Rev 108 | Go to most recent revision | Details | Last modification | View Log | RSS feed
Rev | Author | Line No. | Line |
---|---|---|---|
33 | pmbaty | 1 | #include "chess.h" |
2 | #include "data.h" |
||
3 | /* last modified 02/22/14 */ |
||
4 | /* |
||
5 | ******************************************************************************* |
||
6 | * * |
||
7 | * HashProbe() is used to retrieve entries from the transposition table so * |
||
8 | * this sub-tree won't have to be searched again if we reach a position that * |
||
9 | * has been searched previously. A transposition table position contains * |
||
10 | * the following data packed into 128 bits with each item taking the number * |
||
11 | * of bits given in the table below: * |
||
12 | * * |
||
13 | * shr bits name description * |
||
14 | * 55 9 age search id to identify old trans/ref entries. * |
||
15 | * 53 2 type 0->value is worthless; 1-> value represents a * |
||
16 | * fail-low bound; 2-> value represents a fail-high * |
||
17 | * bound; 3-> value is an exact score. * |
||
18 | * 32 21 move best move from the current position, according to * |
||
19 | * the search at the time this position was stored. * |
||
20 | * 17 15 draft the depth of the search below this position, which * |
||
21 | * is used to see if we can use this entry at the * |
||
22 | * current position. * |
||
23 | * 0 17 value unsigned integer value of this position + 65536. * |
||
24 | * this might be a good score or search bound. * |
||
25 | * 0 64 key 64 bit hash signature, used to verify that this * |
||
26 | * entry goes with the current board position. * |
||
27 | * * |
||
28 | * The underlying scheme here is that we use a "bucket" of N entries. In * |
||
29 | * HashProbe() we simply compare against each of the four entries for a * |
||
30 | * match. Each "bucket" is carefully aligned to a 64-byte boundary so that * |
||
31 | * the bucket fits into a single cache line for efficiency. The bucket size * |
||
32 | * (N) is currently set to 4. * |
||
33 | * * |
||
34 | * Crafty uses the lockless hashing approach to avoid lock overhead in the * |
||
35 | * hash table accessing (reading or writing). What we do is store the key * |
||
36 | * and the information in two successive writes to memory. But since there * |
||
37 | * is nothing that prevents another CPU from interlacing its writes with * |
||
38 | * ours, we want to make sure that the bound/draft/etc really goes with the * |
||
39 | * key. Consider thread 1 trying to store A1 and A2 in two successive 64 * |
||
40 | * words, while thread 2 is trying to store B1 and B2. Since the two cpus * |
||
41 | * are fully independent, we could end up with {A1,A2}, {A1,B2}, {B1,A2} or * |
||
42 | * {B1,B2}. The two cases with one word of entry A and one word of entry B * |
||
43 | * are problematic since the information part does not belong with the * |
||
44 | * signature part, and a hash hit (signature match) will retrieve data that * |
||
45 | * does not match the position. Let's assume that the first word is the * |
||
46 | * signature (A1 or B1) and the second word is the data (A2 or B2). What we * |
||
47 | * do is store A1^A2 (exclusive-or the two parts) in the 1 (key) slot of the * |
||
48 | * entry, and store A2 in the data part. Now, before we try to compare the * |
||
49 | * signatures, we have to "un-corrupt" the stored signature by again using * |
||
50 | * xor, since A1^A2^A2 gives us the original A1 signature again. But if we * |
||
51 | * store A1^A2, and the data part gets replaced by B2, then we try to match * |
||
52 | * against A1^A2^B2 and that won't match unless we are lucky and A2 == B2 * |
||
53 | * which means the match is OK anyway. This eliminates the need to lock the * |
||
54 | * hash table while storing the two values, which would be a big performance * |
||
55 | * hit since hash entries are probed/stored in almost every node of the tree * |
||
56 | * except for the quiescence search. * |
||
57 | * * |
||
58 | ******************************************************************************* |
||
59 | */ |
||
60 | int HashProbe(TREE * RESTRICT tree, int ply, int depth, int side, int alpha, |
||
61 | int beta, int *value) { |
||
62 | HASH_ENTRY *htable; |
||
63 | HPATH_ENTRY *ptable; |
||
64 | uint64_t word1, word2, temp_hashkey; |
||
65 | int type, draft, avoid_null = 0, val, entry, i, j; |
||
66 | |||
67 | /* |
||
68 | ************************************************************ |
||
69 | * * |
||
70 | * All we have to do is loop through four entries to see * |
||
71 | * if there is a signature match. There can only be one * |
||
72 | * instance of any single signature, so the first match is * |
||
73 | * all we need. * |
||
74 | * * |
||
75 | ************************************************************ |
||
76 | */ |
||
77 | tree->hash_move[ply] = 0; |
||
78 | temp_hashkey = (side) ? HashKey : ~HashKey; |
||
79 | htable = trans_ref + (temp_hashkey & hash_mask); |
||
80 | for (entry = 0; entry < 4; entry++, htable++) { |
||
81 | word1 = htable->word1; |
||
82 | word2 = htable->word2 ^ word1; |
||
83 | if (word2 == temp_hashkey) |
||
84 | break; |
||
85 | } |
||
86 | /* |
||
87 | ************************************************************ |
||
88 | * * |
||
89 | * If we found a match, we have to verify that the draft * |
||
90 | * is at least equal to the current depth, if not higher, * |
||
91 | * and that the bound/score will let us terminate the * |
||
92 | * search early. * |
||
93 | * * |
||
94 | * We also return an "avoid_null" status if the matched * |
||
95 | * entry does not have enough draft to terminate the * |
||
96 | * current search but does have enough draft to prove that * |
||
97 | * a null-move search would not fail high. This avoids * |
||
98 | * the null-move search overhead in positions where it is * |
||
99 | * simply a waste of time to try it. * |
||
100 | * * |
||
101 | * If this is an EXACT entry, we are going to store the PV * |
||
102 | * in a safe place so that if we get a hit on this entry, * |
||
103 | * we can recover the PV and see the complete path rather * |
||
104 | * rather than one that is incomplete. * |
||
105 | * * |
||
106 | * One other issue is to update the age field if we get a * |
||
107 | * hit on an old position, so that it won't be replaced * |
||
108 | * just because it came from a previous search. * |
||
109 | * * |
||
110 | ************************************************************ |
||
111 | */ |
||
112 | if (entry < 4) { |
||
113 | if (word1 >> 55 != transposition_age) { |
||
114 | word1 = |
||
115 | (word1 & 0x007fffffffffffffull) | ((uint64_t) transposition_age << |
||
116 | 55); |
||
117 | htable->word1 = word1; |
||
118 | htable->word2 = word1 ^ word2; |
||
119 | } |
||
120 | val = (word1 & 0x1ffff) - 65536; |
||
121 | draft = (word1 >> 17) & 0x7fff; |
||
122 | tree->hash_move[ply] = (word1 >> 32) & 0x1fffff; |
||
123 | type = (word1 >> 53) & 3; |
||
124 | if ((type & UPPER) && depth - null_depth - 1 <= draft && val < beta) |
||
125 | avoid_null = AVOID_NULL_MOVE; |
||
126 | if (depth <= draft) { |
||
127 | if (val > 32000) |
||
128 | val -= ply - 1; |
||
129 | else if (val < -32000) |
||
130 | val += ply - 1; |
||
131 | *value = val; |
||
132 | /* |
||
133 | ************************************************************ |
||
134 | * * |
||
135 | * We have three types of results. An EXACT entry was * |
||
136 | * stored when val > alpha and val < beta, and represents * |
||
137 | * an exact score. An UPPER entry was stored when val < * |
||
138 | * alpha, which represents an upper bound with the score * |
||
139 | * likely being even lower. A LOWER entry was stored when * |
||
140 | * val > beta, which represents alower bound with the * |
||
141 | * score likely being even higher. * |
||
142 | * * |
||
143 | * For EXACT entries, we save the path from the position * |
||
144 | * to the terminal node that produced the backed-up score * |
||
145 | * so that we can complete the PV if we get a hash hit on * |
||
146 | * this entry. * |
||
147 | * * |
||
148 | ************************************************************ |
||
149 | */ |
||
150 | switch (type) { |
||
151 | case EXACT: |
||
152 | if (val > alpha && val < beta) { |
||
153 | SavePV(tree, ply, 1 + (draft == MAX_DRAFT)); |
||
154 | ptable = hash_path + (temp_hashkey & hash_path_mask); |
||
155 | for (i = 0; i < 16; i++, ptable++) |
||
156 | if (ptable->path_sig == temp_hashkey) { |
||
157 | for (j = ply; j < Min(MAXPLY - 1, ptable->hash_pathl + ply); |
||
158 | j++) |
||
159 | tree->pv[ply - 1].path[j] = |
||
160 | ptable->hash_path_moves[j - ply]; |
||
161 | if (draft != MAX_DRAFT && |
||
162 | ptable->hash_pathl + ply < MAXPLY - 1) |
||
163 | tree->pv[ply - 1].pathh = 0; |
||
164 | tree->pv[ply - 1].pathl = |
||
165 | Min(MAXPLY - 1, ply + ptable->hash_pathl); |
||
166 | ptable->hash_path_age = transposition_age; |
||
167 | break; |
||
168 | } |
||
169 | } |
||
170 | return HASH_HIT; |
||
171 | case UPPER: |
||
172 | if (val <= alpha) |
||
173 | return HASH_HIT; |
||
174 | break; |
||
175 | case LOWER: |
||
176 | if (val >= beta) |
||
177 | return HASH_HIT; |
||
178 | break; |
||
179 | } |
||
180 | } |
||
181 | return avoid_null; |
||
182 | } |
||
183 | return HASH_MISS; |
||
184 | } |
||
185 | |||
186 | /* last modified 02/22/14 */ |
||
187 | /* |
||
188 | ******************************************************************************* |
||
189 | * * |
||
190 | * HashStore() is used to store entries into the transposition table so that * |
||
191 | * this sub-tree won't have to be searched again if the same position is * |
||
192 | * reached. We basically store three types of entries: * |
||
193 | * * |
||
194 | * (1) EXACT. This entry is stored when we complete a search at some ply * |
||
195 | * and end up with a score that is greater than alpha and less than * |
||
196 | * beta, which is an exact score, which also has a best move to try if * |
||
197 | * we encounter this position again. * |
||
198 | * * |
||
199 | * (2) LOWER. This entry is stored when we complete a search at some ply * |
||
200 | * and end up with a score that is greater than or equal to beta. We * |
||
201 | * know know that this score should be at least equal to beta and may * |
||
202 | * well be even higher. So this entry represents a lower bound on the * |
||
203 | * score for this node, and we also have a good move to try since it * |
||
204 | * caused the cutoff, although we do not know if it is the best move or * |
||
205 | * not since not all moves were search. * |
||
206 | * * |
||
207 | * (3) UPPER. This entry is stored when we complete a search at some ply * |
||
208 | * and end up with a score that is less than or equal to alpha. We * |
||
209 | * know know that this score should be at least equal to alpha and may * |
||
210 | * well be even lower. So this entry represents an upper bound on the * |
||
211 | * score for this node. We have no idea about which move is best in * |
||
212 | * this position since they all failed low, so we store a best move of * |
||
213 | * zero. * |
||
214 | * * |
||
215 | * For storing, we may require three passes. We make our first pass looking * |
||
216 | * for an entry that matches the current hash signature. If we find a match * |
||
217 | * then we are constrained to overwrite that entry regardless of any other * |
||
218 | * considerations. The second pass looks for entries stored in previous * |
||
219 | * searches (not iterations) and chooses the one with the shallowest draft, * |
||
220 | * if one is found; Otherwise we make a final pass over the bucket and * |
||
221 | * choose the entry with the shallowest draft, period. * |
||
222 | * * |
||
223 | ******************************************************************************* |
||
224 | */ |
||
225 | void HashStore(TREE * RESTRICT tree, int ply, int depth, int side, int type, |
||
226 | int value, int bestmove) { |
||
227 | HASH_ENTRY *htable, *replace = 0; |
||
228 | HPATH_ENTRY *ptable; |
||
229 | uint64_t word1, temp_hashkey; |
||
230 | int entry, draft, age, replace_draft, i, j; |
||
231 | |||
232 | /* |
||
233 | ************************************************************ |
||
234 | * * |
||
235 | * "Fill in the blank" and build a table entry from * |
||
236 | * current search information. * |
||
237 | * * |
||
238 | ************************************************************ |
||
239 | */ |
||
240 | word1 = transposition_age; |
||
241 | word1 = (word1 << 2) | type; |
||
242 | if (value > 32000) |
||
243 | value += ply - 1; |
||
244 | else if (value < -32000) |
||
245 | value -= ply - 1; |
||
246 | word1 = (word1 << 21) | bestmove; |
||
247 | word1 = (word1 << 15) | depth; |
||
248 | word1 = (word1 << 17) | (value + 65536); |
||
249 | temp_hashkey = (side) ? HashKey : ~HashKey; |
||
250 | /* |
||
251 | ************************************************************ |
||
252 | * * |
||
253 | * Now we search for an entry to overwrite in three * |
||
254 | * passes. * |
||
255 | * * |
||
256 | * Pass 1: If any signature in the table matches the * |
||
257 | * current signature, we are going to overwrite this * |
||
258 | * entry, period. It might seem worthwhile to check the * |
||
259 | * draft and not overwrite if the table draft is greater * |
||
260 | * than the current remaining depth, but after you think * |
||
261 | * about it, this is a bad idea. If the draft is * |
||
262 | * greater than or equal the current remaining depth, * |
||
263 | * then we should never get here unless the stored bound * |
||
264 | * or score is unusable because of the current alpha/ * |
||
265 | * beta window. So we are overwriting to avoid losing * |
||
266 | * the current result. * |
||
267 | * * |
||
268 | * Pass 2: If any of the entries come from a previous * |
||
269 | * search (not iteration) then we choose the entry from * |
||
270 | * this set that has the smallest draft, since it is the * |
||
271 | * least potentially usable result. * |
||
272 | * * |
||
273 | * Pass 3: If neither of the above two found an entry to * |
||
274 | * overwrite, we simply choose the entry from the bucket * |
||
275 | * with the smallest draft and overwrite that. * |
||
276 | * * |
||
277 | ************************************************************ |
||
278 | */ |
||
279 | htable = trans_ref + (temp_hashkey & hash_mask); |
||
280 | for (entry = 0; entry < 4; entry++, htable++) { |
||
281 | if (temp_hashkey == (htable->word1 ^ htable->word2)) { |
||
282 | replace = htable; |
||
283 | break; |
||
284 | } |
||
285 | } |
||
286 | if (!replace) { |
||
287 | replace_draft = 99999; |
||
288 | htable = trans_ref + (temp_hashkey & hash_mask); |
||
289 | for (entry = 0; entry < 4; entry++, htable++) { |
||
290 | age = htable->word1 >> 55; |
||
291 | draft = (htable->word1 >> 17) & 0x7fff; |
||
292 | if (age != transposition_age && replace_draft > draft) { |
||
293 | replace = htable; |
||
294 | replace_draft = draft; |
||
295 | } |
||
296 | } |
||
297 | if (!replace) { |
||
298 | htable = trans_ref + (temp_hashkey & hash_mask); |
||
299 | for (entry = 0; entry < 4; entry++, htable++) { |
||
300 | draft = (htable->word1 >> 17) & 0x7fff; |
||
301 | if (replace_draft > draft) { |
||
302 | replace = htable; |
||
303 | replace_draft = draft; |
||
304 | } |
||
305 | } |
||
306 | } |
||
307 | } |
||
308 | /* |
||
309 | ************************************************************ |
||
310 | * * |
||
311 | * Now that we know which entry to replace, we simply * |
||
312 | * stuff the values and exit. Note that the two 64 bit * |
||
313 | * words are xor'ed together and stored as the signature * |
||
314 | * for the "lockless-hash" approach. * |
||
315 | * * |
||
316 | ************************************************************ |
||
317 | */ |
||
318 | replace->word1 = word1; |
||
319 | replace->word2 = temp_hashkey ^ word1; |
||
320 | /* |
||
321 | ************************************************************ |
||
322 | * * |
||
323 | * If this is an EXACT entry, we are going to store the PV * |
||
324 | * in a safe place so that if we get a hit on this entry, * |
||
325 | * we can recover the PV and see the complete path rather * |
||
326 | * rather than one that is incomplete. * |
||
327 | * * |
||
328 | ************************************************************ |
||
329 | */ |
||
330 | if (type == EXACT) { |
||
331 | ptable = hash_path + (temp_hashkey & hash_path_mask); |
||
332 | for (i = 0; i < 16; i++, ptable++) { |
||
333 | if (ptable->path_sig == temp_hashkey || |
||
334 | ((transposition_age - ptable->hash_path_age) > 1)) { |
||
335 | for (j = ply; j < tree->pv[ply - 1].pathl; j++) |
||
336 | ptable->hash_path_moves[j - ply] = tree->pv[ply - 1].path[j]; |
||
337 | ptable->hash_pathl = tree->pv[ply - 1].pathl - ply; |
||
338 | ptable->path_sig = temp_hashkey; |
||
339 | ptable->hash_path_age = transposition_age; |
||
340 | break; |
||
341 | } |
||
342 | } |
||
343 | } |
||
344 | } |
||
345 | |||
346 | /* last modified 02/22/14 */ |
||
347 | /* |
||
348 | ******************************************************************************* |
||
349 | * * |
||
350 | * HashStorePV() is called by Iterate() to insert the PV moves so they will * |
||
351 | * be searched before any other moves. Normally the PV moves would be in * |
||
352 | * the table, but on occasion they can be overwritten, particularly the ones * |
||
353 | * that are a significant distance from the root since those table entries * |
||
354 | * will have a low draft. * |
||
355 | * * |
||
356 | ******************************************************************************* |
||
357 | */ |
||
358 | void HashStorePV(TREE * RESTRICT tree, int side, int ply) { |
||
359 | HASH_ENTRY *htable, *replace; |
||
360 | uint64_t temp_hashkey, word1; |
||
361 | int entry, draft, replace_draft, age; |
||
362 | |||
363 | /* |
||
364 | ************************************************************ |
||
365 | * * |
||
366 | * First, compute the initial hash address and the fake * |
||
367 | * entry we will store if we don't find a valid match * |
||
368 | * already in the table. * |
||
369 | * * |
||
370 | ************************************************************ |
||
371 | */ |
||
372 | temp_hashkey = (side) ? HashKey : ~HashKey; |
||
373 | word1 = transposition_age; |
||
374 | word1 = (word1 << 2) | WORTHLESS; |
||
375 | word1 = (word1 << 21) | tree->pv[0].path[ply]; |
||
376 | word1 = (word1 << 32) | 65536; |
||
377 | /* |
||
378 | ************************************************************ |
||
379 | * * |
||
380 | * Now we search for an entry to overwrite in three * |
||
381 | * passes. * |
||
382 | * * |
||
383 | * Pass 1: If any signature in the table matches the * |
||
384 | * current signature, we are going to overwrite this * |
||
385 | * entry, period. It might seem worthwhile to check the * |
||
386 | * draft and not overwrite if the table draft is greater * |
||
387 | * than the current remaining depth, but after you think * |
||
388 | * about it, this is a bad idea. If the draft is * |
||
389 | * greater than or equal the current remaining depth, * |
||
390 | * then we should never get here unless the stored bound * |
||
391 | * or score is unusable because of the current alpha/ * |
||
392 | * beta window. So we are overwriting to avoid losing * |
||
393 | * the current result. * |
||
394 | * * |
||
395 | * Pass 2: If any of the entries come from a previous * |
||
396 | * search (not iteration) then we choose the entry from * |
||
397 | * this set that has the smallest draft, since it is the * |
||
398 | * least potentially usable result. * |
||
399 | * * |
||
400 | * Pass 3: If neither of the above two found an entry to * |
||
401 | * overwrite, we simply choose the entry from the bucket * |
||
402 | * with the smallest draft and overwrite that. * |
||
403 | * * |
||
404 | ************************************************************ |
||
405 | */ |
||
406 | htable = trans_ref + (temp_hashkey & hash_mask); |
||
407 | for (entry = 0; entry < 4; entry++, htable++) { |
||
408 | if ((htable->word2 ^ htable->word1) == temp_hashkey) { |
||
409 | htable->word1 &= ~((uint64_t) 0x1fffff << 32); |
||
410 | htable->word1 |= (uint64_t) tree->pv[0].path[ply] << 32; |
||
411 | htable->word2 = temp_hashkey ^ htable->word1; |
||
412 | break; |
||
413 | } |
||
414 | } |
||
415 | if (entry == 4) { |
||
416 | htable = trans_ref + (temp_hashkey & hash_mask); |
||
417 | replace = 0; |
||
418 | replace_draft = 99999; |
||
419 | for (entry = 0; entry < 4; entry++, htable++) { |
||
420 | age = htable->word1 >> 55; |
||
421 | draft = (htable->word1 >> 17) & 0x7fff; |
||
422 | if (age != transposition_age && replace_draft > draft) { |
||
423 | replace = htable; |
||
424 | replace_draft = draft; |
||
425 | } |
||
426 | } |
||
427 | if (!replace) { |
||
428 | htable = trans_ref + (temp_hashkey & hash_mask); |
||
429 | for (entry = 0; entry < 4; entry++, htable++) { |
||
430 | draft = (htable->word1 >> 17) & 0x7fff; |
||
431 | if (replace_draft > draft) { |
||
432 | replace = htable; |
||
433 | replace_draft = draft; |
||
434 | } |
||
435 | } |
||
436 | } |
||
437 | replace->word1 = word1; |
||
438 | replace->word2 = temp_hashkey ^ word1; |
||
439 | } |
||
440 | } |