Developing a parallel computer for under $50

XCore Project reviews, ideas, videos and proposals.
User avatar
shawn
XCore Addict
Posts: 238
Joined: Thu Dec 17, 2009 5:15 am

Post by shawn »

I do know what brian means, though. let's exploit C,C++,XC Total enviroment.
This chip we have, screams for it, the computer is the game, really!


User avatar
Berni
Respected Member
Posts: 363
Joined: Thu Dec 10, 2009 10:17 pm
Contact:

Post by Berni »

Well what would realy make it take off is getting a simple x86 emulator running that is capable of running DOS. This opens up the door to a whole mountain of software capable of running on it. You could easily use a SD card as its HDD and the external RAM chip for the virtual machines ram. DOS runs great with 4MB of ram.

Might be a better start to first try emulate a game boy or similar. I heard its not all that complicated to write a emulator for one since it uses a simple CPU with not so much hardware around it.
User avatar
shawn
XCore Addict
Posts: 238
Joined: Thu Dec 17, 2009 5:15 am

Post by shawn »

It could do BIST, SPI, Kermit, etc.
Monitor \ to legacy JIT applications
and then throw to a CSP supervisor.
In middle school among other books,
I used to carry around a 136 columb
bound printout of all the source for
the game programs listed and not,
for the PDP10 TOPS system at Colgate.
The majority of the code was written in
BASIC like Dartmouth, I think. To exec
you had to call OLD to enter interpeter.
On a good system BASIC was a blast,
they used it all over NASA during the
moon shots. B~ASIC as JIT ; )
User avatar
cereal_entrepreneur
Junior Member
Posts: 5
Joined: Fri Feb 19, 2010 9:09 am
Contact:

Post by cereal_entrepreneur »

This is nice, but I wouldn't call it a parallel computer.

The compute engine is just a thread (basically a logical processor) but the interesting thing is the I/O. These are indeed parallel computing tasks - they are running code to emulate peripherals however so effectively it does look just like any micro running running a shell with some I/O hooked up.

I'd agree with heater it's the peripherals that are the cool bit of this, until someone starts using a few of the threads for general computing!
User avatar
shawn
XCore Addict
Posts: 238
Joined: Thu Dec 17, 2009 5:15 am

Post by shawn »

Most Lisps run parallel and lack important features for concurrency.
Most BASIC seems poorley inplemented and not CSP .
Then neather is C , but C has history, forever ranking
as the most narfarious cross compilation hack noted in the jargon files.
A creation without genius other than its auther, "C". A hack so devious and
tricked out and ugly people began to belive it was real and the more real it
became the more they wanted it, that's C. It's a patent language.
My point is don't through the baby out with the basket.
put BASIC in a box and label it OLD, as Proof in CSP.
Let BASIC experts EXEC to like make and extend BASIC \ CSP. who cares
anymore what language you use, use the one you like, make your own PL/1.
Let OCCAM experts spin there CSP and Antifusers burn there preprosess.
The more this XMOS machine gets thrown at it the more, moores law applies.
XC has a real genius, a real author, with a venacular thats purely exploratory.
like an ACE w/afterburners.
User avatar
dave
Member++
Posts: 31
Joined: Thu Dec 10, 2009 10:11 pm
Contact:

Post by dave »

When I made this video, I had in mind the idea that we could make an extensible computer using the XMOS links. When you want to make it more powerful - performance, communications, real-time events, you add another processor - just plug in another board using a link. I think I said this towards the end of the video. In the extreme, you could plug in something like an XMP-64.

Of course, the other interesting thing is the software. For example, we need a shell that can launch a task on another processor. I already have a compiler that can compile itself to run in the XCore - this is the one I use for architectural exploration so its not production quality but might be useful. I'm interested in exploring use of the XMOS chips to build extensible multiprocessor computers based on light-weight energy-efficient software. Note that the L-series chips can be set up so they don't use much power when they're not doing anything - this can be exploited easily using event-driven XC-style programs.
User avatar
otitov
XCore Addict
Posts: 207
Joined: Thu Dec 10, 2009 11:00 pm
Location: Mexico
Contact:

Post by otitov »

is it impossible to achieve same energy savings on G chips as on L chips? why?
User avatar
nieuwhzn
Member++
Posts: 26
Joined: Sat Dec 12, 2009 6:45 am

Post by nieuwhzn »

dave wrote:When I made this video, I had in mind the idea that we could make an extensible computer using the XMOS links. When you want to make it more powerful - performance, communications, real-time events, you add another processor - just plug in another board using a link. I think I said this towards the end of the video. In the extreme, you could plug in something like an XMP-64.

Of course, the other interesting thing is the software. For example, we need a shell that can launch a task on another processor.
Ah, it's good to see the Master posting here too!

Now, this dynamically plugging in new processors and connecting them to the system through XLinks sounds great. However, as I understand it, this is not possible (yet), because the topology needs to be known when booting up the whole system, i.e. no dynamic topologies possible. Is there an effort going on to implement this?

As far as the software is concerned, maybe I'm a bit naive but this sounds a bit like remote procedure calls, which have been around for a long time already. Like I said before, what I would find very interesting is an OS which can 'worm' itself through a topology and establish bits and pieces of itself on the processors it encounters. As the topology changes the OS would adapt. Using XMOS cores and XLinks would guarantee that the OS encounters a homogeneous system, making the implementation much easier.
User avatar
shawn
XCore Addict
Posts: 238
Joined: Thu Dec 17, 2009 5:15 am

Post by shawn »

is it impossible to achieve same energy savings on G chips as on L chips? why?
I'll guess? 90nm, 65nm note the router exist only in the G chip.
User avatar
BrianMiller
Active Member
Posts: 33
Joined: Wed Jan 13, 2010 7:50 am
Location: USA, NC, Raleigh
Contact:

Post by BrianMiller »

nieuwhzn wrote: Like I said before, what I would find very interesting is an OS which can 'worm' itself through a topology and establish bits and pieces of itself on the processors it encounters.
yow. I come from a compsec background and that sounds frightening.
Post Reply