Results 1 to 5 of 5

Thread: Big Picture

  1. #1

    Big Picture

    Hope this clarifies more than it confuses!
    http://www.cogs.susx.ac.uk/users/mmg20/dhe/goals.php

  2. #2
    So why aren't we focussing the power we have at the multi-fault model? Is it that we can't describe it enough to have a genome written for it or do you just think the project is far too small to ever arrive at a solution for that model? It seems like that's _the_ project, no human has been able to do it, if this project did it you and the project would undoubtedly make scientific magazines.


    btw, any new versions and/or further optimizations possible?

  3. #3
    Thanks for your feedback safemode. I've been running the multi fault code on a separate island for a couple of weeks now rounding off the edges. It should be ready for release into the cluster very soon. I've got a feeling we'll need a lot of processing power for these runs because the circuit evaluation time now goes up exponentially with the number of gates - you have to test fault at 0 with fault at 1,2,3,4... 1 with 2,3,4... etc.. Let's see what this baby can do!

    Still, proving we can do better than conventional design on what they already do is also very important. Specially showing that the bigger the circuit, the more advantage there is (the more money you'll save) by using this method.
    Last edited by michaelgarvie; 02-02-2004 at 06:24 AM.

  4. #4

    is this an early win?

    Umm, tell me if i'm mistaken but doesn't a 1 1 1 mean we got a circuit that accomplishes our goals? Cuz that's what i'm seeing on the stat board. If it's not a winner (better than human) circuit then what exactly should we look for on the stat screen for when the goal for finding a testing the tester circuit that works. Is the stats messed up because there is no baseline to compare it to?

  5. #5
    It is a winner! Albeit a small one: it was using 4-input LUTs so instead of needing 5 gates to complete the task it only needed 2: my mistake in the task description.
    So we haven't yet even done one of 5 gates which is the minimum size to tick off the goals section. However this shows the approach works. Now I've launched it again using 2-input LUTs so it needs 5 gates to do the task.
    The main issue here is that the effort for evaluating a circuit increases with O(3^n) where n is the size of the circuit so we'll see how big we can get!

    Oh, if anyone has looked at the evolved Multiple Fault circuit, the E line signals an error when it stops oscillating. You can tell if its oscillating by running the simulator several times, if it ends on a different value each time then its oscillating.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •