Author: Gareth McCaughan
Date: 18:00:52 02/16/04
Go up one level in this thread
Martin Fierz wrote: [Franz Karger:] >> Some studies show that development time in languages >> like Lisp is about half than in C, Java or C++. >> e.g. http://www.algo.be/cl/TEE-lisp/31837187622993390/index.htm ... >i don't buy this, for two reasons: > 1) i started writing my chess program on the 1st of july 2003. it's written in > pure C. it played it's first games within a week. it has been playing on ICC > now for months. it's not a great program by any standard, but it's a decent > amateur engine. > steven edwards has been posting about Symbolic, his LISP-engine, for what > seems like ages to me. it hasn't played a single game yet AFAIK. where > exactly is the reduced development time here?? 1. What Steven's trying to do and what you're trying to do are entirely different things. 2. For reasons I don't pretend to understand, Steven's decided to start by writing his own Lisp interpreter (with a bunch of chess-related primitives built in) rather than gluing his chess-related primitives into an existing Lisp implementation. That means that (a) he's had to spend a certain amount of time not writing in Lisp, but *implementing* Lisp, writing in C (or C++; I forget), and (b) the Lisp he now has to work with doesn't have all the nice things that, say, Common Lisp has. Though it does have some neat chess primitives :-). 3. For all we know, you may be a much better programmer than Steven. >2) once you have your basic engine running, the smallest part in improving > it is actual coding. what you really need to do is look at the games > it plays, find out what it's doing wrong, and find out how to fix it. > this evening, i looked at 70 blitz games my engine played during the day > against other engines. looking at them and drawing conclusions about what > to do took me about 90 minutes (needless to say, i could have spend > much more time on this!). applying the conclusions to my code took > 30 minutes - half of this was changing eval weights and quickly > checking whether the eval in the positions i had in mind was now better. > right now, i'm performing sanity checks which takes another hour. > once they are done, i'm running another 120 blitz games which takes > until tomorrow evening. so out of 24 hours time going into improving > my engine, 30 minutes are spent with actual coding. half of that is > changing weights in the eval, which takes the same amount of time > regardless of programming language. even if i could be 50% more efficient > programming lisp, which i don't believe, i would have saved > exactly 7.5 minutes today. :-) That's a much more convincing reason. -- g
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.