On Sun, May 29, 2011 at 6:47 PM,  <[email protected]> wrote:
> On Mon, 30 May 2011 00:08:02 +0200, "Ing. Daniel Rozsnyó" wrote:
>>
>> You should use a PCI-to-"something" bridge, by "something" I
>> mean GPIO or serial port. For true "native" PCI you will need an FPGA.
>> I doubt that the microcontroller can handle PCI in software way in
>> role of slave (in master it would be possible).
>
> AFAIK the purpose of this project and list is to design a graphic card,
> Renesas stuff are way out of scope and i don't even care about
> looking at the datasheet...

We're interested in all open hardware, although many of us have more
expertise and interest in graphics than perhaps hardware.  Andre and I
are cooking up a project idea that definitely involves graphics, but
most people wouldn't see that as its primary purpose.

The issue with Mr. Jamadhagni is that he appears to be much too
unfamiliar with hardware protocols and is therefore having a very
difficult time communicating his needs.  On this list, we're happy to
educate and to help novices, but as with basically all FOSS groups, we
expect people to google first and ask questions second.  I'm not sure
he's done his homework.

>
>> PC104+ also has an ISA
>> section, but that is 16MHz? also too fast for a software slave..
>
> PC104+ or PCI104 is a PCI bus using the same form factor as PC104
> and most recent boards have dropped the 2.54mm pitch connector
> and keep the 2mm pitch PCI signals, faster.
>
> PC104 (ISA) is very easy to interface with the help of a few TTL parts
> (5V tolerant) so the uC can be slow if needed, using a few latches to
> do the handshaking and data buffering.
>
> And we didn't even scratch the surface. Hardware and software
> are related. PCI(104) is terribly complex.
> In comparison, prototyping with ISA/PC104 is a breeze.
> One reason why I keep industrial PCs around !

Yeah.  I didn't design the OGP's PCI controller over night.  It's
complicated, and there are lots of little details to screw up... for
instance, when you're done asserting a signal, do you just tristate
the line, or do you pull it high for a cycle?  Depends on which
signal.  :)

Actually, I bet PCIe would be a little easier.  All of the low-level
signaling and timing-related stuff would be handled for you in SERDES.
 For one-lane PCIe, the rest is just sending and receiving the right
packets of data.  (I'm doing loads of guessing here, but with many
high-speed serial protocols, the "length" of a bit is less than the
length of the wire it's being transmitted on, which necessarily
implies a certain degree of asynchrony.)

-- 
Timothy Normand Miller
http://www.cse.ohio-state.edu/~millerti
Open Graphics Project
_______________________________________________
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