Training courses

Kernel and Embedded Linux

Bootlin training courses

Embedded Linux, kernel,
Yocto Project, Buildroot, real-time,
graphics, boot time, debugging...

Bootlin logo

Elixir Cross Referencer

   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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
/* Routines to implement minimum-cost maximal flow algorithm used to smooth
   basic block and edge frequency counts.
   Copyright (C) 2008-2020 Free Software Foundation, Inc.
   Contributed by Paul Yuan (yingbo.com@gmail.com) and
                  Vinodha Ramasamy (vinodha@google.com).

This file is part of GCC.
GCC is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free
Software Foundation; either version 3, or (at your option) any later
version.

GCC is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
for more details.

You should have received a copy of the GNU General Public License
along with GCC; see the file COPYING3.  If not see
<http://www.gnu.org/licenses/>.  */

/* References:
   [1] "Feedback-directed Optimizations in GCC with Estimated Edge Profiles
        from Hardware Event Sampling", Vinodha Ramasamy, Paul Yuan, Dehao Chen,
        and Robert Hundt; GCC Summit 2008.
   [2] "Complementing Missing and Inaccurate Profiling Using a Minimum Cost
        Circulation Algorithm", Roy Levin, Ilan Newman and Gadi Haber;
        HiPEAC '08.

   Algorithm to smooth basic block and edge counts:
   1. create_fixup_graph: Create fixup graph by translating function CFG into
      a graph that satisfies MCF algorithm requirements.
   2. find_max_flow: Find maximal flow.
   3. compute_residual_flow: Form residual network.
   4. Repeat:
      cancel_negative_cycle: While G contains a negative cost cycle C, reverse
      the flow on the found cycle by the minimum residual capacity in that
      cycle.
   5. Form the minimal cost flow
      f(u,v) = rf(v, u).
   6. adjust_cfg_counts: Update initial edge weights with corrected weights.
      delta(u.v) = f(u,v) -f(v,u).
      w*(u,v) = w(u,v) + delta(u,v).  */

#include "config.h"
#include "system.h"
#include "coretypes.h"
#include "backend.h"
#include "profile.h"
#include "dumpfile.h"

/* CAP_INFINITY: Constant to represent infinite capacity.  */
#define CAP_INFINITY INTTYPE_MAXIMUM (int64_t)

/* COST FUNCTION.  */
#define K_POS(b)        ((b))
#define K_NEG(b)        (50 * (b))
#define COST(k, w)      ((k) / mcf_ln ((w) + 2))
/* Limit the number of iterations for cancel_negative_cycles() to ensure
   reasonable compile time.  */
#define MAX_ITER(n, e)  10 + (1000000 / ((n) * (e)))
enum edge_type
{
  INVALID_EDGE,
  VERTEX_SPLIT_EDGE,	    /* Edge to represent vertex with w(e) = w(v).  */
  REDIRECT_EDGE,	    /* Edge after vertex transformation.  */
  REVERSE_EDGE,
  SOURCE_CONNECT_EDGE,	    /* Single edge connecting to single source.  */
  SINK_CONNECT_EDGE,	    /* Single edge connecting to single sink.  */
  BALANCE_EDGE,		    /* Edge connecting with source/sink: cp(e) = 0.  */
  REDIRECT_NORMALIZED_EDGE, /* Normalized edge for a redirect edge.  */
  REVERSE_NORMALIZED_EDGE   /* Normalized edge for a reverse edge.  */
};

/* Structure to represent an edge in the fixup graph.  */
struct fixup_edge_type
{
  int src;
  int dest;
  /* Flag denoting type of edge and attributes for the flow field.  */
  edge_type type;
  bool is_rflow_valid;
  /* Index to the normalization vertex added for this edge.  */
  int norm_vertex_index;
  /* Flow for this edge.  */
  gcov_type flow;
  /* Residual flow for this edge - used during negative cycle canceling.  */
  gcov_type rflow;
  gcov_type weight;
  gcov_type cost;
  gcov_type max_capacity;
};

typedef fixup_edge_type *fixup_edge_p;


/* Structure to represent a vertex in the fixup graph.  */
struct fixup_vertex_type
{
  vec<fixup_edge_p> succ_edges;
};

typedef fixup_vertex_type *fixup_vertex_p;

/* Fixup graph used in the MCF algorithm.  */
struct fixup_graph_type
{
  /* Current number of vertices for the graph.  */
  int num_vertices;
  /* Current number of edges for the graph.  */
  int num_edges;
  /* Index of new entry vertex.  */
  int new_entry_index;
  /* Index of new exit vertex.  */
  int new_exit_index;
  /* Fixup vertex list. Adjacency list for fixup graph.  */
  fixup_vertex_p vertex_list;
  /* Fixup edge list.  */
  fixup_edge_p edge_list;
};

struct queue_type
{
  int *queue;
  int head;
  int tail;
  int size;
};

/* Structure used in the maximal flow routines to find augmenting path.  */
struct augmenting_path_type
{
  /* Queue used to hold vertex indices.  */
  queue_type queue_list;
  /* Vector to hold chain of pred vertex indices in augmenting path.  */
  int *bb_pred;
  /* Vector that indicates if basic block i has been visited.  */
  int *is_visited;
};


/* Function definitions.  */

/* Dump routines to aid debugging.  */

/* Print basic block with index N for FIXUP_GRAPH in n' and n'' format.  */

static void
print_basic_block (FILE *file, fixup_graph_type *fixup_graph, int n)
{
  if (n == ENTRY_BLOCK)
    fputs ("ENTRY", file);
  else if (n == ENTRY_BLOCK + 1)
    fputs ("ENTRY''", file);
  else if (n == 2 * EXIT_BLOCK)
    fputs ("EXIT", file);
  else if (n == 2 * EXIT_BLOCK + 1)
    fputs ("EXIT''", file);
  else if (n == fixup_graph->new_exit_index)
    fputs ("NEW_EXIT", file);
  else if (n == fixup_graph->new_entry_index)
    fputs ("NEW_ENTRY", file);
  else
    {
      fprintf (file, "%d", n / 2);
      if (n % 2)
	fputs ("''", file);
      else
	fputs ("'", file);
    }
}


/* Print edge S->D for given fixup_graph with n' and n'' format.
   PARAMETERS:
   S is the index of the source vertex of the edge (input) and
   D is the index of the destination vertex of the edge (input) for the given
   fixup_graph (input).  */

static void
print_edge (FILE *file, fixup_graph_type *fixup_graph, int s, int d)
{
  print_basic_block (file, fixup_graph, s);
  fputs ("->", file);
  print_basic_block (file, fixup_graph, d);
}


/* Dump out the attributes of a given edge FEDGE in the fixup_graph to a
   file.  */
static void
dump_fixup_edge (FILE *file, fixup_graph_type *fixup_graph, fixup_edge_p fedge)
{
  if (!fedge)
    {
      fputs ("NULL fixup graph edge.\n", file);
      return;
    }

  print_edge (file, fixup_graph, fedge->src, fedge->dest);
  fputs (": ", file);

  if (fedge->type)
    {
      fprintf (file, "flow/capacity=%" PRId64 "/",
	       fedge->flow);
      if (fedge->max_capacity == CAP_INFINITY)
	fputs ("+oo,", file);
      else
	fprintf (file, "%" PRId64 ",", fedge->max_capacity);
    }

  if (fedge->is_rflow_valid)
    {
      if (fedge->rflow == CAP_INFINITY)
	fputs (" rflow=+oo.", file);
      else
	fprintf (file, " rflow=%" PRId64 ",", fedge->rflow);
    }

  fprintf (file, " cost=%" PRId64 ".", fedge->cost);

  fprintf (file, "\t(%d->%d)", fedge->src, fedge->dest);

  if (fedge->type)
    {
      switch (fedge->type)
	{
	case VERTEX_SPLIT_EDGE:
	  fputs (" @VERTEX_SPLIT_EDGE", file);
	  break;

	case REDIRECT_EDGE:
	  fputs (" @REDIRECT_EDGE", file);
	  break;

	case SOURCE_CONNECT_EDGE:
	  fputs (" @SOURCE_CONNECT_EDGE", file);
	  break;

	case SINK_CONNECT_EDGE:
	  fputs (" @SINK_CONNECT_EDGE", file);
	  break;

	case REVERSE_EDGE:
	  fputs (" @REVERSE_EDGE", file);
	  break;

	case BALANCE_EDGE:
	  fputs (" @BALANCE_EDGE", file);
	  break;

	case REDIRECT_NORMALIZED_EDGE:
	case REVERSE_NORMALIZED_EDGE:
	  fputs ("  @NORMALIZED_EDGE", file);
	  break;

	default:
	  fputs (" @INVALID_EDGE", file);
	  break;
	}
    }
  fputs ("\n", file);
}


/* Print out the edges and vertices of the given FIXUP_GRAPH, into the dump
   file. The input string MSG is printed out as a heading.  */

static void
dump_fixup_graph (FILE *file, fixup_graph_type *fixup_graph, const char *msg)
{
  int i, j;
  int fnum_vertices, fnum_edges;

  fixup_vertex_p fvertex_list, pfvertex;
  fixup_edge_p pfedge;

  gcc_assert (fixup_graph);
  fvertex_list = fixup_graph->vertex_list;
  fnum_vertices = fixup_graph->num_vertices;
  fnum_edges = fixup_graph->num_edges;

  fprintf (file, "\nDump fixup graph for %s(): %s.\n",
	   current_function_name (), msg);
  fprintf (file,
	   "There are %d vertices and %d edges. new_exit_index is %d.\n\n",
	   fnum_vertices, fnum_edges, fixup_graph->new_exit_index);

  for (i = 0; i < fnum_vertices; i++)
    {
      pfvertex = fvertex_list + i;
      fprintf (file, "vertex_list[%d]: %d succ fixup edges.\n",
	       i, pfvertex->succ_edges.length ());

      for (j = 0; pfvertex->succ_edges.iterate (j, &pfedge);
	   j++)
	{
	  /* Distinguish forward edges and backward edges in the residual flow
             network.  */
	  if (pfedge->type)
	    fputs ("(f) ", file);
	  else if (pfedge->is_rflow_valid)
	    fputs ("(b) ", file);
	  dump_fixup_edge (file, fixup_graph, pfedge);
	}
    }

  fputs ("\n", file);
}


/* Utility routines.  */
/* ln() implementation: approximate calculation. Returns ln of X.  */

static double
mcf_ln (double x)
{
#define E       2.71828
  int l = 1;
  double m = E;

  gcc_assert (x >= 0);

  while (m < x)
    {
      m *= E;
      l++;
    }

  return l;
}


/* sqrt() implementation: based on open source QUAKE3 code (magic sqrt
   implementation) by John Carmack.  Returns sqrt of X.  */

static double
mcf_sqrt (double x)
{
#define MAGIC_CONST1    0x1fbcf800
#define MAGIC_CONST2    0x5f3759df
  union {
    int intPart;
    float floatPart;
  } convertor, convertor2;

  gcc_assert (x >= 0);

  convertor.floatPart = x;
  convertor2.floatPart = x;
  convertor.intPart = MAGIC_CONST1 + (convertor.intPart >> 1);
  convertor2.intPart = MAGIC_CONST2 - (convertor2.intPart >> 1);

  return 0.5f * (convertor.floatPart + (x * convertor2.floatPart));
}


/* Common code shared between add_fixup_edge and add_rfixup_edge. Adds an edge
   (SRC->DEST) to the edge_list maintained in FIXUP_GRAPH with cost of the edge
   added set to COST.  */

static fixup_edge_p
add_edge (fixup_graph_type *fixup_graph, int src, int dest, gcov_type cost)
{
  fixup_vertex_p curr_vertex = fixup_graph->vertex_list + src;
  fixup_edge_p curr_edge = fixup_graph->edge_list + fixup_graph->num_edges;
  curr_edge->src = src;
  curr_edge->dest = dest;
  curr_edge->cost = cost;
  fixup_graph->num_edges++;
  if (dump_file)
    dump_fixup_edge (dump_file, fixup_graph, curr_edge);
  curr_vertex->succ_edges.safe_push (curr_edge);
  return curr_edge;
}


/* Add a fixup edge (src->dest) with attributes TYPE, WEIGHT, COST and
   MAX_CAPACITY to the edge_list in the fixup graph.  */

static void
add_fixup_edge (fixup_graph_type *fixup_graph, int src, int dest,
		edge_type type, gcov_type weight, gcov_type cost,
		gcov_type max_capacity)
{
  fixup_edge_p curr_edge = add_edge (fixup_graph, src, dest, cost);
  curr_edge->type = type;
  curr_edge->weight = weight;
  curr_edge->max_capacity = max_capacity;
}


/* Add a residual edge (SRC->DEST) with attributes RFLOW and COST
   to the fixup graph.  */

static void
add_rfixup_edge (fixup_graph_type *fixup_graph, int src, int dest,
		 gcov_type rflow, gcov_type cost)
{
  fixup_edge_p curr_edge = add_edge (fixup_graph, src, dest, cost);
  curr_edge->rflow = rflow;
  curr_edge->is_rflow_valid = true;
  /* This edge is not a valid edge - merely used to hold residual flow.  */
  curr_edge->type = INVALID_EDGE;
}


/* Return the pointer to fixup edge SRC->DEST or NULL if edge does not
   exist in the FIXUP_GRAPH.  */

static fixup_edge_p
find_fixup_edge (fixup_graph_type *fixup_graph, int src, int dest)
{
  int j;
  fixup_edge_p pfedge;
  fixup_vertex_p pfvertex;

  gcc_assert (src < fixup_graph->num_vertices);

  pfvertex = fixup_graph->vertex_list + src;

  for (j = 0; pfvertex->succ_edges.iterate (j, &pfedge);
       j++)
    if (pfedge->dest == dest)
      return pfedge;

  return NULL;
}


/* Cleanup routine to free structures in FIXUP_GRAPH.  */

static void
delete_fixup_graph (fixup_graph_type *fixup_graph)
{
  int i;
  int fnum_vertices = fixup_graph->num_vertices;
  fixup_vertex_p pfvertex = fixup_graph->vertex_list;

  for (i = 0; i < fnum_vertices; i++, pfvertex++)
    pfvertex->succ_edges.release ();

  free (fixup_graph->vertex_list);
  free (fixup_graph->edge_list);
}


/* Creates a fixup graph FIXUP_GRAPH from the function CFG.  */

static void
create_fixup_graph (fixup_graph_type *fixup_graph)
{
  double sqrt_avg_vertex_weight = 0;
  double total_vertex_weight = 0;
  double k_pos = 0;
  double k_neg = 0;
  /* Vector to hold D(v) = sum_out_edges(v) - sum_in_edges(v).  */
  gcov_type *diff_out_in = NULL;
  gcov_type supply_value = 1, demand_value = 0;
  gcov_type fcost = 0;
  int new_entry_index = 0, new_exit_index = 0;
  int i = 0, j = 0;
  int new_index = 0;
  basic_block bb;
  edge e;
  edge_iterator ei;
  fixup_edge_p pfedge, r_pfedge;
  fixup_edge_p fedge_list;
  int fnum_edges;

  /* Each basic_block will be split into 2 during vertex transformation.  */
  int fnum_vertices_after_transform =  2 * n_basic_blocks_for_fn (cfun);
  int fnum_edges_after_transform =
    n_edges_for_fn (cfun) + n_basic_blocks_for_fn (cfun);

  /* Count the new SOURCE and EXIT vertices to be added.  */
  int fmax_num_vertices =
    (fnum_vertices_after_transform + n_edges_for_fn (cfun)
     + n_basic_blocks_for_fn (cfun) + 2);

  /* In create_fixup_graph: Each basic block and edge can be split into 3
     edges. Number of balance edges = n_basic_blocks. So after
     create_fixup_graph:
     max_edges = 4 * n_basic_blocks + 3 * n_edges
     Accounting for residual flow edges
     max_edges = 2 * (4 * n_basic_blocks + 3 * n_edges)
     = 8 * n_basic_blocks + 6 * n_edges
     < 8 * n_basic_blocks + 8 * n_edges.  */
  int fmax_num_edges = 8 * (n_basic_blocks_for_fn (cfun) +
			    n_edges_for_fn (cfun));

  /* Initial num of vertices in the fixup graph.  */
  fixup_graph->num_vertices = n_basic_blocks_for_fn (cfun);

  /* Fixup graph vertex list.  */
  fixup_graph->vertex_list =
    (fixup_vertex_p) xcalloc (fmax_num_vertices, sizeof (fixup_vertex_type));

  /* Fixup graph edge list.  */
  fixup_graph->edge_list =
    (fixup_edge_p) xcalloc (fmax_num_edges, sizeof (fixup_edge_type));

  diff_out_in =
    (gcov_type *) xcalloc (1 + fnum_vertices_after_transform,
			   sizeof (gcov_type));

  /* Compute constants b, k_pos, k_neg used in the cost function calculation.
     b = sqrt(avg_vertex_weight(cfg)); k_pos = b; k_neg = 50b.  */
  FOR_BB_BETWEEN (bb, ENTRY_BLOCK_PTR_FOR_FN (cfun), NULL, next_bb)
    total_vertex_weight += bb_gcov_count (bb);

  sqrt_avg_vertex_weight = mcf_sqrt (total_vertex_weight /
				     n_basic_blocks_for_fn (cfun));

  k_pos = K_POS (sqrt_avg_vertex_weight);
  k_neg = K_NEG (sqrt_avg_vertex_weight);

  /* 1. Vertex Transformation: Split each vertex v into two vertices v' and v'',
     connected by an edge e from v' to v''. w(e) = w(v).  */

  if (dump_file)
    fprintf (dump_file, "\nVertex transformation:\n");

  FOR_BB_BETWEEN (bb, ENTRY_BLOCK_PTR_FOR_FN (cfun), NULL, next_bb)
  {
    /* v'->v'': index1->(index1+1).  */
    i = 2 * bb->index;
    fcost = (gcov_type) COST (k_pos, bb_gcov_count (bb));
    add_fixup_edge (fixup_graph, i, i + 1, VERTEX_SPLIT_EDGE, bb_gcov_count (bb),
                    fcost, CAP_INFINITY);
    fixup_graph->num_vertices++;

    FOR_EACH_EDGE (e, ei, bb->succs)
    {
      /* Edges with ignore attribute set should be treated like they don't
         exist.  */
      if (EDGE_INFO (e) && EDGE_INFO (e)->ignore)
        continue;
      j = 2 * e->dest->index;
      fcost = (gcov_type) COST (k_pos, edge_gcov_count (e));
      add_fixup_edge (fixup_graph, i + 1, j, REDIRECT_EDGE, edge_gcov_count (e),
		      fcost, CAP_INFINITY);
    }
  }

  /* After vertex transformation.  */
  gcc_assert (fixup_graph->num_vertices == fnum_vertices_after_transform);
  /* Redirect edges are not added for edges with ignore attribute.  */
  gcc_assert (fixup_graph->num_edges <= fnum_edges_after_transform);

  fnum_edges_after_transform = fixup_graph->num_edges;

  /* 2. Initialize D(v).  */
  for (i = 0; i < fnum_edges_after_transform; i++)
    {
      pfedge = fixup_graph->edge_list + i;
      diff_out_in[pfedge->src] += pfedge->weight;
      diff_out_in[pfedge->dest] -= pfedge->weight;
    }

  /* Entry block - vertex indices 0, 1; EXIT block - vertex indices 2, 3.  */
  for (i = 0; i <= 3; i++)
    diff_out_in[i] = 0;

  /* 3. Add reverse edges: needed to decrease counts during smoothing.  */
  if (dump_file)
    fprintf (dump_file, "\nReverse edges:\n");
  for (i = 0; i < fnum_edges_after_transform; i++)
    {
      pfedge = fixup_graph->edge_list + i;
      if ((pfedge->src == 0) || (pfedge->src == 2))
        continue;
      r_pfedge = find_fixup_edge (fixup_graph, pfedge->dest, pfedge->src);
      if (!r_pfedge && pfedge->weight)
	{
	  /* Skip adding reverse edges for edges with w(e) = 0, as its maximum
	     capacity is 0.  */
	  fcost = (gcov_type) COST (k_neg, pfedge->weight);
	  add_fixup_edge (fixup_graph, pfedge->dest, pfedge->src,
			  REVERSE_EDGE, 0, fcost, pfedge->weight);
	}
    }

  /* 4. Create single source and sink. Connect new source vertex s' to function
     entry block. Connect sink vertex t' to function exit.  */
  if (dump_file)
    fprintf (dump_file, "\ns'->S, T->t':\n");

  new_entry_index = fixup_graph->new_entry_index = fixup_graph->num_vertices;
  fixup_graph->num_vertices++;
  /* Set supply_value to 1 to avoid zero count function ENTRY.  */
  add_fixup_edge (fixup_graph, new_entry_index, ENTRY_BLOCK, SOURCE_CONNECT_EDGE,
		  1 /* supply_value */, 0, 1 /* supply_value */);

  /* Create new exit with EXIT_BLOCK as single pred.  */
  new_exit_index = fixup_graph->new_exit_index = fixup_graph->num_vertices;
  fixup_graph->num_vertices++;
  add_fixup_edge (fixup_graph, 2 * EXIT_BLOCK + 1, new_exit_index,
                  SINK_CONNECT_EDGE,
                  0 /* demand_value */, 0, 0 /* demand_value */);

  /* Connect vertices with unbalanced D(v) to source/sink.  */
  if (dump_file)
    fprintf (dump_file, "\nD(v) balance:\n");
  /* Skip vertices for ENTRY (0, 1) and EXIT (2,3) blocks, so start with i = 4.
     diff_out_in[v''] will be 0, so skip v'' vertices, hence i += 2.  */
  for (i = 4; i < new_entry_index; i += 2)
    {
      if (diff_out_in[i] > 0)
	{
	  add_fixup_edge (fixup_graph, i, new_exit_index, BALANCE_EDGE, 0, 0,
			  diff_out_in[i]);
	  demand_value += diff_out_in[i];
	}
      else if (diff_out_in[i] < 0)
	{
	  add_fixup_edge (fixup_graph, new_entry_index, i, BALANCE_EDGE, 0, 0,
			  -diff_out_in[i]);
	  supply_value -= diff_out_in[i];
	}
    }

  /* Set supply = demand.  */
  if (dump_file)
    {
      fprintf (dump_file, "\nAdjust supply and demand:\n");
      fprintf (dump_file, "supply_value=%" PRId64 "\n",
	       supply_value);
      fprintf (dump_file, "demand_value=%" PRId64 "\n",
	       demand_value);
    }

  if (demand_value > supply_value)
    {
      pfedge = find_fixup_edge (fixup_graph, new_entry_index, ENTRY_BLOCK);
      pfedge->max_capacity += (demand_value - supply_value);
    }
  else
    {
      pfedge = find_fixup_edge (fixup_graph, 2 * EXIT_BLOCK + 1, new_exit_index);
      pfedge->max_capacity += (supply_value - demand_value);
    }

  /* 6. Normalize edges: remove anti-parallel edges. Anti-parallel edges are
     created by the vertex transformation step from self-edges in the original
     CFG and by the reverse edges added earlier.  */
  if (dump_file)
    fprintf (dump_file, "\nNormalize edges:\n");

  fnum_edges = fixup_graph->num_edges;
  fedge_list = fixup_graph->edge_list;

  for (i = 0; i < fnum_edges; i++)
    {
      pfedge = fedge_list + i;
      r_pfedge = find_fixup_edge (fixup_graph, pfedge->dest, pfedge->src);
      if (((pfedge->type == VERTEX_SPLIT_EDGE)
	   || (pfedge->type == REDIRECT_EDGE)) && r_pfedge)
	{
	  new_index = fixup_graph->num_vertices;
	  fixup_graph->num_vertices++;

	  if (dump_file)
	    {
	      fprintf (dump_file, "\nAnti-parallel edge:\n");
	      dump_fixup_edge (dump_file, fixup_graph, pfedge);
	      dump_fixup_edge (dump_file, fixup_graph, r_pfedge);
	      fprintf (dump_file, "New vertex is %d.\n", new_index);
	      fprintf (dump_file, "------------------\n");
	    }

	  pfedge->cost /= 2;
	  pfedge->norm_vertex_index = new_index;
	  if (dump_file)
	    {
	      fprintf (dump_file, "After normalization:\n");
	      dump_fixup_edge (dump_file, fixup_graph, pfedge);
	    }

	  /* Add a new fixup edge: new_index->src.  */
	  add_fixup_edge (fixup_graph, new_index, pfedge->src,
			  REVERSE_NORMALIZED_EDGE, 0, r_pfedge->cost,
			  r_pfedge->max_capacity);
	  gcc_assert (fixup_graph->num_vertices <= fmax_num_vertices);

	  /* Edge: r_pfedge->src -> r_pfedge->dest
             ==> r_pfedge->src -> new_index.  */
	  r_pfedge->dest = new_index;
	  r_pfedge->type = REVERSE_NORMALIZED_EDGE;
	  r_pfedge->cost = pfedge->cost;
	  r_pfedge->max_capacity = pfedge->max_capacity;
	  if (dump_file)
	    dump_fixup_edge (dump_file, fixup_graph, r_pfedge);
	}
    }

  if (dump_file)
    dump_fixup_graph (dump_file, fixup_graph, "After create_fixup_graph()");

  /* Cleanup.  */
  free (diff_out_in);
}


/* Allocates space for the structures in AUGMENTING_PATH.  The space needed is
   proportional to the number of nodes in the graph, which is given by
   GRAPH_SIZE.  */

static void
init_augmenting_path (augmenting_path_type *augmenting_path, int graph_size)
{
  augmenting_path->queue_list.queue = (int *)
    xcalloc (graph_size + 2, sizeof (int));
  augmenting_path->queue_list.size = graph_size + 2;
  augmenting_path->bb_pred = (int *) xcalloc (graph_size, sizeof (int));
  augmenting_path->is_visited = (int *) xcalloc (graph_size, sizeof (int));
}

/* Free the structures in AUGMENTING_PATH.  */
static void
free_augmenting_path (augmenting_path_type *augmenting_path)
{
  free (augmenting_path->queue_list.queue);
  free (augmenting_path->bb_pred);
  free (augmenting_path->is_visited);
}


/* Queue routines. Assumes queue will never overflow.  */

static void
init_queue (queue_type *queue_list)
{
  gcc_assert (queue_list);
  queue_list->head = 0;
  queue_list->tail = 0;
}

/* Return true if QUEUE_LIST is empty.  */
static bool
is_empty (queue_type *queue_list)
{
  return (queue_list->head == queue_list->tail);
}

/* Insert element X into QUEUE_LIST.  */
static void
enqueue (queue_type *queue_list, int x)
{
  gcc_assert (queue_list->tail < queue_list->size);
  queue_list->queue[queue_list->tail] = x;
  (queue_list->tail)++;
}

/* Return the first element in QUEUE_LIST.  */
static int
dequeue (queue_type *queue_list)
{
  int x;
  gcc_assert (queue_list->head >= 0);
  x = queue_list->queue[queue_list->head];
  (queue_list->head)++;
  return x;
}


/* Finds a negative cycle in the residual network using
   the Bellman-Ford algorithm. The flow on the found cycle is reversed by the
   minimum residual capacity of that cycle. ENTRY and EXIT vertices are not
   considered.

Parameters:
   FIXUP_GRAPH - Residual graph  (input/output)
   The following are allocated/freed by the caller:
   PI - Vector to hold predecessors in path  (pi = pred index)
   D - D[I] holds minimum cost of path from i to sink
   CYCLE - Vector to hold the minimum cost cycle

Return:
   true if a negative cycle was found, false otherwise.  */

static bool
cancel_negative_cycle (fixup_graph_type *fixup_graph,
		       int *pi, gcov_type *d, int *cycle)
{
  int i, j, k;
  int fnum_vertices, fnum_edges;
  fixup_edge_p fedge_list, pfedge, r_pfedge;
  bool found_cycle = false;
  int cycle_start = 0, cycle_end = 0;
  gcov_type sum_cost = 0, cycle_flow = 0;
  int new_entry_index;
  bool propagated = false;

  gcc_assert (fixup_graph);
  fnum_vertices = fixup_graph->num_vertices;
  fnum_edges = fixup_graph->num_edges;
  fedge_list = fixup_graph->edge_list;
  new_entry_index = fixup_graph->new_entry_index;

  /* Initialize.  */
  /* Skip ENTRY.  */
  for (i = 1; i < fnum_vertices; i++)
    {
      d[i] = CAP_INFINITY;
      pi[i] = -1;
      cycle[i] = -1;
    }
  d[ENTRY_BLOCK] = 0;

  /* Relax.  */
  for (k = 1; k < fnum_vertices; k++)
  {
    propagated = false;
    for (i = 0; i < fnum_edges; i++)
      {
	pfedge = fedge_list + i;
	if (pfedge->src == new_entry_index)
	  continue;
	if (pfedge->is_rflow_valid && pfedge->rflow
            && d[pfedge->src] != CAP_INFINITY
	    && (d[pfedge->dest] > d[pfedge->src] + pfedge->cost))
	  {
	    d[pfedge->dest] = d[pfedge->src] + pfedge->cost;
	    pi[pfedge->dest] = pfedge->src;
            propagated = true;
	  }
      }
    if (!propagated)
      break;
  }

  if (!propagated)
  /* No negative cycles exist.  */
    return 0;

  /* Detect.  */
  for (i = 0; i < fnum_edges; i++)
    {
      pfedge = fedge_list + i;
      if (pfedge->src == new_entry_index)
	continue;
      if (pfedge->is_rflow_valid && pfedge->rflow
          && d[pfedge->src] != CAP_INFINITY
	  && (d[pfedge->dest] > d[pfedge->src] + pfedge->cost))
	{
	  found_cycle = true;
	  break;
	}
    }

  if (!found_cycle)
    return 0;

  /* Augment the cycle with the cycle's minimum residual capacity.  */
  found_cycle = false;
  cycle[0] = pfedge->dest;
  j = pfedge->dest;

  for (i = 1; i < fnum_vertices; i++)
    {
      j = pi[j];
      cycle[i] = j;
      for (k = 0; k < i; k++)
	{
	  if (cycle[k] == j)
	    {
	      /* cycle[k] -> ... -> cycle[i].  */
	      cycle_start = k;
	      cycle_end = i;
	      found_cycle = true;
	      break;
	    }
	}
      if (found_cycle)
	break;
    }

  gcc_assert (cycle[cycle_start] == cycle[cycle_end]);
  if (dump_file)
    fprintf (dump_file, "\nNegative cycle length is %d:\n",
	     cycle_end - cycle_start);

  sum_cost = 0;
  cycle_flow = CAP_INFINITY;
  for (k = cycle_start; k < cycle_end; k++)
    {
      pfedge = find_fixup_edge (fixup_graph, cycle[k + 1], cycle[k]);
      cycle_flow = MIN (cycle_flow, pfedge->rflow);
      sum_cost += pfedge->cost;
      if (dump_file)
	fprintf (dump_file, "%d ", cycle[k]);
    }

  if (dump_file)
    {
      fprintf (dump_file, "%d", cycle[k]);
      fprintf (dump_file,
	       ": (%" PRId64 ", %" PRId64
	       ")\n", sum_cost, cycle_flow);
      fprintf (dump_file,
	       "Augment cycle with %" PRId64 "\n",
	       cycle_flow);
    }

  for (k = cycle_start; k < cycle_end; k++)
    {
      pfedge = find_fixup_edge (fixup_graph, cycle[k + 1], cycle[k]);
      r_pfedge = find_fixup_edge (fixup_graph, cycle[k], cycle[k + 1]);
      pfedge->rflow -= cycle_flow;
      if (pfedge->type)
	pfedge->flow += cycle_flow;
      r_pfedge->rflow += cycle_flow;
      if (r_pfedge->type)
	r_pfedge->flow -= cycle_flow;
    }

  return true;
}


/* Computes the residual flow for FIXUP_GRAPH by setting the rflow field of
   the edges. ENTRY and EXIT vertices should not be considered.  */

static void
compute_residual_flow (fixup_graph_type *fixup_graph)
{
  int i;
  int fnum_edges;
  fixup_edge_p fedge_list, pfedge;

  gcc_assert (fixup_graph);

  if (dump_file)
    fputs ("\ncompute_residual_flow():\n", dump_file);

  fnum_edges = fixup_graph->num_edges;
  fedge_list = fixup_graph->edge_list;

  for (i = 0; i < fnum_edges; i++)
    {
      pfedge = fedge_list + i;
      pfedge->rflow = pfedge->max_capacity - pfedge->flow;
      pfedge->is_rflow_valid = true;
      add_rfixup_edge (fixup_graph, pfedge->dest, pfedge->src, pfedge->flow,
		       -pfedge->cost);
    }
}


/* Uses Edmonds-Karp algorithm - BFS to find augmenting path from SOURCE to
   SINK. The fields in the edge vector in the FIXUP_GRAPH are not modified by
   this routine. The vector bb_pred in the AUGMENTING_PATH structure is updated
   to reflect the path found.
   Returns: 0 if no augmenting path is found, 1 otherwise.  */

static int
find_augmenting_path (fixup_graph_type *fixup_graph,
		      augmenting_path_type *augmenting_path, int source,
		      int sink)
{
  int u = 0;
  int i;
  fixup_vertex_p fvertex_list, pfvertex;
  fixup_edge_p pfedge;
  int *bb_pred, *is_visited;
  queue_type *queue_list;

  gcc_assert (augmenting_path);
  bb_pred = augmenting_path->bb_pred;
  gcc_assert (bb_pred);
  is_visited = augmenting_path->is_visited;
  gcc_assert (is_visited);
  queue_list = &(augmenting_path->queue_list);

  gcc_assert (fixup_graph);

  fvertex_list = fixup_graph->vertex_list;

  for (u = 0; u < fixup_graph->num_vertices; u++)
    is_visited[u] = 0;

  init_queue (queue_list);
  enqueue (queue_list, source);
  bb_pred[source] = -1;

  while (!is_empty (queue_list))
    {
      u = dequeue (queue_list);
      is_visited[u] = 1;
      pfvertex = fvertex_list + u;
      for (i = 0; pfvertex->succ_edges.iterate (i, &pfedge);
	   i++)
	{
	  int dest = pfedge->dest;
	  if ((pfedge->rflow > 0) && (is_visited[dest] == 0))
	    {
	      enqueue (queue_list, dest);
	      bb_pred[dest] = u;
	      is_visited[dest] = 1;
	      if (dest == sink)
		return 1;
	    }
	}
    }

  return 0;
}


/* Routine to find the maximal flow:
   Algorithm:
   1. Initialize flow to 0
   2. Find an augmenting path form source to sink.
   3. Send flow equal to the path's residual capacity along the edges of this path.
   4. Repeat steps 2 and 3 until no new augmenting path is found.

Parameters:
SOURCE: index of source vertex (input)
SINK: index of sink vertex    (input)
FIXUP_GRAPH: adjacency matrix representing the graph. The flow of the edges will be
             set to have a valid maximal flow by this routine. (input)
Return: Maximum flow possible.  */

static gcov_type
find_max_flow (fixup_graph_type *fixup_graph, int source, int sink)
{
  int fnum_edges;
  augmenting_path_type augmenting_path;
  int *bb_pred;
  gcov_type max_flow = 0;
  int i, u;
  fixup_edge_p fedge_list, pfedge, r_pfedge;

  gcc_assert (fixup_graph);

  fnum_edges = fixup_graph->num_edges;
  fedge_list = fixup_graph->edge_list;

  /* Initialize flow to 0.  */
  for (i = 0; i < fnum_edges; i++)
    {
      pfedge = fedge_list + i;
      pfedge->flow = 0;
    }

  compute_residual_flow (fixup_graph);

  init_augmenting_path (&augmenting_path, fixup_graph->num_vertices);

  bb_pred = augmenting_path.bb_pred;
  while (find_augmenting_path (fixup_graph, &augmenting_path, source, sink))
    {
      /* Determine the amount by which we can increment the flow.  */
      gcov_type increment = CAP_INFINITY;
      for (u = sink; u != source; u = bb_pred[u])
	{
	  pfedge = find_fixup_edge (fixup_graph, bb_pred[u], u);
	  increment = MIN (increment, pfedge->rflow);
	}
      max_flow += increment;

      /* Now increment the flow. EXIT vertex index is 1.  */
      for (u = sink; u != source; u = bb_pred[u])
	{
	  pfedge = find_fixup_edge (fixup_graph, bb_pred[u], u);
	  r_pfedge = find_fixup_edge (fixup_graph, u, bb_pred[u]);
	  if (pfedge->type)
	    {
	      /* forward edge.  */
	      pfedge->flow += increment;
	      pfedge->rflow -= increment;
	      r_pfedge->rflow += increment;
	    }
	  else
	    {
	      /* backward edge.  */
	      gcc_assert (r_pfedge->type);
	      r_pfedge->rflow += increment;
	      r_pfedge->flow -= increment;
	      pfedge->rflow -= increment;
	    }
	}

      if (dump_file)
	{
	  fprintf (dump_file, "\nDump augmenting path:\n");
	  for (u = sink; u != source; u = bb_pred[u])
	    {
	      print_basic_block (dump_file, fixup_graph, u);
	      fprintf (dump_file, "<-");
	    }
	  fprintf (dump_file,
		   "ENTRY  (path_capacity=%" PRId64 ")\n",
		   increment);
	  fprintf (dump_file,
		   "Network flow is %" PRId64 ".\n",
		   max_flow);
	}
    }

  free_augmenting_path (&augmenting_path);
  if (dump_file)
    dump_fixup_graph (dump_file, fixup_graph, "After find_max_flow()");
  return max_flow;
}


/* Computes the corrected edge and basic block weights using FIXUP_GRAPH
   after applying the find_minimum_cost_flow() routine.  */

static void
adjust_cfg_counts (fixup_graph_type *fixup_graph)
{
  basic_block bb;
  edge e;
  edge_iterator ei;
  int i, j;
  fixup_edge_p pfedge, pfedge_n;

  gcc_assert (fixup_graph);

  if (dump_file)
    fprintf (dump_file, "\nadjust_cfg_counts():\n");

  FOR_BB_BETWEEN (bb, ENTRY_BLOCK_PTR_FOR_FN (cfun),
		  EXIT_BLOCK_PTR_FOR_FN (cfun), next_bb)
    {
      i = 2 * bb->index;

      /* Fixup BB.  */
      if (dump_file)
        fprintf (dump_file,
                 "BB%d: %" PRId64 "", bb->index, bb_gcov_count (bb));

      pfedge = find_fixup_edge (fixup_graph, i, i + 1);
      if (pfedge->flow)
        {
          bb_gcov_count (bb) += pfedge->flow;
	  if (dump_file)
	    {
	      fprintf (dump_file, " + %" PRId64 "(",
	               pfedge->flow);
	      print_edge (dump_file, fixup_graph, i, i + 1);
	      fprintf (dump_file, ")");
	    }
        }

      pfedge_n =
        find_fixup_edge (fixup_graph, i + 1, pfedge->norm_vertex_index);
      /* Deduct flow from normalized reverse edge.  */
      if (pfedge->norm_vertex_index && pfedge_n->flow)
        {
          bb_gcov_count (bb) -= pfedge_n->flow;
	  if (dump_file)
	    {
	      fprintf (dump_file, " - %" PRId64 "(",
		       pfedge_n->flow);
	      print_edge (dump_file, fixup_graph, i + 1,
			  pfedge->norm_vertex_index);
	      fprintf (dump_file, ")");
	    }
        }
      if (dump_file)
        fprintf (dump_file, " = %" PRId64 "\n", bb_gcov_count (bb));

      /* Fixup edge.  */
      FOR_EACH_EDGE (e, ei, bb->succs)
        {
          /* Treat edges with ignore attribute set as if they don't exist.  */
          if (EDGE_INFO (e) && EDGE_INFO (e)->ignore)
	    continue;

          j = 2 * e->dest->index;
          if (dump_file)
	    fprintf (dump_file, "%d->%d: %" PRId64 "",
		     bb->index, e->dest->index, edge_gcov_count (e));

          pfedge = find_fixup_edge (fixup_graph, i + 1, j);

          if (bb->index != e->dest->index)
	    {
	      /* Non-self edge.  */
	      if (pfedge->flow)
	        {
	          edge_gcov_count (e) += pfedge->flow;
	          if (dump_file)
		    {
		      fprintf (dump_file, " + %" PRId64 "(",
			       pfedge->flow);
		      print_edge (dump_file, fixup_graph, i + 1, j);
		      fprintf (dump_file, ")");
		    }
	        }

	      pfedge_n =
	        find_fixup_edge (fixup_graph, j, pfedge->norm_vertex_index);
	      /* Deduct flow from normalized reverse edge.  */
	      if (pfedge->norm_vertex_index && pfedge_n->flow)
	        {
	          edge_gcov_count (e) -= pfedge_n->flow;
	          if (dump_file)
		    {
		      fprintf (dump_file, " - %" PRId64 "(",
			       pfedge_n->flow);
		      print_edge (dump_file, fixup_graph, j,
			          pfedge->norm_vertex_index);
		      fprintf (dump_file, ")");
		    }
	        }
	    }
          else
	    {
	      /* Handle self edges. Self edge is split with a normalization
                 vertex. Here i=j.  */
	      pfedge = find_fixup_edge (fixup_graph, j, i + 1);
	      pfedge_n =
	        find_fixup_edge (fixup_graph, i + 1, pfedge->norm_vertex_index);
	      edge_gcov_count (e) += pfedge_n->flow;
	      bb_gcov_count (bb) += pfedge_n->flow;
	      if (dump_file)
	        {
	          fprintf (dump_file, "(self edge)");
	          fprintf (dump_file, " + %" PRId64 "(",
		           pfedge_n->flow);
	          print_edge (dump_file, fixup_graph, i + 1,
			      pfedge->norm_vertex_index);
	          fprintf (dump_file, ")");
	        }
	    }

          if (bb_gcov_count (bb))
	    e->probability = profile_probability::probability_in_gcov_type
			 (edge_gcov_count (e), bb_gcov_count (bb));
          if (dump_file)
	    {
	      fprintf (dump_file, " = %" PRId64 "\t",
		       edge_gcov_count (e));
	      e->probability.dump (dump_file);
	      fprintf (dump_file, "\n");
	    }
        }
    }

  bb_gcov_count (ENTRY_BLOCK_PTR_FOR_FN (cfun)) =
		     sum_edge_counts (ENTRY_BLOCK_PTR_FOR_FN (cfun)->succs);
  bb_gcov_count (EXIT_BLOCK_PTR_FOR_FN (cfun)) =
		     sum_edge_counts (EXIT_BLOCK_PTR_FOR_FN (cfun)->preds);

  /* Compute edge probabilities.  */
  FOR_ALL_BB_FN (bb, cfun)
    {
      if (bb_gcov_count (bb))
        {
          FOR_EACH_EDGE (e, ei, bb->succs)
            e->probability = profile_probability::probability_in_gcov_type
				(edge_gcov_count (e), bb_gcov_count (bb));
        }
    }

  if (dump_file)
    {
      fprintf (dump_file, "\nCheck %s() CFG flow conservation:\n",
	       current_function_name ());
      FOR_EACH_BB_FN (bb, cfun)
        {
          if ((bb_gcov_count (bb) != sum_edge_counts (bb->preds))
               || (bb_gcov_count (bb) != sum_edge_counts (bb->succs)))
            {
              fprintf (dump_file,
                       "BB%d(%" PRId64 ")  **INVALID**: ",
                       bb->index, bb_gcov_count (bb));
              fprintf (stderr,
                       "******** BB%d(%" PRId64
                       ")  **INVALID**: \n", bb->index, bb_gcov_count (bb));
              fprintf (dump_file, "in_edges=%" PRId64 " ",
                       sum_edge_counts (bb->preds));
              fprintf (dump_file, "out_edges=%" PRId64 "\n",
                       sum_edge_counts (bb->succs));
            }
         }
    }
}


/* Implements the negative cycle canceling algorithm to compute a minimum cost
   flow.
Algorithm:
1. Find maximal flow.
2. Form residual network
3. Repeat:
  While G contains a negative cost cycle C, reverse the flow on the found cycle
  by the minimum residual capacity in that cycle.
4. Form the minimal cost flow
  f(u,v) = rf(v, u)
Input:
  FIXUP_GRAPH - Initial fixup graph.
  The flow field is modified to represent the minimum cost flow.  */

static void
find_minimum_cost_flow (fixup_graph_type *fixup_graph)
{
  /* Holds the index of predecessor in path.  */
  int *pred;
  /* Used to hold the minimum cost cycle.  */
  int *cycle;
  /* Used to record the number of iterations of cancel_negative_cycle.  */
  int iteration;
  /* Vector d[i] holds the minimum cost of path from i to sink.  */
  gcov_type *d;
  int fnum_vertices;
  int new_exit_index;
  int new_entry_index;

  gcc_assert (fixup_graph);
  fnum_vertices = fixup_graph->num_vertices;
  new_exit_index = fixup_graph->new_exit_index;
  new_entry_index = fixup_graph->new_entry_index;

  find_max_flow (fixup_graph, new_entry_index, new_exit_index);

  /* Initialize the structures for find_negative_cycle().  */
  pred = (int *) xcalloc (fnum_vertices, sizeof (int));
  d = (gcov_type *) xcalloc (fnum_vertices, sizeof (gcov_type));
  cycle = (int *) xcalloc (fnum_vertices, sizeof (int));

  /* Repeatedly find and cancel negative cost cycles, until
     no more negative cycles exist. This also updates the flow field
     to represent the minimum cost flow so far.  */
  iteration = 0;
  while (cancel_negative_cycle (fixup_graph, pred, d, cycle))
    {
      iteration++;
      if (iteration > MAX_ITER (fixup_graph->num_vertices,
                                fixup_graph->num_edges))
        break;
    }

  if (dump_file)
    dump_fixup_graph (dump_file, fixup_graph,
		      "After find_minimum_cost_flow()");

  /* Cleanup structures.  */
  free (pred);
  free (d);
  free (cycle);
}


/* Compute the sum of the edge counts in TO_EDGES.  */

gcov_type
sum_edge_counts (vec<edge, va_gc> *to_edges)
{
  gcov_type sum = 0;
  edge e;
  edge_iterator ei;

  FOR_EACH_EDGE (e, ei, to_edges)
    {
      if (EDGE_INFO (e) && EDGE_INFO (e)->ignore)
        continue;
      sum += edge_gcov_count (e);
    }
  return sum;
}


/* Main routine. Smoothes the initial assigned basic block and edge counts using
   a minimum cost flow algorithm, to ensure that the flow consistency rule is
   obeyed: sum of outgoing edges = sum of incoming edges for each basic
   block.  */

void
mcf_smooth_cfg (void)
{
  fixup_graph_type fixup_graph;
  memset (&fixup_graph, 0, sizeof (fixup_graph));
  create_fixup_graph (&fixup_graph);
  find_minimum_cost_flow (&fixup_graph);
  adjust_cfg_counts (&fixup_graph);
  delete_fixup_graph (&fixup_graph);
}