Re: [oc] CORDIC Processor

[Richard Herveille <richard@asics.ws>]

Keep in mind that besides the two different cordic algorithms 
(rotational/vectorial) you need to load different tables (linear, atan, 
etc) for the different operations modes.

I am not saying it is impossible, but you might be better off writing a 
specialized processor with only a few opcodes, specialized for the cordic 
algorithm.

This is however inherently slower than the pipelined implementation. If you 
choose the pipelined implementation, think about how the code will be 
generated. You will end up with a big multiplexor in front of each 
adder/subtractor unit to select the table for the currently selected 
operating mode. This will also slow down your code.

It's all a matter of choice, but both can be done.

Richard


>Hello,
>I've done some research on CORDIC algorithms, and I pretend to develop
>a CORDIC based processor, in VHDL to run on the new VIRTEX E.
>The main ideia is to make it  general, with a large set of functions, but I
>have a few doubts on the architecture to adopt...
>I'm planning to a have a FSM to control 2 CORDIC, one in rotacional
>mode, and the other in vectorial mode!
>Is this a a good ideia?
>Is the pipelined version possible (for diferent mode of operation (circular,
>linear, hyperbolic??))...
>
>Thanks for your attention!
>
>Joćo Mouro!
>Lisbon
>
>--
>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