I was reading one of those CPU benchmark articles some months back,
and I had the idea that another application area that would be good
for this would be Verilog synthesis (and P&R and all that) and
simulation.  It also occurred to me that if we were to offer some of
OUR Verilog code for the benchmark runs, we could be credited with
that, and it would be good publicity.  I emailed the author of the
article with the idea, and he said he was interested.  Unfortunately,
I have kinda dropped the ball on this with other priorities, so I was
hoping I could ask for some assistance.

I'd like to start a discussion on this as to what pieces that we have
would be good to benchmark with.  We have some larger modules that
take some time to synthesize, like PCI and memory, but none of them
take very long to run through relatively representative simulations.
We should pick modules for these tests that show off what we've done.

One idea I have is to bind together the PCI target, the
nanocontroller, and the memory controller in some reasonably
meaningful way, as though they were on one chip (no need to make this
realistic with respect to OGD1).  This might be large enough to take
some time to for synthesis.  Then, we take the same design and put it
into a simulation environment where the nanocontroller is programmed
through PCI and then made to run a rather long computation from its
microcode memory.

Really, we're going to have to put together a relatively polished
couple of things for the benchmarker to unzip and run.  This isn't
like a test you'd put together yourself, because someone unfamiliar
with chip design will have to be able to run it without breaking it,
and if they do manage to break it, they just reinstall it.  We can
surely have them install the Lattice tools or the Xilinx WebPack, but
then there needs to be a zip file to drop in a specific place so that
they can just open the project and click the right button.

For the simulation, I was wondering what it would take to get a Cygwin
environment set up for easy install, and when it's done, there's a
shell script to run that times an Icarus run on our controller
simulation.

Thoughts?  Anyone want to help?  :)


-- 
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