Author: Gerd Isenberg
Date: 23:09:07 04/14/04
Go up one level in this thread
On April 14, 2004 at 19:34:57, rasjid chan wrote:
>On April 14, 2004 at 15:52:39, Gerd Isenberg wrote:
>
>>Hi Rasjid,
>>
>>i found a slight improvement in one of your three interesting sliding attack
>>macros. It replaces the "inbetween"-expression
>>
>> isqBit(y)-1 & ~( isqBit(x)-1 | isqBit(x) )
>>by
>> isqBit(y) - isqBit(x+1)
>>or
>> isqBit(y) -(isqBit(x)<<1)
>>
>>If one subtracts power of two values, greater minus less:
>> 0100 0000
>> -0000 0100 the result is
>> =0011 1100 a "inner" sequence of ones, except "less" is included.
>>So we start with next greater bit of "less".
>>
>>Couldn't resist to use inlines instead of macros ;-)
>>(not tested!)
>>
>
>Gerd,
>
>How in the world can I see how you see how 0's and 1's interact
>and extract patterns. I'll first cut and paste and learn later.
>
>Thanks
>Rasjid
>
There is even a bit more to squeeze out of the functions. Eg. for 0x88
coordinates the difference between 0x88 coordinates maps to an unique direction
index.
But with 0x88, x+1 may leave the board, so isqBit(x+1) should consider that
fact, otherwise use (isqBit(x)<<1) or *2. On x86-32 i guess simple 32KByte
lookup to get the inbetween set is still faster.
Cheers,
Gerd
bool brq_path_clear(x88sqr gt, x88sqr ls, U64 path) {
return ((isqBit(gt) - 2*isqBit(ls)) & path & all) == 0;
}
bool bstyle_attack(x88sqr x, x88sqr y) {
switch (direction0x88Diff[0x88+x-y]) {
case leftDown:
return brq_path_clear(x, y, diagonal[1][iGetRank(x) + iGetFile(x)]);
case rightUp:
return brq_path_clear(y, x, diagonal[1][iGetRank(x) + iGetFile(x)]);
case leftUp:
return brq_path_clear(x, y, diagonal[0][7 + iGetRank(x) - iGetFile(x)]);
case rightDown:
return brq_path_clear(y, x, diagonal[0][7 + iGetRank(x) - iGetFile(x)]);
}
return false;
}
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.