This file is indexed.

/usr/include/tbb/enumerable_thread_specific.h is in libtbb-dev 4.3~20150611-2.

This file is owned by root:root, with mode 0o644.

The actual contents of the file can be viewed below.

   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
/*
    Copyright 2005-2015 Intel Corporation.  All Rights Reserved.

    This file is part of Threading Building Blocks. Threading Building Blocks is free software;
    you can redistribute it and/or modify it under the terms of the GNU General Public License
    version 2  as  published  by  the  Free Software Foundation.  Threading Building Blocks 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 Threading Building Blocks; if not, write to the
    Free Software Foundation, Inc.,  51 Franklin St,  Fifth Floor,  Boston,  MA 02110-1301 USA

    As a special exception,  you may use this file  as part of a free software library without
    restriction.  Specifically,  if other files instantiate templates  or use macros or inline
    functions from this file, or you compile this file and link it with other files to produce
    an executable,  this file does not by itself cause the resulting executable to be covered
    by the GNU General Public License. This exception does not however invalidate any other
    reasons why the executable file might be covered by the GNU General Public License.
*/

#ifndef __TBB_enumerable_thread_specific_H
#define __TBB_enumerable_thread_specific_H

#include "concurrent_vector.h"
#include "tbb_thread.h"
#include "tbb_allocator.h"
#include "tbb_profiling.h"
#include "cache_aligned_allocator.h"
#include "aligned_space.h"
#include "internal/_template_helpers.h"
#include <string.h>  // for memcpy

#if _WIN32||_WIN64
#include "machine/windows_api.h"
#else
#include <pthread.h>
#endif

#define __TBB_ETS_USE_CPP11 \
    (__TBB_CPP11_RVALUE_REF_PRESENT && __TBB_CPP11_VARIADIC_TEMPLATES_PRESENT \
     && __TBB_CPP11_DECLTYPE_PRESENT && __TBB_CPP11_LAMBDAS_PRESENT)

namespace tbb {

//! enum for selecting between single key and key-per-instance versions
enum ets_key_usage_type { ets_key_per_instance, ets_no_key };

namespace interface6 {

    // Forward declaration to use in internal classes
    template <typename T, typename Allocator, ets_key_usage_type ETS_key_type>
    class enumerable_thread_specific;

    //! @cond
    namespace internal {

        using namespace tbb::internal;

        template<ets_key_usage_type ETS_key_type>
        class ets_base: tbb::internal::no_copy {
        protected:
#if _WIN32||_WIN64
            typedef DWORD key_type;
#else
            typedef pthread_t key_type;
#endif
#if __TBB_PROTECTED_NESTED_CLASS_BROKEN
        public:
#endif
            struct slot;

            struct array {
                array* next;
                size_t lg_size;
                slot& at( size_t k ) {
                    return ((slot*)(void*)(this+1))[k];
                }
                size_t size() const {return (size_t)1<<lg_size;}
                size_t mask() const {return size()-1;}
                size_t start( size_t h ) const {
                    return h>>(8*sizeof(size_t)-lg_size);
                }
            };
            struct slot {
                key_type key;
                void* ptr;
                bool empty() const {return !key;}
                bool match( key_type k ) const {return key==k;}
                bool claim( key_type k ) {
                    __TBB_ASSERT(sizeof(tbb::atomic<key_type>)==sizeof(key_type), NULL);
                    // TODO: maybe claim ptr, because key_type is not guaranteed to match word size
                    return tbb::internal::punned_cast<tbb::atomic<key_type>*>(&key)->compare_and_swap(k,0)==0;
                }
            };
#if __TBB_PROTECTED_NESTED_CLASS_BROKEN
        protected:
#endif

            static key_type key_of_current_thread() {
               // TODO: replace key_type with tbb::tbb_thread::id
               tbb::tbb_thread::id id = tbb::this_tbb_thread::get_id();
               key_type k;
               memcpy( &k, &id, sizeof(k) );
               return k;
            }

            //! Root of linked list of arrays of decreasing size.
            /** NULL if and only if my_count==0.
                Each array in the list is half the size of its predecessor. */
            atomic<array*> my_root;
            atomic<size_t> my_count;
            virtual void* create_local() = 0;
            virtual void* create_array(size_t _size) = 0;  // _size in bytes
            virtual void free_array(void* ptr, size_t _size) = 0; // _size in bytes
            array* allocate( size_t lg_size ) {
                size_t n = 1<<lg_size;
                array* a = static_cast<array*>(create_array( sizeof(array)+n*sizeof(slot) ));
                a->lg_size = lg_size;
                std::memset( a+1, 0, n*sizeof(slot) );
                return a;
            }
            void free(array* a) {
                size_t n = 1<<(a->lg_size);
                free_array( (void *)a, size_t(sizeof(array)+n*sizeof(slot)) );
            }
            static size_t hash( key_type k ) {
                // Multiplicative hashing.  Client should use *upper* bits.
                // casts required for Mac gcc4.* compiler
                // TODO: casting of key_type to uintptr_t is not portable; implement and use hashing of thread::id
                return uintptr_t(k)*tbb::internal::select_size_t_constant<0x9E3779B9,0x9E3779B97F4A7C15ULL>::value;
            }

            ets_base() {my_root=NULL; my_count=0;}
            virtual ~ets_base();  // g++ complains if this is not virtual...
            void* table_lookup( bool& exists );
            void table_clear();
            // table_find is used in copying ETS, so is not used in concurrent context.  So
            // we don't need itt annotations for it.
            slot& table_find( key_type k ) {
                size_t h = hash(k);
                array* r = my_root;
                size_t mask = r->mask();
                for(size_t i = r->start(h);;i=(i+1)&mask) {
                    slot& s = r->at(i);
                    if( s.empty() || s.match(k) )
                        return s;
                }
            }
            void table_reserve_for_copy( const ets_base& other ) {
                __TBB_ASSERT(!my_root,NULL);
                __TBB_ASSERT(!my_count,NULL);
                if( other.my_root ) {
                    array* a = allocate(other.my_root->lg_size);
                    a->next = NULL;
                    my_root = a;
                    my_count = other.my_count;
                }
            }
        };

        template<ets_key_usage_type ETS_key_type>
        ets_base<ETS_key_type>::~ets_base() {
            __TBB_ASSERT(!my_root, NULL);
        }

        template<ets_key_usage_type ETS_key_type>
        void ets_base<ETS_key_type>::table_clear() {
            while( array* r = my_root ) {
                my_root = r->next;
                free(r);
            }
            my_count = 0;
        }

        template<ets_key_usage_type ETS_key_type>
        void* ets_base<ETS_key_type>::table_lookup( bool& exists ) {
            const key_type k = key_of_current_thread();

            __TBB_ASSERT(k!=0,NULL);
            void* found;
            size_t h = hash(k);
            for( array* r=my_root; r; r=r->next ) {
                call_itt_notify(acquired,r);
                size_t mask=r->mask();
                for(size_t i = r->start(h); ;i=(i+1)&mask) {
                    slot& s = r->at(i);
                    if( s.empty() ) break;
                    if( s.match(k) ) {
                        if( r==my_root ) {
                            // Success at top level
                            exists = true;
                            return s.ptr;
                        } else {
                            // Success at some other level.  Need to insert at top level.
                            exists = true;
                            found = s.ptr;
                            goto insert;
                        }
                    }
                }
            }
            // Key does not yet exist.  The density of slots in the table does not exceed 0.5,
            // for if this will occur a new table is allocated with double the current table
            // size, which is swapped in as the new root table.  So an empty slot is guaranteed.
            exists = false;
            found = create_local();
            {
                size_t c = ++my_count;
                array* r = my_root;
                call_itt_notify(acquired,r);
                if( !r || c>r->size()/2 ) {
                    size_t s = r ? r->lg_size : 2;
                    while( c>size_t(1)<<(s-1) ) ++s;
                    array* a = allocate(s);
                    for(;;) {
                        a->next = r;
                        call_itt_notify(releasing,a);
                        array* new_r = my_root.compare_and_swap(a,r);
                        if( new_r==r ) break;
                        call_itt_notify(acquired, new_r);
                        if( new_r->lg_size>=s ) {
                            // Another thread inserted an equal or  bigger array, so our array is superfluous.
                            free(a);
                            break;
                        }
                        r = new_r;
                    }
                }
            }
        insert:
        // Whether a slot has been found in an older table, or if it has been inserted at this level,
        // it has already been accounted for in the total.  Guaranteed to be room for it, and it is
        // not present, so search for empty slot and use it.
            array* ir = my_root;
            call_itt_notify(acquired, ir);
            size_t mask = ir->mask();
            for(size_t i = ir->start(h);;i=(i+1)&mask) {
                slot& s = ir->at(i);
                if( s.empty() ) {
                    if( s.claim(k) ) {
                        s.ptr = found;
                        return found;
                    }
                }
            }
        }

        //! Specialization that exploits native TLS
        template <>
        class ets_base<ets_key_per_instance>: protected ets_base<ets_no_key> {
            typedef ets_base<ets_no_key> super;
#if _WIN32||_WIN64
#if __TBB_WIN8UI_SUPPORT
            typedef DWORD tls_key_t;
            void create_key() { my_key = FlsAlloc(NULL); }
            void destroy_key() { FlsFree(my_key); }
            void set_tls(void * value) { FlsSetValue(my_key, (LPVOID)value); }
            void* get_tls() { return (void *)FlsGetValue(my_key); }
#else
            typedef DWORD tls_key_t;
            void create_key() { my_key = TlsAlloc(); }
            void destroy_key() { TlsFree(my_key); }
            void set_tls(void * value) { TlsSetValue(my_key, (LPVOID)value); }
            void* get_tls() { return (void *)TlsGetValue(my_key); }
#endif
#else
            typedef pthread_key_t tls_key_t;
            void create_key() { pthread_key_create(&my_key, NULL); }
            void destroy_key() { pthread_key_delete(my_key); }
            void set_tls( void * value ) const { pthread_setspecific(my_key, value); }
            void* get_tls() const { return pthread_getspecific(my_key); }
#endif
            tls_key_t my_key;
            virtual void* create_local() = 0;
            virtual void* create_array(size_t _size) = 0;  // _size in bytes
            virtual void free_array(void* ptr, size_t _size) = 0; // size in bytes
        public:
            ets_base() {create_key();}
            ~ets_base() {destroy_key();}
            void* table_lookup( bool& exists ) {
                void* found = get_tls();
                if( found ) {
                    exists=true;
                } else {
                    found = super::table_lookup(exists);
                    set_tls(found);
                }
                return found;
            }
            void table_clear() {
                destroy_key();
                create_key();
                super::table_clear();
            }
        };

        //! Random access iterator for traversing the thread local copies.
        template< typename Container, typename Value >
        class enumerable_thread_specific_iterator
#if defined(_WIN64) && defined(_MSC_VER)
            // Ensure that Microsoft's internal template function _Val_type works correctly.
            : public std::iterator<std::random_access_iterator_tag,Value>
#endif /* defined(_WIN64) && defined(_MSC_VER) */
        {
            //! current position in the concurrent_vector

            Container *my_container;
            typename Container::size_type my_index;
            mutable Value *my_value;

            template<typename C, typename T>
            friend enumerable_thread_specific_iterator<C,T> operator+( ptrdiff_t offset,
                                                                       const enumerable_thread_specific_iterator<C,T>& v );

            template<typename C, typename T, typename U>
            friend bool operator==( const enumerable_thread_specific_iterator<C,T>& i,
                                    const enumerable_thread_specific_iterator<C,U>& j );

            template<typename C, typename T, typename U>
            friend bool operator<( const enumerable_thread_specific_iterator<C,T>& i,
                                   const enumerable_thread_specific_iterator<C,U>& j );

            template<typename C, typename T, typename U>
            friend ptrdiff_t operator-( const enumerable_thread_specific_iterator<C,T>& i, const enumerable_thread_specific_iterator<C,U>& j );

            template<typename C, typename U>
            friend class enumerable_thread_specific_iterator;

            public:

            enumerable_thread_specific_iterator( const Container &container, typename Container::size_type index ) :
                my_container(&const_cast<Container &>(container)), my_index(index), my_value(NULL) {}

            //! Default constructor
            enumerable_thread_specific_iterator() : my_container(NULL), my_index(0), my_value(NULL) {}

            template<typename U>
            enumerable_thread_specific_iterator( const enumerable_thread_specific_iterator<Container, U>& other ) :
                    my_container( other.my_container ), my_index( other.my_index), my_value( const_cast<Value *>(other.my_value) ) {}

            enumerable_thread_specific_iterator operator+( ptrdiff_t offset ) const {
                return enumerable_thread_specific_iterator(*my_container, my_index + offset);
            }

            enumerable_thread_specific_iterator &operator+=( ptrdiff_t offset ) {
                my_index += offset;
                my_value = NULL;
                return *this;
            }

            enumerable_thread_specific_iterator operator-( ptrdiff_t offset ) const {
                return enumerable_thread_specific_iterator( *my_container, my_index-offset );
            }

            enumerable_thread_specific_iterator &operator-=( ptrdiff_t offset ) {
                my_index -= offset;
                my_value = NULL;
                return *this;
            }

            Value& operator*() const {
                Value* value = my_value;
                if( !value ) {
                    value = my_value = (*my_container)[my_index].value();
                }
                __TBB_ASSERT( value==(*my_container)[my_index].value(), "corrupt cache" );
                return *value;
            }

            Value& operator[]( ptrdiff_t k ) const {
               return (*my_container)[my_index + k].value;
            }

            Value* operator->() const {return &operator*();}

            enumerable_thread_specific_iterator& operator++() {
                ++my_index;
                my_value = NULL;
                return *this;
            }

            enumerable_thread_specific_iterator& operator--() {
                --my_index;
                my_value = NULL;
                return *this;
            }

            //! Post increment
            enumerable_thread_specific_iterator operator++(int) {
                enumerable_thread_specific_iterator result = *this;
                ++my_index;
                my_value = NULL;
                return result;
            }

            //! Post decrement
            enumerable_thread_specific_iterator operator--(int) {
                enumerable_thread_specific_iterator result = *this;
                --my_index;
                my_value = NULL;
                return result;
            }

            // STL support
            typedef ptrdiff_t difference_type;
            typedef Value value_type;
            typedef Value* pointer;
            typedef Value& reference;
            typedef std::random_access_iterator_tag iterator_category;
        };

        template<typename Container, typename T>
        enumerable_thread_specific_iterator<Container,T> operator+( ptrdiff_t offset,
                                                                    const enumerable_thread_specific_iterator<Container,T>& v ) {
            return enumerable_thread_specific_iterator<Container,T>( v.my_container, v.my_index + offset );
        }

        template<typename Container, typename T, typename U>
        bool operator==( const enumerable_thread_specific_iterator<Container,T>& i,
                         const enumerable_thread_specific_iterator<Container,U>& j ) {
            return i.my_index==j.my_index && i.my_container == j.my_container;
        }

        template<typename Container, typename T, typename U>
        bool operator!=( const enumerable_thread_specific_iterator<Container,T>& i,
                         const enumerable_thread_specific_iterator<Container,U>& j ) {
            return !(i==j);
        }

        template<typename Container, typename T, typename U>
        bool operator<( const enumerable_thread_specific_iterator<Container,T>& i,
                        const enumerable_thread_specific_iterator<Container,U>& j ) {
            return i.my_index<j.my_index;
        }

        template<typename Container, typename T, typename U>
        bool operator>( const enumerable_thread_specific_iterator<Container,T>& i,
                        const enumerable_thread_specific_iterator<Container,U>& j ) {
            return j<i;
        }

        template<typename Container, typename T, typename U>
        bool operator>=( const enumerable_thread_specific_iterator<Container,T>& i,
                         const enumerable_thread_specific_iterator<Container,U>& j ) {
            return !(i<j);
        }

        template<typename Container, typename T, typename U>
        bool operator<=( const enumerable_thread_specific_iterator<Container,T>& i,
                         const enumerable_thread_specific_iterator<Container,U>& j ) {
            return !(j<i);
        }

        template<typename Container, typename T, typename U>
        ptrdiff_t operator-( const enumerable_thread_specific_iterator<Container,T>& i,
                             const enumerable_thread_specific_iterator<Container,U>& j ) {
            return i.my_index-j.my_index;
        }

    template<typename SegmentedContainer, typename Value >
        class segmented_iterator
#if defined(_WIN64) && defined(_MSC_VER)
        : public std::iterator<std::input_iterator_tag, Value>
#endif
        {
            template<typename C, typename T, typename U>
            friend bool operator==(const segmented_iterator<C,T>& i, const segmented_iterator<C,U>& j);

            template<typename C, typename T, typename U>
            friend bool operator!=(const segmented_iterator<C,T>& i, const segmented_iterator<C,U>& j);

            template<typename C, typename U>
            friend class segmented_iterator;

            public:

                segmented_iterator() {my_segcont = NULL;}

                segmented_iterator( const SegmentedContainer& _segmented_container ) :
                    my_segcont(const_cast<SegmentedContainer*>(&_segmented_container)),
                    outer_iter(my_segcont->end()) { }

                ~segmented_iterator() {}

                typedef typename SegmentedContainer::iterator outer_iterator;
                typedef typename SegmentedContainer::value_type InnerContainer;
                typedef typename InnerContainer::iterator inner_iterator;

                // STL support
                typedef ptrdiff_t difference_type;
                typedef Value value_type;
                typedef typename SegmentedContainer::size_type size_type;
                typedef Value* pointer;
                typedef Value& reference;
                typedef std::input_iterator_tag iterator_category;

                // Copy Constructor
                template<typename U>
                segmented_iterator(const segmented_iterator<SegmentedContainer, U>& other) :
                    my_segcont(other.my_segcont),
                    outer_iter(other.outer_iter),
                    // can we assign a default-constructed iterator to inner if we're at the end?
                    inner_iter(other.inner_iter)
                {}

                // assignment
                template<typename U>
                segmented_iterator& operator=( const segmented_iterator<SegmentedContainer, U>& other) {
                    if(this != &other) {
                        my_segcont = other.my_segcont;
                        outer_iter = other.outer_iter;
                        if(outer_iter != my_segcont->end()) inner_iter = other.inner_iter;
                    }
                    return *this;
                }

                // allow assignment of outer iterator to segmented iterator.  Once it is
                // assigned, move forward until a non-empty inner container is found or
                // the end of the outer container is reached.
                segmented_iterator& operator=(const outer_iterator& new_outer_iter) {
                    __TBB_ASSERT(my_segcont != NULL, NULL);
                    // check that this iterator points to something inside the segmented container
                    for(outer_iter = new_outer_iter ;outer_iter!=my_segcont->end(); ++outer_iter) {
                        if( !outer_iter->empty() ) {
                            inner_iter = outer_iter->begin();
                            break;
                        }
                    }
                    return *this;
                }

                // pre-increment
                segmented_iterator& operator++() {
                    advance_me();
                    return *this;
                }

                // post-increment
                segmented_iterator operator++(int) {
                    segmented_iterator tmp = *this;
                    operator++();
                    return tmp;
                }

                bool operator==(const outer_iterator& other_outer) const {
                    __TBB_ASSERT(my_segcont != NULL, NULL);
                    return (outer_iter == other_outer &&
                            (outer_iter == my_segcont->end() || inner_iter == outer_iter->begin()));
                }

                bool operator!=(const outer_iterator& other_outer) const {
                    return !operator==(other_outer);

                }

                // (i)* RHS
                reference operator*() const {
                    __TBB_ASSERT(my_segcont != NULL, NULL);
                    __TBB_ASSERT(outer_iter != my_segcont->end(), "Dereferencing a pointer at end of container");
                    __TBB_ASSERT(inner_iter != outer_iter->end(), NULL); // should never happen
                    return *inner_iter;
                }

                // i->
                pointer operator->() const { return &operator*();}

            private:
                SegmentedContainer*             my_segcont;
                outer_iterator outer_iter;
                inner_iterator inner_iter;

                void advance_me() {
                    __TBB_ASSERT(my_segcont != NULL, NULL);
                    __TBB_ASSERT(outer_iter != my_segcont->end(), NULL); // not true if there are no inner containers
                    __TBB_ASSERT(inner_iter != outer_iter->end(), NULL); // not true if the inner containers are all empty.
                    ++inner_iter;
                    while(inner_iter == outer_iter->end() && ++outer_iter != my_segcont->end()) {
                        inner_iter = outer_iter->begin();
                    }
                }
        };    // segmented_iterator

        template<typename SegmentedContainer, typename T, typename U>
        bool operator==( const segmented_iterator<SegmentedContainer,T>& i,
                         const segmented_iterator<SegmentedContainer,U>& j ) {
            if(i.my_segcont != j.my_segcont) return false;
            if(i.my_segcont == NULL) return true;
            if(i.outer_iter != j.outer_iter) return false;
            if(i.outer_iter == i.my_segcont->end()) return true;
            return i.inner_iter == j.inner_iter;
        }

        // !=
        template<typename SegmentedContainer, typename T, typename U>
        bool operator!=( const segmented_iterator<SegmentedContainer,T>& i,
                         const segmented_iterator<SegmentedContainer,U>& j ) {
            return !(i==j);
        }

        template<typename T>
        struct construct_by_default: tbb::internal::no_assign {
            void construct(void*where) {new(where) T();} // C++ note: the () in T() ensure zero initialization.
            construct_by_default( int ) {}
        };

        template<typename T>
        struct construct_by_exemplar: tbb::internal::no_assign {
            const T exemplar;
            void construct(void*where) {new(where) T(exemplar);}
            construct_by_exemplar( const T& t ) : exemplar(t) {}
#if __TBB_ETS_USE_CPP11
            construct_by_exemplar( T&& t ) : exemplar(std::move(t)) {}
#endif
        };

        template<typename T, typename Finit>
        struct construct_by_finit: tbb::internal::no_assign {
            Finit f;
            void construct(void* where) {new(where) T(f());}
            construct_by_finit( const Finit& f_ ) : f(f_) {}
#if __TBB_ETS_USE_CPP11
            construct_by_finit( Finit&& f_ ) : f(std::move(f_)) {}
#endif
        };

#if __TBB_ETS_USE_CPP11
        template<typename T, typename... P>
        struct construct_by_args: tbb::internal::no_assign {
            internal::stored_pack<P...> pack;
            void construct(void* where) {
                internal::call( [where](const typename strip<P>::type&... args ){
                   new(where) T(args...);
                }, pack );
            }
            construct_by_args( P&& ... args ) : pack(std::forward<P>(args)...) {}
        };
#endif 

        // storage for initialization function pointer
        template<typename T>
        class callback_base {
        public:
            // Clone *this
            virtual callback_base* clone() const = 0;
            // Destruct and free *this
            virtual void destroy() = 0;
            // Need virtual destructor to satisfy GCC compiler warning
            virtual ~callback_base() { }
            // Construct T at where
            virtual void construct(void* where) = 0;
        };

        template <typename T, typename Constructor>
        class callback_leaf: public callback_base<T>, Constructor {
#if __TBB_ETS_USE_CPP11
            template<typename... P> callback_leaf( P&& ... params ) : Constructor(std::forward<P>(params)...) {}
#else
            template<typename X> callback_leaf( const X& x ) : Constructor(x) {}
#endif

            typedef typename tbb::tbb_allocator<callback_leaf> my_allocator_type;

            /*override*/ callback_base<T>* clone() const {
                return make(*this);
            }

            /*override*/ void destroy() {
                my_allocator_type().destroy(this);
                my_allocator_type().deallocate(this,1);
            }

            /*override*/ void construct(void* where) {
                Constructor::construct(where);
            }
        public:
#if __TBB_ETS_USE_CPP11
            template<typename... P>
            static callback_base<T>* make( P&& ... params ) {
                void* where = my_allocator_type().allocate(1);
                return new(where) callback_leaf( std::forward<P>(params)... );
            }
#else
            template<typename X>
            static callback_base<T>* make( const X& x ) {
                void* where = my_allocator_type().allocate(1);
                return new(where) callback_leaf(x);
            }
#endif
        };

        //! Template for recording construction of objects in table
        /** All maintenance of the space will be done explicitly on push_back,
            and all thread local copies must be destroyed before the concurrent
            vector is deleted.

            The flag is_built is initialized to false.  When the local is
            successfully-constructed, set the flag to true.  If the constructor
            throws, the flag will be false.
        */
        template<typename U>
        struct ets_element {
            tbb::aligned_space<U> my_space;
            bool is_built;
            ets_element() { is_built = false; }  // not currently-built
            U *value() { return my_space.begin(); }
            void unconstruct() { 
                if(is_built) {
                    my_space.begin()->~U();
                    is_built = false;
                }
            }
            ~ets_element() {unconstruct();}
        };

        // A predicate that can be used for a compile-time compatibility check of ETS instances
        // Ideally, it should have been declared inside the ETS class, but unfortunately
        // in that case VS2013 does not enable the variadic constructor.
        template<typename T, typename ETS> struct is_compatible_ets { static const bool value = false; };
        template<typename T, typename U, typename A, ets_key_usage_type C>
        struct is_compatible_ets< T, enumerable_thread_specific<U,A,C> > { static const bool value = internal::is_same_type<T,U>::value; };

#if __TBB_ETS_USE_CPP11
        // A predicate that checks whether, for a variable 'foo' of type T, foo() is a valid expression
        template <typename T>
        class is_callable_no_args {
        private:
            typedef char yes[1];
            typedef char no [2];

            template<typename U> static yes& decide( decltype(declval<U>()())* );
            template<typename U> static no&  decide(...);
        public:
            static const bool value = (sizeof(decide<T>(NULL)) == sizeof(yes));
        };
#endif

    } // namespace internal
    //! @endcond

    //! The enumerable_thread_specific container
    /** enumerable_thread_specific has the following properties:
        - thread-local copies are lazily created, with default, exemplar or function initialization.
        - thread-local copies do not move (during lifetime, and excepting clear()) so the address of a copy is invariant.
        - the contained objects need not have operator=() defined if combine is not used.
        - enumerable_thread_specific containers may be copy-constructed or assigned.
        - thread-local copies can be managed by hash-table, or can be accessed via TLS storage for speed.
        - outside of parallel contexts, the contents of all thread-local copies are accessible by iterator or using combine or combine_each methods

    @par Segmented iterator
        When the thread-local objects are containers with input_iterators defined, a segmented iterator may
        be used to iterate over all the elements of all thread-local copies.

    @par combine and combine_each
        - Both methods are defined for enumerable_thread_specific.
        - combine() requires the type T have operator=() defined.
        - neither method modifies the contents of the object (though there is no guarantee that the applied methods do not modify the object.)
        - Both are evaluated in serial context (the methods are assumed to be non-benign.)

    @ingroup containers */
    template <typename T,
              typename Allocator=cache_aligned_allocator<T>,
              ets_key_usage_type ETS_key_type=ets_no_key >
    class enumerable_thread_specific: internal::ets_base<ETS_key_type> {

        template<typename U, typename A, ets_key_usage_type C> friend class enumerable_thread_specific;

        typedef internal::padded< internal::ets_element<T> > padded_element;

        //! A generic range, used to create range objects from the iterators
        template<typename I>
        class generic_range_type: public blocked_range<I> {
        public:
            typedef T value_type;
            typedef T& reference;
            typedef const T& const_reference;
            typedef I iterator;
            typedef ptrdiff_t difference_type;
            generic_range_type( I begin_, I end_, size_t grainsize_ = 1) : blocked_range<I>(begin_,end_,grainsize_) {}
            template<typename U>
            generic_range_type( const generic_range_type<U>& r) : blocked_range<I>(r.begin(),r.end(),r.grainsize()) {}
            generic_range_type( generic_range_type& r, split ) : blocked_range<I>(r,split()) {}
        };

        typedef typename Allocator::template rebind< padded_element >::other padded_allocator_type;
        typedef tbb::concurrent_vector< padded_element, padded_allocator_type > internal_collection_type;

        internal::callback_base<T> *my_construct_callback;

        internal_collection_type my_locals;

        /*override*/ void* create_local() {
            padded_element* lref = &*my_locals.grow_by(1);
            my_construct_callback->construct(lref->value());
            lref->is_built = true;
            return lref;
        }

        void unconstruct_locals() {
            for(typename internal_collection_type::iterator cvi = my_locals.begin(); cvi != my_locals.end(); ++cvi) {
                cvi->unconstruct();
            }
        }

        typedef typename Allocator::template rebind< uintptr_t >::other array_allocator_type;

        // _size is in bytes
        /*override*/ void* create_array(size_t _size) {
            size_t nelements = (_size + sizeof(uintptr_t) -1) / sizeof(uintptr_t);
            return array_allocator_type().allocate(nelements);
        }

        /*override*/ void free_array( void* _ptr, size_t _size) {
            size_t nelements = (_size + sizeof(uintptr_t) -1) / sizeof(uintptr_t);
            array_allocator_type().deallocate( reinterpret_cast<uintptr_t *>(_ptr),nelements);
        }

    public:

        //! Basic types
        typedef Allocator allocator_type;
        typedef T value_type;
        typedef T& reference;
        typedef const T& const_reference;
        typedef T* pointer;
        typedef const T* const_pointer;
        typedef typename internal_collection_type::size_type size_type;
        typedef typename internal_collection_type::difference_type difference_type;

        // Iterator types
        typedef typename internal::enumerable_thread_specific_iterator< internal_collection_type, value_type > iterator;
        typedef typename internal::enumerable_thread_specific_iterator< internal_collection_type, const value_type > const_iterator;

        // Parallel range types
        typedef generic_range_type< iterator > range_type;
        typedef generic_range_type< const_iterator > const_range_type;

        //! Default constructor.  Each local instance of T is default constructed.
        enumerable_thread_specific() : my_construct_callback(
            internal::callback_leaf<T,internal::construct_by_default<T> >::make(/*dummy argument*/0)
        ){}

        //! Constructor with initializer functor.  Each local instance of T is constructed by T(finit()).
        template <typename Finit
#if __TBB_ETS_USE_CPP11
                  , typename = typename internal::enable_if<internal::is_callable_no_args<typename internal::strip<Finit>::type>::value>::type
#endif
        >
        enumerable_thread_specific( Finit finit ) : my_construct_callback(
            internal::callback_leaf<T,internal::construct_by_finit<T,Finit> >::make( tbb::internal::move(finit) )
        ){}

        //! Constructor with exemplar. Each local instance of T is copy-constructed from the exemplar.
        enumerable_thread_specific( const T& exemplar ) : my_construct_callback(
            internal::callback_leaf<T,internal::construct_by_exemplar<T> >::make( exemplar )
        ){}

#if __TBB_ETS_USE_CPP11
        enumerable_thread_specific( T&& exemplar ) : my_construct_callback(
            internal::callback_leaf<T,internal::construct_by_exemplar<T> >::make( std::move(exemplar) )
        ){}

        //! Variadic constructor with initializer arguments.  Each local instance of T is constructed by T(args...)
        template <typename P1, typename... P,
                  typename = typename internal::enable_if<!internal::is_callable_no_args<typename internal::strip<P1>::type>::value
                                                          && !internal::is_compatible_ets<T, typename internal::strip<P1>::type>::value
                                                          && !internal::is_same_type<T, typename internal::strip<P1>::type>::value
                                                         >::type>
        enumerable_thread_specific( P1&& arg1, P&& ... args ) : my_construct_callback(
            internal::callback_leaf<T,internal::construct_by_args<T,P1,P...> >::make( std::forward<P1>(arg1), std::forward<P>(args)... )
        ){}
#endif

        //! Destructor
        ~enumerable_thread_specific() {
            my_construct_callback->destroy();
            this->clear();  // deallocation before the derived class is finished destructing
            // So free(array *) is still accessible
        }

        //! returns reference to local, discarding exists
        reference local() {
            bool exists;
            return local(exists);
        }

        //! Returns reference to calling thread's local copy, creating one if necessary
        reference local(bool& exists)  {
            void* ptr = this->table_lookup(exists);
            return *(T*)ptr;
        }

        //! Get the number of local copies
        size_type size() const { return my_locals.size(); }

        //! true if there have been no local copies created
        bool empty() const { return my_locals.empty(); }

        //! begin iterator
        iterator begin() { return iterator( my_locals, 0 ); }
        //! end iterator
        iterator end() { return iterator(my_locals, my_locals.size() ); }

        //! begin const iterator
        const_iterator begin() const { return const_iterator(my_locals, 0); }

        //! end const iterator
        const_iterator end() const { return const_iterator(my_locals, my_locals.size()); }

        //! Get range for parallel algorithms
        range_type range( size_t grainsize=1 ) { return range_type( begin(), end(), grainsize ); }

        //! Get const range for parallel algorithms
        const_range_type range( size_t grainsize=1 ) const { return const_range_type( begin(), end(), grainsize ); }

        //! Destroys local copies
        void clear() {
            unconstruct_locals();
            my_locals.clear();
            this->table_clear();
            // callback is not destroyed
            // exemplar is not destroyed
        }

    private:

        template<typename A2, ets_key_usage_type C2>
        void internal_copy( const enumerable_thread_specific<T, A2, C2>& other);

    public:

        template<typename Alloc, ets_key_usage_type Cachetype>
        enumerable_thread_specific( const enumerable_thread_specific<T, Alloc, Cachetype>& other ) : internal::ets_base<ETS_key_type> ()
        {
            internal_copy(other);
        }

        enumerable_thread_specific( const enumerable_thread_specific& other ) : internal::ets_base<ETS_key_type> ()
        {
            internal_copy(other);
        }
        // TODO: add move constructors

    private:

        template<typename A2, ets_key_usage_type C2>
        enumerable_thread_specific &
        internal_assign(const enumerable_thread_specific<T, A2, C2>& other) {
            if(static_cast<void *>( this ) != static_cast<const void *>( &other )) {
                this->clear();
                my_construct_callback->destroy();
                my_construct_callback = 0;
                internal_copy( other );
            }
            return *this;
        }

    public:

        // assignment
        enumerable_thread_specific& operator=(const enumerable_thread_specific& other) {
            return internal_assign(other);
        }

        template<typename Alloc, ets_key_usage_type Cachetype>
        enumerable_thread_specific& operator=(const enumerable_thread_specific<T, Alloc, Cachetype>& other)
        {
            return internal_assign(other);
        }
        // TODO: add move assignments

        // combine_func_t has signature T(T,T) or T(const T&, const T&)
        template <typename combine_func_t>
        T combine(combine_func_t f_combine) {
            if(begin() == end()) {
                internal::ets_element<T> location;
                my_construct_callback->construct(location.value());
                location.is_built = true;
                return *location.value();
            }
            const_iterator ci = begin();
            T my_result = *ci;
            while(++ci != end())
                my_result = f_combine( my_result, *ci );
            return my_result;
        }

        // combine_func_t takes T by value or by [const] reference, and returns nothing
        template <typename combine_func_t>
        void combine_each(combine_func_t f_combine) {
            for(iterator ci = begin(); ci != end(); ++ci) {
                f_combine( *ci );
            }
        }

    }; // enumerable_thread_specific

    template <typename T, typename Allocator, ets_key_usage_type ETS_key_type>
    template<typename A2, ets_key_usage_type C2>
    void enumerable_thread_specific<T,Allocator,ETS_key_type>::internal_copy( const enumerable_thread_specific<T, A2, C2>& other) {
#if __TBB_ETS_USE_CPP11
        __TBB_STATIC_ASSERT( (internal::is_compatible_ets<T, typename internal::strip<decltype(other)>::type>::value), "Maybe is_compatible_ets works incorrectly" );
#endif
        // Initialize my_construct_callback first, so that it is valid even if rest of this routine throws an exception.
        my_construct_callback = other.my_construct_callback->clone();

        typedef internal::ets_base<ets_no_key> base;
        __TBB_ASSERT(my_locals.size()==0,NULL);
        this->table_reserve_for_copy( other );
        for( base::array* r=other.my_root; r; r=r->next ) {
            for( size_t i=0; i<r->size(); ++i ) {
                base::slot& s1 = r->at(i);
                if( !s1.empty() ) {
                    base::slot& s2 = this->table_find(s1.key);
                    if( s2.empty() ) {
                        void* lref = &*my_locals.grow_by(1);
                        s2.ptr = new(lref) T(*(T*)s1.ptr);
                        s2.key = s1.key;
                    } else {
                        // Skip the duplicate
                    }
                }
            }
        }
    }

    template< typename Container >
    class flattened2d {

        // This intermediate typedef is to address issues with VC7.1 compilers
        typedef typename Container::value_type conval_type;

    public:

        //! Basic types
        typedef typename conval_type::size_type size_type;
        typedef typename conval_type::difference_type difference_type;
        typedef typename conval_type::allocator_type allocator_type;
        typedef typename conval_type::value_type value_type;
        typedef typename conval_type::reference reference;
        typedef typename conval_type::const_reference const_reference;
        typedef typename conval_type::pointer pointer;
        typedef typename conval_type::const_pointer const_pointer;

        typedef typename internal::segmented_iterator<Container, value_type> iterator;
        typedef typename internal::segmented_iterator<Container, const value_type> const_iterator;

        flattened2d( const Container &c, typename Container::const_iterator b, typename Container::const_iterator e ) :
            my_container(const_cast<Container*>(&c)), my_begin(b), my_end(e) { }

        flattened2d( const Container &c ) :
            my_container(const_cast<Container*>(&c)), my_begin(c.begin()), my_end(c.end()) { }

        iterator begin() { return iterator(*my_container) = my_begin; }
        iterator end() { return iterator(*my_container) = my_end; }
        const_iterator begin() const { return const_iterator(*my_container) = my_begin; }
        const_iterator end() const { return const_iterator(*my_container) = my_end; }

        size_type size() const {
            size_type tot_size = 0;
            for(typename Container::const_iterator i = my_begin; i != my_end; ++i) {
                tot_size += i->size();
            }
            return tot_size;
        }

    private:

        Container *my_container;
        typename Container::const_iterator my_begin;
        typename Container::const_iterator my_end;

    };

    template <typename Container>
    flattened2d<Container> flatten2d(const Container &c, const typename Container::const_iterator b, const typename Container::const_iterator e) {
        return flattened2d<Container>(c, b, e);
    }

    template <typename Container>
    flattened2d<Container> flatten2d(const Container &c) {
        return flattened2d<Container>(c);
    }

} // interface6

namespace internal {
using interface6::internal::segmented_iterator;
}

using interface6::enumerable_thread_specific;
using interface6::flattened2d;
using interface6::flatten2d;

} // namespace tbb

#endif