RE: [pci] IRDY & TRDY

["Charles Krinke" <ckrinke@pulselink.net>]

Dear Miha:
	Thank you for replying to my question. I heard you were on vacation and
hope that it was restful for you. Let me try to get into the meat of the
questions and see where we go. First of all, let me describe my setup. I
have a 33Mhz oscillator coming into the VirtexE on GCK1. It then drives the
internal system controller (to generate PCI RST), the "blue beaver" arbiter,
PCI_BRIDGE32 and all of the PCI interface signals come out on two Mictor
connectors to a daughter board with trace lengths of about 3 inches to a PCI
connector into which I have plugged a PCI extender card (4 inch trace
lengths, logic analyzer connected here) and then a Cyclone PCI730 containing
an XScale 80200 microprocessor and 80312 PCI bridge. It is the 80312 that is
generating PCI configuration transactions.
	I have copied the pci_crt.ucf file from the app directory and put the
relevant in before and out after statements into my TOP.ucf where I am
instantiating the PCI_BRIDGE32. I have defined IOSTANDARD as PCI33_3 for the
PCI interface pins (except for PCICLK, which is TTL). Timing constraints are
tough to meet as you know, but near as I can tell, they are met as the
Xilinx ISE toolchain says "all timing constraits met".
	What I suspect is happening is that the 80312 de-asserts IRDY right at the
rising edge of the CLK or within 2NS after the rising edge. My analyzer time
resolution is 2NS so I have that much uncertaintity in the measurement. In
looking at the simulation, I can see that the IRDY signal is asserted longer
then the 80312 asserts. I am suspecting that the newer 80312 is pushing the
PCI spec more then the slightly older chips that you might have used in your
testing.
	My current thinking is going along the direction that line 711 in
pci_target32_sm.v may not need the and qualifier ~pci_irdy_reg_in and I can
say that commenting that one input to the 5 input and gate that creates the
load_to_conf_out pulse does allow writes to proceed. I am most interested in
your thoughts regarding this change to the state machine.
	I will make another pdf file on Friday and forward it to you with all of
the control signals shown between the Intel 80312 and the PCI_BRIDGE32 in
the FPGA. Its just that I need to wait until then so I can be away from the
heat of battle. Basically, I work in my home office on Friday and live in a
cubicle Monday-Thursday where it is difficult to concentrate.

Charles


--
To unsubscribe from pci mailing list please visit http://www.opencores.org/mailinglists.shtml