Reply by Dave Hansen ●February 8, 20052005-02-08
On Mon, 07 Feb 2005 18:31:39 +0200, Anton Erasmus
<nobody@spam.prevent.net> wrote:


>On Mon, 07 Feb 2005 14:27:03 GMT, iddw@hotmail.com (Dave Hansen)
>wrote:
>

[...]

>>FWIW, at the turn of the century I (briefly) worked on an application
>>using a (mumblemumble 500?) MHz Pentium running QNX and hosting 256 MB
>>RAM.  The application itself wasn't that big (< 2 MB, IIRC), but it
>>sucked in huge volumes of (mostly image, but other kinds of mapping)
>>data for processing in real time.  I won't claim it used all its
>>memory, but I know it used more than 128 MB.
>
>Yes, but most of the stuff was data. To have that much code is
>absolutely huge. What sort of application needs these massive amounts
>of code space. I am sure that many databases can basically grow to any
>size datawize.


Yes, that was my point.  Memory holds more than programs.  

In terms of conventional "embedded programming," we paid essentially
_no_ attention to the size of the application.  

Regards,

                               -=Dave
-- 
Change is inevitable, progress is not.
Reply by Anton Erasmus ●February 7, 20052005-02-07
On Mon, 07 Feb 2005 14:27:03 GMT, iddw@hotmail.com (Dave Hansen)
wrote:


>On Sat, 05 Feb 2005 08:43:02 +0200, Anton Erasmus
><nobody@spam.prevent.net> wrote:
>
>>On Thu, 03 Feb 2005 20:57:34 -0500, "Michael N. Moran"
>><mnmoran@bellsouth.net> wrote:
>[...]
>>>My current project is a PowerPC 8555 based telecom product
>>>with 128MB of RAM.
>>>
>>>You guys are sooooo jealous ;-)
>>
>>Only if we do not have to actually fill that much space with usefull
>>efficient code. An application that actually uses that much RAM
>>efficiently is absolutely massive.
>
>FWIW, at the turn of the century I (briefly) worked on an application
>using a (mumblemumble 500?) MHz Pentium running QNX and hosting 256 MB
>RAM.  The application itself wasn't that big (< 2 MB, IIRC), but it
>sucked in huge volumes of (mostly image, but other kinds of mapping)
>data for processing in real time.  I won't claim it used all its
>memory, but I know it used more than 128 MB.


Yes, but most of the stuff was data. To have that much code is
absolutely huge. What sort of application needs these massive amounts
of code space. I am sure that many databases can basically grow to any
size datawize.

Talking of big applications. What is the executable size of the very
big Windows applications compared to the big Unix/Linux applications ?

Regards
   Anton Erasmus
Reply by Dave Hansen ●February 7, 20052005-02-07
On Sat, 05 Feb 2005 08:43:02 +0200, Anton Erasmus
<nobody@spam.prevent.net> wrote:


>On Thu, 03 Feb 2005 20:57:34 -0500, "Michael N. Moran"
><mnmoran@bellsouth.net> wrote:

[...]

>>My current project is a PowerPC 8555 based telecom product
>>with 128MB of RAM.
>>
>>You guys are sooooo jealous ;-)
>
>Only if we do not have to actually fill that much space with usefull
>efficient code. An application that actually uses that much RAM
>efficiently is absolutely massive.


FWIW, at the turn of the century I (briefly) worked on an application
using a (mumblemumble 500?) MHz Pentium running QNX and hosting 256 MB
RAM.  The application itself wasn't that big (< 2 MB, IIRC), but it
sucked in huge volumes of (mostly image, but other kinds of mapping)
data for processing in real time.  I won't claim it used all its
memory, but I know it used more than 128 MB.

But I haven't worked on a product with more than 32k program memory
since then.

Regards,

                               -=Dave
-- 
Change is inevitable, progress is not.
Reply by Anton Erasmus ●February 5, 20052005-02-05
On Thu, 03 Feb 2005 20:57:34 -0500, "Michael N. Moran"
<mnmoran@bellsouth.net> wrote:


>Richard wrote:
>> Ed Beroset <beroset@mindspring.com> wrote in news:xfTLd.6653$Ix.1482
>>>Richard wrote:
>>>
>>>The SCALE of the embedded device has a lot to do with whether 
>>>C++ might be a viable choice or not.
>
>Bingo.
>
>> No doubt.  In my 20+ years of experience with embedded systems, 16MB is 
>> absolutely humongous, far and away larger than than any I have ever 
>> worked on going through a fair variety of companies working on a variety 
>> of embedded systems.  So to me, that is off the scale huge.
>
>My current project is a PowerPC 8555 based telecom product
>with 128MB of RAM.
>
>You guys are sooooo jealous ;-)


Only if we do not have to actually fill that much space with usefull
efficient code. An application that actually uses that much RAM
efficiently is absolutely massive.

Regards
   Anton Erasmus
Reply by R Adsett ●February 5, 20052005-02-05
In article <4203f33f$0$25966$9b622d9e@news.freenet.de>, mur@awite.de 
says...

> The device (called FRED) is from www.kws.de. I looked at their homepage, 
> but it is only on the german version:
> 
> http://www.kws-computer.de/framesets/frame215.html

 
<snip>

> 
> I have also found a second company which offer a similar system 
> (www.ultratronik.de), but they are some weeks behind with the 
> development (and I need it now).
> 
> I also would not have a problem to share my experiences or source code 
> (as I use open source software).


Thanks, I may take you up on that offer at some point.  At the moment I'm 
still in the process of figuring out the spec. for my client.

Robert
Reply by Guy Macon ●February 4, 20052005-02-04
Andreas Huber wrote:

>
>Guy Macon wrote:
>
>> Andreas Huber wrote:
>>
>>> I might agree if the saturation value is INF (which is often
>>> supported with FP arithmetic). Then, further calculation will yield
>>> *obviously* wrong results and you also don't need to hand-propagate
>>> errors.
>>
>> Hand-propagate errors?  That's what signaling NANs an silent NANs
>> were invented for.
>
>Well, INF is propagated automatically as well, isn't it? Moreover, if 
>you make further calculations with INF values (e.g. INF*0), NAN values 
>often result...


Exactly so.  In many cases, no need to hand-propagate errors.
Reply by Ernst Murnleitner ●February 4, 20052005-02-04
R Adsett wrote:

> In article <420160fe$0$25942$9b622d9e@news.freenet.de>, mur@awite.de 
> says...
> 
>>Currently, I am replacing this system (it has not 16 but 8 MB, I was 
>>wrong) by a system with 64 MB, 32bit CPU and 32 MB RAM, colour TFT touch 
>>screen (480 x 234 Pixels), CF slot, ethernet interface - WHY? Because I 
>>get it for the same price and for 50 to 100 peaces per year, the 
>>software development costs are the highest costs - so it is very nice to 
>>test the software on a normal pc. Such a system cost me about 500 EUR 
>>(600 USD), which is in a device for 7000. 
> 
> 
> What/whose system are you using for that?  I have a customer who might be 
> interested in a similiar system.
> 


I wrote, that you can test it on a PC. But you have to compile your 
software with a cross compiler, as the system includes a MIPS32 
compatible CPU. (If you use Linux; I have no experience with Windows 
CE). I think this is OK? (IA32 (i86) compatible systems cost twice the 
price). We already downloaded some utilitys from a 2.4.20 mipsel-debian 
and it worked.

The device (called FRED) is from www.kws.de. I looked at their homepage, 
but it is only on the german version:

http://www.kws-computer.de/framesets/frame215.html

Somewhere you can also download a price list. But I think, the 7 " FRED 
is not on the list now, because it is only available soon (but it should 
be available in the next few weeks). I got a prototype from them. The 8 
" version is available, but costs 50 % more (but has a high resolution 
display for this size: 800 x 600).

I run, amongst my applications, the processviewbrowser on it 
(www.pvbrowser.org). The device (500 MHz AMD Alchemy Au1100) seems to be 
faster than a Panel PC with 700 MHz Celeron. And faster than StrongARM 
or XScale, which I also took into account. I use Linux (2.4.20, from 
Linux Division).

I have also found a second company which offer a similar system 
(www.ultratronik.de), but they are some weeks behind with the 
development (and I need it now).

I also would not have a problem to share my experiences or source code 
(as I use open source software).

Greetings

Ernst
Reply by Ernst Murnleitner ●February 4, 20052005-02-04
R Adsett wrote:

> In article <420160fe$0$25942$9b622d9e@news.freenet.de>, mur@awite.de 
> says...
> 
>>Currently, I am replacing this system (it has not 16 but 8 MB, I was 
>>wrong) by a system with 64 MB, 32bit CPU and 32 MB RAM, colour TFT touch 
>>screen (480 x 234 Pixels), CF slot, ethernet interface - WHY? Because I 
>>get it for the same price and for 50 to 100 peaces per year, the 
>>software development costs are the highest costs - so it is very nice to 
>>test the software on a normal pc. Such a system cost me about 500 EUR 
>>(600 USD), which is in a device for 7000. 
> 
> 
> What/whose system are you using for that?  I have a customer who might be 
> interested in a similiar system.
> 


I wrote, that you can test it on a PC. But you have to compile your 
software with a cross compiler, as the system includes a MIPS32 
compatible CPU. (If you use Linux; I have no experience with Windows 
CE). I think this is OK? (IA32 (i86) compatible systems cost twice the 
price). We already downloaded some utilitys from a 2.4.20 mipsel-debian 
and it worked.

The device (called FRED) is from www.kws.de. I looked at their homepage, 
but it is only on the german version:

http://www.kws-computer.de/framesets/frame215.html

Somewhere you can also download a price list. But I think, the 7 " FRED 
is not on the list now, because it is only available soon (but it should 
be available in the next few weeks). I got a prototype from them. The 8 
" version is available, but costs 50 % more (but has a high resolution 
display for this size: 800 x 600).

I run, amongst my applications, the processviewbrowser on it 
(www.pvbrowser.org). The device (500 MHz AMD Alchemy Au1100) seems to be 
faster than a Panel PC with 700 MHz Celeron. And faster than StrongARM 
or XScale, which I also took into account. I use Linux (2.4.20, from 
Linux Division).

I have also found a second company which offer a similar system 
(www.ultratronik.de), but they are some weeks behind with the 
development (and I need it now).

I also would not have a problem to share my experiences or source code 
(as I use open source software).

Greetings

Ernst
Reply by Ernst Murnleitner ●February 4, 20052005-02-04
R Adsett wrote:

> In article <420160fe$0$25942$9b622d9e@news.freenet.de>, mur@awite.de 
> says...
> 
>>Currently, I am replacing this system (it has not 16 but 8 MB, I was 
>>wrong) by a system with 64 MB, 32bit CPU and 32 MB RAM, colour TFT touch 
>>screen (480 x 234 Pixels), CF slot, ethernet interface - WHY? Because I 
>>get it for the same price and for 50 to 100 peaces per year, the 
>>software development costs are the highest costs - so it is very nice to 
>>test the software on a normal pc. Such a system cost me about 500 EUR 
>>(600 USD), which is in a device for 7000. 
> 
> 
> What/whose system are you using for that?  I have a customer who might be 
> interested in a similiar system.
> 


I wrote, that you can test it on a PC. But you have to compile your 
software with a cross compiler, as the system includes a MIPS32 
compatible CPU. (If you use Linux; I have no experience with Windows 
CE). I think this is OK? (IA32 (i86) compatible systems cost twice the 
price). We already downloaded some utilitys from a 2.4.20 mipsel-debian 
and it worked.

The device (called FRED) is from www.kws.de. I looked at their homepage, 
but it is only on the german version:

http://www.kws-computer.de/framesets/frame215.html

Somewhere you can also download a price list. But I think, the 7 " FRED 
is not on the list now, because it is only available soon (but it should 
be available in the next few weeks). I got a prototype from them. The 8 
" version is available, but costs 50 % more (but has a high resolution 
display for this size: 800 x 600).

I run, amongst my applications, the processviewbrowser on it 
(www.pvbrowser.org). The device (500 MHz AMD Alchemy Au1100) seems to be 
faster than a Panel PC with 700 MHz Celeron. And faster than StrongARM 
or XScale, which I also took into account. I use Linux (2.4.20, from 
Linux Division).

I have also found a second company which offer a similar system 
(www.ultratronik.de), but they are some weeks behind with the 
development (and I need it now).

I also would not have a problem to share my experiences or source code 
(as I use open source software).

Greetings

Ernst
Reply by Andreas Huber ●February 4, 20052005-02-04
Guy Macon wrote:

> Andreas Huber wrote:
>
>> I might agree if the saturation value is INF (which is often
>> supported with FP arithmetic). Then, further calculation will yield
>> *obviously* wrong results and you also don't need to hand-propagate
>> errors.
>
> Hand-propagate errors?  That's what signaling NANs an silent NANs
> were invented for.


Well, INF is propagated automatically as well, isn't it? Moreover, if 
you make further calculations with INF values (e.g. INF*0), NAN values 
often result...

Regards,

-- 
Andreas Huber

When replying by private email, please remove the words spam and trap
from the address shown in the header.