2008/6/16 Timothy Normand Miller <[EMAIL PROTECTED]>:
> On Mon, Jun 16, 2008 at 8:47 AM, Kenneth Ostby <[EMAIL PROTECTED]> wrote:
>> Timothy Normand Miller:
>>>On Fri, Jun 13, 2008 at 6:54 PM, Jared Putnam <[EMAIL PROTECTED]> wrote:
>>>> --- On Fri, 6/13/08, Timothy Normand Miller <[EMAIL PROTECTED]> wrote:
>>>>> These may be nothing, but they should be investigated:
>>>> [unused pins]
>>>>
>>>> Did you ever get a tool put together to ensure that all module 
>>>> instatiations use the right number of wires for each port?
>>>
>>>Not sure.  I think someone wrote one in Perl or was going to, but I
>>>can't remember.  That would be a good sanity check.
>>
>> I did write a simple one in perl quite some time ago, however, when I
>> saw that the icarus had a switch for it, I discontinued the development
>> of it. Atm. I can't remember the name of the switch, but should be
>> easy to find it in the icarus man page.
>
> Well, anyhow, I've managed to fix all the warnings that are actually
> problems.  Also, there were some problems with entries in the
> constraints file not finding the nets they were associated with.
> Turns out that some signals were getting ripped out because they were
> duplicates of other signals.  I put in some metacomments to force them
> not to be ripped out, and voila.
>
> So, wrt the XP10, we still are stuck on the PCI clock skew issue, and
> for the S3, we're blowing the memory clock period constraint.  The
> latter is entirely an internal problem.  I have four different logic
> blocks trying to pull from a single fifo, so when it's time to
> dequeue, I have to make sure only to dequeue when the intended target
> is ready to get the data.  That results in a loop from fifo out data
> back into the dequeue signal.  The simplest fix is to have four
> separate command fifos, but that's a massive waste of logic area.  But
> then again, so are most of the the other options.
>
>

Can't you use a pipeline ? Duplicate only the last fifo stage ? Or
sometime, in a FSM, you could create a specific register for the bit
in the critical path. So you avoid any output decoding logic.

Regards,
Nicolas
_______________________________________________
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