Author: Gerd Isenberg
Date: 08:00:35 07/01/04
Go up one level in this thread
On July 01, 2004 at 09:20:31, Tony Werten wrote:
>On July 01, 2004 at 04:39:24, Gerd Isenberg wrote:
>
>>On July 01, 2004 at 02:50:35, Tony Werten wrote:
>>
>>>Hi all,
>>>
>>>although I like the principle of bitboards, it really bothers me that I can't
>>>seem to find a decent/fast way to evaluate weighted safe squares.
>>>
>>>Suppose I want to (simple) evaluate a rook, I generate a bitboard with all
>>>reachable squares and mask off the squares attacked by lower pieces (that's no
>>>problem).
>>>
>>>(This doesn't exacly generate safe squares, only the ones that aren't attacked
>>>at all by opponents pieces are, for the remaining squares one would need a SEE,
>>>but that's not the point )
>>>
>>>Now I can use this bitboard ( say rook on e4 ), mask the rank state, and look in
>>>a precomputed table how this rankstate scores on an e rank. No problem.
>>>
>>>But how to do the files ? If I use the rotated board, I need to have the
>>>opponents attackboard in this rotated board as well, wich would be very costly
>>>to compute (ie also for the bishops,queens ) and very complicated.
>>>
>>>Any ideas ? Am I missing something ?
>>>
>>>BTW, doing a popcount isn't a solution, since it violates the elegance of
>>>bitboards ( and is slow ?)
>>
>>Ok, bitscan traversing and psq-lookup is a even more violation IMHO ;-)
>
>:) That's what I meant. Popcount will only give an unweighted value.
>
>>
>>For safe rook attacks one may look for boolean pattern/properties (depending on
>>the game state). Intersect the attack set with several small areas of the board,
>>like center, remaining extended center, squares near opposite or own king,
>>passers, squares behind passers or other movable pawns e.g. to support a
>>minority attack, squares on open/halfopen files, squares on some rook
>>trajetories to opposite king and what ever else.
>>
>>With those sets, look whether they are empty and if not, probably look whether
>>the population count is greater one.
>
>OK, that's a different approach. Rook, knight and bishop can at most attack 2
>squares in the (centre) centre. That gives some possibilities to play with.
>
>
>>
>>That may be implemented loopless and with some preconditions:
>>
>>if ( safeRookAttacks )
>>{
>> set = safeRookAttacks & someAreaOfInterest;
>> bunus += f1*(set!=0) + f2*((set&(set-1))!=0);
>
>I know about conditional moves, but are there conditional multiplies ?
Yes "and" with -1|0 masks. See my reply to Vincent's post.
>
>But even with some conditional stuff fiddling, I have to agree with Vincent on
>this, this would leave a horrible amount of patterns. That would definately have
>to be "write once, never touch it again" code.
>
>Tony
A matter of taste. Currently i really do weighted popcount for mobility.
But many (piece interaction) patterns are simply boolean or 0,1,2.
Gerd
>
>> ...
>>}
>>
>>Of course such statements "cry" for MMX,SSE2,Itanium or AltiVec SIMD
>>instructions like pcmpeq to build -1|0 masks to "and" with. Itanium has popcnt
>>instruction, and there are rumors that AMD64 has undocumented popcnt too,
>>somewhere in the bt opcode range.
>>
>>SIMD within a register (SWAR) popcount looks relative expensive, but up to four
>>or even more popcounts may be processed simultaniously, only with a little more
>>time than one, by keeping all units busy. One may combine attack sets, e.g. from
>>knights(2)/bishops(1) to "odd" and "atLeasstTwo" bitboards with five bitwise
>>operations to safe one additional popcount per disjoint set too.
>>
>>Gerd
>>
>>>
>>>Tony
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.