On 8/19/07, Dieter <[EMAIL PROTECTED]> wrote: > > On the other hand, we could also develop something that takes > > advantage of clusters; imagine if Google or someone were to get > > interested and offered the ability to submit synthesis jobs to their > > cluster. > > Would it be practical to do floss-chip-design-at-home similar to > folding-at-home, seti-at-home, ... ? > > If so, it would then be easy to run the final design on multiple > OSes and multiple CPU arches to catch FDIV type bugs. Testing the > app itself enough to trust it and have the ma$k built would be a lot > of work. Any wild guesses on how much code it would be? > > Or is testing the FPGA version sufficient?
I don't know what it would take. But I think we could look into it. I might suggest an approach. How about we begin by developing some code that, given a fully-routed design, could perform an analysis of timing and other characteristics? This would be the part of a GA, for instance, that calculates the "objective function." (If we're going to take the cluster approach, let's use a genetic algorithm, why not. But we can also consider other types of evolutionary algorithms.) For school, I've developed a pluggable GA framework for Ruby. For the moment, what's implemented is an experimental multicriterial algorithm. But shortly, I have to implement SPEA2 so I can have a point of comparison for a paper I'm writing. We could adapt this framework a bit and implement the FPGA evaluation block. (Ruby is slow for this kind of performance critical thing like P&R, but we can move important blocks into C extensions, leaving the high-level stuff in Ruby. Or we can use the prototype as a reference and rewrite the whole thing in C/C++ later.) Also, we could consider using the PISA framework. http://www.tik.ee.ethz.ch/sop/pisa/ -- 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)
