, you have to type in more VHDL to create a test harness. You 
can't just click a button that says "please attach a logic probe here", 
you have to actually code it all by hand.

I'll have a look at this and see if it works any better...

Post a reply to this message

On 5/18/2010 3:12 AM, Invisible wrote:

>> I mean... basically to build a custom, application specific, chunk of
>> logic. If it's a customized CPU (er, say a DSP...), then all the more
>> fun.
>>
>> It's tempting to get something like an FPGA test board and poke
>> something into it, though.... but I feel this will take some of the "I
>> built it myself" satisfaction from the project.
>
>  From time to time, people try to build ray tracers using FPGA. And,
> occasionally, they make one that out-performs a software ray tracer
> running on a quad-core Xeon or something. So don't go thinking that just
> because it's FPGA it can't be fast; special-purpose machines can and
> sometimes do out-perform general-purpose machines. ;-)
>

I didn't say anything about the speed. Some of them clock at hundreds of 
MHz... which is why they would excel in applications such as a 
customized DSP.

>> .... I still have to figure out how to pare this sucker down so I can
>> do the big clacking pile of relays version of it ;)
>
> Uh, yeah. Good luck with that.
>
> You know that the first computers filled entire buildings, right? ;-)

Right. Which is why it must be as simple as possible, because I don't 
need to dedicate a 20'x24' room to the thing ;) Thankfully memory is 
nice and solid-state, nowadays!, no need for ferrite cores, drums, relay 
latches or any of that nonesense.

Of course, the relay-based machine may be simplified to the point where 
it can't do anything but play tic-tac-toe...

more than space, my biggest concern is power usage. I need to be able to 
operate off of a 15-amp breaker without tripping the thing. (That's 15a 
@120v, btw... not 240v!



-- 
~Mike

Post a reply to this message

>>> It's tempting to get something like an FPGA test board and poke
>>> something into it, though.... but I feel this will take some of the "I
>>> built it myself" satisfaction from the project.
>>
>>  From time to time, people try to build ray tracers using FPGA. And,
>> occasionally, they make one that out-performs a software ray tracer
>> running on a quad-core Xeon or something. So don't go thinking that just
>> because it's FPGA it can't be fast; special-purpose machines can and
>> sometimes do out-perform general-purpose machines. ;-)
> 
> I didn't say anything about the speed. Some of them clock at hundreds of 
> MHz... which is why they would excel in applications such as a 
> customized DSP.

I was thinking more along the lines of watching a ray-tracer running in 
near-realtime, and knowing you couldn't possibly be doing that unless 
you'd built the hardware yourself. But yeah, I guess it's not the same. 
(It's also presumably not very easy...!)

>> You know that the first computers filled entire buildings, right? ;-)
> 
> Right. Which is why it must be as simple as possible, because I don't 
> need to dedicate a 20'x24' room to the thing ;)

With the reinforced flooring, and the custom air conditioning system, 
and the specialist PSU, and the team of technitions to tend to the 
beast's needs? ;-)

> Thankfully memory is 
> nice and solid-state, nowadays!, no need for ferrite cores, drums, relay 
> latches or any of that nonesense.

It strikes me that if you had a relay without a spring, you could use a 
single relay as a kind of mechanical memory.

It also strikes me that you could really easily "cheat" and hide a small 
Z80 under one of the relays and fake the whole exercise! :-D

> Of course, the relay-based machine may be simplified to the point where 
> it can't do anything but play tic-tac-toe...

No, you only need some marbles in matchboxes for that. ;-)

> more than space, my biggest concern is power usage. I need to be able to 
> operate off of a 15-amp breaker without tripping the thing. (That's 15a 
> @120v, btw... not 240v!

Hmm. Most of your current is going to come from the relays held open. 
The closed ones won't use any power. I have no idea how much power a 
relay uses, but I shouldn't have thought it's that much... Then again, 
just how many are you expecting to have? ;-)

Post a reply to this message

On 5/18/2010 8:52 AM, Invisible wrote:

>> Of course, the relay-based machine may be simplified to the point
>> where it can't do anything but play tic-tac-toe...
>
> No, you only need some marbles in matchboxes for that. ;-)
>

Haha. No. Only if you want to do something complicated like have a 
genetic algorithm to solve the tic-tac-toe board ;)

> Hmm. Most of your current is going to come from the relays held open.
> The closed ones won't use any power. I have no idea how much power a
> relay uses, but I shouldn't have thought it's that much... Then again,
> just how many are you expecting to have? ;-)

Looking at the spec for a small 5v spdt relay (3A current rating) the 
coil draws 40ma.

25 would draw 1 amp.

Using the 5v rail of a cheap PC power supply I could fit 50.

Using the 12v (16a @ 12 v) and using a regulator to obtain 5v, I can fit 
quite a few more. But, then there is also money.
-- 
~Mike

Post a reply to this message