flat assembler
Message board for the users of flat assembler.

Index > Heap > Building a mini-CPU (just wanted to let you know of Logisim)

Goto page Previous  1, 2, 3, 4  Next
Author
Thread Post new topic Reply to topic
Madis731



Joined: 25 Sep 2003
Posts: 2141
Location: Estonia
Madis731
/me needs to do more research. I guess Itanium ~= ARM to me because they're both more distant to me than x86 Smile
...and they're both fixed instruction word length?! Wink
Post 28 Dec 2009, 17:50
View user's profile Send private message Visit poster's website Yahoo Messenger MSN Messenger Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 17271
Location: In your JS exploiting you and your system
revolution
X86 is CISC architecture.
ARM is RISC architecture.
Itanium is EPIC/VLIW architecture.

Madis731, your proposed architecture, TTA, is much closer to Itanium than to x86 or ARM.

I'm not saying that is good or bad. Just that from the little that I have seen it looks a lot like Itanium.
Post 29 Dec 2009, 00:22
View user's profile Send private message Visit poster's website Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2466
Location: Bucharest, Romania
Borsuc
CISC is best for a desktop computer normal apps -- preserves size and performance not that critical. Itanium-like architecture may be very well suited to a specialized tasks, maybe in Larabee? It would go very well with vectorization and such.

But for the general-purpose instructions, nah, it's horrible. Razz

_________________
Previously known as The_Grey_Beast
Post 29 Dec 2009, 01:14
View user's profile Send private message Reply with quote
bitRAKE



Joined: 21 Jul 2003
Posts: 2913
Location: [RSP+8*5]
bitRAKE
Large instruction words seem too fragile - without sufficient redundancy, no optimalization is capable of efficiently using the processor resources - any speed gain achieved is lost. Or rather, this has been the problem with previous attempts. Look at the decode stage of modern x86 cores - there has been no need to expand them in several generations! It hasn't been possible to sufficiently optimize instructions to the point where decoding is a bottleneck. Instead, much effort has been invested in optimizing decoded instructions. Almost any encoding could be presented to the decoder - nothing is gained from large instruction words.
Post 29 Dec 2009, 19:38
View user's profile Send private message Visit poster's website Reply with quote
Madis731



Joined: 25 Sep 2003
Posts: 2141
Location: Estonia
Madis731
Yet I am surprised that I can discover more programs that Google didn't suggest. Here's one that does everything free of charge and is only ~1MB.
http://ghdl.free.fr/
The capabilities? I'm still trying it out, but it has a manual with "Hello World" and I applaud to this learning curve. http://ghdl.free.fr/ghdl/The-hello-word-program.html#The-hello-word-program
LogisSim is for kids! Smile
Post 08 Jan 2010, 12:16
View user's profile Send private message Visit poster's website Yahoo Messenger MSN Messenger Reply with quote
LocoDelAssembly
Your code has a bug


Joined: 06 May 2005
Posts: 4633
Location: Argentina
LocoDelAssembly
Madis731, if you are going to learn a HDL, don't forget to also give a look at Verilog before entering to deeply in VHLD.

A nice site with has all the examples in Verilog: http://www.fpga4fun.com/
Post 08 Jan 2010, 16:17
View user's profile Send private message Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2466
Location: Bucharest, Romania
Borsuc
hmm all these are scripting languages, anyone know of a "graphical" or "schematical" circuit simulator, but obviously complex enough to have features like modules/groups and other hierarchical stuff (not just to make very simple simulations)?

Obviously I prefer it being free and not bloated.
Post 08 Jan 2010, 17:11
View user's profile Send private message Reply with quote
rocketsoft



Joined: 26 Jan 2010
Posts: 189
rocketsoft
I dont believe in parralel cpu's... i believe in single pipeline single clock cpu's
the x586 cpu i am designing has a 10 times faster clock than the pentium
and is source code compatible
It does everything in one clock but streches the clock if needed
its a variable-size clock, its not an asynchroneous cpu
2 architectural innovations were used to make it 10 times faster:
-The digital signal amplifier tree
-The delta add/sub ripple carry block
(a normal ripple carry unit with delta detectors on it!) Smile
Post 27 Jan 2010, 00:56
View user's profile Send private message Visit poster's website Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 17271
Location: In your JS exploiting you and your system
revolution
rocketsoft: I have to take you to task over your claims. Vague claims like yours are seen all the time in the Internet so some proof would be more convincing than a few sentences in a forum.

The Pentium came in various speeds, from 60MHz to 300MHz. Although I think the memory interface was kinda slow ~66MHz max. So your CPU at "10 times faster" (faster than what btw?) would not be difficult to do with modern silicon today. But have you ever actually realised your chip in a real hardware system? Or is this all just theory? How is your metric of 10 times obtained? Simple clock speed comparison? Or some other objective measure?
Post 27 Jan 2010, 01:58
View user's profile Send private message Visit poster's website Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2466
Location: Bucharest, Romania
Borsuc
rocketsoft wrote:
I dont believe in parralel cpu's... i believe in single pipeline single clock cpu's
Well I think you got it the wrong way from the start. I'm not saying your CPU, if true, is bad -- actually if it's faster than a Pentium it means it's better than what Intel did back then.

However, how are you going to make it competitive with today's CPUs?

The simplest answer is, of course, to stack multiple cores of your CPU, say 128 cores... then you'll have a monster.

But you see, parallel is needed.

_________________
Previously known as The_Grey_Beast
Post 27 Jan 2010, 17:06
View user's profile Send private message Reply with quote
rocketsoft



Joined: 26 Jan 2010
Posts: 189
rocketsoft
Offcource i believe in MULTIPLE cpu's but not 'parallel as in superscallar'
My x586 is gonna have a 30GHz clock -on average- using 90nm fabbing
it has about 250,000 mosfets in its integer core (like the 486)
It has only 3 pipeline stages
U should check out my RASM package if u dont believe me
im not gonna explain it all here!
those 2 discoveries were the mayor reasons why i wrote a silicon compiler! Smile
Post 27 Jan 2010, 17:53
View user's profile Send private message Visit poster's website Reply with quote
LocoDelAssembly
Your code has a bug


Joined: 06 May 2005
Posts: 4633
Location: Argentina
LocoDelAssembly
Quote:

My x586 is gonna have a 30GHz clock -on average- using 90nm fabbing
Do you mean with silicon technology that exists today? Any particular reason for choosing 90nm instead of 60nm or 45nm?
Post 28 Jan 2010, 02:37
View user's profile Send private message Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 17271
Location: In your JS exploiting you and your system
revolution
LocoDelAssembly: I think that 90nm is a current tech. The 45nm and 65nm are the very high end that only the bigger companies can afford at present.

rocketsoft: I am very sceptical about the claim of 30GHz Confused Do I have to ask again if this has actually been realised in real silicon? What sort of heat will this thing generate at such a high speed?
Post 28 Jan 2010, 04:51
View user's profile Send private message Visit poster's website Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2466
Location: Bucharest, Romania
Borsuc
While I'm also skeptical of such amazing results, I think that he mentioned his CPU is not "100% synchronous" so 30Ghz is more or less an 'average' figure, not the "real" clock. (because it doesn't even have one central clock)

_________________
Previously known as The_Grey_Beast
Post 28 Jan 2010, 17:24
View user's profile Send private message Reply with quote
rocketsoft



Joined: 26 Jan 2010
Posts: 189
rocketsoft
It is indeed average
Post 28 Jan 2010, 19:47
View user's profile Send private message Visit poster's website Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 17271
Location: In your JS exploiting you and your system
revolution
rocketsoft: If you don't show some proof then what you are saying is not true. If you can show proof then you will have my full respect. Which will it be? Evil or Very Mad
Post 28 Jan 2010, 19:53
View user's profile Send private message Visit poster's website Reply with quote
LocoDelAssembly
Your code has a bug


Joined: 06 May 2005
Posts: 4633
Location: Argentina
LocoDelAssembly
I'm out of luck finding this: What is the maximum transistor switching speed of good quality 90 nm chips? I mean, not theoretical "good", so some capacitance is expected (besides other delaying factors that I don't know)

I found transistors that apparently are able to switch at hundreds of GHz, but those are not MOSFETs (and is the Wikipedia so I don't know if the information is correct).
Post 28 Jan 2010, 20:29
View user's profile Send private message Reply with quote
rocketsoft



Joined: 26 Jan 2010
Posts: 189
rocketsoft
Mosfets at 90nm switch at over a 1000Ghz... but u cant find that kinda info on the internet
Post 28 Jan 2010, 21:06
View user's profile Send private message Visit poster's website Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 17271
Location: In your JS exploiting you and your system
revolution
rocketsoft wrote:
Mosfets at 90nm switch at over a 1000Ghz... but u cant find that kinda info on the internet
How do you reconcile that with the following quote?

The data is from 2005 which is around the time of 90nm.
Feng and Hafez developed a transistor less than half a millionth of a metre long, with a maximum operating speed of 604 GHz, meaning it can carry out 604 billion operations every second.

"This establishes a new benchmark for transistor performance," says Doug Barlage at North Carolina State University, US. "It is probably three times faster than the fastest silicon-based device."
So that is 3 times faster than 90nm silicon, i.e. silicon 90nm is ~200GHz maximum under ideal conditions. And, of course, since a single transistor does not make a CPU then once we add it into a full circuit with gates and things, plus heat considerations and timing guard bands for reliable operations etc. etc., then we get a ~3GHz CPU called a P4.

And still no proof from you for 30GHz? Hmm ...
Post 29 Jan 2010, 05:19
View user's profile Send private message Visit poster's website Reply with quote
rocketsoft



Joined: 26 Jan 2010
Posts: 189
rocketsoft
cuote "half a milliionth of a meter"... thats 500nm those were the
days of 200Mhz pentiums... if u scale it to 90nm it would be
2400GHz transistors!... also it means that a modern pentiums
clock runs at 1/1000 the transistor speed, which is exactly what i estimated... my design runs at 1/100 transistor speed! Smile
If INTEL or AMD made the same discoveries as i did and implemented them
then i would have expected a speed jump somewhere in cpu history of a factor 10... but that never hapened...
If u want proof... read my 'amdcall.txt' file... it describes the signal amplifier in detail


Last edited by rocketsoft on 29 Jan 2010, 08:30; edited 2 times in total
Post 29 Jan 2010, 08:19
View user's profile Send private message Visit poster's website Reply with quote
Display posts from previous:
Post new topic Reply to topic

Jump to:  
Goto page Previous  1, 2, 3, 4  Next

< Last Thread | Next Thread >
Forum Rules:
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 vote in polls in this forum
You can attach files in this forum
You can download files in this forum


Copyright © 1999-2020, Tomasz Grysztar.

Powered by rwasa.