Gene,
We have been producing a cards based on the Xilnx PCI macro for several years. The Xilnx macro uses one clock of address stepping to ease the timing to turn on the output buffers. I have taken the designs to 3 compliance workshops and have not found a system that did not support the address stepping. I also have not run into any systems in the field that do not support the feature.
I understand that this macro is used by lots of folks, so I would expect that you could run into lots of problems not supporting address stepping. In general it is a bad idea to deviate from the spec, unless you have an embedded system that you can control all of the elements in the system.
Tony Clark
EFI
> -----Original Message-----
> From: Gene Morgan [mailto:morgan@equator.com]
> Sent: Sunday, August 20, 2000 10:17 PM
> To: pci-sig@znyx.com
> Subject: Address Stepping
>
>
> Does anyone know of PCI devices that use address stepping?
>
> Does anyone have experiences with PCI devices that use
> address stepping? In
> particular, do most arbiters handle them correctly? Is it
> permissible for
> arbiters to be programmed to not support address stepping?
> How many cycles do
> such devices generally take between receiving GNT# and
> asserting FRAME#? Is
> there a maximum number of cycles that such devices can take?
>
> Thanks,
>
> Gene Morgan
> Equator Technologies, Inc.
> morgan@equator.com
>