Re: [oc] complex division in Hardware

[Richard Herveille <richard@asics.ws>]

There's a hardware division unit used by the 'Quantization Unit' in the 
video sytems project.
You can find it in the 'common' directory.

Richard


>Hi,
>
>I am trying to implement an HDL Module that performs a 16b/16b division
>and i can't find efficient algorith or strategy, when you write
>tmp00 = 1 / tmp00 , it is not staitforward, but how do you perform such
>an operation, of course it is not a simple shifter .
>could you please tell how to perform it.
>
>Kind regards
>
>
>----- Original Message -----
>From: Johan Klockars <rand@c... >
>To: cores@o...
>Date: Tue, 8 Jan 2002 11:50:51 +0100
>Subject: Re: [oc] complex division in Hardware
>
> >
> >
> > > I am implementing complex division in hardware. Is
> > > there any easy method to do that?
> >
> > Do you need this to be done in a single (very long) cycle, or is
> > multi-cycle operation OK?
> >
> > > As you know complex division can be done like this:
> > > a+bi      a+bi      c-di      (a+bi) * (c-di)
> > > ------ = ------- * ------- = --------------------
> > > c+di      c+di      c-di          c^2 + d^2
> >
> > Equivalent to the following (unless I made some mistake):
> > tmp00 = c * c
> > tmp01 = d * d
> > tmp00 = tmp00 + tmp01
> > tmp00 = 1 / tmp00
> > tmp01 = a * c
> > tmp02 = b * d
> > tmp01 = tmp01 + tmp02
> > real = tmp01 * tmp00
> > tmp01 = a * d
> > tmp02 = b * c
> > tmp01 = tmp02 - tmp01
> > imag = tmp01 * tmp00
> >
> > Possibly the output registers could be reused as temporaries too.
> >
> > > But it takes lot of hardware if we directly implement
> >
> > Obviously, both the multiplications and the division can be
> > implemented
> > using an adder/subtractor if the capability to shift values is
> > added.
> >
> > That is, you can make do with a single adder/subtractor, a couple
> > of
> > registers, a couple of multiplexers and a state machine.
> > For better performance, add a multiplier, divider, extra
> > adder/subtractor,
> > etc.
> >
> > --
> >   Chalmers University   | Why are these |  e-mail:
> > rand@c...
> >      of Technology      |  .signatures  |
> > johank@o...
> >                         | so hard to do |  WWW:
> > rand.thn.htu.se
> >    Gothenburg, Sweden   |     well?     |            (fVDI, MGIFv5,
> > QLem)
> >
>--
>To unsubscribe from cores mailing list please visit 
>http://www.opencores.org/mailinglists.shtml

--
To unsubscribe from cores mailing list please visit http://www.opencores.org/mailinglists.shtml