View Single Post
  #2 (permalink)  
Old 04-12-2006, 07:48 AM
Jim Granville
Guest
 
Posts: n/a
Default Re: State Machine and Area Estimate Question

[email protected] wrote:
> Hi,
> I have a design with 2 components, each of which implement a small
> but non trivial state machine. Each one has an associated area estimate
> reported by Xilinx, say A1 and A2. Now I realized that both of these
> components would never execute their state machine concurrently so I
> combined both state machines into one larger state machine, with the
> idea that I might be able to save some area since states in both
> individual state machine might use the same resource. At the very least
> I thought the new area of the larger state machine would be A1 + A2. I
> was actually expecting something smaller than an additive area ( due to
> the resource sharing), but instead I was surprised to note that Xilinx
> reported an area larger than (A1 + A2). Whats going on here? Are
> additional LUT components being used for routing purposes? Im somewhat
> confused as it did the exact opposite of what I expected. Any help or
> explanation would be appreciated!


What exactly did you expect to share ?
Think of a state machine as a Set of registers, and a ROM.

The registers that hold the present state, clearly cannot share.
The ROMs will individually logic-reduce to pack into logic fabric,
but how would sharing help shrink 2 ROMs ?

Suppose a portion of the 2 roms does match, then you will need
decoders and MUXs to switch in the shared portion - so you go
backwards before you go forwards ( and expect a very smart synthesis
tool !)

Then a later state-edit could undo all this....

Generally, you make state-machines smaller by doing the opposite of
what you seem to propose : ie sometimes nested machines can be smaller.

-jg


Reply With Quote