Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.forth > #15024 > unrolled thread

Green Arrays presentation at Dark Silicon Workshop

Back to article view | Back to comp.lang.forth

Contents

  Green Arrays presentation at Dark Silicon Workshop Paul Rubin <no.email@nospam.invalid> - 2012-08-18 17:44 -0700
    Re: Green Arrays presentation at Dark Silicon Workshop mhx@iae.nl (Marcel Hendrix) - 2012-08-19 10:04 +0200
      Re: Green Arrays presentation at Dark Silicon Workshop RR <freedomspyder@gmail.com> - 2012-08-19 07:58 -0700
        Re: Green Arrays presentation at Dark Silicon Workshop Paul Rubin <no.email@nospam.invalid> - 2012-08-19 08:27 -0700
          Re: Green Arrays presentation at Dark Silicon Workshop Richard Owlett <rowlett@pcnetinc.com> - 2012-08-20 09:05 -0500
            Re: Green Arrays presentation at Dark Silicon Workshop rickman <gnuarm@gmail.com> - 2012-08-20 15:55 -0400
              Re: Green Arrays presentation at Dark Silicon Workshop Richard Owlett <rowlett@pcnetinc.com> - 2012-08-20 15:42 -0500
                Re: Green Arrays presentation at Dark Silicon Workshop forther <forther@gmail.com> - 2012-08-20 14:17 -0700
                  Re: Green Arrays presentation at Dark Silicon Workshop Paul Rubin <no.email@nospam.invalid> - 2012-08-20 14:26 -0700
                    Re: Green Arrays presentation at Dark Silicon Workshop Jason Damisch <jasondamisch@yahoo.com> - 2012-08-20 15:32 -0700
                  Re: Green Arrays presentation at Dark Silicon Workshop rickman <gnuarm@gmail.com> - 2012-08-20 20:58 -0400
      Re: Green Arrays presentation at Dark Silicon Workshop Brad Eckert <hwfwguy@gmail.com> - 2012-09-20 11:54 -0700
      Re: Green Arrays presentation at Dark Silicon Workshop rickman <gnuarm@gmail.com> - 2012-09-21 00:22 -0400
    Re: Green Arrays presentation at Dark Silicon Workshop Jason Damisch <jasondamisch@yahoo.com> - 2012-08-19 12:17 -0700
      Re: Green Arrays presentation at Dark Silicon Workshop Andrew Haley <andrew29@littlepinkcloud.invalid> - 2012-09-01 05:04 -0500
    Re: Green Arrays presentation at Dark Silicon Workshop dambere@web.de - 2012-08-20 01:34 -0700
      Re: Green Arrays presentation at Dark Silicon Workshop rickman <gnuarm@gmail.com> - 2012-08-20 15:53 -0400
    Re: Green Arrays presentation at Dark Silicon Workshop gavino_himself <visploveslisp@gmail.com> - 2012-09-05 02:18 -0700
      Re: Green Arrays presentation at Dark Silicon Workshop John Passaniti <john.passaniti@gmail.com> - 2012-09-05 13:07 -0700
        Re: Green Arrays presentation at Dark Silicon Workshop gavino_himself <visploveslisp@gmail.com> - 2012-09-12 18:24 -0700
          Re: Green Arrays presentation at Dark Silicon Workshop Mark Wills <forthfreak@gmail.com> - 2012-09-13 08:31 -0700
          Re: Green Arrays presentation at Dark Silicon Workshop Jason Damisch <jasondamisch@yahoo.com> - 2012-09-13 11:25 -0700
            Re: Green Arrays presentation at Dark Silicon Workshop "Elizabeth D. Rather" <erather@forth.com> - 2012-09-13 14:03 -1000
              Re: Green Arrays presentation at Dark Silicon Workshop Mark Wills <forthfreak@gmail.com> - 2012-09-14 00:39 -0700

Page 1 of 2  [1] 2  Next page →

#15024 — Green Arrays presentation at Dark Silicon Workshop

From June 2012 but I just found the slides.

http://darksilicon.ucsd.edu/2012/assets/slides/5

Title: Illuminating Dark Silicon with a Fabric of Simple Computers
Author: Charles H. Moore, Greg Bailey
GreenArrays, Inc.
File size: 935.32 KB
Format: PDF 

[toc] | [next] | [standalone]

#15028

Paul Rubin <no.email@nospam.invalid> writes Re: Green Arrays presentation at Dark Silicon Workshop

> From June 2012 but I just found the slides.

> http://darksilicon.ucsd.edu/2012/assets/slides/5
[..]

They answer the question why the chip has almost no RAM.

To me, in the limit the authors' expressed philosophy leads 
to the design of e.g. a dedicated video processor, when such 
a device is needed. That will be simpler, cheaper, and 
higher performance than what the GA144 does now.

Apparently they are working on a 32-bit design. Unexpected,
given the above philosophy.

It is strange that there is no mention of the analog
hardware that is available in the GA144. 

-marcel

[toc] | [prev] | [next] | [standalone]

#15036

"10 Mb full duplex Ethernet network interface controller has been built in software using 27 nodes connecting pins with external SRAM, and organized into two pipelines for transmit and receive with two nodes between them to implement link negotiation."

Have they ever released / published their software NIC?

[toc] | [prev] | [next] | [standalone]

#15038

RR <freedomspyder@gmail.com> writes:
> "10 Mb full duplex Ethernet network interface controller has been
> built in software using 27 nodes..."
> Have they ever released / published their software NIC?

I don't think they have released it.  I don't know if they plan to.  I
do remember seeing some mention of it but didn't realize it was on 27
nodes.

It does make me a bit sad, thinking of how much programming effort must
have gone into that, only to end up with a 10 mbit NIC.  I don't even
remember how long it's been since I last saw any 10 mbit ethernet gear.
Even 100 mbit is kind of creaky these days.  Most users want 1000 mbit,
which is completely out of reach for that software approach. And 10000
mbit or even 25000 mbit is starting to make itself felt.  (100 gbit is
actually four bonded 25 gbit channels, I think).

[toc] | [prev] | [next] | [standalone]

#15053

Paul Rubin wrote:
> RR<freedomspyder@gmail.com>  writes:
>> "10 Mb full duplex Ethernet network interface controller has been
>> built in software using 27 nodes..."
>> Have they ever released / published their software NIC?
>
> I don't think they have released it.  I don't know if they plan to.  I
> do remember seeing some mention of it but didn't realize it was on 27
> nodes.
>
> It does make me a bit sad, thinking of how much programming effort must
> have gone into that, only to end up with a 10 mbit NIC.  I don't even
> remember how long it's been since I last saw any 10 mbit ethernet gear.
> Even 100 mbit is kind of creaky these days.  Most users want 1000 mbit,
> which is completely out of reach for that software approach. And 10000
> mbit or even 25000 mbit is starting to make itself felt.  (100 gbit is
> actually four bonded 25 gbit channels, I think).


For the market where the GA144 is likely to appear, is more 
than 10 mbit/s relavant?

[toc] | [prev] | [next] | [standalone]

#15055

On 8/20/2012 10:05 AM, Richard Owlett wrote:
> Paul Rubin wrote:
>> RR<freedomspyder@gmail.com> writes:
>>> "10 Mb full duplex Ethernet network interface controller has been
>>> built in software using 27 nodes..."
>>> Have they ever released / published their software NIC?
>>
>> I don't think they have released it. I don't know if they plan to. I
>> do remember seeing some mention of it but didn't realize it was on 27
>> nodes.
>>
>> It does make me a bit sad, thinking of how much programming effort must
>> have gone into that, only to end up with a 10 mbit NIC. I don't even
>> remember how long it's been since I last saw any 10 mbit ethernet gear.
>> Even 100 mbit is kind of creaky these days. Most users want 1000 mbit,
>> which is completely out of reach for that software approach. And 10000
>> mbit or even 25000 mbit is starting to make itself felt. (100 gbit is
>> actually four bonded 25 gbit channels, I think).
>
>
> For the market where the GA144 is likely to appear, is more than 10
> mbit/s relavant?
>

What market would that be?

I agree that there are many apps that will work well with 10 Mbps 
Ethernet.  But there are lots more that could use even just 100 Mbps.

Rick

[toc] | [prev] | [next] | [standalone]

#15056

rickman wrote:
> On 8/20/2012 10:05 AM, Richard Owlett wrote:
>> Paul Rubin wrote:
>>> RR<freedomspyder@gmail.com> writes:
>>>> "10 Mb full duplex Ethernet network interface controller
>>>> has been
>>>> built in software using 27 nodes..."
>>>> Have they ever released / published their software NIC?
>>>
>>> I don't think they have released it. I don't know if they
>>> plan to. I
>>> do remember seeing some mention of it but didn't realize
>>> it was on 27
>>> nodes.
>>>
>>> It does make me a bit sad, thinking of how much
>>> programming effort must
>>> have gone into that, only to end up with a 10 mbit NIC. I
>>> don't even
>>> remember how long it's been since I last saw any 10 mbit
>>> ethernet gear.
>>> Even 100 mbit is kind of creaky these days. Most users
>>> want 1000 mbit,
>>> which is completely out of reach for that software
>>> approach. And 10000
>>> mbit or even 25000 mbit is starting to make itself felt.
>>> (100 gbit is
>>> actually four bonded 25 gbit channels, I think).
>>
>>
>> For the market where the GA144 is likely to appear, is
>> more than 10
>> mbit/s relevant?
>>
>
> What market would that be?
>
> I agree that there are many apps that will work well with 10
> Mbps Ethernet. But there are lots more that could use even
> just 100 Mbps.
>
> Rick

What market? I don't know.
I was extrapolating from comments here about low cost and 
low power (watts, not MIPS).
Though not as old as Jerry, I remember when 300 Baud was 
high speed.
And for more perspective, my father operated a _legal_ 
*land* based spark gap transmitter.

Besides the MS Mantra of "bigger and faster - the only 
metric of progress" ruffles my feathers. The OWLett ;)

[toc] | [prev] | [next] | [standalone]

#15057

http://cdn.intechopen.com/pdfs/37775/InTech-Wavefront_systolic_algorithms_for_implementation_of_stereo_vision_and_obstacle_avoidance_computations_on_a_very_low_power_mimd_many_core_parallel_architecture_applications_for_mobile_systems_and_wearable_visual_guidance_.pdf

It may answers some questions, raised in this thread. 

[toc] | [prev] | [next] | [standalone]

#15058

forther <forther@gmail.com> writes:
> [very long url snipped]
> It may answers some questions, raised in this thread. 

Thanks.  I notice the much friendlier url

   http://cdn.intechopen.com/pdfs/37775/InTech.pdf

also works.  The numeric part and the .pdf extension identifies the file
and the rest is search keywords.

[toc] | [prev] | [next] | [standalone]

#15060

well the Mars rovers are in an environment which won't change, or won't change rapidly, that is to say, the Martian environment is static because it is not yet inhabited by people and their machinery. 

[toc] | [prev] | [next] | [standalone]

#15062

On 8/20/2012 5:17 PM, forther wrote:
> http://cdn.intechopen.com/pdfs/37775/InTech-Wavefront_systolic_algorithms_for_implementation_of_stereo_vision_and_obstacle_avoidance_computations_on_a_very_low_power_mimd_many_core_parallel_architecture_applications_for_mobile_systems_and_wearable_visual_guidance_.pdf
>
> It may answers some questions, raised in this thread.

I took a look and don't have nearly enough time at the moment to read 
the full paper in any depth.  Anyone have any ideas on which questions 
this paper answers and what those answers are?  In other words, can I 
get the Cliff Notes version?

Rick

[toc] | [prev] | [next] | [standalone]

#15718

On Sunday, August 19, 2012 1:04:03 AM UTC-7, Marcel Hendrix wrote:
> Apparently they are working on a 32-bit design. Unexpected, 
> given the above philosophy.
> 

Too many requests for 32-bit. Maybe they should have a mix of core types, rather than a uniform array.

[toc] | [prev] | [next] | [standalone]

#15720

On 8/19/2012 4:04 AM, Marcel Hendrix wrote:
> Paul Rubin<no.email@nospam.invalid>  writes Re: Green Arrays presentation at Dark Silicon Workshop
>
>>  From June 2012 but I just found the slides.
>
>> http://darksilicon.ucsd.edu/2012/assets/slides/5
> [..]
>
> They answer the question why the chip has almost no RAM.

Ok, I give up, why almost no RAM?  I can't find this.


> To me, in the limit the authors' expressed philosophy leads
> to the design of e.g. a dedicated video processor, when such
> a device is needed. That will be simpler, cheaper, and
> higher performance than what the GA144 does now.
>
> Apparently they are working on a 32-bit design. Unexpected,
> given the above philosophy.

After looking at some of the limitations of the instructions being 
packed into an 18 bit word, I have wondered why not a 20 bit word?  If 
the data size matches 20 bits will work better for audio apps which 
aren't typically 16 bits anymore.  18 bits is a bit tight for audio.  20 
bits gives a lot more headroom and each word provides for a FULL 4 
instructions of 5 bits each.


> It is strange that there is no mention of the analog
> hardware that is available in the GA144.

Yes, that is a bit odd.  I guess that just didn't fit into their focus 
on low engergy.

Rick

[toc] | [prev] | [next] | [standalone]

#15042

–Must every widget be a unix machine with the resources and versatility of a minicomputer?

–Do we really believe our lamps and toasters require software updates?

LOL

[toc] | [prev] | [next] | [standalone]

#15359

Jason Damisch <jasondamisch@yahoo.com> wrote:

> * Must every widget be a unix machine with the resources and
> versatility of a minicomputer?
> 
> * Do we really believe our lamps and toasters require software
> updates?
> 
> LOL

Right!  That reminds me of a request to tender many years ago.  It was
for a controller for the display boards at a London railway terminus.
I thought of tendering for the job, but then I saw running on UNIX as
a *requirement*.  Why on Earth would a display controller need UNIX?
Years later, I'm fascinated (and appalled) that the railway stations
still have different display boards and, presumably, different custom
software.  These days they're probably using a 16-node server with 128
gigabytes of memory...

Andrew.

[toc] | [prev] | [next] | [standalone]

#15046

Am Sonntag, 19. August 2012 02:44:02 UTC+2 schrieb Paul Rubin:
> From June 2012 but I just found the slides.
> 
> 
> 
> http://darksilicon.ucsd.edu/2012/assets/slides/5
> 
> 
> 
> Title: Illuminating Dark Silicon with a Fabric of Simple Computers
> 
> Author: Charles H. Moore, Greg Bailey
> 
> GreenArrays, Inc.
> 
> File size: 935.32 KB
> 
> Format: PDF

The concepts there are presenting are important in my opinion but I am curious about the market there want to target. Probably the GA144 can offer flexibility against CPLD's ?!??

[toc] | [prev] | [next] | [standalone]

#15054

On 8/20/2012 4:34 AM, dambere@web.de wrote:
> Am Sonntag, 19. August 2012 02:44:02 UTC+2 schrieb Paul Rubin:
>>  From June 2012 but I just found the slides.
>>
>>
>>
>> http://darksilicon.ucsd.edu/2012/assets/slides/5
>>
>>
>>
>> Title: Illuminating Dark Silicon with a Fabric of Simple Computers
>>
>> Author: Charles H. Moore, Greg Bailey
>>
>> GreenArrays, Inc.
>>
>> File size: 935.32 KB
>>
>> Format: PDF
>
> The concepts there are presenting are important in my opinion but I am curious about the market there want to target. Probably the GA144 can offer flexibility against CPLD's ?!??

I think the fact that their presentations create more questions than 
they answer is important.  They are pitching their technology without 
mentioning where it applies.  That is not a great way to convince 
customers.

I don't think you need to dwell on how you designed your product when 
potential customers have so many questions about how to use the product. 
  I've been banging on this for several weeks now and I still haven't 
been able to prove to myself that my application is practical on their 
device.  But I am still working on it.

Rick

[toc] | [prev] | [next] | [standalone]

#15476

On Saturday, August 18, 2012 5:44:02 PM UTC-7, Paul Rubin wrote:
> From June 2012 but I just found the slides.
> 
> 
> 
> http://darksilicon.ucsd.edu/2012/assets/slides/5
> 
> 
> 
> Title: Illuminating Dark Silicon with a Fabric of Simple Computers
> 
> Author: Charles H. Moore, Greg Bailey
> 
> GreenArrays, Inc.
> 
> File size: 935.32 KB
> 
> Format: PDF

bring out a pc done in forth and hurry up, one with working web browser and internet :) and file manager in gui as well, but can be inside browser if u want

of course everything runs in 500k ram so rest of 4096M ram can be data right?
do it already

enuf blowing smoke

[toc] | [prev] | [next] | [standalone]

#15484

On Wednesday, September 5, 2012 5:18:28 AM UTC-4, gavino_himself wrote:
> bring out a pc done in forth and hurry up, one with 
> working web browser and internet :) and file manager 
> in gui as well, but can be inside browser if u want

wont evar happen , dude!!! as u know forth is 2 good and if someone was to make a pc with web browser and internet and file manager and gui and persistance layer (like prevayler !!!!) it would kill industry. larry ellison has to get the money to buy that hawaiian island from somewhere!!!!

real question is : can chicken scheme have higher levels of abstraction than forth or would a window manager in Forth be 10000 time faster and use less resources? could bernd's web server in forth let me write blog in 1 lines of code?  how about storing everything in ram?  what about if I replaced database with a global variable??

[toc] | [prev] | [next] | [standalone]

#15623

On Wednesday, September 5, 2012 1:07:13 PM UTC-7, John Passaniti wrote:
> On Wednesday, September 5, 2012 5:18:28 AM UTC-4, gavino_himself wrote:
> 
> > bring out a pc done in forth and hurry up, one with 
> 
> > working web browser and internet :) and file manager 
> 
> > in gui as well, but can be inside browser if u want
> 
> 
> 
> wont evar happen , dude!!! as u know forth is 2 good and if someone was to make a pc with web browser and internet and file manager and gui and persistance layer (like prevayler !!!!) it would kill industry. larry ellison has to get the money to buy that hawaiian island from somewhere!!!!
> 
> 
> 
> real question is : can chicken scheme have higher levels of abstraction than forth or would a window manager in Forth be 10000 time faster and use less resources? could bernd's web server in forth let me write blog in 1 lines of code?  how about storing everything in ram?  what about if I replaced database with a global variable??




a pc than runs as fast on 1 m ram as the current ones do in 1000 would be impressive

can changes to such global variables be logged to disk? if so it seem bam then all good

you seem like a moron when u word salad my logical questons john and no one thinks of you as smart rather jsut a complainer, show us working apps, thats what we ask!! working apps!!

yeah!!

your not me!

[toc] | [prev] | [next] | [standalone]

Page 1 of 2  [1] 2  Next page →

Back to top | Article view | comp.lang.forth

csiph-web