MATLAB BIT PRECISION!!

Hi all !!

I would like to know if it is possible to run Matlab on 16-bit and 8
-bit processor !!

Since it is already running on 32 bit !!

what is the possible way to quantize to lower bit precision !!!

Thanking you 

Ali

MuhammadAli.AA@gmail.com wrote:

> 
> Hi all !!
> 
> I would like to know if it is possible to run Matlab on 16-bit and 8
> -bit processor !!
> 
> Since it is already running on 32 bit !!
> 
> what is the possible way to quantize to lower bit precision !!!
> 
> Thanking you 
> 
> Ali
> 

Does it have anything with the US/EU ban on the export of the high 
performance CPUs to some countries?

VLV

MuhammadAli.AA@gmail.com wrote:
> Hi all !!
>
> I would like to know if it is possible to run Matlab on 16-bit and 8
> -bit processor !!
>
> Since it is already running on 32 bit !!
>
> what is the possible way to quantize to lower bit precision !!!
>
> Thanking you
>
> Ali

Do you want to:

1) Run the software Matlab on a 16-bit or 8-bit processor,

or

2) Use Matlab to simulate a 16-bit or 8-bit processor?

For (1), you should ask the company Mathworks if they have a version
that
runs on whatever processor you want.

For (2), write a quantizer or define the appropriate data type:

http://www.mathworks.com/access/helpdesk/help/techdoc/matlab_prog/index.html?/access/helpdesk/help/techdoc/matlab_prog/f2-43934.html&http://www.google.com/search?hl=en&lr=&client=firefox-a&rls=org.mozilla%3Aen-US%3Aofficial&hs=Wcy&q=matlab+variable+type&btnG=Search

Hope that helps,
Julius

>
>
>Hi all !!
>
>I would like to know if it is possible to run Matlab on 16-bit and 8
>-bit processor !!
>
>Since it is already running on 32 bit !!
>
>what is the possible way to quantize to lower bit precision !!!
>
>Thanking you 
>
>Ali
>
>

Ali,
Matlab has a toolbox called "Fixed Point Toolbox" that makes things like
this very easy.  If you don't, and you wouldn't be alone, you have to cast
your Matlab variables using int16.  It's kind of a pain because every time
you do any math to it you have to re-cast.

EX.

IIR_Output(j) = int16(dfrY);
IIR_Output(j) = int16(IIR_Output(j)*Coeff(j));

If you don't keep re-casting it will default back to the native data type
which is actually 64 bits.

Jeff