Re: Some basic queries

["Swapnajit Mittra" <mittra@my-deja.com> (by way of Daniel Weaver <danw@znyx.com>)]

>2. PCI-X says that Retry should be avoided, because it is not guaranteed
>that the cycle will come again, but this constraint is not in Conventional
>PCI. Why?

   I believe this has been done to avoid 
   "retry thrashing" (a particular transaction 
   clogging all the bandwidth while other 
   requests are pending) that has been a problem
   in PCI. Most PCI system have some upper count
   of retry attempts to avoid this. In PCIX, 
   since retry is not guaranteed, master has 
   a choice of starting this transaction 
   later as a new transaction if other 
   transactions are pending.

>
>3. Some of the terminations in PCI such as disconnect without data,
>disconnect with data after n data phases have been altogether removed from
>PCI-X. Why?
>

   The new set of PCIX target initiated 
   terminations have effectively removed the 
   scopes of these terminations. In PCIX now
   we have "Disconnect at Next ADB" which 
   means that only the terminations at 
   cacheline boundaries are permitted. This 
   has been done so that data transmission takes
   place in chunks and a good throughput is 
   maintained. 

   IMHO, this has largely simplified the 
   internal address calculations in a design. 

   Rgds,
   - Swapnajit. 

--
"Principles of Verilog PLI" by Swapnajit Mittra.
Kluwer Academic Publishers. ISBN: 0-7923-8477-6
Available in all online bookstores.

--== Sent via Deja.com http://www.deja.com/ ==--
Share what you know. Learn what you don't.