Marco Costalba
cf2bafb661
Relax time constraints
...
Allow a seacrh to take a bit more time if needed.
This reduces the chanches of wast all the search time
for the last iteration and also allow to start the last
iteration when we have less time remaining.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-11-01 12:46:53 +01:00
Marco Costalba
79513e3a43
Material: micro optimize map reading
...
Do only one map walk per read instead of two.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-11-01 12:46:52 +01:00
Marco Costalba
bd3ec6af15
Material: protect global map access with locks
...
STL library is not guaranteed to be thread safe, even for
read-access.
So because these global maps are accessed by all the threads
we need to protect them.
This fixes a random crash experienced during testing.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-11-01 12:46:52 +01:00
Marco Costalba
b2b86cfd27
MovePicker::score_captures() order with SEE when pv
...
Order PV nodes by SEE instead of MVV/LVA.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-11-01 12:46:52 +01:00
Marco Costalba
7d717df4e4
Revert "Last minute surprise" for now
...
It defenitly needs more testing. Just postponed.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-11-01 12:46:52 +01:00
Marco Costalba
5e906ea10e
Finish material.cpp cleanup
...
Hopefully no regression this time!
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-11-01 12:46:52 +01:00
Marco Costalba
73cce873de
Start to simplify material.cpp
...
It is posisble to simplify a lot here!
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-11-01 12:46:51 +01:00
Marco Costalba
038235ba35
Factor out Position::do_capture_move()
...
Start to slim line count i position.cpp
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-11-01 12:46:51 +01:00
Marco Costalba
5e2fc2aa16
Fix a bug in generate_evasions()
...
Introduced in the patch "movegen: prefer (*mlist++) to mlist[n++]"
This was nasty because due to a mismerge the repo in one PC had the bug,
but the testing one did not, so I had non reproducible results according
to which machine I used for testing.
This hopefully closes a more then one week regression that made me go crazy!
It was found by accident comparing, for other reasons, the sources of the
two PC's.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-11-01 12:46:51 +01:00
Marco Costalba
74fd57220e
Refine "few moves" in "last minute surprise"
...
It seems that "few moves" works because we extend the good
captures at the last ply of PV, so code it directly.
This version seems defenitly stronger then previous one.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-11-01 12:46:51 +01:00
Marco Costalba
74f1efee26
Manual merge
2008-10-26 21:44:58 +01:00
Marco Costalba
0936a96a0e
Space inflate extension() code
...
No functional change.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-26 21:30:21 +01:00
Marco Costalba
13d1776a98
TEST: extend when few moves available
...
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-26 21:30:21 +01:00
Marco Costalba
a56e1c662c
Revert see shortcut when only one attacker
...
It does not seem to work after a little testing.
Perhaps it works on the long terms, but it is also
ugly because not correct, so revert for now.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-26 21:30:21 +01:00
Marco Costalba
fcc49561a1
Prefer out of loop variables in MovePicker::score_captures()
...
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-26 21:30:21 +01:00
Marco Costalba
714069e248
Effectively use MVV/LVA in MovePicker::score_captures()
...
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-26 21:30:20 +01:00
Marco Costalba
88885399f4
Reintroduce piece/square tables to score non-captures
...
Was removed after original movepick restore. But proved
to be useful.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-26 21:30:20 +01:00
Marco Costalba
a5c1b3e8f6
Position: fix a couple of Intel compiler warnings
...
Plus usual trailing whitespace.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-26 13:23:27 +01:00
Marco Costalba
8a85007023
Test with SEE shortcut
...
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-26 10:11:13 +01:00
Marco Costalba
4397e6c03e
Better naming of pseudo-legality and legality testing
...
No functional change.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-26 10:11:13 +01:00
Marco Costalba
4f14bd5032
Try to cleanup movepick
...
Hopefully without regressions.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-26 10:11:13 +01:00
Marco Costalba
f8bc38a006
Restore original movepick modulo space inflation
...
We have a regression somewhere here so restart from zero
and proceed one change at a time.
With this modification we have the same strenght of
"Introduce Stockfish" patch that is our strongest to date.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-26 10:11:13 +01:00
Marco Costalba
5dd9159106
Space inflate position: complete!
...
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-26 10:10:56 +01:00
Marco Costalba
ad956ef00a
Space inflate position until do_promotion_move()
...
We will end some day ;-)
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-24 21:10:05 +02:00
Marco Costalba
d155cd88d1
Start to space inflate position.cpp
...
It's a big file!
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-24 21:10:05 +02:00
Marco Costalba
2aebf8eb55
Fix a performance bug in generate_move_if_legal
...
Use the pinned argument in pos.move_is_legal()
No functional change, simply use pos.move_is_legal() as
was meant to be.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-24 21:10:05 +02:00
Marco Costalba
1ac2f50145
Unify pinned and discovery checks code
...
Templates are our friends here. No functional change.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-24 21:10:04 +02:00
Marco Costalba
af59cb1d63
Refactor Position::pinned_pieces() to use templates
...
Also better document this interesting function.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-24 21:10:04 +02:00
Marco Costalba
2f8961beef
movegen: add SERIALIZE_MOVES and hides a bunch of loops
...
Only syntactic sugar, perhaps we should leave as is, anyhow...
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-24 21:10:04 +02:00
Marco Costalba
af5743837d
Another generate_piece_moves() micro optimization
...
This time on the for loop.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-24 21:10:04 +02:00
Marco Costalba
5e768e4b0a
Fix another template conversion bug in movegen
...
Hopefully the last one.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-24 21:09:51 +02:00
Marco Costalba
2d867109d9
movegen: prefer (*mlist++) to mlist[n++]
...
Teoretically faster, practically it helps to
removes some more lines.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-23 07:17:04 +02:00
Marco Costalba
b145e99559
Fix a bug in generate_pawn_captures()
...
Introduced in "movegen: Introduce generate_pawn_captures()"
when unifiying black and white functions.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-23 07:17:04 +02:00
Marco Costalba
8f2c1c59eb
Fully templetize pawn move generators
...
A little bit more syntax heavuer but surely faster.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-23 07:17:04 +02:00
Marco Costalba
c40249e9d2
Micro optimize generate_piece_moves()
...
It is a time critical path. The biggest in move generation.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-21 00:10:27 +02:00
Marco Costalba
ce93a202b5
Another Intel warning sqeezed
...
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:57:13 +02:00
Marco Costalba
279ed7ed48
Quiet a warning on Intel compiler
...
Plus usual trailing whitespace noise.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:55:43 +02:00
Marco Costalba
af6571856e
Shrink arguments in move generation functions
...
Perhaps no speedup, but it is more readable.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:47:21 +02:00
Marco Costalba
0c8659721f
Fix a bug in king discoveries checks
...
Introduced in "Add a generate_piece_checks() specialization for the king"
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:47:20 +02:00
Marco Costalba
5dc2312121
Update copyright info
...
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:47:20 +02:00
Marco Costalba
aa94f2f4c2
Last little touches to movegen
...
No functional change.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:47:19 +02:00
Marco Costalba
d7161c1ce6
Rename PawnOffsets in PawnParams
...
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:47:19 +02:00
Marco Costalba
e2af0e775b
Pawn move generator: dispatch at compile time
...
Instead of function pointers use templates to
dispatch shift operations.
It is more clear and possibly also faster because
branches are removed at compile time.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:47:19 +02:00
Marco Costalba
ad1bb084dd
Start to templetize pawn move generators
...
Still very soft, we will see if compiler is
enough or we need more aggressive templetization.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:47:18 +02:00
Marco Costalba
760f77872f
Remove Positions::xxx_count() functions
...
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:47:18 +02:00
Marco Costalba
257689dec7
Remove white/black_pawn_attacks_square()
...
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:47:18 +02:00
Marco Costalba
ff211469ba
Templetize Position::xxx_attacks_square()
...
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:47:17 +02:00
Marco Costalba
d4f14a8e83
Remove Position::xxx_list() functions
...
No useful, only obfuscating.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:47:17 +02:00
Marco Costalba
7eb290a509
Add a generate_piece_checks() specialization for the king
...
Also reshuffle the code a bit.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:47:17 +02:00
Marco Costalba
d316b02771
Remove white/black_pawn_attacks()
...
Unuseful syntactic sugar, obfuscates the
real code.
Signed-off-by: Marco Costalba <mcostalba@gmail.com>
2008-10-20 21:47:17 +02:00