Author: Vincent Diepeveen
Date: 07:06:31 08/08/03
Go up one level in this thread
On August 08, 2003 at 09:39:51, Bo Persson wrote: >On August 08, 2003 at 08:43:49, Uri Blass wrote: > >>On August 08, 2003 at 08:22:30, Bo Persson wrote: >> >>> >>>That move generation tests are silly, because they compare apples to organges? >> >>Did not you learn that it is better to ignore everything that vincent say? > >No, because I am totally out of my mind. Vincent told me that the other day. :-) > >>He attacks personally people who disagree with him so it is better to say >>nothing if you disagree with him. > >So then I shouldn't say this: > >I once tried to avoid the "slow" bitboard operations by saving a list of From >squares. While generating capture moves, I also stored the squares in an array. >Then I could use this array "for free" when generating non-capturing moves. >Saved me one bitboard scan! > >Guess what - it made my move generation slower. Scanning the same MovingPieces >bitboard twice, once for captures and once for non-captures was faster. > > >>My post was not about the importance of the tests but a simple question to sune. >>Sune said that he is using incremental move generator > >Yes, if you use bitboards you can do that. Easily. > >>It is unimportant to discuss about subjects that vincent talk about unless he >>apologizes about his bad behaviour. > >Why should he apologize, when he is always right? > >I was very happy to read about how he does a branchless version of > > if (piece == pawn) you can get rid of that. i didn't in move generation because it is not important at 0.6% system time. the important thing is that this code is only having a loop for all kind of evaluation scans which eat majority of the system time when checking for patterns of course. Bitboards you can't combine all that knowledge in as i already proved a year ago, so i am really amazed by Joost Buijs comments. Hyatt said he didn't need such sophisticated features in evaluation. Joost didn't get that far yet it seems which amazes me. Most important thing is that future processors and already a few existing processors which are important to me, can use this way of coding very well. This where at bitboard code they suck ass. Even though they are or will be 64 bits processors. >earlier in this thread. You learn something every day! > >> >>Uri > > >Bo Persson
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.