Just a note that we think we may have figured out what was going
wrong.  We've not completed the simulations to be sure.  Anyhow,
there's a cycle when PCI deasserts FRAME and IRDY, and we've
deasserted TRDY.  We're not sure if it's our fault or the host, but
since this cycle is the end of the read transaction, we should
probably stop driving the bus.  We were still driving it for that
cycle.  My fix should eliminate the contention.

How Howard found it was interesting.  I think it was AD19.  It gets
the input register clock from one bank of logic in the FPGA and the
output register clock from another bank.  This resulted in odd timing
skews.  If we add additional load to that signal on the bus, the data
corruption goes away.  Remove the load, and things break.  I'm not
sure how Howard got from there to figuring out what simulation to run,
but he found the contention, and I modified the PCI target as
described above.
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to