[Open-graphics] A really good immediate use for OGD1: PCI bus analyzer

[Dieter]

Tim> Since we can't count on the user's UART being fast...  why not ship
Tim> our own?  Most inexpensive USB to Serial adapters claim they can do >
Tim> 1Mbps.  Roughly 128k/s, or 34 minutes for a full 256Mb ram dump.  We
Tim> get the simplicity of serial on the FPGA-side, the compatibility and
Tim> speed of USB on the PC side.  And all for $5 - $15.
Tim> 
Tim> See: http://www.newegg.com/Product/Product.asp?Item=N82E16812189123

$7.99 + $4.99 shipping

http://www.link-depot.com/USB-DB9.htm

Question is, what chip does this one use?  We need device driver support
for the various OSes.

I'm told that increasing the speed of a UART is just a matter of using a
faster crystal.  So any RS-232 port that we can change the crystal on should
work.

USB has its problems, but one advantage to a USB-to-RS232 is that we can make
the rs232 cable very very short.  Wire the OGD as a DCE and plug it in directly,
no rs232 cable at all.

-------------------------------

Timothy> I want to make sure we can identify that this serial
Timothy> device is the one associated with our board, as opposed to some random
Timothy> serial device.  This just makes it easier for our tracer software to
Timothy> figure out which serial device it wants to talk to.

Figure out what /dev entry it is manually, then

	ln -s /dev/tty037  /ogd_pci_analyzer/control_port

and have the software use /ogd_pci_analyzer/control_port

-------------------------------

Timothy> I'm inspired by the Itanium instruction set.

Timothy> the effect will be that we report that a signal changed 2.5ns
Timothy> later than it actually did.

Yep, getting incorrect answers sounds like something inspired by Intel.  :-(
Itanium is especially bad.

-------------------------------

Timothy> Yeah.  Excepting the case of a 64-bit PCI bus (which we could
Timothy> support), we have fewer than 64 signals to track.

I thought you were debugging a 64-bit PCI bus?

-------------------------------

Daniel> It's a Microsoft spec

Isn't that enough of a warning?