On Mon, 10 Aug 2009, Dave Tweed wrote:

> Sergio wrote:
> > Ok, further to Dave's asertion that we could use a flag to allow us to
> > completely fill the FIFO. I have revised the FIFO code I posted and come
> > up with the following (appended to the end of this post). It uses two
> > extra counts instead of the one flag proposed by Dave.
> 
> Actually, that gives me an idea. If you simply allow the "head" and "tail"
> variables to count to twice the size of the FIFO, then they can perform
> both the indexing function and the counting function, with very little
> additional overhead, and no added thread-safety issues.

Very good!

But I would have got there in another 20 years or so :-)

> 
> If FIFO_SIZE is a power of two, then all of the '%' (modulus) operators
> can become '&' (bitwise AND) operators, efficient even on tiny processors.
> 

The only problem with this is that FIFO size needs to be a constant for 
the '%' to be optimised at compile time. That said all your FIFOs end up 
being the same size (because the size of each FIFO is not maintained per 
FIFO). I guess one way around that would be to declare different FIFOs 
each with their own predefined size (e.g. FIFO_16 has FIFO_SIZE_16
declared as its constant size). However this does mean that you'd also 
need special versions of fifo_add and fifo_remove for each predefined size 
(e.g. fifo_16_add and fifo_16_remove)

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