Author: Uri Blass
Date: 16:25:11 12/17/02
Go up one level in this thread
On December 17, 2002 at 18:57:20, Gerd Isenberg wrote:
><snip>
>>>Hi Uri,
>>>
>>>OK you have pawnBB[2] with white pawns pawnBB[WHITE] and black pawns
>>>pawnBB[BLACK]. "allPawns" combines both sets by or.
>>>
>>>Some pawn-patterns on the fly (I'm at work and can't look to my sources):
>>>
>>
>>Thanks for the information
>>
>>I thoughtabout your post and understood part of it.
>>Here are my comments when I tried to explain it to myself.
>>At this moment I did not understand what is the meaning of duo
>>and there were some other things that I did not understand.
>
>
>OK, i made some mistakes on the fly and try to correct or comment.
>
>Some of the terms are german introduced by the austrian Chess Master H.Kmoch in
>"Die Kunst der Bauernführung".
>
>Widder, Hebel and Duo:
>
>[D] 8/6pp/8/1p2p3/1P1P4/6PP/8/8 w - -
>
>
>
>>
>>>pawnAttacksAH[WHITE] = (pawnBB[WHITE] << 9) & 0xfefefefefefefefe;
>>
>>
>>for me a1=0... h1=7 a2=8...h8=63
>
>Yes.
>
>pawnAttacksAH[WHITE] = (pawnBB[WHITE] << 9) & NOTFILEABB;
>
>... all white pawn attacks from a to h direction. Due to wrap around of pawns
>from h to a-file by shift up (<<8) and left (<<1, means board right here from
>whites point of view!) it is necessary to clear these possible wrapped bits by
>anding with a not a-file bitboard.
>
>With mmx-registers anding with NOTFILEABB is not necessary due to bytewise
>(rankwise) add (paddb mm1,mm1) instead of shift left one.
>
>
>>
>>suppose we are in the opening position
>>I guess pawnBB[WHITE]=(1<<8)+(1<<9)+...(1<<15)
>>
>>pawnsBB[white]<<9=(1<<17)+(1<<18)+...(1<<24)
>>0xfefefefefefefefe helps to get rid of the 1<<24 because fe mean 1 only at files
>>b-h
>>>pawnAttacksHA[WHITE] = (pawnBB[WHITE] << 7) & 0x7f7f7f7f7f7f7f7f;
>>
>>here we need to get rid of h4 that is not attacked
>>
>>7f is 1 in files a-g so it help for this target.
>>
>>>pawnAttacks[WHITE] = pawnAttacksAH[WHITE] | pawnAttacksHA[WHITE];
>>>pawnDblAttacks[WHITE] = pawnAttacksAH[WHITE] & pawnAttacksHA[WHITE];
>>
>>understood
>>
>>squares that are attacked by 1 pawn and squares that are attacked by 2 pawns.
>>
>>>
>>>hebel[WHITE] = pawnAttacks[WHITE] & pawnBB[BLACK];
>>
>>black pawns that are attacked by white pawns.
>
>
>
>Yes, but my definition was wrong. That are the black "hebel"-pawns.
>
>hebel[WHITE] = pawnBB[WHITE] & pawnAttacks[BLACK];
>hebel[BLACK] = pawnBB[BLACK] & pawnAttacks[WHITE];
>
>
>
>
>>
>>>widder[WHITE] = (pawnBB[WHITE] << 8) & pawnBB[BLACK]
>>
>>
>>If I understand correctly blocked white pawns by black pawns with distance of
>>one square.
>
>Yes, i also changed colors here
>
>widder[WHITE] = pawnBB[WHITE] & (pawnBB[BLACK] >> 8);
>widder[BLACK] = pawnBB[BLACK] & (pawnBB[WHITE] << 8);
>
>>
>>> /* & ~hebel[WHITE] */;
>>>defended[WHITE] = pawnAttacks[WHITE] & pawnBB[WHITE];
>>
>>understood.
>>>duo[WHITE] = ((pawnBB[WHITE]<<8)& pawnAttacks[WHITE]
>>> & ~pawnDblAttacks[WHITE]) >> 8;
>>
>>What does it mean?
>
>
>Duos are two pawns of same color on neighboared files and same rank.
>A duo-pawn is a pawn whose advance square is controlled exactly by one (but not
>two) neighboared pawn.
>
>
>
>>>
>>>Some helpfull routines:
>>>
>>>BitBoard fillUp(BitBoard bb)
>>>{
>>> bb |= (bb<<8);
>>> bb |= (bb<<16);
>>> bb |= (bb<<32);
>>> return bb;
>>>}
>>>
>>>BitBoard fillDown(BitBoard bb)
>>>{
>>> bb |= (bb>>8);
>>> bb |= (bb>>16);
>>> bb |= (bb>>32);
>>> return bb;
>>>}
>
>Kogge-Stone generator routines introduced by Steffan Westcott.
>
>I give two epd-strings representing the source bb and the return bitboard for
>fillUp:
>
>[D] 8/8/8/1P6/4PP2/8/7P/8 w - -
>[D] 1Q2QQ1Q/1Q2QQ1Q/1Q2QQ1Q/1P2QQ1Q/4PP1Q/7Q/7P/8 w - -
>
>
>>
>>I guess that fillup(pawnBB[WHITE]) gives me all the squares that white pawn can
>>visit without captures by pawns
>>I will say later in this post that these are the squares that the white pawns
>>"see" when filldown(pawnBB[BLACK]) are the square that black pawns "see"
>>correct?
>>
>>
><snip>
>>I do not understand why white pawns cannot be in these files.
>>
>>Uri
>
>OK, forgot the last errornous routines, this should be better and easier:
>
>openPawns[WHITE] = pawnBB[WHITE] & ~filldown(allPawns);
>passedPawns[WHITE] = pawnBB[WHITE] & ~filldown(allPawns|pawnAttacks[BLACK]);
>notDefendable[WHITE] = pawnBB[WHITE] & fillup(pawnAttacks[WHITE]);
>
>OpenPawns: All pawns without own or opposite pawns infront on same file.
>
>Backward pawns:
>
>First we need all squares dominated by sides pawn attacks:
>
>pawnDomination[side] = (pawnDblAttacks[side] & ~pawnDblAttacks[other(side)])
> | (pawnAttacks[side] & ~pawnAttacks[other(side)]);
>
>backWard1[WHITE] = notDefendable[WHITE] & (pawnDomination[BLACK] >> 8);
>backWard2[WHITE] = backWard1[WHITE] & openPawns[WHITE];
>
>
>I hope this helps a bit and gives you an impression of the compution of
>pawn-bitboard pattern. Your questions 4) affects more singular properties of a
>pawn.
>But even here Kogge-Stone-Up/Down fill (the passers as generator g, own attacks
>as propagator p) may be helpfull to get a set of the closest possible defenders
>(for each passer simultaniously).
>
>Steffan Westcott's routines for this purpose:
>
>BitBoard FillUpOccluded(BitBoard g, BitBoard p)
>{
> g |= p & (g << 8);
> p &= (p << 8);
> g |= p & (g << 16);
> p &= (p << 16);
> return g |= p & (g << 32);
>}
>
>BitBoard FillDownOccluded(BitBoard g, BitBoard p)
>{
> g |= p & (g >> 8);
> p &= (p >> 8);
> g |= p & (g >> 16);
> p &= (p >> 16);
> return g |= p & (g >> 32);
>}
>
>Question 5 is a bit unclear to me, i guess you mean candidates?
>May you post some sample epd-strings?
>
>Gerd
I admit that it is hard to define it for specific pawns and I guess that it may
be better to evaluate it only for a set of pawns.
I meant cadidares to be passer pawns
Here are some examples:
[D]k5K1/8/8/2p5/1p1p4/1P6/2P4P/8 b - - 0 1
black can generate a passed pawn but needs to sacrifice 2 pawns for that target.
c4 bxc4 d3 or c4 dxc4 b3
[D]k5K1/8/8/2pp4/1p6/1P6/2P4P/8 b - - 0 1
black does not need to sacrific to generate a passed pawn(again I see no way to
define exactly which pawn is the candidate to be passed pawn for evaluation).
Uri
Uri
This page took 0.01 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.