Author: Stuart Cracraft
Date: 16:32:26 09/19/04
Go up one level in this thread
On September 19, 2004 at 17:49:09, Andrew Williams wrote: >On September 19, 2004 at 17:28:47, Jon Dart wrote: > >>On September 19, 2004 at 15:18:47, Andrew Williams wrote: >> >>>That may be true, but I would reiterate that looking at its performance in WAC >>>is not going to help Stuart much in improving it. I don't even think it will >>>help much in improving its performance on other tactical tests, but that is just >>>a guess. I would strongly re-state my point: to learn what is wrong with a chess >>>program, it is better to play games than to test over and over on a test suite. >>>Even testing over and over on several test suites is not a good idea, in my >>>opinion. >> >>Test suites have some value. I'd add, that few programs are bug free. Finding >>and fixing bugs is beneficial over the long run, even if in the short run such >>fixes sometimes actually hurt performance. It is easy to have code that plays >>legal chess and even wins games and still have it do horrible wrong things >>internally--buffer overruns, memory corruption, you name it. That's why Arasan >>has ridiculous amounts of optional debugging and assert checking code. I also >>use Bounds Checker. >> > >I'd certainly agree about the use of asserts (PM should have more and I should >enable them more often in testing) and Bounds checking (I use valgrind, which is >fantastic). I think of those things as finding bugs, rather than improving my >program as such. So I wouldn't be looking at how many solutions I got, so much >as whether any asserts failed or valgrind saw some problem. > >I *do* use test sets sometimes (and I like your Arasan suites a lot), but it's >more for my amusement than because I think I'm going to learn anything >particularly interesting. Perhaps I'm just using them wrong... > >Andrew I have zillions of asserts. I used them to do some heavy debugging about 1/2 of the time ago to the program inception in June. Those plus this board got the jump up from about 160 out of 300 on WAC to 250/300. One of the exponents of program provability and advocate of asserts was Bob Floyd at Stanford. He was my initial introduction to programming and I remember the strong feelings he had about the subject. Bob has since passed on, regrettably, but he left an enormous legacy. Here for some details: http://www.fact-index.com/r/ro/robert_floyd.html and here for his primary collaborator, Knuth, with thoughts about Bob: http://sigact.acm.org/floyd/ Stuart
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.