-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathligra.h
1477 lines (1368 loc) · 44.8 KB
/
ligra.h
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
// -*- C++ -*-
// This code is part of the project "Ligra: A Lightweight Graph Processing
// Framework for Shared Memory", presented at Principles and Practice of
// Parallel Programming, 2013.
// Copyright (c) 2013 Julian Shun and Guy Blelloch
//
// This code has been extended in the project "GraphGrind: Addressing Load
// Imbalance of Graph Partitioning", presented at International Symposium
// on Supercomputing, 2017.
// Copyright (c) 2017 Jiawen Sun, Hans Vandierendonck, Dimitrios S. Nikolopoulos
//
// Permission is hereby granted, free of charge, to any person obtaining a
// copy of this software and associated documentation files (the
// "Software"), to deal in the Software without restriction, including
// without limitation the rights (to use, copy, modify, merge, publish,
// distribute, sublicense, and/or sell copies of the Software, and to
// permit persons to whom the Software is furnished to do so, subject to
// the following conditions:
//
// The above copyright notice and this permission notice shall be included
// in all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
// OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
// MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
// NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
// LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
// OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
// WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
#include <iostream>
#include <fstream>
#include <stdlib.h>
#include <cstring>
#include <string>
#include <assert.h>
#include <algorithm>
#include <sys/mman.h>
#include "parallel.h"
#include "gettime.h"
#include "utils.h"
#include "graph.h"
#include "IO.h"
#include "parseCommandLine.h"
#include "gettime.h"
#include "papi_code.h"
#if REUSE_DIST
#include "reuse.h"
RDLogger<intT> * rd_bwd_src = 0; // old frontier
RDLogger<intT> * rd_bwd_data = 0; // old data
RDLogger<intT> * rd_bwd_next = 0; // next frontier
RDLogger<intT> * rd_fwd_dest = 0; // next frontier
RDLogger<intT> * rd_fwd_new = 0; // new data
#endif
//Timer design
using namespace std;
static double tm_edgemap_ = 0;
static double tm_edgemap_tosparse_alloc_ = 0;
static double tm_edgemap_tosparse_collect_ = 0;
static double tm_edgemap_tosparse_itself_ = 0;
static double tm_edgemap_tosparse_reduce_ = 0;
static double tm_edgemap_todense_ = 0;
static double tm_edgemap_dense_fwd_ = 0;
static double tm_edgemap_dense_ = 0;
static double tm_edgemap_dense_bwd_ = 0;
static double tm_edgemap_dense_sync_ = 0;
static double tm_edgemap_sparse_ = 0;
static double tm_edgemap_sparse_sync_ = 0;
static double dm_dense_each=0;
static double sm_sparse_each=0;
static double fm_forward_each=0;
static double tm_edgemap_setup_ = 0;
static double tm_edgemap_map_ = 0;
static double tm_edgemap_collect_ = 0;
double tmlog( timer & tm, double & cnt )
{
double val = tm.next();
cnt += val;
return val;
}
void timePrint()
{
std::cerr << "\ntotal edgemap setup: " << tm_edgemap_setup_
<< "\ntotal edgemap map(dense): " << tm_edgemap_dense_
<< "\ntotal edgemap map(sparse): " << tm_edgemap_sparse_
<< "\ntotal edgemap collect: " << tm_edgemap_collect_
<< std::endl;
}
void timeprint()
{
std::cerr << "edgemap: " << tm_edgemap_
<< " edgemap_tosparse_alloc: " << tm_edgemap_tosparse_alloc_ <<"\n"
<< " edgemap_tosparse_collect: " << tm_edgemap_tosparse_collect_ <<"\n"
<< " edgemap_tosparse_itself: " << tm_edgemap_tosparse_itself_ <<"\n"
<< " edgemap_tosparse_reduce: " << tm_edgemap_tosparse_reduce_ <<"\n"
<< " edgemap_todense: " << tm_edgemap_todense_ <<"\n"
<< " edgemap_dense_fwd: " << tm_edgemap_dense_fwd_ <<"\n"
<< " edgemap_dense_bwd: " << tm_edgemap_dense_bwd_ <<"\n"
<< " edgemap_dense_sync: " << tm_edgemap_dense_sync_ <<"\n"
<< " edgemap_sparse: " << tm_edgemap_sparse_ <<"\n"
<< " edgemap_sparse_sync_: " << tm_edgemap_sparse_sync_ <<"\n"
<<std::endl;
}
typedef pair<intT, intT> intTpair;
//#define STEP (2048)
#define STEP (4096)
//*****START FRAMEWORK*****
//Design for partitioned_graph (dense, sparse)
//Dense contains a whole boolean array for each partitioned graph
//to share and update
//Sparse uses the vertices structure, m is the active vertex number
// in each partitioned
//intT * s is the intT array for update the out-degree array
class partitioned_vertices
{
public:
intT numVertices;
intT d_m; // active vertices number for dense part
bool *d; // dense part, boolean array
bool has_sparse; // flag if sparse representation is present
bool has_dense; // flag if dense representation is present
intT *s; //sparse active vertices array
intT num_out_edges; //acitve vertices's out-degree for sparse/dense selection
//Create the initial partitioned frontier with one start vertex.
//Used for traversal algorithm, such as BFS,BC and BellmanFord
static partitioned_vertices create(intT n, intT v, intT initialOutdegree)
{
partitioned_vertices pv;
pv.numVertices=n;
pv.d = 0;
pv.d_m = 1;
pv.has_dense = false;
pv.has_sparse = false;
pv.s = new intT [1];
pv.s[0] = v;
pv.num_out_edges = initialOutdegree;
return pv;
}
//Empty frontier to stop the programming
static partitioned_vertices empty()
{
partitioned_vertices pv;
pv.d = 0;
pv.d_m = 0;
pv.numVertices=0;
pv.has_dense = false;
pv.has_sparse = false;
pv.s=NULL;
pv.num_out_edges = 0;
return pv;
}
//Create dense frontier for the dense part
static partitioned_vertices dense(intT n, const partitioner & part)
{
partitioned_vertices pv;
pv.numVertices=n;
pv.d = new bool [n];
pv.d_m = 0;
pv.has_dense = true;
pv.has_sparse = false;
pv.s = NULL;
pv.num_out_edges = 0;
return pv;
}
//Create sparse frontier for the sparse part
static partitioned_vertices sparse(intT n)
{
partitioned_vertices pv;
pv.numVertices=n;
pv.d = 0;
pv.d_m = 0;
pv.has_dense = false;
pv.has_sparse = true;
pv.s = NULL;
pv.num_out_edges = 0;
return pv;
}
//create of frontier with n from the boolean array as d[i]
//vertices(intT _n,intT _m, bool* bits)
//:n(_n),m(_m),s(NULL),isDense(1){}
//and create the frontier boolean array inside the bits()
//not to copy in each algorithm defined
//Use for Components, BC, BellmanFord and PR, PRDelta
static partitioned_vertices bits(const partitioner & part, intT n, intT initialOutdegree)
{
partitioned_vertices pv;
pv.num_out_edges = initialOutdegree;
pv.s = NULL;
pv.numVertices=n;
pv.d = new bool [n];
parallel_for(intT i=0; i<n; i++) pv.d[i]=1;
pv.d_m = n;
pv.has_dense = true;
pv.has_sparse = false;
return pv;
}
//Used for vertexFilter to create the array with number of vertices and boolean array
//This is using for the radii algorithm, and VertexFilter
static partitioned_vertices boolean(intT n, bool* bits,intT activeM, intT out_edges)
{
partitioned_vertices pv;
pv.numVertices=n;
pv.num_out_edges = out_edges;
pv.d_m = activeM;
pv.s = NULL;
pv.d = bits;
pv.has_dense = true;
pv.has_sparse = false;
return pv;
}
void del()
{
if(s!=NULL) delete [] s;
//if(s!=NULL) s.del();
if( d )
delete [] d;
//d.del();
}
//get the intT array for sparse iteration
intT* get_partition( unsigned p )
{
return s;
}
//check the empty boolean function for while loop
bool isEmpty()
{
return d_m == 0;
}
//Represent the frontier as the dense format (boolean dense array [0,1,0,1,0,1,1,1,.....]), used for the edgeMapDense function
void toDense()
{
// The dense pull operation requires to know for every vertex whether
// it is active. So we convert all sparse partitions. The dense push
// only needs to know for vertices within partitions traversed in
// dense format. In those cases we could avoid creating the dense
// representation, e.g., if the flag is set to use DENSE_FORWARD.
if (!has_dense)
{
intT n = numVertices;
if (!d) d = new bool [n];
{
parallel_for(intT i=0; i<n; i++) d[i] = 0;
}
{
parallel_for(intT i=0; i<d_m; i++) d[s[i]] = 1;
}
has_dense = true;
}
}
// Represent the frontier as the sparse format (only active vertex [1,5,8,..]), used for the edgeMapSparse Function
void toSparse()
{
// Convert the partition p to sparse. Only one of them!
if( has_dense && !has_sparse )
{
_seq<intT> R = sequence::packIndex(d,numVertices);
if (d_m != R.n)
{
cout<<"bad stored value of m"<<endl;
abort();
}
s=R.A;
d_m = R.n;
has_sparse = true;
has_dense = false;
}
}
// parttioned_vertices: numRows(); return n (all vertices number of frontier);
intT numRows()
{
return numVertices;
}
// partitioned_vertices: return non active vertices
intT numNonzeros()
{
return d_m;
}
};
//To collect all outdegree for partitioned graph
template<class vertex>
class GoutDegree
{
public:
partitioned_graph<vertex> &PG;
bool* dense;
GoutDegree(partitioned_graph<vertex>& pg, bool* pdense):PG(pg),dense(pdense) {}
pair<intT, intT> operator()(intT i)
{
return make_pair((intT)dense[i], dense[i]? PG.getFullOutDegree(i):(intT)0);
}
};
//To collect part outdegree for partitioned graph
template<class vertex>
class GoutDegreePart
{
public:
partitioned_graph<vertex> &PG;
bool* dense;
intT p;
GoutDegreePart(partitioned_graph<vertex>& pg, bool* pdense, intT p_)
: PG(pg), dense(pdense), p(p_) {}
pair<intT, intT> operator()(intT i)
{
bool inc = dense[i]; // && PG.get_partitioner().partition_of(i) == p;
return make_pair((intT)inc, inc ? PG.get_partition(p).V[i].getOutDegree():(intT)0);
}
};
//Get part out-degree of partitioned graph
template<class vertex>
class GoutDegreeV
{
public:
graph<vertex> G;
intT *s;
GoutDegreeV(graph<vertex> G_, intT* s_) : G(G_), s(s_) { }
intT operator()( intT i )
{
return G.V[s[i]].getOutDegree();
}
};
//Get whole out-degree of partitioned graph
template<class vertex>
class GoutDegreeVFull
{
public:
partitioned_graph<vertex> G;
intT *s;
GoutDegreeVFull(partitioned_graph<vertex> G_, intT* s_) : G(G_), s(s_) { }
intT operator()( intT i )
{
return G.getFullOutDegree(s[i]);
}
};
//Use for filter function for set the out-degree array without -1 for sparse
struct nonNegF
{
bool operator() (intT a)
{
return (a>=0);
}
};
//options to edgeMap for different versions of dense edgeMap (default is DENSE)
enum options { DENSE, DENSE_FORWARD};
//remove duplicate integers in [0,...,n-1]
void remDuplicates(intT* indices, intT* flags, intT m, intT n)
{
//make flags for first time
if(flags == NULL)
{
flags = new intT [n];
{
parallel_for(intT i=0; i<n; i++) flags[i]=-1;
}
}
{
parallel_for(intT i=0; i<m; i++)
{
if(indices[i] != -1 && flags[indices[i]] == -1)
CAS(&flags[indices[i]],(intT)-1,i);
}
}
//reset flags
{
parallel_for(intT i=0; i<m; i++)
{
if(indices[i] != -1)
{
if(flags[indices[i]] == i) //win
{
flags[indices[i]] = -1; //reset
}
else indices[i] = -1; //lost
}
}
}
}
//*****EDGE FUNCTIONS*****
// Apply an edgeMapDense(Backward) to all vertices in the frontier.
// By construction of the graph, only vertices within the partition
// corresponding to GA will be set. Pull operator, walk all vertices,
// obtain values from active vertices following in-edges, update value
//and state (active) of vertex for next round. A pull operator writes
//the value of the active vertex by updating it with the neighbours' values.
//This is out-degree id check, shuttle different start vertex
// Note: pos could be a generic "edge index" for BP
//edgeOpFwd: Forward operator, traversal active source vertices (u) src->id
//update destination dst->ngh (V.getOutNeighbor(j))
template<class F>
inline bool edgeOpFwd( intT src, intT pos, intT dst, intE w, F f )
{
#ifdef MORE_ARG
return f.cond(dst) && f.updateAtomic(src,dst,pos);
#else // MORE_ARG
#ifndef WEIGHTED
return f.cond(dst) && f.updateAtomic(src,dst);
#else
return f.cond(dst) && f.updateAtomic(src,dst,w);
#endif // WEIGHTED
#endif // MORE_ARG
}
//edgeOpBwd: Backward operator, src->ngh (destination) , dst->id (source)
template<class F>
inline bool edgeOpBwd( intT src, intT pos, intT dst, intE w, F f,
bool *frontier )
{
#ifdef MORE_ARG
return frontier[src] && f.update(src,dst,pos);
#else // MORE_ARG
#ifndef WEIGHTED
return frontier[src] && f.update(src,dst);
#else
return frontier[src] && f.update(src,dst,w);
#endif // WEIGHTED
#endif // MORE_ARG
}
template<class F>
inline bool edgeOpBwdAtomic( intT src, intT pos, intT dst, intE w, F f,
bool *frontier )
{
#ifdef MORE_ARG
return frontier[src] && f.updateAtomic(src,dst,pos);
#else // MORE_ARG
#ifndef WEIGHTED
return frontier[src] && f.updateAtomic(src,dst);
#else
return frontier[src] && f.updateAtomic(src,dst,w);
#endif // WEIGHTED
#endif // MORE_ARG
}
template<class F>
inline bool edgeOpIn( intT src, intT pos, intT dst, intE w, F f,
bool *frontier, bool *next )
{
#if REUSE_DIST
rd_bwd_src->access(src);
if( frontier[src] )
rd_bwd_data->access(src);
#endif
if( edgeOpBwd( src, pos, dst, w, f, frontier ) )
{
#if REUSE_DIST
rd_bwd_next->access(dst);
#endif
next[dst]=1;
return f.cond(dst);
}
return true;
}
template<class F>
inline bool edgeOpInAtomic( intT src, intT pos, intT dst, intE w, F f,
bool *frontier, bool *next )
{
// std::cerr << src << " -> " << dst << '\n';
#if REUSE_DIST
rd_bwd_src->access(src);
if( frontier[src] )
rd_bwd_data->access(src);
#endif
if( edgeOpBwdAtomic( src, pos, dst, w, f, frontier ) )
{
#if REUSE_DIST
rd_bwd_next->access(dst);
#endif
next[dst]=1;
// return f.cond(dst);
}
return true;
}
template<class F>
inline bool edgeOpIn( intT src, typename F::cache_t & cache,
intT pos, intT dst, intE w, F f,
bool *frontier, bool *next )
{
#ifdef MORE_ARG
if (frontier[src] && f.update(cache,src,pos))
{
next[dst]=1;
return f.cond(dst);
}
#else
#ifndef WEIGHTED
if( frontier[src] && f.update(cache,src) )
{
next[dst]=1;
return f.cond(dst);
}
#else
if( frontier[src] && f.update(cache,src,w) )
{
next[dst]=1;
return f.cond(dst);
}
#endif
#endif
return true;
}
#if PULL_EDGES
#if HILBERT_EDGES || DEST_EDGES || 1
template<class F, class Edge>
bool* edgeMapDense(const EdgeList<Edge> & EL, bool* vertices, F f, bool *next,
bool parallel = false)
{
typename EdgeList<Edge>::const_iterator E=EL.cend();
parallel_for( typename EdgeList<Edge>::const_iterator
I=EL.cbegin(); I != E; ++I )
{
const Edge &eref = *I;
intT src = eref.getSource();
intT dst = eref.getDestination();
intE wgh = eref.getWeight();
/* When operating on partitions that have been created such that
* all incoming edges to a node are located in the current partition,
* then we can execute a non-atomic edge-op. */
if( f.cond(dst) )
{
edgeOpInAtomic( src, -1, dst, wgh, f, vertices, next );
//edgeOpIn( src, /*unused*/-1, dst, wgh, f, vertices, next );
}
}
return next;
}
#else
template<class F, class Edge>
bool* edgeMapDense(const EdgeList<Edge> & EL, bool* vertices, F f, bool *next,
bool parallel = false)
{
// This version of the code is valid when edgelist is sorted by destination
static const size_t step_size = 2048;
static const size_t m = EL.get_num_edges();
parallel_for( size_t si=0; si < m; si += step_size )
{
size_t e = std::min( si + step_size, m );
intT prevv = si == 0 ? ~intT(0) : EL[si-1].getDestination();
intT nextv = e == m ? EL.get_num_vertices() : EL[e].getDestination();
size_t i;
for( i=si; i < e; ++i )
{
const Edge &eref = EL[i];
intT src = eref.getSource();
intT dst = eref.getDestination();
intE wgh = eref.getWeight();
if( f.cond(dst) )
{
if( dst != prevv )
{
edgeOpIn( src, /*unused*/-1, dst, wgh, f, vertices, next );
++i;
break;
}
else
edgeOpInAtomic( src, -1, dst, wgh, f, vertices, next );
}
}
for( ; i < e; ++i )
{
const Edge &eref = EL[i];
intT src = eref.getSource();
intT dst = eref.getDestination();
intE wgh = eref.getWeight();
if( f.cond(dst) )
{
if( dst == nextv )
{
edgeOpInAtomic( src, -1, dst, wgh, f, vertices, next );
++i;
break;
}
else
edgeOpIn( src, /*unused*/-1, dst, wgh, f, vertices, next );
}
}
for( ; i < e; ++i )
{
const Edge &eref = EL[i];
intT src = eref.getSource();
intT dst = eref.getDestination();
intE wgh = eref.getWeight();
if( f.cond(dst) )
edgeOpInAtomic( src, -1, dst, wgh, f, vertices, next );
}
}
return next;
}
#endif
#endif // PULL_EDGES
//If partitioning by source vertices, avoiding data race
//using the atomic update function
#if COMPRESSED_CSCCSR
template<class F, class vertex>
bool* edgeMapDenseCSC(graph<vertex> GA, bool * vertices,
F f, bool *next, bool source, bool parallel = 0)
{
// timer dm;
// dm.start();
pair<intT,vertex> *G=GA.CSCV;
intT CSCVn = GA.CSCVn;
parallel_for (intT i=0; i<CSCVn; i++)
{
intT id = G[i].first;
if (f.cond(id))
{
vertex V = G[i].second;
intT d = V.getInDegree();
if (source)
{
parallel_for(intT j=0; j<d; j++)
{
intT ngh = V.getInNeighbor(j);
edgeOpInAtomic( ngh, j, id, V.getInWeight(j), f, vertices, next );
}
}
else
{
if(!parallel || d < 1000)
{
if(f.use_cache)
{
typename F::cache_t cache;
f.create_cache(cache,id);
for(intT j=0; j<d; j++)
{
intT ngh = V.getInNeighbor(j);
if( !edgeOpIn( ngh, cache, j, id, V.getInWeight(j), f, vertices, next ) )
break;
}
f.commit_cache(cache,id);
}
else //use_cache ==false
{
for(intT j=0; j<d; j++)
{
intT ngh = V.getInNeighbor(j);
if( !edgeOpIn( ngh, j, id, V.getInWeight(j), f, vertices, next ) )
break;
}
}
}
else //parallel&&d>1000
{
parallel_for(intT j=0; j<d; j++)
{
intT ngh = V.getInNeighbor(j);
edgeOpInAtomic( ngh, j, id, V.getInWeight(j), f, vertices, next );
}
}
}
}
}
// dm_dense_each=dm.next();
// cerr << "MapDensePullCompressed: "<< dm_dense_each << endl;
return next;
}
template <class F, class vertex>
bool* edgeMapDenseForwardCSR(graph<vertex> GA, bool *vertices, F f, bool* next) // partitioned_vertices instead of vertices
{
// timer fm;
// fm.start();
// intT end = GA.n_src_end;
// intT start = GA.n_src_start;
pair<intT,vertex> *G = GA.CSRV;
intT CSRVn=GA.CSRVn;
// #define pragma cilk grainsize = 1
parallel_for (intT i=0; i<CSRVn; i++)
{
intT id = G[i].first;
if (vertices[id])
{
vertex V = G[i].second;
intT d = V.getOutDegree();
if(d < 1000)
{
for(intT j=0; j<d; j++)
{
uintT ngh = V.getOutNeighbor(j);
#if REUSE_DIST
rd_fwd_dest->access(ngh);
#endif
if( edgeOpFwd( id, j, ngh, V.getOutWeight(j), f ) )
{
#if REUSE_DIST
rd_fwd_new->access(ngh);
#endif
next[ngh] = 1;
}
}
}
else //d>1000
{
parallel_for(intT j=0; j<d; j++)
{
uintT ngh = V.getOutNeighbor(j);
#if REUSE_DIST
rd_fwd_dest->access(ngh);
#endif
if( edgeOpFwd( id, j, ngh, V.getOutWeight(j), f ) )
{
#if REUSE_DIST
rd_fwd_new->access(ngh);
#endif
next[ngh] = 1;
}
}
}
}
}
// fm_forward_each=fm.next();
return next;
}
#else //COMPRESSED_CSCCSR
#if !COMPRESSED_VERTICES
template<class F, class vertex>
bool* edgeMapDense(graph<vertex> GA, bool * vertices,
F f, bool *next, bool source, bool parallel = 0)
{
// timer dm;
// dm.start();
vertex *G=GA.V;
intT start = GA.n_dst_start;
intT end = GA.n_dst_end;
// if ( source ) { traverse atomically only } else { traverse with non-atomic and cache opt; }
parallel_for (intT id=start; id<end; id++)
{
vertex & V = G[id];
if (f.cond(id))
{
intT d = V.getInDegree();
if (source)
{
parallel_for(intT j=0; j<d; j++)
{
intT ngh = V.getInNeighbor(j);
edgeOpInAtomic( ngh, j, id, V.getInWeight(j), f, vertices, next );
}
}
else
{
if(!parallel || d < 1000)
{
if(f.use_cache && d > 1)
{
typename F::cache_t cache;
f.create_cache(cache,id);
for(intT j=0; j<d; j++)
{
intT ngh = V.getInNeighbor(j);
if( !edgeOpIn( ngh, cache, j, id, V.getInWeight(j), f, vertices, next ) )
break;
}
f.commit_cache(cache,id);
}
else //use_cache ==false
{
for(intT j=0; j<d; j++)
{
intT ngh = V.getInNeighbor(j);
if( !edgeOpIn( ngh, j, id, V.getInWeight(j), f, vertices, next ) )
break;
}
}
}
else //parallel&&d>1000
{
parallel_for(intT j=0; j<d; j++)
{
intT ngh = V.getInNeighbor(j);
edgeOpInAtomic( ngh, j, id, V.getInWeight(j), f, vertices, next );
}
}
}
}
}
return next;
}
template<class F, class vertex>
bool* edgeMapDenseWithSkip(graph<vertex> GA, bool * vertices,
F f, bool *next, bool source, bool parallel = 0)
{
timer dm;
dm.start();
vertex *G=GA.V;
intT start = GA.n_dst_start;
intT end = GA.n_dst_end;
parallel_for (intT big_id=start; big_id<end; big_id+=STEP)
{
intT end_id = std::min(big_id+STEP, end);
intT id = big_id;
while( id < end_id )
{
vertex & V = G[id];
if (f.cond(id))
{
intT d = V.getInDegree();
if (source)
{
parallel_for(intT j=0; j<d; j++)
{
intT ngh = V.getInNeighbor(j);
edgeOpInAtomic( ngh, j, id, V.getInWeight(j), f, vertices, next );
}
}
else
{
if(!parallel || d < 1000)
{
if(f.use_cache && d > 1)
{
typename F::cache_t cache;
f.create_cache(cache,id);
for(intT j=0; j<d; j++)
{
intT ngh = V.getInNeighbor(j);
if( !edgeOpIn( ngh, cache, j, id, V.getInWeight(j), f, vertices, next ) )
break;
}
f.commit_cache(cache,id);
}
else //use_cache ==false
{
for(intT j=0; j<d; j++)
{
intT ngh = V.getInNeighbor(j);
if( !edgeOpIn( ngh, j, id, V.getInWeight(j), f, vertices, next ) )
break;
}
}
}
else //parallel&&d>1000
{
parallel_for(intT j=0; j<d; j++)
{
intT ngh = V.getInNeighbor(j);
edgeOpInAtomic( ngh, j, id, V.getInWeight(j), f, vertices, next );
}
}
}
}
id = V.getInDegId();
}
}
return next;
}
//Apply an edgemapDense(forward) to all vertices in the frontier.
//by construction of the graph, only vertices within the partition
//corresponding to GA will be set. Push operator, walk
//active vertex array, push updated value to all target vertices
//by following out-edges, and enables those vertices for the next round.
//Push operator reads the value of the active vertex and updates its neighbours.
template <class F, class vertex>
bool* edgeMapDenseForward(graph<vertex> GA, bool *vertices, F f, bool* next)
{
// timer fm;
// fm.start();
vertex *G = GA.V;
intT end = GA.n_src_end;
intT start = GA.n_src_start;
parallel_for (intT big_id=start; big_id<end; big_id+=STEP)
{
intT id=big_id;//initial the non-zero-out-degree vertex id
intT max_id = min(big_id+STEP,end);
while (id<max_id)
{
vertex V = G[id];
if (vertices[id])
{
intT d = V.getOutDegree();
if(d < 1000)
{
for(intT j=0; j<d; j++)
{
uintT ngh = V.getOutNeighbor(j);
#if REUSE_DIST
rd_fwd_dest->access(ngh);
#endif
if( edgeOpFwd( id, j, ngh, V.getOutWeight(j), f ) )
{
#if REUSE_DIST
rd_fwd_new->access(ngh);
#endif
next[ngh] = 1;
}
}
}
else //d>1000
{
parallel_for(intT j=0; j<d; j++)
{
uintT ngh = V.getOutNeighbor(j);
#if REUSE_DIST
rd_fwd_dest->access(ngh);
#endif
if( edgeOpFwd( id, j, ngh, V.getOutWeight(j), f ) )
{
#if REUSE_DIST
rd_fwd_new->access(ngh);
#endif
next[ngh] = 1;
}
}
}
}
id = V.getOutDegId();
}
}
// fm_forward_each=fm.next();
return next;
}
#else //COMPRESSED_VERTICES
template<class F, class vertex>
bool* edgeMapDenseCompressed(graph<vertex> GA, bool * vertices,
F f, bool *next, bool source, bool parallel = 0)
{
// timer dm;
// dm.start();
pair<intT,vertex> *G=GA.CV;
intT CVn = GA.CVn;
parallel_for (intT i=0; i<CVn; i++)
{
intT id = G[i].first;
if (f.cond(id))
{
vertex V = G[i].second;
intT d = V.getInDegree();
if (source)
{
parallel_for(intT j=0; j<d; j++)
{
intT ngh = V.getInNeighbor(j);
edgeOpInAtomic( ngh, j, id, V.getInWeight(j), f, vertices, next );
}
}
else
{
if(!parallel || d < 1000)
{
if(f.use_cache)
{