>>> operator with c langage

[NaltaP312]

Hi,

i'm trying to code the bishopAttack and rookAttack.
to compute index with rotated bitboard i use this with java

Code: Select all

int idx_down = (int) (pieces_45 >>> DIA_DOWN_SHIFT[sq]) & 63;
int idx_up = (int) (pieces_315 >>> DIA_UP_SHIFT[sq]) & 63;

but in c, it is different.
i think i must divide by 2 my integer the number return by DIA_DOWN_SHIFT[sq].

Thanks for any help.
Best regards
Yves

[Gerd Isenberg]

Hi,

i'm trying to code the bishopAttack and rookAttack.
to compute index with rotated bitboard i use this with java

Code: Select all

int idx_down = (int) (pieces_45 >>> DIA_DOWN_SHIFT[sq]) & 63;
int idx_up = (int) (pieces_315 >>> DIA_UP_SHIFT[sq]) & 63;

but in c, it is different.
i think i must divide by 2 my integer the number return by DIA_DOWN_SHIFT[sq].

Thanks for any help.
Best regards
Yves

While all scalar data types (char, short, int, long) in Java are signed, C distinguish signed and unsigned data types. In order to shift right with zeros from left, like Java >>>, you need to declare (or cast) the data to shift unsigned and use the same operator >> as for signed shifts, where the most significant bit (sign bit) is shifted from left.
Typically one uses a typedefinition:

Code: Select all

typedef unsigned long long  BitBoard;

or

Code: Select all

typedef unsigned long long  U64;

[Desperado]

i dont know java, so i justed googled the _>>>_ operator.
So is there a reason why you dont use _>>_ operator in java.
this would also be the same in c.

_div by 2_ you can reach with c using this: _value >> 1_

[Desperado]

uhps, gerd was too fast for me

he already answered what i want to know,before i was able
to post the question...

[NaltaP312]

Hello Gerd,

yes i have defined like this
typedef unsigned long long u64;

so if i have well defined i think my bug is in another place
i continue to debug.

Best regards
Yves

[bob]

Hi,

i'm trying to code the bishopAttack and rookAttack.
to compute index with rotated bitboard i use this with java

Code: Select all

int idx_down = (int) (pieces_45 >>> DIA_DOWN_SHIFT[sq]) & 63;
int idx_up = (int) (pieces_315 >>> DIA_UP_SHIFT[sq]) & 63;

but in c, it is different.
i think i must divide by 2 my integer the number return by DIA_DOWN_SHIFT[sq].

Thanks for any help.
Best regards
Yves

>>> is not a shift operator. I suspect you want ">>".

2 >> 1 turns into 1, for example.

Make certain you know what you are doing when you use signed vs unsigned.

Unsigned fills from the left with zeroes, while signed fills from the left using the sign bit itself.

[NaltaP312]

Hello Robert,

Thanks for your reply.
what i want is to use the following:
to have a rotation right filled with zeros.

Shifts bits to the right (divide by 2 for each shift). Zeros emerging on the right are lost, while zeros are inserted to the left.

At this time i used unsigned for u64.
i suspect to have a bug in my rookAttack and bishopAttack.

Code: Select all

u64 bishopAttacks(int sq) {
	int idx_down = (int) (pieces_45 >> DIA_DOWN_SHIFT[sq]) & 63;
	int idx_up = (int) (pieces_315 >> DIA_UP_SHIFT[sq]) & 63;
	return	DIA_UP_SLIDES&#91;&#40;sq<<6&#41; | idx_up&#93; | DIA_DOWN_SLIDES&#91;&#40;sq<<6&#41; | idx_down&#93;;
&#125;

/* Returns rook-style attacks bitboard from a given square.*/
u64 rookAttacks&#40;int sq&#41; &#123;
	int idx_0 = &#40;int&#41; &#40;pieces >> (&#40;sq & 0x38&#41; | 1&#41;) & 63;
	int idx_90 = &#40;int&#41; &#40;pieces_90 >> FILE_SHIFT&#91;sq&#93;) & 63;
	return FILE_SLIDES&#91;&#40;sq << 6&#41; | idx_90&#93; | RANK_SLIDES&#91;&#40;sq << 6&#41; | idx_0&#93;;
&#125;

with this position
R2KR2R/PP1Q2NP/2PPnN2/3P1B2/3p4/2N4q/ppp2ppp/r3rk2 w - - 0 1

my result is the following for level one depth.
1 : The white king on d8 move to c8

2 : The white pawn on d5 capture the black knight at e6

3 : The white queen on d7 capture the black knight at e6

4 : The white knight on g7 capture the black knight at e6

5 : The white rook on a8 capture the black knight at e6

6 : The white rook on e8 capture the black knight at e6

so a bug for rook and a bug for king move.
but i will continue to debug and see if my precomputed table are not wrong.

Thanks
Best Regards
Yves

[Sven]

Code: Select all

/* Returns rook-style attacks bitboard from a given square.*/
u64 rookAttacks(int sq) {
	int idx_0 = (int) (pieces >> ((sq & 0x38) | 1)) & 63;
	int idx_90 = (int) (pieces_90 >> FILE_SHIFT[sq]) & 63;
	return FILE_SLIDES&#91;&#40;sq << 6&#41; | idx_90&#93; | RANK_SLIDES&#91;&#40;sq << 6&#41; | idx_0&#93;;
&#125;

In the expression "pieces >> ((sq & 0x38) | 1)", you are shifting by either 1, 9, 49, or 57, depending on which bits are set in 'sq'. Is this your intent?

Sven

[Sven]

Code: Select all

/* Returns rook-style attacks bitboard from a given square.*/
u64 rookAttacks(int sq) {
	int idx_0 = (int) (pieces >> ((sq & 0x38) | 1)) & 63;
	int idx_90 = (int) (pieces_90 >> FILE_SHIFT[sq]) & 63;
	return FILE_SLIDES&#91;&#40;sq << 6&#41; | idx_90&#93; | RANK_SLIDES&#91;&#40;sq << 6&#41; | idx_0&#93;;
&#125;

In the expression "pieces >> ((sq & 0x38) | 1)", you are shifting by either 1, 9, 49, or 57, depending on which bits are set in 'sq'. Is this your intent?

Sven

Sorry, I meant:
"you are shifting by either 1, 9, 17, 25, 33, 41, 49, or 57 ..."

(no EDIT allowed anymore )

Sven

[Gerd Isenberg]

Code: Select all

/* Returns rook-style attacks bitboard from a given square.*/
u64 rookAttacks(int sq) {
	int idx_0 = (int) (pieces >> ((sq & 0x38) | 1)) & 63;
	int idx_90 = (int) (pieces_90 >> FILE_SHIFT[sq]) & 63;
	return FILE_SLIDES&#91;&#40;sq << 6&#41; | idx_90&#93; | RANK_SLIDES&#91;&#40;sq << 6&#41; | idx_0&#93;;
&#125;

In the expression "pieces >> ((sq & 0x38) | 1)", you are shifting by either 1, 9, 49, or 57, depending on which bits are set in 'sq'. Is this your intent?

Sven

Sorry, I meant:
"you are shifting by either 1, 9, 17, 25, 33, 41, 49, or 57 ..."

(no EDIT allowed anymore )

Sven

Seems correct to me, to extract the six inner bits of each rank. In First Rank Attacks I shift by 0,8,16,..56 to extract 2*innersixbits in place by and with 2*63, and use a one-dimensional array to mul the already doubled index by 4 rather than 8:

Code: Select all

BYTE arrFirstRankAttacks64x8[64*8]; // 512 Bytes = 1/2KByte
 
U64 rankAttacks(U64 occ, enumSuare sq) {
   unsigned int file = sq &  7;
   unsigned int rkx8 = sq & 56; // rank * 8
   occ = (occ >> rkx8) & 2*63;
   U64 attacks = arrFirstRankAttacks64x8[4*occ + file];
   return attacks << rkx8;
&#125;