help needed on asynchromous FIFO............

[ANIRUDDHA13]

Hi everybody...i'm relatively new in vhdl...& i,wid my frnds hav decided to do a project on Asynchronous Fifo... can u help me out wid some good suggestns??

---------- Post added at 19:36 ---------- Previous post was at 19:35 ----------

any kind of help,like key points,logic,precautions etc.. shall be greatly apreiated... pls help...

 

[Chops211]

you can create one using handshaking. So a readyRead and readyWrite signal outputs used to tell whether something else is using the FIFO. Handshaking between the FIFO and whatever is talking to it is the most important thing in the design. I would also use a data valid signal that remains high while the data on the input port is valid.

Hope this helps

 

[supercat]

Hi everybody...i'm relatively new in vhdl...& i,wid my frnds hav decided to do a project on Asynchronous Fifo... can u help me out wid some good suggestns??

While it's possible to generate write-ready and read-ready signals asynchronously, it's really not very useful, since the only way those signals can be used for decision-making is if they're synchronized. If a write-ready signal is generated asynchronously and then synchronized, one will be limited to one write every 2-3 cycles because after each write one will have to wait for the synchronizer to find out whether the write filled up the queue.

A better approach to generating a write-ready signal is to synchronize a gray-coded read pointer with the writer's clock, and then compare that synchronously with the write pointer. The net effect will be that if the queue is full and data is read, it will take 2-3 clocks for the writer to find out that it can safely write data, but when the writer outputs data which fills the queue, it will discover that immediately. Read-ready should be generated with similar logic, synchronized to the read clock.

 

[Chops211]

I was under the assumption that the writer would be asynchronous as well. I have experience with RSFQ superconductors which must be asynchronous in order to work and the way we have implemented FIFOs has been with handshaking. I guess when talking about petaflops you think differently since there is really no such thing as a clock.

Supercat is correct if the writer is synchronous.