Author: Dann Corbit
Date: 09:20:21 03/30/99
Go up one level in this thread
On March 30, 1999 at 01:14:10, David Blackman wrote: [snip] >I'd suggest step 1 TSCP is a good idea. But try to really understand it well, >including adding improvements of various kinds. Then scrap it and write a >completely new program of your own, from scratch. Don't try to copy crafty >bitboards. If you want bitboards, invent your own method. There is at least one >better method already in use, and further improvement should be possible. I would be keenly interested to know about the better method in use. Do you know of a code sample I could examine? By the way, I agree 100% with your method of study and emulation of technique. That's the only way you'll really understand it anyway. >You'll >learn more by reading ideas from Bob (if we can persuade him to return) and all >the others programmers, and trying to implement them yourself, than from looking >at the code. Yet there are sometimes when looking at an example and tracing through it will be much faster than reading a paper on the topic. >One problem is that good well tuned programs like Crafty are very hard to >improve. Another is that the "understand crafty" approach ignores alternative >methods that might seem quite promising, but don't easily fit in with crafty's >existing methods. To know that your new idea is better, you must understand the old idea first. The sort of thing to do with crafty is look at the advancements that nobody else has and features that are very intricate and difficult. For instance, you might examine his SMP code. You might also consider the way he does eval. The EPD stuff by Steven Edwards and the tablebase interface are also all very interesting.
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.