What would you do with a dynamic/reconfigurable ISA??

[Bryan Bishop]

Hey all,

I have somewhat of a weird question. I am working on a conceptual
problem, toying around with some ideas on automated manufacturing,
kind of like apt-get but for physical projects (manufacturing
processes -- lathes, ingots, smelting, all of it). So to do this I
figure there'd be a set of instruments that you have, and these
instruments have some sort of defined functionality, physical routines
that you know they are capable of doing. For example, you know that
the arm can rotate 180 degrees from an axis, and that it can lift a
certain amount of weight. What I'm wondering about is the
generalizability aspects of this. It's very much like an ISA on a pic,
the whole 'instrument network' that you have up in your machine shop,
since all of the automated tools already have microcontrollers onboard
in the first place. And you can add/remove tools from the shop too, so
it's a reconfigurable instruction set architecture. So, how do the
compiler projects do it -- like gcc? Do they parse VHDL/Verilog files
and then generate compiler modifications to let gcc target that
platform? Are there ways to 'specify' the functionality that parts
(such as an ALU, or in my case, such as an arm or lathe) provide? And
would it be meaningful to try to generate both the ISAs and the
compiler at the same time as a possible solution?

Willing to clarify.

- Bryan
http://heybryan.org/
--
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist

[Shawn Tan Ser Ngiap]

On Friday 09 May 2008 00:00:20 Bryan Bishop wrote:
> Willing to clarify.

It may be 1am here, but I'm still totally confused.

--
with metta,
Shawn Tan

Aeste Works (M) Sdn Bhd - Engineering Elegance
http://www.aeste.net

--
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist

[Bryan Bishop]

On Thu, May 8, 2008 at 7:06 PM, Shawn Tan <shawn.tan@...> wrote:
> On Friday 09 May 2008 00:00:20 Bryan Bishop wrote:
>> Willing to clarify.
>
> It may be 1am here, but I'm still totally confused.

Imagine an FPGA that can do more changes than certain pathways in
logic circuits being flipped around; more than just logic circuits
being on/off, but enture physical movements and manufacturing tools.
Does this help?

- Bryan
--
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist

[Xiaofan Chen]

On 5/9/08, Bryan Bishop <kanzure@...> wrote:
>  And you can add/remove tools from the shop too, so
> it's a reconfigurable instruction set architecture. So, how do the
> compiler projects do it -- like gcc? Do they parse VHDL/Verilog files
> and then generate compiler modifications to let gcc target that
> platform? Are there ways to 'specify' the functionality that parts
> (such as an ALU, or in my case, such as an arm or lathe) provide? And
> would it be meaningful to try to generate both the ISAs and the
> compiler at the same time as a possible solution?
>

I do not quite understand you but let me try.

You can use the same MCU (same ISA), same compiler, but you
write different firmwares for different functions.

You can use the same FPGA, same tools, but different VHDL/
Verilog codes to have different logic fuctions.

You can have the same manufacturing lines, same PLC,
same motor drive, etc, but you write different PLC programs
to have different manufacturing process.


Xiaofan
--
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist

[Bryan Bishop]

On 5/8/08, Xiaofan Chen <xiaofanc@...> wrote:
>  You can have the same manufacturing lines, same PLC,
>  same motor drive, etc, but you write different PLC programs
>  to have different manufacturing process.

Yes, but what if the manufacturing line is reconfigurable to some
extent? Suppose, for example, that we give wheels and components to
the tools, or that two users (Bryan and Xiaofan) opt to make
installations, and Xiaofan opts to implement his own makeshift arm
that, more or less, follows most of the standards and specifications?
How can this be accounted for? Would a total refactoring of all of the
PLC-programs have to be rewritten? Or could something dynamic be done?

- Bryan
--
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist

[Xiaofan Chen]

On 5/9/08, Bryan Bishop <kanzure@...> wrote:

> On 5/8/08, Xiaofan Chen <xiaofanc@...> wrote:
> >  You can have the same manufacturing lines, same PLC,
> >  same motor drive, etc, but you write different PLC programs
> >  to have different manufacturing process.
>
> Yes, but what if the manufacturing line is reconfigurable to some
> extent? Suppose, for example, that we give wheels and components to
> the tools, or that two users (Bryan and Xiaofan) opt to make
> installations, and Xiaofan opts to implement his own makeshift arm
> that, more or less, follows most of the standards and specifications?
> How can this be accounted for? Would a total refactoring of all of the
> PLC-programs have to be rewritten? Or could something dynamic be done?
>

Since PLC can be programmable, you can 'dynamicly' change the
program if you want. But you may need to reconfigure the
related I/O modules, sensors, actuators, machines. So
in reality, it is not as simple to implement this kind of system.
FMS (flexibible manufacturing system,
http://en.wikipedia.org/wiki/Flexible_manufacturing)
and CIMS (Computer-Integrated Manufacturing Systems)
used to be popular terms but it seems to be fading away.

Xiaofan
--
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist

[Bryan Bishop]

On 5/8/08, Xiaofan Chen <xiaofanc@...> wrote:
> Since PLC can be programmable, you can 'dynamicly' change the
>  program if you want. But you may need to reconfigure the
>  related I/O modules, sensors, actuators, machines. So
>  in reality, it is not as simple to implement this kind of system.
>  FMS (flexibible manufacturing system,
>  http://en.wikipedia.org/wiki/Flexible_manufacturing)
>  and CIMS (Computer-Integrated Manufacturing Systems)
>  used to be popular terms but it seems to be fading away.

Interesting. Particularly my project is based around the idea of
automating manufacturing 'knowledge', kind of like apt-get for
software except now for more than just software -- packages of all
sorts, constructable objects and so on. But there's a wide range of
possible instruments that you can use, and having to re-code every
single item in the database every time a new tool comes along, seems
rather inefficient. I was wondering if there'd be a way that I could
avoid this. Apparently not? Though flexible manufacturing looks ...
close. No free lunch.

- Bryan
--
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist

[Xiaofan Chen]

On 5/9/08, Bryan Bishop <kanzure@...> wrote:

> On 5/8/08, Xiaofan Chen <xiaofanc@...> wrote:
> > Since PLC can be programmable, you can 'dynamicly' change the
> >  program if you want. But you may need to reconfigure the
> >  related I/O modules, sensors, actuators, machines. So
> >  in reality, it is not as simple to implement this kind of system.
> >  FMS (flexibible manufacturing system,
> >  http://en.wikipedia.org/wiki/Flexible_manufacturing)
> >  and CIMS (Computer-Integrated Manufacturing Systems)
> >  used to be popular terms but it seems to be fading away.
>
> Interesting. Particularly my project is based around the idea of
> automating manufacturing 'knowledge', kind of like apt-get for
> software except now for more than just software -- packages of all
> sorts, constructable objects and so on. But there's a wide range of
> possible instruments that you can use, and having to re-code every
> single item in the database every time a new tool comes along, seems
> rather inefficient. I was wondering if there'd be a way that I could
> avoid this. Apparently not? Though flexible manufacturing looks ...
> close. No free lunch.

You do not recode for everything. You have your existing
code base. It takes time to build your code base. Once
you have it, you can construct your program by assemble
the modules (I know this is oversimplification).

For the PLC side, some vendors will provide a feature
called "Add-on instruction", i.e., user defined function.
You also have subroutines, etc.

In the C world, you have the standard language
library.

In the higher level system world, you have all
kinds of abstraction. You can use C library,
RTOS, DotNet assembly, Delphi components,
etc.

In the end, you are using your existing
sub-assemblies to assemble a system.
But you need to assemble them. Software
may help (say expert system or wizard) but it
still need you, a human to do the job.

Robots can be quite good in some assembly
lines and can be kind of teach to do flexible
things. But in the end you need to teach them.


Xiaofan
--
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist