From: Don Smith, May 5, 2004 Progress Report: I found a major bug that had been causing focussing+dyckhoff ("D") to be incomplete. Previously it had clearly failed on a couple of formulas ('30b' and 31). Now D succeeds on both 30b and 31. The bug was in the the code that I used for removing duplicate assumptions (it was too aggressive, because of the way I standardized free variables incorrectly). With this change, I am more confident that focussing+Dyckhoff is complete. Running time for the test suite without D = 41.1 seconds. With D = 7.4. There are a couple of formulas: t10,t11, and synth1 which take orders of magnitude longer without D. Some formulas are faster without D. Timing results vary more now. I notice that the time to prove a formula can change from hours to seconds by increasing the start depth for iterative deepening from, say, 8 to 26. For example, for formula 31 it takes at least a half an hour when it starts with iterative deepening from depth 8. But it takes < 1 mls when it starts from depth 38. I automated the production of (formatted) benchmark results and added a timeout facility for the harder problems. I also log query and results incrementally to file. I added more examples, mostly from Gandalf. Plan: I want to do a thorough comparison with Gandalf. If I tweak the starting depth of iterative deepening, I can sometimes speed up finding proofs by orders of magnitude. Probably time sharing among multiple proof attempts would be the fastest.