hamster Posted March 3, 2014 Report Share Posted March 3, 2014 Just dropping by to say I've finished an FPGA-based solver for Eternity II style tile matching puzzles. (see http://en.wikipedia.org/wiki/Eternity_II_puzzle). It runs on the Spartan 6 LX9 @ 200 MHz with a four-stage pipeline, allowing four 'threads' to share the logic resources and work on the same puzzle at any time. Even better than that, I can fit four of these four stage pipelines in the LX9, giving 16 threads that each can place up to 50 M tiles per second. If I use distribututed RAM for some of the smaller RAM instances I can fit 12 threads @ but at a slower clock of 160MHz. Each thread is about 10% the speed of a highly optimised software solver running on a 3.5GHz i7 core, making the total LX9 about as fast as a single i7 core. However, it is very, very unlikely that it will ever find a solution to the puzzle - even at > 100 million search nodes per second it takes a very long time to brute-force the the estimated 10^52 search nodes. Link to comment Share on other sites More sharing options...
This topic is now archived and is closed to further replies.