@@ -21,11 +21,6 @@ using namespace NKikimr;
21
21
using namespace NMiniKQL ;
22
22
using namespace NUdf ;
23
23
24
- inline ui64 SpreadHash (ui64 hash) {
25
- // https://probablydance.com/2018/06/16/fibonacci-hashing-the-optimization-that-the-world-forgot-or-a-better-alternative-to-integer-modulo/
26
- return ((unsigned __int128)hash * 11400714819323198485llu) >> 64 ;
27
- }
28
-
29
24
30
25
class TDqOutputMultiConsumer : public IDqOutputConsumer {
31
26
public:
@@ -195,9 +190,6 @@ class TDqOutputHashPartitionConsumer : public IDqOutputConsumer {
195
190
hash = CombineHashes (hash, HashColumn (keyId, columnValue));
196
191
}
197
192
198
-
199
- hash = SpreadHash (hash);
200
-
201
193
return hash % Outputs.size ();
202
194
}
203
195
@@ -209,8 +201,6 @@ class TDqOutputHashPartitionConsumer : public IDqOutputConsumer {
209
201
hash = CombineHashes (hash, HashColumn (keyId, values[KeyColumns[keyId].Index ]));
210
202
}
211
203
212
- hash = SpreadHash (hash);
213
-
214
204
return hash % Outputs.size ();
215
205
}
216
206
@@ -314,8 +304,6 @@ class TDqOutputHashPartitionConsumerScalar : public IDqOutputConsumer {
314
304
hash = CombineHashes (hash, HashColumn (keyId, values[KeyColumns_[keyId].Index ]));
315
305
}
316
306
317
- hash = SpreadHash (hash);
318
-
319
307
return hash % Outputs_.size ();
320
308
}
321
309
@@ -513,9 +501,6 @@ class TDqOutputHashPartitionConsumerBlock : public IDqOutputConsumer {
513
501
}
514
502
hash = CombineHashes (hash, keyHash);
515
503
}
516
-
517
- hash = SpreadHash (hash);
518
-
519
504
return hash % Outputs_.size ();
520
505
}
521
506
0 commit comments