Xilinx IP core drivers for RTEMS- Keith Robertson still around?

gregory.menke at gsfc.nasa.gov gregory.menke at gsfc.nasa.gov
Wed Nov 8 19:47:02 UTC 2006

Keith Robertson writes:
 > Hi Greg.
 > I'd be interested in any/all of these updates if you're able to provide 
 > them at this point.  I've also significantly changed the soft core 
 > xilemac driver (32/64 bit fifo autodetection, much better code design, 
 > bug fixes), although it sounds like your changes incorporate that plus more.

I will deinitely contribute the fixes back, they've been stable for a
few weeks now.

 > I'm particularly interested in your rtems infrastructure changes, 
 > (makefiles, workspace/heap configuration, and the like) I was never 
 > particularly happy with the way the gen405 did all that.

I think it makes some more sense, and can be controlled via the
linkscript similarly to how other bsp's do it.  I don't presume to say
the few moves of files is entirely right, but the moves made sense at
the time...  I did take the first steps in decoupling the xilemac driver
from the gen405 bsp though.

 > Might we consider a new xilinx or virtex bsp rather than submitting to 
 > the gen405 one?

I think that would be a great idea, some clarification of the role of
libcpu/ppc_403 would be of considerable help.

 > A couple of questions as well:
 > 1) Is your updated port still using the old exception mechanism?

Yes- its simplicity was quite attractive.

 > 2) If you are using the new exception mechanism, do you handle/use the 
 > critical interrupt simultaneously with the external interrupt?  If not, 
 > have you given any consideration as to how to support it?

The external interrupt is stand-alone, how the gen405 did it.  The
clumsy part is that it consequently has its own vector table.  Its also
an advantage though- the Virtex interrupt handling is made somewhat more
modular with respect to the rest of the ppc exception model, that being
nice in this particular case.

The ppc exception setup in RTEMS looks so vastly complicated I didn't
want to get sucked into it given our short deadlines...



More information about the users mailing list