_ZN7roaring7RoaringC2EmPKj: 85| 12.9k| Roaring(size_t n, const uint32_t *data) : Roaring() { 86| 12.9k| api::roaring_bitmap_add_many(&roaring, n, data); 87| 12.9k| } _ZN7roaring7RoaringC2Ev: 76| 19.4k| Roaring() : roaring{} { 77| | // The empty constructor roaring{} silences warnings from pedantic 78| | // static analyzers. 79| 19.4k| api::roaring_bitmap_init_cleared(&roaring); 80| 19.4k| } _ZN7roaring7RoaringD2Ev: 913| 58.3k| ~Roaring() { 914| 58.3k| if (!(roaring.high_low_container.flags & ROARING_FLAG_FROZEN)) { ------------------ | | 47| 58.3k|#define ROARING_FLAG_FROZEN UINT8_C(0x2) ------------------ | Branch (914:13): [True: 58.3k, False: 0] ------------------ 915| 58.3k| api::roaring_bitmap_clear(&roaring); 916| 58.3k| } else { 917| | // The roaring member variable copies the `roaring_bitmap_t` and 918| | // nested `roaring_array_t` structures by value and is freed in the 919| | // constructor, however the underlying memory arena used for the 920| | // container data is not freed with it. Here we derive the arena 921| | // pointer from the second arena allocation in 922| | // `roaring_bitmap_frozen_view` and free it as well. 923| 0| roaring_bitmap_free( 924| 0| (roaring_bitmap_t *)((char *) 925| 0| roaring.high_low_container.containers - 926| 0| sizeof(roaring_bitmap_t))); 927| 0| } 928| 58.3k| } _ZN7roaring7Roaring11runOptimizeEv: 493| 12.9k| bool runOptimize() noexcept { 494| 12.9k| return api::roaring_bitmap_run_optimize(&roaring); 495| 12.9k| } _ZN7roaring7Roaring11shrinkToFitEv: 501| 6.48k| size_t shrinkToFit() noexcept { 502| 6.48k| return api::roaring_bitmap_shrink_to_fit(&roaring); 503| 6.48k| } _ZN7roaring7Roaring3addEj: 198| 6.48k| void add(uint32_t x) noexcept { api::roaring_bitmap_add(&roaring, x); } _ZN7roaring7Roaring10addCheckedEj: 205| 6.48k| bool addChecked(uint32_t x) noexcept { 206| 6.48k| return api::roaring_bitmap_add_checked(&roaring, x); 207| 6.48k| } _ZN7roaring7Roaring8addRangeEmm: 212| 6.48k| void addRange(const uint64_t min, const uint64_t max) noexcept { 213| 6.48k| return api::roaring_bitmap_add_range(&roaring, min, max); 214| 6.48k| } _ZN7roaring7Roaring7addManyEmPKj: 226| 6.48k| void addMany(size_t n_args, const uint32_t *vals) noexcept { 227| 6.48k| api::roaring_bitmap_add_many(&roaring, n_args, vals); 228| 6.48k| } _ZN7roaring7Roaring6removeEj: 258| 6.48k| void remove(uint32_t x) noexcept { 259| 6.48k| api::roaring_bitmap_remove(&roaring, x); 260| 6.48k| } _ZN7roaring7Roaring13removeCheckedEj: 267| 6.48k| bool removeChecked(uint32_t x) noexcept { 268| 6.48k| return api::roaring_bitmap_remove_checked(&roaring, x); 269| 6.48k| } _ZN7roaring7Roaring11removeRangeEmm: 274| 6.48k| void removeRange(uint64_t min, uint64_t max) noexcept { 275| 6.48k| return api::roaring_bitmap_remove_range(&roaring, min, max); 276| 6.48k| } _ZN7roaring7Roaring17removeRangeClosedEjj: 281| 6.48k| void removeRangeClosed(uint32_t min, uint32_t max) noexcept { 282| 6.48k| return api::roaring_bitmap_remove_range_closed(&roaring, min, max); 283| 6.48k| } _ZNK7roaring7Roaring7maximumEv: 315| 6.48k| uint32_t maximum() const noexcept { 316| 6.48k| return api::roaring_bitmap_maximum(&roaring); 317| 6.48k| } _ZNK7roaring7Roaring7minimumEv: 322| 6.48k| uint32_t minimum() const noexcept { 323| 6.48k| return api::roaring_bitmap_minimum(&roaring); 324| 6.48k| } _ZNK7roaring7Roaring8containsEj: 329| 6.48k| bool contains(uint32_t x) const noexcept { 330| 6.48k| return api::roaring_bitmap_contains(&roaring, x); 331| 6.48k| } _ZNK7roaring7Roaring13containsRangeEmm: 336| 6.48k| bool containsRange(const uint64_t x, const uint64_t y) const noexcept { 337| 6.48k| return api::roaring_bitmap_contains_range(&roaring, x, y); 338| 6.48k| } _ZNK7roaring7Roaring6selectEjPj: 526| 6.48k| bool select(uint32_t rnk, uint32_t *element) const noexcept { 527| 6.48k| return api::roaring_bitmap_select(&roaring, rnk, element); 528| 6.48k| } _ZNK7roaring7Roaring9intersectERKS0_: 540| 6.48k| bool intersect(const Roaring &r) const noexcept { 541| 6.48k| return api::roaring_bitmap_intersect(&roaring, &r.roaring); 542| 6.48k| } _ZNK7roaring7Roaring13jaccard_indexERKS0_: 551| 6.48k| double jaccard_index(const Roaring &r) const noexcept { 552| 6.48k| return api::roaring_bitmap_jaccard_index(&roaring, &r.roaring); 553| 6.48k| } _ZNK7roaring7Roaring14or_cardinalityERKS0_: 558| 6.48k| uint64_t or_cardinality(const Roaring &r) const noexcept { 559| 6.48k| return api::roaring_bitmap_or_cardinality(&roaring, &r.roaring); 560| 6.48k| } _ZNK7roaring7Roaring18andnot_cardinalityERKS0_: 565| 6.48k| uint64_t andnot_cardinality(const Roaring &r) const noexcept { 566| 6.48k| return api::roaring_bitmap_andnot_cardinality(&roaring, &r.roaring); 567| 6.48k| } _ZNK7roaring7Roaring15xor_cardinalityERKS0_: 573| 6.48k| uint64_t xor_cardinality(const Roaring &r) const noexcept { 574| 6.48k| return api::roaring_bitmap_xor_cardinality(&roaring, &r.roaring); 575| 6.48k| } _ZNK7roaring7Roaring4rankEj: 585| 6.48k| uint64_t rank(uint32_t x) const noexcept { 586| 6.48k| return api::roaring_bitmap_rank(&roaring, x); 587| 6.48k| } _ZNK7roaring7Roaring14getSizeInBytesEb: 732| 12.9k| size_t getSizeInBytes(bool portable = true) const noexcept { 733| 12.9k| if (portable) { ------------------ | Branch (733:13): [True: 12.9k, False: 0] ------------------ 734| 12.9k| return api::roaring_bitmap_portable_size_in_bytes(&roaring); 735| 12.9k| } else { 736| 0| return api::roaring_bitmap_size_in_bytes(&roaring); 737| 0| } 738| 12.9k| } _ZNK7roaring7RoaringanERKS0_: 794| 6.48k| Roaring operator&(const Roaring &o) const { 795| 6.48k| roaring_bitmap_t *r = api::roaring_bitmap_and(&roaring, &o.roaring); 796| 6.48k| if (r == NULL) { ------------------ | Branch (796:13): [True: 0, False: 6.48k] ------------------ 797| 0| ROARING_TERMINATE("failed materialization in and"); ------------------ | | 27| 0|#define ROARING_TERMINATE(_s) throw std::runtime_error(_s) ------------------ 798| 0| } 799| 6.48k| return Roaring(r); 800| 6.48k| } _ZN7roaring7RoaringC2EPNS_3api16roaring_bitmap_sE: 102| 32.4k| explicit Roaring(roaring_bitmap_t *s) noexcept : roaring(*s) { 103| 32.4k| roaring_free(s); // deallocate the passed-in pointer 104| 32.4k| } _ZNK7roaring7RoaringmiERKS0_: 807| 6.48k| Roaring operator-(const Roaring &o) const { 808| 6.48k| roaring_bitmap_t *r = api::roaring_bitmap_andnot(&roaring, &o.roaring); 809| 6.48k| if (r == NULL) { ------------------ | Branch (809:13): [True: 0, False: 6.48k] ------------------ 810| 0| ROARING_TERMINATE("failed materialization in andnot"); ------------------ | | 27| 0|#define ROARING_TERMINATE(_s) throw std::runtime_error(_s) ------------------ 811| 0| } 812| 6.48k| return Roaring(r); 813| 6.48k| } _ZNK7roaring7RoaringorERKS0_: 820| 6.48k| Roaring operator|(const Roaring &o) const { 821| 6.48k| roaring_bitmap_t *r = api::roaring_bitmap_or(&roaring, &o.roaring); 822| 6.48k| if (r == NULL) { ------------------ | Branch (822:13): [True: 0, False: 6.48k] ------------------ 823| 0| ROARING_TERMINATE("failed materialization in or"); ------------------ | | 27| 0|#define ROARING_TERMINATE(_s) throw std::runtime_error(_s) ------------------ 824| 0| } 825| 6.48k| return Roaring(r); 826| 6.48k| } _ZNK7roaring7RoaringeoERKS0_: 833| 6.48k| Roaring operator^(const Roaring &o) const { 834| 6.48k| roaring_bitmap_t *r = api::roaring_bitmap_xor(&roaring, &o.roaring); 835| 6.48k| if (r == NULL) { ------------------ | Branch (835:13): [True: 0, False: 6.48k] ------------------ 836| 0| ROARING_TERMINATE("failed materialization in xor"); ------------------ | | 27| 0|#define ROARING_TERMINATE(_s) throw std::runtime_error(_s) ------------------ 837| 0| } 838| 6.48k| return Roaring(r); 839| 6.48k| } _ZN7roaring7RoaringoRERKS0_: 378| 6.48k| Roaring &operator|=(const Roaring &r) noexcept { 379| 6.48k| api::roaring_bitmap_or_inplace(&roaring, &r.roaring); 380| 6.48k| return *this; 381| 6.48k| } _ZN7roaring7RoaringaNERKS0_: 356| 6.48k| Roaring &operator&=(const Roaring &r) noexcept { 357| 6.48k| api::roaring_bitmap_and_inplace(&roaring, &r.roaring); 358| 6.48k| return *this; 359| 6.48k| } _ZN7roaring7RoaringmIERKS0_: 366| 6.48k| Roaring &operator-=(const Roaring &r) noexcept { 367| 6.48k| api::roaring_bitmap_andnot_inplace(&roaring, &r.roaring); 368| 6.48k| return *this; 369| 6.48k| } _ZN7roaring7RoaringeOERKS0_: 388| 6.48k| Roaring &operator^=(const Roaring &r) noexcept { 389| 6.48k| api::roaring_bitmap_xor_inplace(&roaring, &r.roaring); 390| 6.48k| return *this; 391| 6.48k| } _ZNK7roaring7RoaringeqERKS0_: 458| 12.9k| bool operator==(const Roaring &r) const noexcept { 459| 12.9k| return api::roaring_bitmap_equals(&roaring, &r.roaring); 460| 12.9k| } _ZNK7roaring7Roaring11cardinalityEv: 401| 6.48k| uint64_t cardinality() const noexcept { 402| 6.48k| return api::roaring_bitmap_get_cardinality(&roaring); 403| 6.48k| } _ZNK7roaring7Roaring7isEmptyEv: 408| 12.9k| bool isEmpty() const noexcept { 409| 12.9k| return api::roaring_bitmap_is_empty(&roaring); 410| 12.9k| } _ZNK7roaring7Roaring13toUint32ArrayEPj: 441| 6.48k| void toUint32Array(uint32_t *ans) const noexcept { 442| 6.48k| api::roaring_bitmap_to_uint32_array(&roaring, ans); 443| 6.48k| } _ZNK7roaring7Roaring8isSubsetERKS0_: 425| 6.48k| bool isSubset(const Roaring &r) const noexcept { 426| 6.48k| return api::roaring_bitmap_is_subset(&roaring, &r.roaring); 427| 6.48k| } _ZNK7roaring7Roaring14isStrictSubsetERKS0_: 432| 6.48k| bool isStrictSubset(const Roaring &r) const noexcept { 433| 6.48k| return api::roaring_bitmap_is_strict_subset(&roaring, &r.roaring); 434| 6.48k| } _ZN7roaring7Roaring4flipEmm: 466| 6.48k| void flip(uint64_t range_start, uint64_t range_end) noexcept { 467| 6.48k| api::roaring_bitmap_flip_inplace(&roaring, range_start, range_end); 468| 6.48k| } _ZN7roaring7Roaring10flipClosedEjj: 474| 6.48k| void flipClosed(uint32_t range_start, uint32_t range_end) noexcept { 475| 6.48k| api::roaring_bitmap_flip_inplace_closed(&roaring, range_start, 476| 6.48k| range_end); 477| 6.48k| } _ZN7roaring7Roaring20removeRunCompressionEv: 483| 6.48k| bool removeRunCompression() noexcept { 484| 6.48k| return api::roaring_bitmap_remove_run_compression(&roaring); 485| 6.48k| } _ZN7roaring7RoaringC2ERKS0_: 110| 6.48k| Roaring(const Roaring &r) : Roaring() { 111| 6.48k| if (!api::roaring_bitmap_overwrite(&roaring, &r.roaring)) { ------------------ | Branch (111:13): [True: 0, False: 6.48k] ------------------ 112| 0| ROARING_TERMINATE("failed roaring_bitmap_overwrite in constructor"); ------------------ | | 27| 0|#define ROARING_TERMINATE(_s) throw std::runtime_error(_s) ------------------ 113| 0| } 114| 6.48k| api::roaring_bitmap_set_copy_on_write( 115| 6.48k| &roaring, api::roaring_bitmap_get_copy_on_write(&r.roaring)); 116| 6.48k| } _ZN7roaring7RoaringC2EOS0_: 122| 6.48k| Roaring(Roaring &&r) noexcept : roaring(r.roaring) { 123| | // 124| | // !!! This clones the bits of the roaring structure to a new location 125| | // and then overwrites the old bits...assuming that this will still 126| | // work. There are scenarios where this could break; e.g. if some of 127| | // those bits were pointers into the structure memory itself. If such 128| | // things were possible, a roaring_bitmap_move() API would be needed. 129| | // 130| 6.48k| api::roaring_bitmap_init_cleared(&r.roaring); 131| 6.48k| } _ZN7roaring7RoaringaSERKS0_: 152| 6.48k| Roaring &operator=(const Roaring &r) { 153| 6.48k| if (!api::roaring_bitmap_overwrite(&roaring, &r.roaring)) { ------------------ | Branch (153:13): [True: 0, False: 6.48k] ------------------ 154| 0| ROARING_TERMINATE("failed memory alloc in assignment"); ------------------ | | 27| 0|#define ROARING_TERMINATE(_s) throw std::runtime_error(_s) ------------------ 155| 0| } 156| 6.48k| api::roaring_bitmap_set_copy_on_write( 157| 6.48k| &roaring, api::roaring_bitmap_get_copy_on_write(&r.roaring)); 158| 6.48k| return *this; 159| 6.48k| } _ZN7roaring7RoaringaSEOS0_: 165| 6.48k| Roaring &operator=(Roaring &&r) noexcept { 166| 6.48k| api::roaring_bitmap_clear(&roaring); // free this class's allocations 167| | 168| | // !!! See notes in the Move Constructor regarding roaring_bitmap_move() 169| | // 170| 6.48k| roaring = r.roaring; 171| 6.48k| api::roaring_bitmap_init_cleared(&r.roaring); 172| | 173| 6.48k| return *this; 174| 6.48k| } _ZN7roaring7Roaring8readSafeEPKcm: 706| 12.9k| static Roaring readSafe(const char *buf, size_t maxbytes) { 707| 12.9k| roaring_bitmap_t *r = 708| 12.9k| api::roaring_bitmap_portable_deserialize_safe(buf, maxbytes); 709| 12.9k| if (r == NULL) { ------------------ | Branch (709:13): [True: 6.46k, False: 6.49k] ------------------ 710| 6.46k| ROARING_TERMINATE("failed alloc while reading"); ------------------ | | 27| 6.46k|#define ROARING_TERMINATE(_s) throw std::runtime_error(_s) ------------------ 711| 6.46k| } 712| 6.49k| return Roaring(r); 713| 12.9k| } _ZNK7roaring7Roaring5writeEPcb: 649| 6.48k| size_t write(char *buf, bool portable = true) const noexcept { 650| 6.48k| if (portable) { ------------------ | Branch (650:13): [True: 6.48k, False: 0] ------------------ 651| 6.48k| return api::roaring_bitmap_portable_serialize(&roaring, buf); 652| 6.48k| } else { 653| 0| return api::roaring_bitmap_serialize(&roaring, buf); 654| 0| } 655| 6.48k| } _ZNK7roaring7Roaring8toStringEv: 856| 6.48k| std::string toString() const noexcept { 857| 6.48k| struct iter_data { 858| 6.48k| std::string str{}; // The empty constructor silences warnings from 859| | // pedantic static analyzers. 860| 6.48k| char first_char = '{'; 861| 6.48k| } outer_iter_data; 862| 6.48k| if (!isEmpty()) { ------------------ | Branch (862:13): [True: 6.46k, False: 15] ------------------ 863| 6.46k| iterate( 864| 6.46k| [](uint32_t value, void *inner_iter_data) -> bool { 865| 6.46k| ((iter_data *)inner_iter_data)->str += 866| 6.46k| ((iter_data *)inner_iter_data)->first_char; 867| 6.46k| ((iter_data *)inner_iter_data)->str += 868| 6.46k| std::to_string(value); 869| 6.46k| ((iter_data *)inner_iter_data)->first_char = ','; 870| 6.46k| return true; 871| 6.46k| }, 872| 6.46k| (void *)&outer_iter_data); 873| 6.46k| } else 874| 15| outer_iter_data.str = '{'; 875| 6.48k| outer_iter_data.str += '}'; 876| 6.48k| return outer_iter_data.str; 877| 6.48k| } _ZNK7roaring7Roaring7iterateEPFbjPvES1_: 514| 6.46k| void iterate(api::roaring_iterator iterator, void *ptr) const { 515| 6.46k| api::roaring_iterate(&roaring, iterator, ptr); 516| 6.46k| } _ZZNK7roaring7Roaring8toStringEvENKUljPvE_clEjS1_: 864| 3.98G| [](uint32_t value, void *inner_iter_data) -> bool { 865| 3.98G| ((iter_data *)inner_iter_data)->str += 866| 3.98G| ((iter_data *)inner_iter_data)->first_char; 867| 3.98G| ((iter_data *)inner_iter_data)->str += 868| 3.98G| std::to_string(value); 869| 3.98G| ((iter_data *)inner_iter_data)->first_char = ','; 870| 3.98G| return true; 871| 3.98G| }, _ZNK7roaring7Roaring5beginEv: 1073| 12.9k|inline RoaringSetBitBiDirectionalIterator Roaring::begin() const { 1074| 12.9k| return RoaringSetBitBiDirectionalIterator(*this); 1075| 12.9k|} _ZN7roaring34RoaringSetBitBiDirectionalIteratorC2ERKNS_7RoaringEb: 967| 12.9k| bool exhausted = false) { 968| 12.9k| if (exhausted) { ------------------ | Branch (968:13): [True: 1, False: 12.9k] ------------------ 969| 1| i.parent = &parent.roaring; 970| 1| i.container_index = INT32_MAX; 971| 1| i.has_value = false; 972| 1| i.current_value = UINT32_MAX; 973| 12.9k| } else { 974| 12.9k| api::roaring_iterator_init(&parent.roaring, &i); 975| 12.9k| } 976| 12.9k| } _ZNK7roaring34RoaringSetBitBiDirectionalIteratorneERKS0_: 1064| 342k| bool operator!=(const RoaringSetBitBiDirectionalIterator &o) const { 1065| 342k| return i.current_value != *o || i.has_value != o.i.has_value; ------------------ | Branch (1065:16): [True: 335k, False: 6.48k] | Branch (1065:41): [True: 0, False: 6.48k] ------------------ 1066| 342k| } _ZNK7roaring34RoaringSetBitBiDirectionalIteratordeEv: 981| 342k| value_type operator*() const { return i.current_value; } _ZNK7roaring7Roaring3endEv: 1077| 342k|inline RoaringSetBitBiDirectionalIterator &Roaring::end() const { 1078| 342k| static RoaringSetBitBiDirectionalIterator e(*this, true); 1079| 342k| return e; 1080| 342k|} _ZN7roaring34RoaringSetBitBiDirectionalIteratorppEi: 1012| 335k| type_of_iterator operator++(int) { // i++, must return orig. value 1013| 335k| RoaringSetBitBiDirectionalIterator orig(*this); 1014| 335k| api::roaring_uint32_iterator_advance(&i); 1015| 335k| return orig; 1016| 335k| } _ZN7roaring34RoaringSetBitBiDirectionalIterator13equalorlargerEj: 1027| 6.48k| CROARING_DEPRECATED void equalorlarger(uint32_t val) { 1028| 6.48k| api::roaring_uint32_iterator_move_equalorlarger(&i, val); 1029| 6.48k| } binarySearch: 52| 7.41M| uint16_t ikey) { 53| 7.41M| const int32_t gap = 16; 54| 7.41M| if (lenarray < gap) { ------------------ | Branch (54:9): [True: 2.82M, False: 4.58M] ------------------ 55| 4.32M| for (int32_t j = 0; j < lenarray; j++) { ------------------ | Branch (55:29): [True: 4.31M, False: 7.53k] ------------------ 56| 4.31M| if (array[j] >= ikey) { ------------------ | Branch (56:17): [True: 2.81M, False: 1.49M] ------------------ 57| 2.81M| return (array[j] == ikey) ? j : -(j + 1); ------------------ | Branch (57:24): [True: 2.66M, False: 153k] ------------------ 58| 2.81M| } 59| 4.31M| } 60| 7.53k| return -(lenarray + 1); 61| 2.82M| } 62| 4.58M| const int32_t num_blocks = lenarray / gap; 63| 4.58M| int32_t base = 0; 64| 4.58M| int32_t n = num_blocks; 65| 9.25M| while (n > 3) { ------------------ | Branch (65:12): [True: 4.66M, False: 4.58M] ------------------ 66| 4.66M| int32_t quarter = n >> 2; 67| | 68| 4.66M| int32_t k1 = array[(base + quarter + 1) * gap - 1]; 69| 4.66M| int32_t k2 = array[(base + 2 * quarter + 1) * gap - 1]; 70| 4.66M| int32_t k3 = array[(base + 3 * quarter + 1) * gap - 1]; 71| | 72| 4.66M| int32_t c1 = (k1 < ikey); 73| 4.66M| int32_t c2 = (k2 < ikey); 74| 4.66M| int32_t c3 = (k3 < ikey); 75| | 76| 4.66M| base += (c1 + c2 + c3) * quarter; 77| 4.66M| n -= 3 * quarter; 78| 4.66M| } 79| 8.50M| while (n > 1) { ------------------ | Branch (79:12): [True: 3.92M, False: 4.58M] ------------------ 80| 3.92M| int32_t half = n >> 1; 81| 3.92M| base = (array[(base + half + 1) * gap - 1] < ikey) ? base + half : base; ------------------ | Branch (81:16): [True: 550k, False: 3.37M] ------------------ 82| 3.92M| n -= half; 83| 3.92M| } 84| 4.58M| int32_t lo = (array[(base + 1) * gap - 1] < ikey) ? base + 1 : base; ------------------ | Branch (84:18): [True: 2.14M, False: 2.43M] ------------------ 85| | 86| 4.58M| if (lo < num_blocks) { ------------------ | Branch (86:9): [True: 4.25M, False: 325k] ------------------ 87| 4.25M| const int32_t start = lo * gap; 88| 4.25M|#if defined(CROARING_IS_X64) 89| | // SSE2: subs_epu16 yields zero where lane >= ikey. movemask of an 90| | // epi16 compare gives 2 bits per lane; ctz>>1 = lane index. Scan 91| | // the first 8 lanes first and exit early when they contain the 92| | // answer; otherwise the block-narrowing invariant guarantees the 93| | // second-half mask is non-zero. 94| 4.25M| __m128i needle = _mm_set1_epi16((short)ikey); 95| 4.25M| __m128i zero = _mm_setzero_si128(); 96| 4.25M| __m128i v0 = _mm_loadu_si128((const __m128i *)(array + start)); 97| 4.25M| __m128i ge0 = _mm_cmpeq_epi16(_mm_subs_epu16(needle, v0), zero); 98| 4.25M| unsigned m0 = (unsigned)_mm_movemask_epi8(ge0); 99| 4.25M| if (m0 != 0) { ------------------ | Branch (99:13): [True: 3.02M, False: 1.23M] ------------------ 100| 3.02M| int32_t j = start + (int32_t)(roaring_trailing_zeroes(m0) >> 1); 101| 3.02M| return (array[j] == ikey) ? j : -(j + 1); ------------------ | Branch (101:20): [True: 2.24M, False: 786k] ------------------ 102| 3.02M| } 103| 1.23M| __m128i v1 = _mm_loadu_si128((const __m128i *)(array + start + 8)); 104| 1.23M| __m128i ge1 = _mm_cmpeq_epi16(_mm_subs_epu16(needle, v1), zero); 105| 1.23M| unsigned m1 = (unsigned)_mm_movemask_epi8(ge1); 106| 1.23M| int32_t j = start + 8 + (int32_t)(roaring_trailing_zeroes(m1) >> 1); 107| 1.23M| return (array[j] == ikey) ? j : -(j + 1); ------------------ | Branch (107:16): [True: 533k, False: 698k] ------------------ 108| |#else 109| | const int32_t end = start + gap; 110| | for (int32_t j = start; j < end; j++) { 111| | if (array[j] >= ikey) { 112| | return (array[j] == ikey) ? j : -(j + 1); 113| | } 114| | } 115| | // Unreachable: the narrowing guarantees the last element of the 116| | // selected block is >= ikey. 117| | return -(end + 1); 118| |#endif 119| 4.25M| } 120| | 121| 1.86M| for (int32_t j = num_blocks * gap; j < lenarray; j++) { ------------------ | Branch (121:40): [True: 1.85M, False: 11.0k] ------------------ 122| 1.85M| if (array[j] >= ikey) { ------------------ | Branch (122:13): [True: 314k, False: 1.54M] ------------------ 123| 314k| return (array[j] == ikey) ? j : -(j + 1); ------------------ | Branch (123:20): [True: 172k, False: 141k] ------------------ 124| 314k| } 125| 1.85M| } 126| 11.0k| return -(lenarray + 1); 127| 325k|} roaring.c:advanceUntil: 137| 518| int32_t length, uint16_t min) { 138| 518| int32_t lower = pos + 1; 139| | 140| 518| if ((lower >= length) || (array[lower] >= min)) { ------------------ | Branch (140:9): [True: 518, False: 0] | Branch (140:30): [True: 0, False: 0] ------------------ 141| 518| return lower; 142| 518| } 143| | 144| 0| int32_t spansize = 1; 145| | 146| 0| while ((lower + spansize < length) && (array[lower + spansize] < min)) { ------------------ | Branch (146:12): [True: 0, False: 0] | Branch (146:43): [True: 0, False: 0] ------------------ 147| 0| spansize <<= 1; 148| 0| } 149| 0| int32_t upper = (lower + spansize < length) ? lower + spansize : length - 1; ------------------ | Branch (149:21): [True: 0, False: 0] ------------------ 150| | 151| 0| if (array[upper] == min) { ------------------ | Branch (151:9): [True: 0, False: 0] ------------------ 152| 0| return upper; 153| 0| } 154| 0| if (array[upper] < min) { ------------------ | Branch (154:9): [True: 0, False: 0] ------------------ 155| | // means 156| | // array 157| | // has no 158| | // item 159| | // >= min 160| | // pos = array.length; 161| 0| return length; 162| 0| } 163| | 164| | // we know that the next-smallest span was too small 165| 0| lower += (spansize >> 1); 166| | 167| 0| int32_t mid = 0; 168| 0| while (lower + 1 != upper) { ------------------ | Branch (168:12): [True: 0, False: 0] ------------------ 169| 0| mid = (lower + upper) >> 1; 170| 0| if (array[mid] == min) { ------------------ | Branch (170:13): [True: 0, False: 0] ------------------ 171| 0| return mid; 172| 0| } else if (array[mid] < min) { ------------------ | Branch (172:20): [True: 0, False: 0] ------------------ 173| 0| lower = mid; 174| 0| } else { 175| 0| upper = mid; 176| 0| } 177| 0| } 178| 0| return upper; 179| 0|} roaring.c:count_greater: 197| 9.36k| uint16_t ikey) { 198| 9.36k| if (lenarray == 0) return 0; ------------------ | Branch (198:9): [True: 0, False: 9.36k] ------------------ 199| 9.36k| int32_t pos = binarySearch(array, lenarray, ikey); 200| 9.36k| if (pos >= 0) { ------------------ | Branch (200:9): [True: 2.04k, False: 7.31k] ------------------ 201| 2.04k| return lenarray - (pos + 1); 202| 7.31k| } else { 203| 7.31k| return lenarray - (-pos - 1); 204| 7.31k| } 205| 9.36k|} roaring.c:count_less: 186| 15.9k| uint16_t ikey) { 187| 15.9k| if (lenarray == 0) return 0; ------------------ | Branch (187:9): [True: 3.07k, False: 12.8k] ------------------ 188| 12.8k| int32_t pos = binarySearch(array, lenarray, ikey); 189| 12.8k| return pos >= 0 ? pos : -(pos + 1); ------------------ | Branch (189:12): [True: 9.72k, False: 3.12k] ------------------ 190| 15.9k|} array_util.c:advanceUntil: 137| 20.3k| int32_t length, uint16_t min) { 138| 20.3k| int32_t lower = pos + 1; 139| | 140| 20.3k| if ((lower >= length) || (array[lower] >= min)) { ------------------ | Branch (140:9): [True: 25, False: 20.3k] | Branch (140:30): [True: 6.97k, False: 13.3k] ------------------ 141| 7.00k| return lower; 142| 7.00k| } 143| | 144| 13.3k| int32_t spansize = 1; 145| | 146| 58.1k| while ((lower + spansize < length) && (array[lower + spansize] < min)) { ------------------ | Branch (146:12): [True: 56.7k, False: 1.41k] | Branch (146:43): [True: 44.7k, False: 11.9k] ------------------ 147| 44.7k| spansize <<= 1; 148| 44.7k| } 149| 13.3k| int32_t upper = (lower + spansize < length) ? lower + spansize : length - 1; ------------------ | Branch (149:21): [True: 11.9k, False: 1.41k] ------------------ 150| | 151| 13.3k| if (array[upper] == min) { ------------------ | Branch (151:9): [True: 3.06k, False: 10.3k] ------------------ 152| 3.06k| return upper; 153| 3.06k| } 154| 10.3k| if (array[upper] < min) { ------------------ | Branch (154:9): [True: 213, False: 10.0k] ------------------ 155| | // means 156| | // array 157| | // has no 158| | // item 159| | // >= min 160| | // pos = array.length; 161| 213| return length; 162| 213| } 163| | 164| | // we know that the next-smallest span was too small 165| 10.0k| lower += (spansize >> 1); 166| | 167| 10.0k| int32_t mid = 0; 168| 31.9k| while (lower + 1 != upper) { ------------------ | Branch (168:12): [True: 27.0k, False: 4.92k] ------------------ 169| 27.0k| mid = (lower + upper) >> 1; 170| 27.0k| if (array[mid] == min) { ------------------ | Branch (170:13): [True: 5.17k, False: 21.8k] ------------------ 171| 5.17k| return mid; 172| 21.8k| } else if (array[mid] < min) { ------------------ | Branch (172:20): [True: 11.0k, False: 10.8k] ------------------ 173| 11.0k| lower = mid; 174| 11.0k| } else { 175| 10.8k| upper = mid; 176| 10.8k| } 177| 27.0k| } 178| 4.92k| return upper; 179| 10.0k|} mixed_intersection.c:advanceUntil: 137| 142k| int32_t length, uint16_t min) { 138| 142k| int32_t lower = pos + 1; 139| | 140| 142k| if ((lower >= length) || (array[lower] >= min)) { ------------------ | Branch (140:9): [True: 520, False: 142k] | Branch (140:30): [True: 80.7k, False: 61.4k] ------------------ 141| 81.2k| return lower; 142| 81.2k| } 143| | 144| 61.4k| int32_t spansize = 1; 145| | 146| 147k| while ((lower + spansize < length) && (array[lower + spansize] < min)) { ------------------ | Branch (146:12): [True: 145k, False: 1.27k] | Branch (146:43): [True: 85.5k, False: 60.2k] ------------------ 147| 85.5k| spansize <<= 1; 148| 85.5k| } 149| 61.4k| int32_t upper = (lower + spansize < length) ? lower + spansize : length - 1; ------------------ | Branch (149:21): [True: 60.2k, False: 1.27k] ------------------ 150| | 151| 61.4k| if (array[upper] == min) { ------------------ | Branch (151:9): [True: 28.7k, False: 32.7k] ------------------ 152| 28.7k| return upper; 153| 28.7k| } 154| 32.7k| if (array[upper] < min) { ------------------ | Branch (154:9): [True: 611, False: 32.1k] ------------------ 155| | // means 156| | // array 157| | // has no 158| | // item 159| | // >= min 160| | // pos = array.length; 161| 611| return length; 162| 611| } 163| | 164| | // we know that the next-smallest span was too small 165| 32.1k| lower += (spansize >> 1); 166| | 167| 32.1k| int32_t mid = 0; 168| 55.4k| while (lower + 1 != upper) { ------------------ | Branch (168:12): [True: 37.4k, False: 18.0k] ------------------ 169| 37.4k| mid = (lower + upper) >> 1; 170| 37.4k| if (array[mid] == min) { ------------------ | Branch (170:13): [True: 14.0k, False: 23.3k] ------------------ 171| 14.0k| return mid; 172| 23.3k| } else if (array[mid] < min) { ------------------ | Branch (172:20): [True: 8.64k, False: 14.7k] ------------------ 173| 8.64k| lower = mid; 174| 14.7k| } else { 175| 14.7k| upper = mid; 176| 14.7k| } 177| 37.4k| } 178| 18.0k| return upper; 179| 32.1k|} mixed_subset.c:advanceUntil: 137| 504| int32_t length, uint16_t min) { 138| 504| int32_t lower = pos + 1; 139| | 140| 504| if ((lower >= length) || (array[lower] >= min)) { ------------------ | Branch (140:9): [True: 0, False: 504] | Branch (140:30): [True: 262, False: 242] ------------------ 141| 262| return lower; 142| 262| } 143| | 144| 242| int32_t spansize = 1; 145| | 146| 1.18k| while ((lower + spansize < length) && (array[lower + spansize] < min)) { ------------------ | Branch (146:12): [True: 1.08k, False: 94] | Branch (146:43): [True: 940, False: 148] ------------------ 147| 940| spansize <<= 1; 148| 940| } 149| 242| int32_t upper = (lower + spansize < length) ? lower + spansize : length - 1; ------------------ | Branch (149:21): [True: 148, False: 94] ------------------ 150| | 151| 242| if (array[upper] == min) { ------------------ | Branch (151:9): [True: 0, False: 242] ------------------ 152| 0| return upper; 153| 0| } 154| 242| if (array[upper] < min) { ------------------ | Branch (154:9): [True: 10, False: 232] ------------------ 155| | // means 156| | // array 157| | // has no 158| | // item 159| | // >= min 160| | // pos = array.length; 161| 10| return length; 162| 10| } 163| | 164| | // we know that the next-smallest span was too small 165| 232| lower += (spansize >> 1); 166| | 167| 232| int32_t mid = 0; 168| 842| while (lower + 1 != upper) { ------------------ | Branch (168:12): [True: 610, False: 232] ------------------ 169| 610| mid = (lower + upper) >> 1; 170| 610| if (array[mid] == min) { ------------------ | Branch (170:13): [True: 0, False: 610] ------------------ 171| 0| return mid; 172| 610| } else if (array[mid] < min) { ------------------ | Branch (172:20): [True: 320, False: 290] ------------------ 173| 320| lower = mid; 174| 320| } else { 175| 290| upper = mid; 176| 290| } 177| 610| } 178| 232| return upper; 179| 232|} mixed_andnot.c:advanceUntil: 137| 101k| int32_t length, uint16_t min) { 138| 101k| int32_t lower = pos + 1; 139| | 140| 101k| if ((lower >= length) || (array[lower] >= min)) { ------------------ | Branch (140:9): [True: 4.56k, False: 97.1k] | Branch (140:30): [True: 82.6k, False: 14.5k] ------------------ 141| 87.2k| return lower; 142| 87.2k| } 143| | 144| 14.5k| int32_t spansize = 1; 145| | 146| 29.3k| while ((lower + spansize < length) && (array[lower + spansize] < min)) { ------------------ | Branch (146:12): [True: 28.8k, False: 490] | Branch (146:43): [True: 14.8k, False: 14.0k] ------------------ 147| 14.8k| spansize <<= 1; 148| 14.8k| } 149| 14.5k| int32_t upper = (lower + spansize < length) ? lower + spansize : length - 1; ------------------ | Branch (149:21): [True: 14.0k, False: 490] ------------------ 150| | 151| 14.5k| if (array[upper] == min) { ------------------ | Branch (151:9): [True: 7.12k, False: 7.39k] ------------------ 152| 7.12k| return upper; 153| 7.12k| } 154| 7.39k| if (array[upper] < min) { ------------------ | Branch (154:9): [True: 441, False: 6.94k] ------------------ 155| | // means 156| | // array 157| | // has no 158| | // item 159| | // >= min 160| | // pos = array.length; 161| 441| return length; 162| 441| } 163| | 164| | // we know that the next-smallest span was too small 165| 6.94k| lower += (spansize >> 1); 166| | 167| 6.94k| int32_t mid = 0; 168| 10.7k| while (lower + 1 != upper) { ------------------ | Branch (168:12): [True: 6.26k, False: 4.45k] ------------------ 169| 6.26k| mid = (lower + upper) >> 1; 170| 6.26k| if (array[mid] == min) { ------------------ | Branch (170:13): [True: 2.49k, False: 3.77k] ------------------ 171| 2.49k| return mid; 172| 3.77k| } else if (array[mid] < min) { ------------------ | Branch (172:20): [True: 1.30k, False: 2.47k] ------------------ 173| 1.30k| lower = mid; 174| 2.47k| } else { 175| 2.47k| upper = mid; 176| 2.47k| } 177| 6.26k| } 178| 4.45k| return upper; 179| 6.94k|} roaring.c:bitset_lenrange_cardinality: 63| 1.31k| uint32_t lenminusone) { 64| 1.31k| uint32_t firstword = start / 64; 65| 1.31k| uint32_t endword = (start + lenminusone) / 64; 66| 1.31k| if (firstword == endword) { ------------------ | Branch (66:9): [True: 967, False: 349] ------------------ 67| 967| return roaring_hamming(words[firstword] & 68| 967| ((~UINT64_C(0)) >> ((63 - lenminusone) % 64)) 69| 967| << (start % 64)); 70| 967| } 71| 349| int answer = 72| 349| roaring_hamming(words[firstword] & ((~UINT64_C(0)) << (start % 64))); 73| 151k| for (uint32_t i = firstword + 1; i < endword; i++) { ------------------ | Branch (73:38): [True: 151k, False: 349] ------------------ 74| 151k| answer += roaring_hamming(words[i]); 75| 151k| } 76| 349| answer += roaring_hamming(words[endword] & 77| | (~UINT64_C(0)) >> 78| 349| (((~start + 1) - lenminusone - 1) % 64)); 79| 349| return answer; 80| 1.31k|} roaring.c:bitset_reset_range: 147| 1.28k| uint32_t end) { 148| 1.28k| if (start == end) return; ------------------ | Branch (148:9): [True: 0, False: 1.28k] ------------------ 149| 1.28k| uint32_t firstword = start / 64; 150| 1.28k| uint32_t endword = (end - 1) / 64; 151| 1.28k| if (firstword == endword) { ------------------ | Branch (151:9): [True: 967, False: 317] ------------------ 152| 967| words[firstword] &= ~(((~UINT64_C(0)) << (start % 64)) & 153| 967| ((~UINT64_C(0)) >> ((~end + 1) % 64))); 154| 967| return; 155| 967| } 156| 317| words[firstword] &= ~((~UINT64_C(0)) << (start % 64)); 157| 125k| for (uint32_t i = firstword + 1; i < endword; i++) { ------------------ | Branch (157:38): [True: 124k, False: 317] ------------------ 158| 124k| words[i] = UINT64_C(0); 159| 124k| } 160| | words[endword] &= ~((~UINT64_C(0)) >> ((~end + 1) % 64)); 161| 317|} roaring.c:bitset_set_lenrange: 112| 1.54k| uint32_t lenminusone) { 113| 1.54k| uint32_t firstword = start / 64; 114| 1.54k| uint32_t endword = (start + lenminusone) / 64; 115| 1.54k| if (firstword == endword) { ------------------ | Branch (115:9): [True: 0, False: 1.54k] ------------------ 116| 0| words[firstword] |= ((~UINT64_C(0)) >> ((63 - lenminusone) % 64)) 117| 0| << (start % 64); 118| 0| return; 119| 0| } 120| 1.54k| uint64_t temp = words[endword]; 121| 1.54k| words[firstword] |= (~UINT64_C(0)) << (start % 64); 122| 452k| for (uint32_t i = firstword + 1; i < endword; i += 2) ------------------ | Branch (122:38): [True: 451k, False: 1.54k] ------------------ 123| 451k| words[i] = words[i + 1] = ~UINT64_C(0); 124| 1.54k| words[endword] = 125| | temp | (~UINT64_C(0)) >> (((~start + 1) - lenminusone - 1) % 64); 126| 1.54k|} bitset.c:avx2_harley_seal_popcount256: 340| 3.60k| const uint64_t size) { 341| 3.60k| __m256i total = _mm256_setzero_si256(); 342| 3.60k| __m256i ones = _mm256_setzero_si256(); 343| 3.60k| __m256i twos = _mm256_setzero_si256(); 344| 3.60k| __m256i fours = _mm256_setzero_si256(); 345| 3.60k| __m256i eights = _mm256_setzero_si256(); 346| 3.60k| __m256i sixteens = _mm256_setzero_si256(); 347| 3.60k| __m256i twosA, twosB, foursA, foursB, eightsA, eightsB; 348| | 349| 3.60k| const uint64_t limit = size - size % 16; 350| 3.60k| uint64_t i = 0; 351| | 352| 61.3k| for (; i < limit; i += 16) { ------------------ | Branch (352:12): [True: 57.7k, False: 3.60k] ------------------ 353| 57.7k| CSA(&twosA, &ones, ones, _mm256_lddqu_si256(data + i), 354| 57.7k| _mm256_lddqu_si256(data + i + 1)); 355| 57.7k| CSA(&twosB, &ones, ones, _mm256_lddqu_si256(data + i + 2), 356| 57.7k| _mm256_lddqu_si256(data + i + 3)); 357| 57.7k| CSA(&foursA, &twos, twos, twosA, twosB); 358| 57.7k| CSA(&twosA, &ones, ones, _mm256_lddqu_si256(data + i + 4), 359| 57.7k| _mm256_lddqu_si256(data + i + 5)); 360| 57.7k| CSA(&twosB, &ones, ones, _mm256_lddqu_si256(data + i + 6), 361| 57.7k| _mm256_lddqu_si256(data + i + 7)); 362| 57.7k| CSA(&foursB, &twos, twos, twosA, twosB); 363| 57.7k| CSA(&eightsA, &fours, fours, foursA, foursB); 364| 57.7k| CSA(&twosA, &ones, ones, _mm256_lddqu_si256(data + i + 8), 365| 57.7k| _mm256_lddqu_si256(data + i + 9)); 366| 57.7k| CSA(&twosB, &ones, ones, _mm256_lddqu_si256(data + i + 10), 367| 57.7k| _mm256_lddqu_si256(data + i + 11)); 368| 57.7k| CSA(&foursA, &twos, twos, twosA, twosB); 369| 57.7k| CSA(&twosA, &ones, ones, _mm256_lddqu_si256(data + i + 12), 370| 57.7k| _mm256_lddqu_si256(data + i + 13)); 371| 57.7k| CSA(&twosB, &ones, ones, _mm256_lddqu_si256(data + i + 14), 372| 57.7k| _mm256_lddqu_si256(data + i + 15)); 373| 57.7k| CSA(&foursB, &twos, twos, twosA, twosB); 374| 57.7k| CSA(&eightsB, &fours, fours, foursA, foursB); 375| 57.7k| CSA(&sixteens, &eights, eights, eightsA, eightsB); 376| | 377| 57.7k| total = _mm256_add_epi64(total, popcount256(sixteens)); 378| 57.7k| } 379| | 380| 3.60k| total = _mm256_slli_epi64(total, 4); // * 16 381| 3.60k| total = _mm256_add_epi64( 382| 3.60k| total, _mm256_slli_epi64(popcount256(eights), 3)); // += 8 * ... 383| 3.60k| total = _mm256_add_epi64( 384| 3.60k| total, _mm256_slli_epi64(popcount256(fours), 2)); // += 4 * ... 385| 3.60k| total = _mm256_add_epi64( 386| 3.60k| total, _mm256_slli_epi64(popcount256(twos), 1)); // += 2 * ... 387| 3.60k| total = _mm256_add_epi64(total, popcount256(ones)); 388| 3.60k| for (; i < size; i++) ------------------ | Branch (388:12): [True: 0, False: 3.60k] ------------------ 389| 0| total = 390| 0| _mm256_add_epi64(total, popcount256(_mm256_lddqu_si256(data + i))); 391| | 392| 3.60k| return (uint64_t)(_mm256_extract_epi64(total, 0)) + 393| 3.60k| (uint64_t)(_mm256_extract_epi64(total, 1)) + 394| 3.60k| (uint64_t)(_mm256_extract_epi64(total, 2)) + 395| | (uint64_t)(_mm256_extract_epi64(total, 3)); 396| 3.60k|} bitset.c:CSA: 328| 1.60M| __m256i c) { 329| 1.60M| const __m256i u = _mm256_xor_si256(a, b); 330| 1.60M| *h = _mm256_or_si256(_mm256_and_si256(a, b), _mm256_and_si256(u, c)); 331| 1.60M| *l = _mm256_xor_si256(u, c); 332| 1.60M|} bitset.c:popcount256: 292| 134k|static inline __m256i popcount256(__m256i v) { 293| 134k| const __m256i lookuppos = _mm256_setr_epi8( 294| 134k| /* 0 */ 4 + 0, /* 1 */ 4 + 1, /* 2 */ 4 + 1, /* 3 */ 4 + 2, 295| 134k| /* 4 */ 4 + 1, /* 5 */ 4 + 2, /* 6 */ 4 + 2, /* 7 */ 4 + 3, 296| 134k| /* 8 */ 4 + 1, /* 9 */ 4 + 2, /* a */ 4 + 2, /* b */ 4 + 3, 297| 134k| /* c */ 4 + 2, /* d */ 4 + 3, /* e */ 4 + 3, /* f */ 4 + 4, 298| | 299| | /* 0 */ 4 + 0, /* 1 */ 4 + 1, /* 2 */ 4 + 1, /* 3 */ 4 + 2, 300| 134k| /* 4 */ 4 + 1, /* 5 */ 4 + 2, /* 6 */ 4 + 2, /* 7 */ 4 + 3, 301| 134k| /* 8 */ 4 + 1, /* 9 */ 4 + 2, /* a */ 4 + 2, /* b */ 4 + 3, 302| 134k| /* c */ 4 + 2, /* d */ 4 + 3, /* e */ 4 + 3, /* f */ 4 + 4); 303| 134k| const __m256i lookupneg = _mm256_setr_epi8( 304| 134k| /* 0 */ 4 - 0, /* 1 */ 4 - 1, /* 2 */ 4 - 1, /* 3 */ 4 - 2, 305| 134k| /* 4 */ 4 - 1, /* 5 */ 4 - 2, /* 6 */ 4 - 2, /* 7 */ 4 - 3, 306| 134k| /* 8 */ 4 - 1, /* 9 */ 4 - 2, /* a */ 4 - 2, /* b */ 4 - 3, 307| 134k| /* c */ 4 - 2, /* d */ 4 - 3, /* e */ 4 - 3, /* f */ 4 - 4, 308| | 309| | /* 0 */ 4 - 0, /* 1 */ 4 - 1, /* 2 */ 4 - 1, /* 3 */ 4 - 2, 310| 134k| /* 4 */ 4 - 1, /* 5 */ 4 - 2, /* 6 */ 4 - 2, /* 7 */ 4 - 3, 311| 134k| /* 8 */ 4 - 1, /* 9 */ 4 - 2, /* a */ 4 - 2, /* b */ 4 - 3, 312| 134k| /* c */ 4 - 2, /* d */ 4 - 3, /* e */ 4 - 3, /* f */ 4 - 4); 313| 134k| const __m256i low_mask = _mm256_set1_epi8(0x0f); 314| | 315| 134k| const __m256i lo = _mm256_and_si256(v, low_mask); 316| 134k| const __m256i hi = _mm256_and_si256(_mm256_srli_epi16(v, 4), low_mask); 317| 134k| const __m256i popcnt1 = _mm256_shuffle_epi8(lookuppos, lo); 318| 134k| const __m256i popcnt2 = _mm256_shuffle_epi8(lookupneg, hi); 319| 134k| return _mm256_sad_epu8(popcnt1, popcnt2); 320| 134k|} convert.c:bitset_set_lenrange: 112| 112k| uint32_t lenminusone) { 113| 112k| uint32_t firstword = start / 64; 114| 112k| uint32_t endword = (start + lenminusone) / 64; 115| 112k| if (firstword == endword) { ------------------ | Branch (115:9): [True: 37.7k, False: 74.6k] ------------------ 116| 37.7k| words[firstword] |= ((~UINT64_C(0)) >> ((63 - lenminusone) % 64)) 117| 37.7k| << (start % 64); 118| 37.7k| return; 119| 37.7k| } 120| 74.6k| uint64_t temp = words[endword]; 121| 74.6k| words[firstword] |= (~UINT64_C(0)) << (start % 64); 122| 36.8M| for (uint32_t i = firstword + 1; i < endword; i += 2) ------------------ | Branch (122:38): [True: 36.7M, False: 74.6k] ------------------ 123| 36.7M| words[i] = words[i + 1] = ~UINT64_C(0); 124| 74.6k| words[endword] = 125| | temp | (~UINT64_C(0)) >> (((~start + 1) - lenminusone - 1) % 64); 126| 74.6k|} mixed_intersection.c:bitset_lenrange_cardinality: 63| 225k| uint32_t lenminusone) { 64| 225k| uint32_t firstword = start / 64; 65| 225k| uint32_t endword = (start + lenminusone) / 64; 66| 225k| if (firstword == endword) { ------------------ | Branch (66:9): [True: 219k, False: 6.15k] ------------------ 67| 219k| return roaring_hamming(words[firstword] & 68| 219k| ((~UINT64_C(0)) >> ((63 - lenminusone) % 64)) 69| 219k| << (start % 64)); 70| 219k| } 71| 6.15k| int answer = 72| 6.15k| roaring_hamming(words[firstword] & ((~UINT64_C(0)) << (start % 64))); 73| 6.44k| for (uint32_t i = firstword + 1; i < endword; i++) { ------------------ | Branch (73:38): [True: 288, False: 6.15k] ------------------ 74| 288| answer += roaring_hamming(words[i]); 75| 288| } 76| 6.15k| answer += roaring_hamming(words[endword] & 77| | (~UINT64_C(0)) >> 78| 6.15k| (((~start + 1) - lenminusone - 1) % 64)); 79| 6.15k| return answer; 80| 225k|} mixed_intersection.c:bitset_lenrange_empty: 86| 4.12k| uint32_t lenminusone) { 87| 4.12k| uint32_t firstword = start / 64; 88| 4.12k| uint32_t endword = (start + lenminusone) / 64; 89| 4.12k| if (firstword == endword) { ------------------ | Branch (89:9): [True: 3.80k, False: 319] ------------------ 90| 3.80k| return (words[firstword] & ((~UINT64_C(0)) >> ((63 - lenminusone) % 64)) 91| 3.80k| << (start % 64)) == 0; 92| 3.80k| } 93| 319| if (((words[firstword] & ((~UINT64_C(0)) << (start % 64)))) != 0) { ------------------ | Branch (93:9): [True: 108, False: 211] ------------------ 94| 108| return false; 95| 108| } 96| 221| for (uint32_t i = firstword + 1; i < endword; i++) { ------------------ | Branch (96:38): [True: 11, False: 210] ------------------ 97| 11| if (words[i] != 0) { ------------------ | Branch (97:13): [True: 1, False: 10] ------------------ 98| 1| return false; 99| 1| } 100| 11| } 101| 210| if ((words[endword] & ------------------ | Branch (101:9): [True: 119, False: 91] ------------------ 102| 210| (~UINT64_C(0)) >> (((~start + 1) - lenminusone - 1) % 64)) != 0) { 103| 119| return false; 104| 119| } 105| 91| return true; 106| 210|} mixed_union.c:bitset_set_lenrange: 112| 112k| uint32_t lenminusone) { 113| 112k| uint32_t firstword = start / 64; 114| 112k| uint32_t endword = (start + lenminusone) / 64; 115| 112k| if (firstword == endword) { ------------------ | Branch (115:9): [True: 109k, False: 3.07k] ------------------ 116| 109k| words[firstword] |= ((~UINT64_C(0)) >> ((63 - lenminusone) % 64)) 117| 109k| << (start % 64); 118| 109k| return; 119| 109k| } 120| 3.07k| uint64_t temp = words[endword]; 121| 3.07k| words[firstword] |= (~UINT64_C(0)) << (start % 64); 122| 3.18k| for (uint32_t i = firstword + 1; i < endword; i += 2) ------------------ | Branch (122:38): [True: 106, False: 3.07k] ------------------ 123| 106| words[i] = words[i + 1] = ~UINT64_C(0); 124| 3.07k| words[endword] = 125| | temp | (~UINT64_C(0)) >> (((~start + 1) - lenminusone - 1) % 64); 126| 3.07k|} mixed_negation.c:bitset_flip_range: 132| 1.57k| uint32_t end) { 133| 1.57k| if (start == end) return; ------------------ | Branch (133:9): [True: 0, False: 1.57k] ------------------ 134| 1.57k| uint32_t firstword = start / 64; 135| 1.57k| uint32_t endword = (end - 1) / 64; 136| 1.57k| words[firstword] ^= ~((~UINT64_C(0)) << (start % 64)); 137| 208k| for (uint32_t i = firstword; i < endword; i++) { ------------------ | Branch (137:34): [True: 207k, False: 1.57k] ------------------ 138| 207k| words[i] = ~words[i]; 139| 207k| } 140| | words[endword] ^= ((~UINT64_C(0)) >> ((~end + 1) % 64)); 141| 1.57k|} mixed_xor.c:bitset_flip_range: 132| 59.8k| uint32_t end) { 133| 59.8k| if (start == end) return; ------------------ | Branch (133:9): [True: 0, False: 59.8k] ------------------ 134| 59.8k| uint32_t firstword = start / 64; 135| 59.8k| uint32_t endword = (end - 1) / 64; 136| 59.8k| words[firstword] ^= ~((~UINT64_C(0)) << (start % 64)); 137| 127k| for (uint32_t i = firstword; i < endword; i++) { ------------------ | Branch (137:34): [True: 67.5k, False: 59.8k] ------------------ 138| 67.5k| words[i] = ~words[i]; 139| 67.5k| } 140| | words[endword] ^= ((~UINT64_C(0)) >> ((~end + 1) % 64)); 141| 59.8k|} array_container_contains: 343| 3.36k| uint16_t pos) { 344| | /** 345| | * SIMD Quad algorithm 346| | * Daniel Lemire, "You can beat the binary search," in Daniel Lemire's blog, 347| | * April 27, 2026, 348| | * https://lemire.me/blog/2026/04/27/you-can-beat-the-binary-search/. 349| | */ 350| 3.36k| const int32_t gap = 16; 351| 3.36k| const uint16_t *carr = arr->array; 352| 3.36k| int32_t cardinality = arr->cardinality; 353| 3.36k| if (cardinality < gap) { ------------------ | Branch (353:9): [True: 946, False: 2.42k] ------------------ 354| 1.06k| for (int32_t j = 0; j < cardinality; j++) { ------------------ | Branch (354:29): [True: 1.03k, False: 31] ------------------ 355| 1.03k| if (carr[j] >= pos) return carr[j] == pos; ------------------ | Branch (355:17): [True: 915, False: 122] ------------------ 356| 1.03k| } 357| 31| return false; 358| 946| } 359| 2.42k| int32_t num_blocks = cardinality / gap; 360| 2.42k| int32_t base = 0; 361| 2.42k| int32_t n = num_blocks; 362| 5.65k| while (n > 3) { ------------------ | Branch (362:12): [True: 3.23k, False: 2.42k] ------------------ 363| 3.23k| int32_t quarter = n >> 2; 364| | 365| 3.23k| int32_t k1 = carr[(base + quarter + 1) * gap - 1]; 366| 3.23k| int32_t k2 = carr[(base + 2 * quarter + 1) * gap - 1]; 367| 3.23k| int32_t k3 = carr[(base + 3 * quarter + 1) * gap - 1]; 368| | 369| 3.23k| int32_t c1 = (k1 < pos); 370| 3.23k| int32_t c2 = (k2 < pos); 371| 3.23k| int32_t c3 = (k3 < pos); 372| | 373| 3.23k| base += (c1 + c2 + c3) * quarter; 374| 3.23k| n -= 3 * quarter; 375| 3.23k| } 376| 4.28k| while (n > 1) { ------------------ | Branch (376:12): [True: 1.86k, False: 2.42k] ------------------ 377| 1.86k| int32_t half = n >> 1; 378| 1.86k| base = (carr[(base + half + 1) * gap - 1] < pos) ? base + half : base; ------------------ | Branch (378:16): [True: 72, False: 1.79k] ------------------ 379| 1.86k| n -= half; 380| 1.86k| } 381| 2.42k| int32_t lo = (carr[(base + 1) * gap - 1] < pos) ? base + 1 : base; ------------------ | Branch (381:18): [True: 112, False: 2.30k] ------------------ 382| | 383| 2.42k| if (lo < num_blocks) { ------------------ | Branch (383:9): [True: 2.34k, False: 78] ------------------ 384| 2.34k| const uint16_t *blk = carr + lo * gap; 385| |#ifdef CROARING_USENEON 386| | uint16x8_t needle = vdupq_n_u16(pos); 387| | uint16x8_t v0 = vld1q_u16(blk); 388| | uint16x8_t v1 = vld1q_u16(blk + 8); 389| | uint16x8_t hit = 390| | vorrq_u16(vceqq_u16(v0, needle), vceqq_u16(v1, needle)); 391| | return vmaxvq_u16(hit) != 0; 392| |#elif defined(CROARING_IS_X64) 393| | __m128i needle = _mm_set1_epi16((short)pos); 394| 2.34k| __m128i v0 = _mm_loadu_si128((const __m128i *)blk); 395| 2.34k| __m128i v1 = _mm_loadu_si128((const __m128i *)(blk + 8)); 396| 2.34k| __m128i hit = _mm_or_si128(_mm_cmpeq_epi16(v0, needle), 397| 2.34k| _mm_cmpeq_epi16(v1, needle)); 398| 2.34k| return _mm_movemask_epi8(hit) != 0; 399| |#else 400| | for (int32_t j = 0; j < gap; j++) { 401| | if (blk[j] >= pos) return blk[j] == pos; 402| | } 403| | return false; 404| |#endif 405| 2.34k| } 406| | 407| 606| for (int32_t j = num_blocks * gap; j < cardinality; j++) { ------------------ | Branch (407:40): [True: 556, False: 50] ------------------ 408| 556| uint16_t v = carr[j]; 409| 556| if (v >= pos) return (v == pos); ------------------ | Branch (409:13): [True: 28, False: 528] ------------------ 410| 556| } 411| 50| return false; 412| 78|} array_container_minimum: 445| 3.63k|inline uint16_t array_container_minimum(const array_container_t *arr) { 446| 3.63k| if (arr->cardinality == 0) return 0; ------------------ | Branch (446:9): [True: 0, False: 3.63k] ------------------ 447| 3.63k| return arr->array[0]; 448| 3.63k|} array_container_maximum: 451| 13.3k|inline uint16_t array_container_maximum(const array_container_t *arr) { 452| 13.3k| if (arr->cardinality == 0) return 0; ------------------ | Branch (452:9): [True: 0, False: 13.3k] ------------------ 453| 13.3k| return arr->array[arr->cardinality - 1]; 454| 13.3k|} array_container_rank: 457| 3.80k|inline int array_container_rank(const array_container_t *arr, uint16_t x) { 458| 3.80k| const int32_t idx = binarySearch(arr->array, arr->cardinality, x); 459| 3.80k| const bool is_present = idx >= 0; 460| 3.80k| if (is_present) { ------------------ | Branch (460:9): [True: 3.25k, False: 553] ------------------ 461| 3.25k| return idx + 1; 462| 3.25k| } else { 463| 553| return -idx - 1; 464| 553| } 465| 3.80k|} array_container_index_equalorlarger: 507| 4.46k| uint16_t x) { 508| 4.46k| const int32_t idx = binarySearch(arr->array, arr->cardinality, x); 509| 4.46k| const bool is_present = idx >= 0; 510| 4.46k| if (is_present) { ------------------ | Branch (510:9): [True: 4.06k, False: 407] ------------------ 511| 4.06k| return idx; 512| 4.06k| } else { 513| 407| int32_t candidate = -idx - 1; 514| 407| if (candidate < arr->cardinality) return candidate; ------------------ | Branch (514:13): [True: 407, False: 0] ------------------ 515| 0| return -1; 516| 407| } 517| 4.46k|} roaring.c:array_container_add_range_nvals: 527| 645| int32_t nvals_greater) { 528| 645| int32_t union_cardinality = nvals_less + (max - min + 1) + nvals_greater; 529| 645| if (union_cardinality > array->capacity) { ------------------ | Branch (529:9): [True: 599, False: 46] ------------------ 530| 599| array_container_grow(array, union_cardinality, true); 531| 599| } 532| 645| memmove(&(array->array[union_cardinality - nvals_greater]), 533| 645| &(array->array[array->cardinality - nvals_greater]), 534| 645| nvals_greater * sizeof(uint16_t)); 535| 1.56M| for (uint32_t i = 0; i <= max - min; i++) { ------------------ | Branch (535:26): [True: 1.56M, False: 645] ------------------ 536| 1.56M| array->array[nvals_less + i] = (uint16_t)(min + i); 537| 1.56M| } 538| 645| array->cardinality = union_cardinality; 539| 645|} roaring.c:array_container_remove_range: 557| 3.23k| uint32_t pos, uint32_t count) { 558| 3.23k| if (count != 0) { ------------------ | Branch (558:9): [True: 314, False: 2.92k] ------------------ 559| 314| memmove(&(array->array[pos]), &(array->array[pos + count]), 560| 314| (array->cardinality - pos - count) * sizeof(uint16_t)); 561| 314| array->cardinality -= count; 562| 314| } 563| 3.23k|} roaring.c:array_container_cardinality: 82| 72.4k|static inline int array_container_cardinality(const array_container_t *array) { 83| 72.4k| return array->cardinality; 84| 72.4k|} roaring.c:array_container_try_add: 293| 7.39M| int32_t max_cardinality) { 294| 7.39M| const int32_t cardinality = arr->cardinality; 295| | 296| | // best case, we can append. 297| 7.39M| if ((array_container_empty(arr) || arr->array[cardinality - 1] < value) && ------------------ | Branch (297:10): [True: 16.3k, False: 7.37M] | Branch (297:40): [True: 52.4k, False: 7.32M] ------------------ 298| 68.8k| cardinality < max_cardinality) { ------------------ | Branch (298:9): [True: 68.8k, False: 0] ------------------ 299| 68.8k| array_container_append(arr, value); 300| 68.8k| return 1; 301| 68.8k| } 302| | 303| 7.32M| const int32_t loc = binarySearch(arr->array, cardinality, value); 304| | 305| 7.32M| if (loc >= 0) { ------------------ | Branch (305:9): [True: 5.56M, False: 1.75M] ------------------ 306| 5.56M| return 0; 307| 5.56M| } else if (cardinality < max_cardinality) { ------------------ | Branch (307:16): [True: 1.75M, False: 131] ------------------ 308| 1.75M| if (array_container_full(arr)) { ------------------ | Branch (308:13): [True: 78.1k, False: 1.67M] ------------------ 309| 78.1k| array_container_grow(arr, arr->capacity + 1, true); 310| 78.1k| } 311| 1.75M| const int32_t insert_idx = -loc - 1; 312| 1.75M| memmove(arr->array + insert_idx + 1, arr->array + insert_idx, 313| 1.75M| (cardinality - insert_idx) * sizeof(uint16_t)); 314| 1.75M| arr->array[insert_idx] = value; 315| 1.75M| arr->cardinality++; 316| 1.75M| return 1; 317| 1.75M| } else { 318| 131| return -1; 319| 131| } 320| 7.32M|} roaring.c:array_container_empty: 100| 7.39M|static inline bool array_container_empty(const array_container_t *array) { 101| 7.39M| return array->cardinality == 0; 102| 7.39M|} roaring.c:array_container_append: 274| 68.8k| uint16_t pos) { 275| 68.8k| const int32_t capacity = arr->capacity; 276| | 277| 68.8k| if (array_container_full(arr)) { ------------------ | Branch (277:9): [True: 32.9k, False: 35.9k] ------------------ 278| 32.9k| array_container_grow(arr, capacity + 1, true); 279| 32.9k| } 280| | 281| 68.8k| arr->array[arr->cardinality++] = pos; 282| 68.8k|} roaring.c:array_container_full: 106| 1.82M|static inline bool array_container_full(const array_container_t *array) { 107| 1.82M| return array->cardinality == array->capacity; 108| 1.82M|} roaring.c:array_container_remove: 329| 6.21k| uint16_t pos) { 330| 6.21k| const int32_t idx = binarySearch(arr->array, arr->cardinality, pos); 331| 6.21k| const bool is_present = idx >= 0; 332| 6.21k| if (is_present) { ------------------ | Branch (332:9): [True: 3.11k, False: 3.09k] ------------------ 333| 3.11k| memmove(arr->array + idx, arr->array + idx + 1, 334| 3.11k| (arr->cardinality - idx - 1) * sizeof(uint16_t)); 335| 3.11k| arr->cardinality--; 336| 3.11k| } 337| | 338| 6.21k| return is_present; 339| 6.21k|} roaring.c:array_container_nonzero_cardinality: 87| 107k| const array_container_t *array) { 88| 107k| return array->cardinality > 0; 89| 107k|} roaring.c:array_container_equals: 230| 9.61k| const array_container_t *container2) { 231| 9.61k| if (container1->cardinality != container2->cardinality) { ------------------ | Branch (231:9): [True: 2.26k, False: 7.34k] ------------------ 232| 2.26k| return false; 233| 2.26k| } 234| 7.34k| return memequals(container1->array, container2->array, 235| 7.34k| container1->cardinality * 2); 236| 9.61k|} roaring.c:array_container_select: 252| 3.80k| uint32_t *element) { 253| 3.80k| int card = array_container_cardinality(container); 254| 3.80k| if (*start_rank + card <= rank) { ------------------ | Branch (254:9): [True: 321, False: 3.48k] ------------------ 255| 321| *start_rank += card; 256| 321| return false; 257| 3.48k| } else { 258| 3.48k| *element = container->array[rank - *start_rank]; 259| | return true; 260| 3.48k| } 261| 3.80k|} roaring.c:array_container_contains_range: 421| 122| uint32_t range_end) { 422| 122| const int32_t range_count = range_end - range_start; 423| 122| const uint16_t rs_included = (uint16_t)range_start; 424| 122| const uint16_t re_included = (uint16_t)(range_end - 1); 425| | 426| | // Empty range is always included 427| 122| if (range_count <= 0) { ------------------ | Branch (427:9): [True: 0, False: 122] ------------------ 428| 0| return true; 429| 0| } 430| 122| if (range_count > arr->cardinality) { ------------------ | Branch (430:9): [True: 53, False: 69] ------------------ 431| 53| return false; 432| 53| } 433| | 434| 69| const int32_t start = 435| 69| binarySearch(arr->array, arr->cardinality, rs_included); 436| | // If this sorted array contains all items in the range: 437| | // * the start item must be found 438| | // * the last item in range range_count must exist, and be the expected end 439| | // value 440| 69| return (start >= 0) && (arr->cardinality >= start + range_count) && ------------------ | Branch (440:12): [True: 45, False: 24] | Branch (440:28): [True: 40, False: 5] ------------------ 441| 40| (arr->array[start + range_count - 1] == re_included); ------------------ | Branch (441:12): [True: 12, False: 28] ------------------ 442| 122|} roaring_array.c:array_container_size_in_bytes: 221| 18.2k| const array_container_t *container) { 222| 18.2k| return container->cardinality * sizeof(uint16_t); 223| 18.2k|} roaring_array.c:array_container_cardinality: 82| 7.20k|static inline int array_container_cardinality(const array_container_t *array) { 83| 7.20k| return array->cardinality; 84| 7.20k|} array.c:array_container_size_in_bytes: 221| 14.4k| const array_container_t *container) { 222| 14.4k| return container->cardinality * sizeof(uint16_t); 223| 14.4k|} convert.c:array_container_cardinality: 82| 15.3k|static inline int array_container_cardinality(const array_container_t *array) { 83| 15.3k| return array->cardinality; 84| 15.3k|} convert.c:array_container_serialized_size_in_bytes: 173| 154k|static inline int32_t array_container_serialized_size_in_bytes(int32_t card) { 174| 154k| return card * sizeof(uint16_t); 175| 154k|} mixed_intersection.c:array_container_empty: 100| 34|static inline bool array_container_empty(const array_container_t *array) { 101| 34| return array->cardinality == 0; 102| 34|} mixed_andnot.c:array_container_cardinality: 82| 381|static inline int array_container_cardinality(const array_container_t *array) { 83| 381| return array->cardinality; 84| 381|} bitset_container_get: 193| 1.12M| uint16_t pos) { 194| 1.12M| const uint64_t word = bitset->words[pos >> 6]; 195| 1.12M| return (word >> (pos & 63)) & 1; 196| 1.12M|} bitset_container_contains: 232| 974k| uint16_t pos) { 233| 974k| return bitset_container_get(bitset, pos); 234| 974k|} roaring.c:bitset_container_cardinality: 248| 11.3k| const bitset_container_t *bitset) { 249| 11.3k| return bitset->cardinality; 250| 11.3k|} roaring.c:bitset_container_set: 147| 326k| uint16_t pos) { 148| 326k| const uint64_t old_word = bitset->words[pos >> 6]; 149| 326k| const int index = pos & 63; 150| | const uint64_t new_word = old_word | (UINT64_C(1) << index); 151| 326k| bitset->cardinality += (uint32_t)((old_word ^ new_word) >> index); 152| 326k| bitset->words[pos >> 6] = new_word; 153| 326k|} roaring.c:bitset_container_add: 168| 131| uint16_t pos) { 169| 131| const uint64_t old_word = bitset->words[pos >> 6]; 170| 131| const int index = pos & 63; 171| | const uint64_t new_word = old_word | (UINT64_C(1) << index); 172| 131| const uint64_t increment = (old_word ^ new_word) >> index; 173| 131| bitset->cardinality += (uint32_t)increment; 174| 131| bitset->words[pos >> 6] = new_word; 175| 131| return increment > 0; 176| 131|} roaring.c:bitset_container_remove: 181| 2.32k| uint16_t pos) { 182| 2.32k| const uint64_t old_word = bitset->words[pos >> 6]; 183| 2.32k| const int index = pos & 63; 184| | const uint64_t new_word = old_word & (~(UINT64_C(1) << index)); 185| 2.32k| const uint64_t increment = (old_word ^ new_word) >> index; 186| 2.32k| bitset->cardinality -= (uint32_t)increment; 187| 2.32k| bitset->words[pos >> 6] = new_word; 188| 2.32k| return increment > 0; 189| 2.32k|} roaring.c:bitset_container_const_nonzero_cardinality: 281| 7.16k| const bitset_container_t *bitset) { 282| 7.16k| return !bitset_container_empty(bitset); 283| 7.16k|} roaring.c:bitset_container_empty: 268| 7.16k|static inline bool bitset_container_empty(const bitset_container_t *bitset) { 269| 7.16k| if (bitset->cardinality == BITSET_UNKNOWN_CARDINALITY) { ------------------ | Branch (269:9): [True: 0, False: 7.16k] ------------------ 270| 0| for (int i = 0; i < BITSET_CONTAINER_SIZE_IN_WORDS; i++) { ------------------ | Branch (270:25): [True: 0, False: 0] ------------------ 271| 0| if ((bitset->words[i]) != 0) return false; ------------------ | Branch (271:17): [True: 0, False: 0] ------------------ 272| 0| } 273| 0| return true; 274| 0| } 275| 7.16k| return bitset->cardinality == 0; 276| 7.16k|} roaring.c:bitset_container_get_range: 206| 263| uint32_t pos_end) { 207| 263| const uint32_t start = pos_start >> 6; 208| 263| const uint32_t end = pos_end >> 6; 209| | 210| 263| const uint64_t first = ~((1ULL << (pos_start & 0x3F)) - 1); 211| 263| const uint64_t last = (1ULL << (pos_end & 0x3F)) - 1; 212| | 213| 263| if (start == end) ------------------ | Branch (213:9): [True: 60, False: 203] ------------------ 214| 60| return ((bitset->words[end] & first & last) == (first & last)); 215| 203| if ((bitset->words[start] & first) != first) return false; ------------------ | Branch (215:9): [True: 75, False: 128] ------------------ 216| | 217| 128| if ((end < BITSET_CONTAINER_SIZE_IN_WORDS) && ------------------ | Branch (217:9): [True: 113, False: 15] ------------------ 218| 113| ((bitset->words[end] & last) != last)) { ------------------ | Branch (218:9): [True: 96, False: 17] ------------------ 219| 96| return false; 220| 96| } 221| | 222| 32| for (uint32_t i = start + 1; 223| 9.23k| (i < BITSET_CONTAINER_SIZE_IN_WORDS) && (i < end); ++i) { ------------------ | Branch (223:10): [True: 9.23k, False: 7] | Branch (223:50): [True: 9.21k, False: 14] ------------------ 224| 9.21k| if (bitset->words[i] != UINT64_C(0xFFFFFFFFFFFFFFFF)) return false; ------------------ | Branch (224:13): [True: 11, False: 9.20k] ------------------ 225| 9.21k| } 226| | 227| 21| return true; 228| 32|} roaring_array.c:bitset_container_size_in_bytes: 468| 147k| const bitset_container_t *container) { 469| 147k| (void)container; 470| 147k| return BITSET_CONTAINER_SIZE_IN_WORDS * sizeof(uint64_t); 471| 147k|} roaring_array.c:bitset_container_cardinality: 248| 73.9k| const bitset_container_t *bitset) { 249| 73.9k| return bitset->cardinality; 250| 73.9k|} bitset.c:bitset_container_size_in_bytes: 468| 147k| const bitset_container_t *container) { 469| 147k| (void)container; 470| 147k| return BITSET_CONTAINER_SIZE_IN_WORDS * sizeof(uint64_t); 471| 147k|} convert.c:bitset_container_set: 147| 2.46M| uint16_t pos) { 148| 2.46M| const uint64_t old_word = bitset->words[pos >> 6]; 149| 2.46M| const int index = pos & 63; 150| | const uint64_t new_word = old_word | (UINT64_C(1) << index); 151| 2.46M| bitset->cardinality += (uint32_t)((old_word ^ new_word) >> index); 152| 2.46M| bitset->words[pos >> 6] = new_word; 153| 2.46M|} convert.c:bitset_container_serialized_size_in_bytes: 427| 141k|static inline int32_t bitset_container_serialized_size_in_bytes(void) { 428| 141k| return BITSET_CONTAINER_SIZE_IN_WORDS * 8; 429| 141k|} container_iterator_next: 2456| 335k| uint16_t *value) { 2457| 335k| switch (typecode) { 2458| 0| case BITSET_CONTAINER_TYPE: { ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (2458:9): [True: 0, False: 335k] ------------------ 2459| 0| const bitset_container_t *bc = const_CAST_bitset(c); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2460| 0| it->index++; 2461| | 2462| 0| uint32_t wordindex = it->index / 64; 2463| 0| if (wordindex >= BITSET_CONTAINER_SIZE_IN_WORDS) { ------------------ | Branch (2463:17): [True: 0, False: 0] ------------------ 2464| 0| return false; 2465| 0| } 2466| | 2467| 0| uint64_t word = 2468| 0| bc->words[wordindex] & (UINT64_MAX << (it->index % 64)); 2469| | // next part could be optimized/simplified 2470| 0| while (word == 0 && ------------------ | Branch (2470:20): [True: 0, False: 0] ------------------ 2471| 0| (wordindex + 1 < BITSET_CONTAINER_SIZE_IN_WORDS)) { ------------------ | Branch (2471:20): [True: 0, False: 0] ------------------ 2472| 0| wordindex++; 2473| 0| word = bc->words[wordindex]; 2474| 0| } 2475| 0| if (word != 0) { ------------------ | Branch (2475:17): [True: 0, False: 0] ------------------ 2476| 0| it->index = wordindex * 64 + roaring_trailing_zeroes(word); 2477| 0| *value = it->index; 2478| 0| return true; 2479| 0| } 2480| 0| return false; 2481| 0| } 2482| 293k| case ARRAY_CONTAINER_TYPE: { ------------------ | | 49| 293k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (2482:9): [True: 293k, False: 42.2k] ------------------ 2483| 293k| const array_container_t *ac = const_CAST_array(c); ------------------ | | 55| 293k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 293k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2484| 293k| it->index++; 2485| 293k| if (it->index < ac->cardinality) { ------------------ | Branch (2485:17): [True: 287k, False: 5.72k] ------------------ 2486| 287k| *value = ac->array[it->index]; 2487| 287k| return true; 2488| 287k| } 2489| 5.72k| return false; 2490| 293k| } 2491| 42.2k| case RUN_CONTAINER_TYPE: { ------------------ | | 50| 42.2k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (2491:9): [True: 42.2k, False: 293k] ------------------ 2492| 42.2k| if (*value == UINT16_MAX) { // Avoid overflow to zero ------------------ | Branch (2492:17): [True: 0, False: 42.2k] ------------------ 2493| 0| return false; 2494| 0| } 2495| | 2496| 42.2k| const run_container_t *rc = const_CAST_run(c); ------------------ | | 78| 42.2k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 42.2k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2497| 42.2k| uint32_t limit = 2498| 42.2k| rc->runs[it->index].value + rc->runs[it->index].length; 2499| 42.2k| if (*value < limit) { ------------------ | Branch (2499:17): [True: 26.9k, False: 15.3k] ------------------ 2500| 26.9k| (*value)++; 2501| 26.9k| return true; 2502| 26.9k| } 2503| | 2504| 15.3k| it->index++; 2505| 15.3k| if (it->index < rc->n_runs) { ------------------ | Branch (2505:17): [True: 14.8k, False: 503] ------------------ 2506| 14.8k| *value = rc->runs[it->index].value; 2507| 14.8k| return true; 2508| 14.8k| } 2509| 503| return false; 2510| 15.3k| } 2511| 0| default: ------------------ | Branch (2511:9): [True: 0, False: 335k] ------------------ 2512| 0| assert(false); 2513| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2514| 0| return false; 2515| 335k| } 2516| 335k|} roaring.c:container_from_range: 317| 73.2k| uint32_t max, uint16_t step) { 318| 73.2k| if (step == 0) return NULL; // being paranoid ------------------ | Branch (318:9): [True: 0, False: 73.2k] ------------------ 319| 73.2k| if (step == 1) { ------------------ | Branch (319:9): [True: 73.2k, False: 0] ------------------ 320| 73.2k| return container_range_of_ones(min, max, type); 321| | // Note: the result is not always a run (need to check the cardinality) 322| | //*type = RUN_CONTAINER_TYPE; 323| | // return run_container_create_range(min, max); 324| 73.2k| } 325| 0| int size = (max - min + step - 1) / step; 326| 0| if (size <= DEFAULT_MAX_SIZE) { // array container ------------------ | Branch (326:9): [True: 0, False: 0] ------------------ 327| 0| *type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 328| 0| array_container_t *array = array_container_create_given_capacity(size); 329| 0| array_container_add_from_range(array, min, max, step); 330| 0| assert(array->cardinality == size); 331| 0| return array; 332| 0| } else { // bitset container 333| 0| *type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 334| 0| bitset_container_t *bitset = bitset_container_create(); 335| 0| bitset_container_add_from_range(bitset, min, max, step); 336| | assert(bitset->cardinality == size); 337| 0| return bitset; 338| 0| } 339| 0|} roaring.c:container_range_of_ones: 302| 91.5k| uint8_t *result_type) { 303| 91.5k| assert(range_end >= range_start); 304| 91.5k| uint64_t cardinality = range_end - range_start + 1; 305| 91.5k| if (cardinality <= 2) { ------------------ | Branch (305:9): [True: 144, False: 91.3k] ------------------ 306| 144| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 144|#define ARRAY_CONTAINER_TYPE 2 ------------------ 307| 144| return array_container_create_range(range_start, range_end); 308| 91.3k| } else { 309| 91.3k| *result_type = RUN_CONTAINER_TYPE; ------------------ | | 50| 91.3k|#define RUN_CONTAINER_TYPE 3 ------------------ 310| 91.3k| return run_container_create_range(range_start, range_end); 311| 91.3k| } 312| 91.5k|} roaring.c:container_add_range: 2282| 3.34k| uint8_t *result_type) { 2283| | // NB: when selecting new container type, we perform only inexpensive checks 2284| 3.34k| switch (type) { 2285| 0| case BITSET_CONTAINER_TYPE: { ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (2285:9): [True: 0, False: 3.34k] ------------------ 2286| 0| bitset_container_t *bitset = CAST_bitset(c); ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2287| | 2288| 0| int32_t union_cardinality = 0; 2289| 0| union_cardinality += bitset->cardinality; 2290| 0| union_cardinality += max - min + 1; 2291| 0| union_cardinality -= 2292| 0| bitset_lenrange_cardinality(bitset->words, min, max - min); 2293| | 2294| 0| if (union_cardinality == INT32_C(0x10000)) { ------------------ | Branch (2294:17): [True: 0, False: 0] ------------------ 2295| 0| *result_type = RUN_CONTAINER_TYPE; ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ 2296| 0| return run_container_create_range(0, INT32_C(0x10000)); 2297| 0| } else { 2298| 0| *result_type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 2299| 0| bitset_set_lenrange(bitset->words, min, max - min); 2300| 0| bitset->cardinality = union_cardinality; 2301| 0| return bitset; 2302| 0| } 2303| 0| } 2304| 2.88k| case ARRAY_CONTAINER_TYPE: { ------------------ | | 49| 2.88k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (2304:9): [True: 2.88k, False: 462] ------------------ 2305| 2.88k| array_container_t *array = CAST_array(c); ------------------ | | 54| 2.88k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 2.88k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2306| | 2307| 2.88k| int32_t nvals_greater = 2308| 2.88k| count_greater(array->array, array->cardinality, (uint16_t)max); 2309| 2.88k| int32_t nvals_less = 2310| 2.88k| count_less(array->array, array->cardinality - nvals_greater, 2311| 2.88k| (uint16_t)min); 2312| 2.88k| int32_t union_cardinality = 2313| 2.88k| nvals_less + (max - min + 1) + nvals_greater; 2314| | 2315| 2.88k| if (union_cardinality == INT32_C(0x10000)) { ------------------ | Branch (2315:17): [True: 696, False: 2.18k] ------------------ 2316| 696| *result_type = RUN_CONTAINER_TYPE; ------------------ | | 50| 696|#define RUN_CONTAINER_TYPE 3 ------------------ 2317| 696| return run_container_create_range(0, INT32_C(0x10000)); 2318| 2.18k| } else if (union_cardinality <= DEFAULT_MAX_SIZE) { ------------------ | Branch (2318:24): [True: 645, False: 1.54k] ------------------ 2319| 645| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 645|#define ARRAY_CONTAINER_TYPE 2 ------------------ 2320| 645| array_container_add_range_nvals(array, min, max, nvals_less, 2321| 645| nvals_greater); 2322| 645| return array; 2323| 1.54k| } else { 2324| 1.54k| *result_type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 1.54k|#define BITSET_CONTAINER_TYPE 1 ------------------ 2325| 1.54k| bitset_container_t *bitset = bitset_container_from_array(array); 2326| 1.54k| bitset_set_lenrange(bitset->words, min, max - min); 2327| 1.54k| bitset->cardinality = union_cardinality; 2328| 1.54k| return bitset; 2329| 1.54k| } 2330| 2.88k| } 2331| 462| case RUN_CONTAINER_TYPE: { ------------------ | | 50| 462|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (2331:9): [True: 462, False: 2.88k] ------------------ 2332| 462| run_container_t *run = CAST_run(c); ------------------ | | 77| 462|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 462|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2333| | 2334| 462| int32_t nruns_greater = 2335| 462| rle16_count_greater(run->runs, run->n_runs, (uint16_t)max); 2336| 462| int32_t nruns_less = rle16_count_less( 2337| 462| run->runs, run->n_runs - nruns_greater, (uint16_t)min); 2338| | 2339| 462| int32_t run_size_bytes = 2340| 462| (nruns_less + 1 + nruns_greater) * sizeof(rle16_t); 2341| 462| int32_t bitset_size_bytes = 2342| 462| BITSET_CONTAINER_SIZE_IN_WORDS * sizeof(uint64_t); 2343| | 2344| 462| if (run_size_bytes <= bitset_size_bytes) { ------------------ | Branch (2344:17): [True: 462, False: 0] ------------------ 2345| 462| run_container_add_range_nruns(run, min, max, nruns_less, 2346| 462| nruns_greater); 2347| 462| *result_type = RUN_CONTAINER_TYPE; ------------------ | | 50| 462|#define RUN_CONTAINER_TYPE 3 ------------------ 2348| 462| return run; 2349| 462| } else { 2350| 0| return container_from_run_range(run, min, max, result_type); 2351| 0| } 2352| 462| } 2353| 0| default: ------------------ | Branch (2353:9): [True: 0, False: 3.34k] ------------------ 2354| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2355| 3.34k| } 2356| 3.34k|} roaring.c:container_remove_range: 2371| 20.0k| uint8_t *result_type) { 2372| 20.0k| switch (type) { 2373| 1.31k| case BITSET_CONTAINER_TYPE: { ------------------ | | 48| 1.31k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (2373:9): [True: 1.31k, False: 18.7k] ------------------ 2374| 1.31k| bitset_container_t *bitset = CAST_bitset(c); ------------------ | | 52| 1.31k|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.31k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2375| | 2376| 1.31k| int32_t result_cardinality = 2377| 1.31k| bitset->cardinality - 2378| 1.31k| bitset_lenrange_cardinality(bitset->words, min, max - min); 2379| | 2380| 1.31k| if (result_cardinality == 0) { ------------------ | Branch (2380:17): [True: 32, False: 1.28k] ------------------ 2381| 32| return NULL; 2382| 1.28k| } else if (result_cardinality <= DEFAULT_MAX_SIZE) { ------------------ | Branch (2382:24): [True: 110, False: 1.17k] ------------------ 2383| 110| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 110|#define ARRAY_CONTAINER_TYPE 2 ------------------ 2384| 110| bitset_reset_range(bitset->words, min, max + 1); 2385| 110| bitset->cardinality = result_cardinality; 2386| 110| return array_container_from_bitset(bitset); 2387| 1.17k| } else { 2388| 1.17k| *result_type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 1.17k|#define BITSET_CONTAINER_TYPE 1 ------------------ 2389| 1.17k| bitset_reset_range(bitset->words, min, max + 1); 2390| 1.17k| bitset->cardinality = result_cardinality; 2391| 1.17k| return bitset; 2392| 1.17k| } 2393| 1.31k| } 2394| 3.59k| case ARRAY_CONTAINER_TYPE: { ------------------ | | 49| 3.59k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (2394:9): [True: 3.59k, False: 16.4k] ------------------ 2395| 3.59k| array_container_t *array = CAST_array(c); ------------------ | | 54| 3.59k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 3.59k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2396| | 2397| 3.59k| int32_t nvals_greater = 2398| 3.59k| count_greater(array->array, array->cardinality, (uint16_t)max); 2399| 3.59k| int32_t nvals_less = 2400| 3.59k| count_less(array->array, array->cardinality - nvals_greater, 2401| 3.59k| (uint16_t)min); 2402| 3.59k| int32_t result_cardinality = nvals_less + nvals_greater; 2403| | 2404| 3.59k| if (result_cardinality == 0) { ------------------ | Branch (2404:17): [True: 360, False: 3.23k] ------------------ 2405| 360| return NULL; 2406| 3.23k| } else { 2407| 3.23k| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 3.23k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 2408| 3.23k| array_container_remove_range( 2409| 3.23k| array, nvals_less, array->cardinality - result_cardinality); 2410| 3.23k| return array; 2411| 3.23k| } 2412| 3.59k| } 2413| 15.1k| case RUN_CONTAINER_TYPE: { ------------------ | | 50| 15.1k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (2413:9): [True: 15.1k, False: 4.91k] ------------------ 2414| 15.1k| run_container_t *run = CAST_run(c); ------------------ | | 77| 15.1k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 15.1k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2415| | 2416| 15.1k| if (run->n_runs == 0) { ------------------ | Branch (2416:17): [True: 0, False: 15.1k] ------------------ 2417| 0| return NULL; 2418| 0| } 2419| 15.1k| if (min <= run_container_minimum(run) && ------------------ | Branch (2419:17): [True: 14.8k, False: 367] ------------------ 2420| 14.8k| max >= run_container_maximum(run)) { ------------------ | Branch (2420:17): [True: 13.4k, False: 1.39k] ------------------ 2421| 13.4k| return NULL; 2422| 13.4k| } 2423| | 2424| 1.75k| run_container_remove_range(run, min, max); 2425| 1.75k| return convert_run_to_efficient_container(run, result_type); 2426| 15.1k| } 2427| 0| default: ------------------ | Branch (2427:9): [True: 0, False: 20.0k] ------------------ 2428| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2429| 20.0k| } 2430| 20.0k|} roaring.c:container_get_cardinality: 244| 399k| uint8_t typecode) { 245| 399k| c = container_unwrap_shared(c, &typecode); 246| 399k| switch (typecode) { ------------------ | Branch (246:13): [True: 399k, False: 0] ------------------ 247| 9.90k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 9.90k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (247:9): [True: 9.90k, False: 390k] ------------------ 248| 9.90k| return bitset_container_cardinality(const_CAST_bitset(c)); ------------------ | | 53| 9.90k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 9.90k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 249| 64.7k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 64.7k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (249:9): [True: 64.7k, False: 335k] ------------------ 250| 64.7k| return array_container_cardinality(const_CAST_array(c)); ------------------ | | 55| 64.7k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 64.7k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 251| 325k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 325k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (251:9): [True: 325k, False: 74.7k] ------------------ 252| 325k| return run_container_cardinality(const_CAST_run(c)); ------------------ | | 78| 325k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 325k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 253| 399k| } 254| 399k| assert(false); 255| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 256| 0| return 0; // unreached 257| 399k|} roaring.c:get_container_type: 129| 81.1k|static inline uint8_t get_container_type(const container_t *c, uint8_t type) { 130| 81.1k| if (type == SHARED_CONTAINER_TYPE) { ------------------ | | 51| 81.1k|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (130:9): [True: 0, False: 81.1k] ------------------ 131| 0| return const_CAST_shared(c)->typecode; ------------------ | | 80| 0|#define const_CAST_shared(c) CAST(const shared_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 132| 81.1k| } else { 133| 81.1k| return type; 134| 81.1k| } 135| 81.1k|} roaring.c:get_writable_copy_if_shared: 146| 8.37M| uint8_t *type) { 147| 8.37M| if (*type == SHARED_CONTAINER_TYPE) { // shared, return enclosed container ------------------ | | 51| 8.37M|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (147:9): [True: 0, False: 8.37M] ------------------ 148| 0| return shared_container_extract_copy(CAST_shared(c), type); ------------------ | | 79| 0|#define CAST_shared(c) CAST(shared_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 149| 8.37M| } else { 150| 8.37M| return c; // not shared, so return as-is 151| 8.37M| } 152| 8.37M|} roaring.c:container_add: 492| 8.13M| uint8_t *new_typecode) { 493| 8.13M| c = get_writable_copy_if_shared(c, &typecode); 494| 8.13M| switch (typecode) { 495| 326k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 326k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (495:9): [True: 326k, False: 7.80M] ------------------ 496| 326k| bitset_container_set(CAST_bitset(c), val); ------------------ | | 52| 326k|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 326k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 497| 326k| *new_typecode = BITSET_CONTAINER_TYPE; ------------------ | | 48| 326k|#define BITSET_CONTAINER_TYPE 1 ------------------ 498| 326k| return c; 499| 7.39M| case ARRAY_CONTAINER_TYPE: { ------------------ | | 49| 7.39M|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (499:9): [True: 7.39M, False: 740k] ------------------ 500| 7.39M| array_container_t *ac = CAST_array(c); ------------------ | | 54| 7.39M|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 7.39M|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 501| 7.39M| if (array_container_try_add(ac, val, DEFAULT_MAX_SIZE) != -1) { ------------------ | Branch (501:17): [True: 7.39M, False: 131] ------------------ 502| 7.39M| *new_typecode = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 7.39M|#define ARRAY_CONTAINER_TYPE 2 ------------------ 503| 7.39M| return ac; 504| 7.39M| } else { 505| 131| bitset_container_t *bitset = bitset_container_from_array(ac); 506| 131| bitset_container_add(bitset, val); 507| 131| *new_typecode = BITSET_CONTAINER_TYPE; ------------------ | | 48| 131|#define BITSET_CONTAINER_TYPE 1 ------------------ 508| 131| return bitset; 509| 131| } 510| 7.39M| } break; 511| 413k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 413k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (511:9): [True: 413k, False: 7.71M] ------------------ 512| | // per Java, no container type adjustments are done (revisit?) 513| 413k| run_container_add(CAST_run(c), val); ------------------ | | 77| 413k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 413k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 514| 413k| *new_typecode = RUN_CONTAINER_TYPE; ------------------ | | 50| 413k|#define RUN_CONTAINER_TYPE 3 ------------------ 515| 413k| return c; 516| 0| default: ------------------ | Branch (516:9): [True: 0, False: 8.13M] ------------------ 517| 0| assert(false); 518| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 519| 0| return NULL; 520| 8.13M| } 521| 8.13M|} roaring.c:container_remove: 533| 11.6k| uint8_t *new_typecode) { 534| 11.6k| c = get_writable_copy_if_shared(c, &typecode); 535| 11.6k| switch (typecode) { 536| 2.32k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 2.32k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (536:9): [True: 2.32k, False: 9.35k] ------------------ 537| 2.32k| if (bitset_container_remove(CAST_bitset(c), val)) { ------------------ | | 52| 2.32k|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 2.32k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (537:17): [True: 1.41k, False: 911] ------------------ 538| 1.41k| int card = bitset_container_cardinality(CAST_bitset(c)); ------------------ | | 52| 1.41k|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.41k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 539| 1.41k| if (card <= DEFAULT_MAX_SIZE) { ------------------ | Branch (539:21): [True: 14, False: 1.40k] ------------------ 540| 14| *new_typecode = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 14|#define ARRAY_CONTAINER_TYPE 2 ------------------ 541| 14| return array_container_from_bitset(CAST_bitset(c)); ------------------ | | 52| 14|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 14|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 542| 14| } 543| 1.41k| } 544| 2.31k| *new_typecode = typecode; 545| 2.31k| return c; 546| 6.21k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 6.21k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (546:9): [True: 6.21k, False: 5.46k] ------------------ 547| 6.21k| *new_typecode = typecode; 548| 6.21k| array_container_remove(CAST_array(c), val); ------------------ | | 54| 6.21k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 6.21k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 549| 6.21k| return c; 550| 3.14k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 3.14k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (550:9): [True: 3.14k, False: 8.53k] ------------------ 551| | // per Java, no container type adjustments are done (revisit?) 552| 3.14k| run_container_remove(CAST_run(c), val); ------------------ | | 77| 3.14k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 3.14k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 553| 3.14k| *new_typecode = RUN_CONTAINER_TYPE; ------------------ | | 50| 3.14k|#define RUN_CONTAINER_TYPE 3 ------------------ 554| 3.14k| return c; 555| 0| default: ------------------ | Branch (555:9): [True: 0, False: 11.6k] ------------------ 556| 0| assert(false); 557| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 558| 0| return NULL; 559| 11.6k| } 560| 11.6k|} roaring.c:container_nonzero_cardinality: 437| 181k| uint8_t typecode) { 438| 181k| c = container_unwrap_shared(c, &typecode); 439| 181k| switch (typecode) { ------------------ | Branch (439:13): [True: 181k, False: 0] ------------------ 440| 7.16k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 7.16k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (440:9): [True: 7.16k, False: 174k] ------------------ 441| 7.16k| return bitset_container_const_nonzero_cardinality( 442| 7.16k| const_CAST_bitset(c)); ------------------ | | 53| 7.16k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 7.16k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 443| 107k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 107k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (443:9): [True: 107k, False: 74.2k] ------------------ 444| 107k| return array_container_nonzero_cardinality(const_CAST_array(c)); ------------------ | | 55| 107k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 107k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 445| 67.0k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 67.0k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (445:9): [True: 67.0k, False: 114k] ------------------ 446| 67.0k| return run_container_nonzero_cardinality(const_CAST_run(c)); ------------------ | | 78| 67.0k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 67.0k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 447| 181k| } 448| 181k| assert(false); 449| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 450| 0| return 0; // unreached 451| 181k|} roaring.c:container_and: 728| 6.36k| uint8_t *result_type) { 729| 6.36k| c1 = container_unwrap_shared(c1, &type1); 730| 6.36k| c2 = container_unwrap_shared(c2, &type2); 731| 6.36k| container_t *result = NULL; 732| 6.36k| switch (PAIR_CONTAINER_TYPES(type1, type2)) { ------------------ | | 62| 6.36k|#define PAIR_CONTAINER_TYPES(type1, type2) (4 * (type1) + (type2)) ------------------ 733| 0| case CONTAINER_PAIR(BITSET, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (733:9): [True: 0, False: 6.36k] ------------------ 734| 0| *result_type = 735| 0| bitset_bitset_container_intersection( ------------------ | Branch (735:17): [True: 0, False: 0] ------------------ 736| 0| const_CAST_bitset(c1), const_CAST_bitset(c2), &result) ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_bitset(c1), const_CAST_bitset(c2), &result) ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 737| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 738| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 739| 0| return result; 740| | 741| 2.32k| case CONTAINER_PAIR(ARRAY, ARRAY): ------------------ | | 65| 2.32k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.32k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.32k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (741:9): [True: 2.32k, False: 4.03k] ------------------ 742| 2.32k| result = array_container_create(); 743| 2.32k| array_container_intersection( 744| 2.32k| const_CAST_array(c1), const_CAST_array(c2), CAST_array(result)); ------------------ | | 55| 2.32k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.32k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c1), const_CAST_array(c2), CAST_array(result)); ------------------ | | 55| 2.32k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.32k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c1), const_CAST_array(c2), CAST_array(result)); ------------------ | | 54| 2.32k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 2.32k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 745| 2.32k| *result_type = ARRAY_CONTAINER_TYPE; // never bitset ------------------ | | 49| 2.32k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 746| 2.32k| return result; 747| | 748| 680| case CONTAINER_PAIR(RUN, RUN): ------------------ | | 65| 680| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 680|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 680|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (748:9): [True: 680, False: 5.68k] ------------------ 749| 680| result = run_container_create(); 750| 680| run_container_intersection(const_CAST_run(c1), const_CAST_run(c2), ------------------ | | 78| 680|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 680|#define CAST(type, value) ((type)value) | | ------------------ ------------------ run_container_intersection(const_CAST_run(c1), const_CAST_run(c2), ------------------ | | 78| 680|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 680|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 751| 680| CAST_run(result)); ------------------ | | 77| 680|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 680|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 752| 680| return convert_run_to_efficient_container_and_free(CAST_run(result), ------------------ | | 77| 680|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 680|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 753| 680| result_type); 754| | 755| 0| case CONTAINER_PAIR(BITSET, ARRAY): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (755:9): [True: 0, False: 6.36k] ------------------ 756| 0| result = array_container_create(); 757| 0| array_bitset_container_intersection(const_CAST_array(c2), ------------------ | | 55| 0|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 758| 0| const_CAST_bitset(c1), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 759| 0| CAST_array(result)); ------------------ | | 54| 0|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 760| 0| *result_type = ARRAY_CONTAINER_TYPE; // never bitset ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 761| 0| return result; 762| | 763| 575| case CONTAINER_PAIR(ARRAY, BITSET): ------------------ | | 65| 575| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 575|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 575|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (763:9): [True: 575, False: 5.78k] ------------------ 764| 575| result = array_container_create(); 765| 575| *result_type = ARRAY_CONTAINER_TYPE; // never bitset ------------------ | | 49| 575|#define ARRAY_CONTAINER_TYPE 2 ------------------ 766| 575| array_bitset_container_intersection(const_CAST_array(c1), ------------------ | | 55| 575|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 575|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 767| 575| const_CAST_bitset(c2), ------------------ | | 53| 575|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 575|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 768| 575| CAST_array(result)); ------------------ | | 54| 575|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 575|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 769| 575| return result; 770| | 771| 0| case CONTAINER_PAIR(BITSET, RUN): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (771:9): [True: 0, False: 6.36k] ------------------ 772| 0| *result_type = 773| 0| run_bitset_container_intersection( ------------------ | Branch (773:17): [True: 0, False: 0] ------------------ 774| 0| const_CAST_run(c2), const_CAST_bitset(c1), &result) ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c2), const_CAST_bitset(c1), &result) ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 775| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 776| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 777| 0| return result; 778| | 779| 691| case CONTAINER_PAIR(RUN, BITSET): ------------------ | | 65| 691| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 691|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 691|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (779:9): [True: 691, False: 5.66k] ------------------ 780| 691| *result_type = 781| 691| run_bitset_container_intersection( ------------------ | Branch (781:17): [True: 0, False: 691] ------------------ 782| 691| const_CAST_run(c1), const_CAST_bitset(c2), &result) ------------------ | | 78| 691|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c1), const_CAST_bitset(c2), &result) ------------------ | | 53| 691|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 783| 691| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 784| 691| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 1.38k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 785| 691| return result; 786| | 787| 814| case CONTAINER_PAIR(ARRAY, RUN): ------------------ | | 65| 814| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 814|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 814|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (787:9): [True: 814, False: 5.54k] ------------------ 788| 814| result = array_container_create(); 789| 814| *result_type = ARRAY_CONTAINER_TYPE; // never bitset ------------------ | | 49| 814|#define ARRAY_CONTAINER_TYPE 2 ------------------ 790| 814| array_run_container_intersection( 791| 814| const_CAST_array(c1), const_CAST_run(c2), CAST_array(result)); ------------------ | | 55| 814|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 814|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c1), const_CAST_run(c2), CAST_array(result)); ------------------ | | 78| 814|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 814|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c1), const_CAST_run(c2), CAST_array(result)); ------------------ | | 54| 814|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 814|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 792| 814| return result; 793| | 794| 1.27k| case CONTAINER_PAIR(RUN, ARRAY): ------------------ | | 65| 1.27k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 1.27k|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 1.27k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (794:9): [True: 1.27k, False: 5.08k] ------------------ 795| 1.27k| result = array_container_create(); 796| 1.27k| *result_type = ARRAY_CONTAINER_TYPE; // never bitset ------------------ | | 49| 1.27k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 797| 1.27k| array_run_container_intersection( 798| 1.27k| const_CAST_array(c2), const_CAST_run(c1), CAST_array(result)); ------------------ | | 55| 1.27k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c2), const_CAST_run(c1), CAST_array(result)); ------------------ | | 78| 1.27k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c2), const_CAST_run(c1), CAST_array(result)); ------------------ | | 54| 1.27k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 799| 1.27k| return result; 800| | 801| 0| default: ------------------ | Branch (801:9): [True: 0, False: 6.36k] ------------------ 802| 0| assert(false); 803| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 804| 0| return NULL; 805| 6.36k| } 806| 6.36k|} roaring.c:container_iand: 923| 69.2k| uint8_t *result_type) { 924| 69.2k| c1 = get_writable_copy_if_shared(c1, &type1); 925| 69.2k| c2 = container_unwrap_shared(c2, &type2); 926| 69.2k| container_t *result = NULL; 927| 69.2k| switch (PAIR_CONTAINER_TYPES(type1, type2)) { ------------------ | | 62| 69.2k|#define PAIR_CONTAINER_TYPES(type1, type2) (4 * (type1) + (type2)) ------------------ 928| 1.51k| case CONTAINER_PAIR(BITSET, BITSET): ------------------ | | 65| 1.51k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 1.51k|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 1.51k|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (928:9): [True: 1.51k, False: 67.7k] ------------------ 929| 1.51k| *result_type = bitset_bitset_container_intersection_inplace( ------------------ | Branch (929:28): [True: 1.51k, False: 0] ------------------ 930| 1.51k| CAST_bitset(c1), const_CAST_bitset(c2), &result) ------------------ | | 52| 1.51k|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.51k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_bitset(c1), const_CAST_bitset(c2), &result) ------------------ | | 53| 1.51k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 1.51k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 931| 1.51k| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 1.51k|#define BITSET_CONTAINER_TYPE 1 ------------------ 932| 1.51k| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 1.51k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 933| 1.51k| return result; 934| | 935| 4.86k| case CONTAINER_PAIR(ARRAY, ARRAY): ------------------ | | 65| 4.86k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 4.86k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 4.86k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (935:9): [True: 4.86k, False: 64.4k] ------------------ 936| 4.86k| array_container_intersection_inplace(CAST_array(c1), ------------------ | | 54| 4.86k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 4.86k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 937| 4.86k| const_CAST_array(c2)); ------------------ | | 55| 4.86k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 4.86k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 938| 4.86k| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 4.86k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 939| 4.86k| return c1; 940| | 941| 61.5k| case CONTAINER_PAIR(RUN, RUN): ------------------ | | 65| 61.5k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 61.5k|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 61.5k|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (941:9): [True: 61.5k, False: 7.69k] ------------------ 942| 61.5k| result = run_container_create(); 943| 61.5k| run_container_intersection(const_CAST_run(c1), const_CAST_run(c2), ------------------ | | 78| 61.5k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 61.5k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ run_container_intersection(const_CAST_run(c1), const_CAST_run(c2), ------------------ | | 78| 61.5k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 61.5k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 944| 61.5k| CAST_run(result)); ------------------ | | 77| 61.5k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 61.5k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 945| | // as of January 2016, Java code used non-in-place intersection for 946| | // two runcontainers 947| 61.5k| return convert_run_to_efficient_container_and_free(CAST_run(result), ------------------ | | 77| 61.5k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 61.5k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 948| 61.5k| result_type); 949| | 950| 52| case CONTAINER_PAIR(BITSET, ARRAY): ------------------ | | 65| 52| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 52|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 52|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (950:9): [True: 52, False: 69.2k] ------------------ 951| | // c1 is a bitmap so no inplace possible 952| 52| result = array_container_create(); 953| 52| array_bitset_container_intersection(const_CAST_array(c2), ------------------ | | 55| 52|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 52|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 954| 52| const_CAST_bitset(c1), ------------------ | | 53| 52|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 52|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 955| 52| CAST_array(result)); ------------------ | | 54| 52|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 52|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 956| 52| *result_type = ARRAY_CONTAINER_TYPE; // never bitset ------------------ | | 49| 52|#define ARRAY_CONTAINER_TYPE 2 ------------------ 957| 52| return result; 958| | 959| 0| case CONTAINER_PAIR(ARRAY, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (959:9): [True: 0, False: 69.2k] ------------------ 960| 0| *result_type = ARRAY_CONTAINER_TYPE; // never bitset ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 961| 0| array_bitset_container_intersection( 962| 0| const_CAST_array(c1), const_CAST_bitset(c2), ------------------ | | 55| 0|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c1), const_CAST_bitset(c2), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 963| 0| CAST_array(c1)); // result is allowed to be same as c1 ------------------ | | 54| 0|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 964| 0| return c1; 965| | 966| 0| case CONTAINER_PAIR(BITSET, RUN): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (966:9): [True: 0, False: 69.2k] ------------------ 967| | // will attempt in-place computation 968| 0| *result_type = run_bitset_container_intersection( ------------------ | Branch (968:28): [True: 0, False: 0] ------------------ 969| 0| const_CAST_run(c2), const_CAST_bitset(c1), &c1) ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c2), const_CAST_bitset(c1), &c1) ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 970| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 971| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 972| 0| return c1; 973| | 974| 0| case CONTAINER_PAIR(RUN, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (974:9): [True: 0, False: 69.2k] ------------------ 975| 0| *result_type = 976| 0| run_bitset_container_intersection( ------------------ | Branch (976:17): [True: 0, False: 0] ------------------ 977| 0| const_CAST_run(c1), const_CAST_bitset(c2), &result) ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c1), const_CAST_bitset(c2), &result) ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 978| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 979| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 980| 0| return result; 981| | 982| 193| case CONTAINER_PAIR(ARRAY, RUN): ------------------ | | 65| 193| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 193|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 193|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (982:9): [True: 193, False: 69.0k] ------------------ 983| 193| result = array_container_create(); 984| 193| *result_type = ARRAY_CONTAINER_TYPE; // never bitset ------------------ | | 49| 193|#define ARRAY_CONTAINER_TYPE 2 ------------------ 985| 193| array_run_container_intersection( 986| 193| const_CAST_array(c1), const_CAST_run(c2), CAST_array(result)); ------------------ | | 55| 193|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 193|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c1), const_CAST_run(c2), CAST_array(result)); ------------------ | | 78| 193|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 193|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c1), const_CAST_run(c2), CAST_array(result)); ------------------ | | 54| 193|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 193|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 987| 193| return result; 988| | 989| 1.07k| case CONTAINER_PAIR(RUN, ARRAY): ------------------ | | 65| 1.07k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 1.07k|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 1.07k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (989:9): [True: 1.07k, False: 68.2k] ------------------ 990| 1.07k| result = array_container_create(); 991| 1.07k| *result_type = ARRAY_CONTAINER_TYPE; // never bitset ------------------ | | 49| 1.07k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 992| 1.07k| array_run_container_intersection( 993| 1.07k| const_CAST_array(c2), const_CAST_run(c1), CAST_array(result)); ------------------ | | 55| 1.07k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 1.07k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c2), const_CAST_run(c1), CAST_array(result)); ------------------ | | 78| 1.07k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 1.07k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c2), const_CAST_run(c1), CAST_array(result)); ------------------ | | 54| 1.07k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.07k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 994| 1.07k| return result; 995| | 996| 0| default: ------------------ | Branch (996:9): [True: 0, False: 69.2k] ------------------ 997| 0| assert(false); 998| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 999| 0| return NULL; 1000| 69.2k| } 1001| 69.2k|} roaring.c:container_or: 1010| 6.36k| uint8_t *result_type) { 1011| 6.36k| c1 = container_unwrap_shared(c1, &type1); 1012| 6.36k| c2 = container_unwrap_shared(c2, &type2); 1013| 6.36k| container_t *result = NULL; 1014| 6.36k| switch (PAIR_CONTAINER_TYPES(type1, type2)) { ------------------ | | 62| 6.36k|#define PAIR_CONTAINER_TYPES(type1, type2) (4 * (type1) + (type2)) ------------------ 1015| 0| case CONTAINER_PAIR(BITSET, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1015:9): [True: 0, False: 6.36k] ------------------ 1016| 0| result = bitset_container_create(); 1017| 0| bitset_container_or(const_CAST_bitset(c1), const_CAST_bitset(c2), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ bitset_container_or(const_CAST_bitset(c1), const_CAST_bitset(c2), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1018| 0| CAST_bitset(result)); ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1019| 0| *result_type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1020| 0| return result; 1021| | 1022| 2.32k| case CONTAINER_PAIR(ARRAY, ARRAY): ------------------ | | 65| 2.32k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.32k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.32k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1022:9): [True: 2.32k, False: 4.03k] ------------------ 1023| 2.32k| *result_type = 1024| 2.32k| array_array_container_union(const_CAST_array(c1), ------------------ | | 55| 2.32k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.32k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1024:17): [True: 52, False: 2.27k] ------------------ 1025| 2.32k| const_CAST_array(c2), &result) ------------------ | | 55| 2.32k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.32k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1026| 2.32k| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 52|#define BITSET_CONTAINER_TYPE 1 ------------------ 1027| 2.32k| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 4.59k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1028| 2.32k| return result; 1029| | 1030| 680| case CONTAINER_PAIR(RUN, RUN): ------------------ | | 65| 680| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 680|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 680|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1030:9): [True: 680, False: 5.68k] ------------------ 1031| 680| result = run_container_create(); 1032| 680| run_container_union(const_CAST_run(c1), const_CAST_run(c2), ------------------ | | 78| 680|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 680|#define CAST(type, value) ((type)value) | | ------------------ ------------------ run_container_union(const_CAST_run(c1), const_CAST_run(c2), ------------------ | | 78| 680|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 680|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1033| 680| CAST_run(result)); ------------------ | | 77| 680|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 680|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1034| 680| *result_type = RUN_CONTAINER_TYPE; ------------------ | | 50| 680|#define RUN_CONTAINER_TYPE 3 ------------------ 1035| | // todo: could be optimized since will never convert to array 1036| 680| result = convert_run_to_efficient_container_and_free( 1037| 680| CAST_run(result), result_type); ------------------ | | 77| 680|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 680|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1038| 680| return result; 1039| | 1040| 0| case CONTAINER_PAIR(BITSET, ARRAY): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1040:9): [True: 0, False: 6.36k] ------------------ 1041| 0| result = bitset_container_create(); 1042| 0| array_bitset_container_union(const_CAST_array(c2), ------------------ | | 55| 0|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1043| 0| const_CAST_bitset(c1), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1044| 0| CAST_bitset(result)); ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1045| 0| *result_type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1046| 0| return result; 1047| | 1048| 575| case CONTAINER_PAIR(ARRAY, BITSET): ------------------ | | 65| 575| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 575|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 575|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1048:9): [True: 575, False: 5.78k] ------------------ 1049| 575| result = bitset_container_create(); 1050| 575| array_bitset_container_union(const_CAST_array(c1), ------------------ | | 55| 575|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 575|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1051| 575| const_CAST_bitset(c2), ------------------ | | 53| 575|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 575|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1052| 575| CAST_bitset(result)); ------------------ | | 52| 575|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 575|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1053| 575| *result_type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 575|#define BITSET_CONTAINER_TYPE 1 ------------------ 1054| 575| return result; 1055| | 1056| 0| case CONTAINER_PAIR(BITSET, RUN): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1056:9): [True: 0, False: 6.36k] ------------------ 1057| 0| if (run_container_is_full(const_CAST_run(c2))) { ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1057:17): [True: 0, False: 0] ------------------ 1058| 0| result = run_container_create(); 1059| 0| *result_type = RUN_CONTAINER_TYPE; ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ 1060| 0| run_container_copy(const_CAST_run(c2), CAST_run(result)); ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ run_container_copy(const_CAST_run(c2), CAST_run(result)); ------------------ | | 77| 0|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1061| 0| return result; 1062| 0| } 1063| 0| result = bitset_container_create(); 1064| 0| run_bitset_container_union( 1065| 0| const_CAST_run(c2), const_CAST_bitset(c1), CAST_bitset(result)); ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c2), const_CAST_bitset(c1), CAST_bitset(result)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c2), const_CAST_bitset(c1), CAST_bitset(result)); ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1066| 0| *result_type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1067| 0| return result; 1068| | 1069| 691| case CONTAINER_PAIR(RUN, BITSET): ------------------ | | 65| 691| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 691|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 691|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1069:9): [True: 691, False: 5.66k] ------------------ 1070| 691| if (run_container_is_full(const_CAST_run(c1))) { ------------------ | | 78| 691|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1070:17): [True: 0, False: 691] ------------------ 1071| 0| result = run_container_create(); 1072| 0| *result_type = RUN_CONTAINER_TYPE; ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ 1073| 0| run_container_copy(const_CAST_run(c1), CAST_run(result)); ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ run_container_copy(const_CAST_run(c1), CAST_run(result)); ------------------ | | 77| 0|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1074| 0| return result; 1075| 0| } 1076| 691| result = bitset_container_create(); 1077| 691| run_bitset_container_union( 1078| 691| const_CAST_run(c1), const_CAST_bitset(c2), CAST_bitset(result)); ------------------ | | 78| 691|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c1), const_CAST_bitset(c2), CAST_bitset(result)); ------------------ | | 53| 691|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c1), const_CAST_bitset(c2), CAST_bitset(result)); ------------------ | | 52| 691|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1079| 691| *result_type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 691|#define BITSET_CONTAINER_TYPE 1 ------------------ 1080| 691| return result; 1081| | 1082| 814| case CONTAINER_PAIR(ARRAY, RUN): ------------------ | | 65| 814| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 814|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 814|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1082:9): [True: 814, False: 5.54k] ------------------ 1083| 814| result = run_container_create(); 1084| 814| array_run_container_union(const_CAST_array(c1), const_CAST_run(c2), ------------------ | | 55| 814|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 814|#define CAST(type, value) ((type)value) | | ------------------ ------------------ array_run_container_union(const_CAST_array(c1), const_CAST_run(c2), ------------------ | | 78| 814|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 814|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1085| 814| CAST_run(result)); ------------------ | | 77| 814|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 814|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1086| 814| result = convert_run_to_efficient_container_and_free( 1087| 814| CAST_run(result), result_type); ------------------ | | 77| 814|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 814|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1088| 814| return result; 1089| | 1090| 1.27k| case CONTAINER_PAIR(RUN, ARRAY): ------------------ | | 65| 1.27k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 1.27k|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 1.27k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1090:9): [True: 1.27k, False: 5.08k] ------------------ 1091| 1.27k| result = run_container_create(); 1092| 1.27k| array_run_container_union(const_CAST_array(c2), const_CAST_run(c1), ------------------ | | 55| 1.27k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ array_run_container_union(const_CAST_array(c2), const_CAST_run(c1), ------------------ | | 78| 1.27k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1093| 1.27k| CAST_run(result)); ------------------ | | 77| 1.27k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1094| 1.27k| result = convert_run_to_efficient_container_and_free( 1095| 1.27k| CAST_run(result), result_type); ------------------ | | 77| 1.27k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1096| 1.27k| return result; 1097| | 1098| 0| default: ------------------ | Branch (1098:9): [True: 0, False: 6.36k] ------------------ 1099| 0| assert(false); 1100| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 1101| 0| return NULL; // unreached 1102| 6.36k| } 1103| 6.36k|} roaring.c:container_is_full: 262| 7.34k|static inline bool container_is_full(const container_t *c, uint8_t typecode) { 263| 7.34k| c = container_unwrap_shared(c, &typecode); 264| 7.34k| switch (typecode) { ------------------ | Branch (264:13): [True: 7.34k, False: 0] ------------------ 265| 0| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (265:9): [True: 0, False: 7.34k] ------------------ 266| 0| return bitset_container_cardinality(const_CAST_bitset(c)) == ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 267| 0| (1 << 16); 268| 3.80k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 3.80k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (268:9): [True: 3.80k, False: 3.53k] ------------------ 269| 3.80k| return array_container_cardinality(const_CAST_array(c)) == ------------------ | | 55| 3.80k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 3.80k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 270| 3.80k| (1 << 16); 271| 3.53k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 3.53k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (271:9): [True: 3.53k, False: 3.80k] ------------------ 272| 3.53k| return run_container_is_full(const_CAST_run(c)); ------------------ | | 78| 3.53k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 3.53k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 273| 7.34k| } 274| 7.34k| assert(false); 275| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 276| 0| return 0; // unreached 277| 7.34k|} roaring.c:container_ior: 1228| 6.48k| uint8_t *result_type) { 1229| 6.48k| c1 = get_writable_copy_if_shared(c1, &type1); 1230| 6.48k| c2 = container_unwrap_shared(c2, &type2); 1231| 6.48k| container_t *result = NULL; 1232| 6.48k| switch (PAIR_CONTAINER_TYPES(type1, type2)) { ------------------ | | 62| 6.48k|#define PAIR_CONTAINER_TYPES(type1, type2) (4 * (type1) + (type2)) ------------------ 1233| 0| case CONTAINER_PAIR(BITSET, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1233:9): [True: 0, False: 6.48k] ------------------ 1234| 0| bitset_container_or(const_CAST_bitset(c1), const_CAST_bitset(c2), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ bitset_container_or(const_CAST_bitset(c1), const_CAST_bitset(c2), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1235| 0| CAST_bitset(c1)); ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1236| 0|#ifdef OR_BITSET_CONVERSION_TO_FULL 1237| 0| if (CAST_bitset(c1)->cardinality == (1 << 16)) { // we convert ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1237:17): [True: 0, False: 0] ------------------ 1238| 0| result = run_container_create_range(0, (1 << 16)); 1239| 0| *result_type = RUN_CONTAINER_TYPE; ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ 1240| 0| return result; 1241| 0| } 1242| 0|#endif 1243| 0| *result_type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1244| 0| return c1; 1245| | 1246| 2.51k| case CONTAINER_PAIR(ARRAY, ARRAY): ------------------ | | 65| 2.51k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.51k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.51k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1246:9): [True: 2.51k, False: 3.96k] ------------------ 1247| 2.51k| *result_type = array_array_container_inplace_union( ------------------ | Branch (1247:28): [True: 0, False: 2.51k] ------------------ 1248| 2.51k| CAST_array(c1), const_CAST_array(c2), &result) ------------------ | | 54| 2.51k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 2.51k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_array(c1), const_CAST_array(c2), &result) ------------------ | | 55| 2.51k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.51k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1249| 2.51k| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1250| 2.51k| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 5.02k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1251| 2.51k| if ((result == NULL) && (*result_type == ARRAY_CONTAINER_TYPE)) { ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (1251:17): [True: 0, False: 2.51k] | Branch (1251:37): [True: 0, False: 0] ------------------ 1252| 0| return c1; // the computation was done in-place! 1253| 0| } 1254| 2.51k| return result; 1255| | 1256| 1.73k| case CONTAINER_PAIR(RUN, RUN): ------------------ | | 65| 1.73k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 1.73k|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 1.73k|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1256:9): [True: 1.73k, False: 4.74k] ------------------ 1257| 1.73k| run_container_union_inplace(CAST_run(c1), const_CAST_run(c2)); ------------------ | | 77| 1.73k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.73k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ run_container_union_inplace(CAST_run(c1), const_CAST_run(c2)); ------------------ | | 78| 1.73k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 1.73k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1258| 1.73k| return convert_run_to_efficient_container(CAST_run(c1), ------------------ | | 77| 1.73k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.73k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1259| 1.73k| result_type); 1260| | 1261| 0| case CONTAINER_PAIR(BITSET, ARRAY): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1261:9): [True: 0, False: 6.48k] ------------------ 1262| 0| array_bitset_container_union( 1263| 0| const_CAST_array(c2), const_CAST_bitset(c1), CAST_bitset(c1)); ------------------ | | 55| 0|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c2), const_CAST_bitset(c1), CAST_bitset(c1)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c2), const_CAST_bitset(c1), CAST_bitset(c1)); ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1264| 0| *result_type = BITSET_CONTAINER_TYPE; // never array ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1265| 0| return c1; 1266| | 1267| 627| case CONTAINER_PAIR(ARRAY, BITSET): ------------------ | | 65| 627| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 627|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 627|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1267:9): [True: 627, False: 5.85k] ------------------ 1268| | // c1 is an array, so no in-place possible 1269| 627| result = bitset_container_create(); 1270| 627| *result_type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 627|#define BITSET_CONTAINER_TYPE 1 ------------------ 1271| 627| array_bitset_container_union(const_CAST_array(c1), ------------------ | | 55| 627|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 627|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1272| 627| const_CAST_bitset(c2), ------------------ | | 53| 627|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 627|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1273| 627| CAST_bitset(result)); ------------------ | | 52| 627|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 627|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1274| 627| return result; 1275| | 1276| 0| case CONTAINER_PAIR(BITSET, RUN): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1276:9): [True: 0, False: 6.48k] ------------------ 1277| 0| if (run_container_is_full(const_CAST_run(c2))) { ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1277:17): [True: 0, False: 0] ------------------ 1278| 0| result = run_container_create(); 1279| 0| *result_type = RUN_CONTAINER_TYPE; ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ 1280| 0| run_container_copy(const_CAST_run(c2), CAST_run(result)); ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ run_container_copy(const_CAST_run(c2), CAST_run(result)); ------------------ | | 77| 0|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1281| 0| return result; 1282| 0| } 1283| 0| run_bitset_container_union(const_CAST_run(c2), ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1284| 0| const_CAST_bitset(c1), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1285| 0| CAST_bitset(c1)); // allowed ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1286| 0| *result_type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1287| 0| return c1; 1288| | 1289| 691| case CONTAINER_PAIR(RUN, BITSET): ------------------ | | 65| 691| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 691|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 691|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1289:9): [True: 691, False: 5.78k] ------------------ 1290| 691| if (run_container_is_full(const_CAST_run(c1))) { ------------------ | | 78| 691|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1290:17): [True: 0, False: 691] ------------------ 1291| 0| *result_type = RUN_CONTAINER_TYPE; ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ 1292| 0| return c1; 1293| 0| } 1294| 691| result = bitset_container_create(); 1295| 691| run_bitset_container_union( 1296| 691| const_CAST_run(c1), const_CAST_bitset(c2), CAST_bitset(result)); ------------------ | | 78| 691|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c1), const_CAST_bitset(c2), CAST_bitset(result)); ------------------ | | 53| 691|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c1), const_CAST_bitset(c2), CAST_bitset(result)); ------------------ | | 52| 691|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1297| 691| *result_type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 691|#define BITSET_CONTAINER_TYPE 1 ------------------ 1298| 691| return result; 1299| | 1300| 665| case CONTAINER_PAIR(ARRAY, RUN): ------------------ | | 65| 665| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 665|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 665|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1300:9): [True: 665, False: 5.81k] ------------------ 1301| 665| result = run_container_create(); 1302| 665| array_run_container_union(const_CAST_array(c1), const_CAST_run(c2), ------------------ | | 55| 665|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 665|#define CAST(type, value) ((type)value) | | ------------------ ------------------ array_run_container_union(const_CAST_array(c1), const_CAST_run(c2), ------------------ | | 78| 665|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 665|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1303| 665| CAST_run(result)); ------------------ | | 77| 665|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 665|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1304| 665| result = convert_run_to_efficient_container_and_free( 1305| 665| CAST_run(result), result_type); ------------------ | | 77| 665|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 665|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1306| 665| return result; 1307| | 1308| 248| case CONTAINER_PAIR(RUN, ARRAY): ------------------ | | 65| 248| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 248|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 248|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1308:9): [True: 248, False: 6.23k] ------------------ 1309| 248| array_run_container_inplace_union(const_CAST_array(c2), ------------------ | | 55| 248|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 248|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1310| 248| CAST_run(c1)); ------------------ | | 77| 248|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 248|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1311| 248| c1 = convert_run_to_efficient_container(CAST_run(c1), result_type); ------------------ | | 77| 248|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 248|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1312| 248| return c1; 1313| | 1314| 0| default: ------------------ | Branch (1314:9): [True: 0, False: 6.48k] ------------------ 1315| 0| assert(false); 1316| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 1317| 0| return NULL; 1318| 6.48k| } 1319| 6.48k|} roaring.c:container_xor: 1451| 6.36k| uint8_t *result_type) { 1452| 6.36k| c1 = container_unwrap_shared(c1, &type1); 1453| 6.36k| c2 = container_unwrap_shared(c2, &type2); 1454| 6.36k| container_t *result = NULL; 1455| 6.36k| switch (PAIR_CONTAINER_TYPES(type1, type2)) { ------------------ | | 62| 6.36k|#define PAIR_CONTAINER_TYPES(type1, type2) (4 * (type1) + (type2)) ------------------ 1456| 0| case CONTAINER_PAIR(BITSET, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1456:9): [True: 0, False: 6.36k] ------------------ 1457| 0| *result_type = 1458| 0| bitset_bitset_container_xor(const_CAST_bitset(c1), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1458:17): [True: 0, False: 0] ------------------ 1459| 0| const_CAST_bitset(c2), &result) ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1460| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1461| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1462| 0| return result; 1463| | 1464| 2.32k| case CONTAINER_PAIR(ARRAY, ARRAY): ------------------ | | 65| 2.32k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.32k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.32k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1464:9): [True: 2.32k, False: 4.03k] ------------------ 1465| 2.32k| *result_type = 1466| 2.32k| array_array_container_xor(const_CAST_array(c1), ------------------ | | 55| 2.32k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.32k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1466:17): [True: 8, False: 2.31k] ------------------ 1467| 2.32k| const_CAST_array(c2), &result) ------------------ | | 55| 2.32k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.32k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1468| 2.32k| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 8|#define BITSET_CONTAINER_TYPE 1 ------------------ 1469| 2.32k| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 4.63k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1470| 2.32k| return result; 1471| | 1472| 680| case CONTAINER_PAIR(RUN, RUN): ------------------ | | 65| 680| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 680|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 680|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1472:9): [True: 680, False: 5.68k] ------------------ 1473| 680| *result_type = (uint8_t)run_run_container_xor( 1474| 680| const_CAST_run(c1), const_CAST_run(c2), &result); ------------------ | | 78| 680|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 680|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c1), const_CAST_run(c2), &result); ------------------ | | 78| 680|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 680|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1475| 680| return result; 1476| | 1477| 0| case CONTAINER_PAIR(BITSET, ARRAY): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1477:9): [True: 0, False: 6.36k] ------------------ 1478| 0| *result_type = 1479| 0| array_bitset_container_xor(const_CAST_array(c2), ------------------ | | 55| 0|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1479:17): [True: 0, False: 0] ------------------ 1480| 0| const_CAST_bitset(c1), &result) ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1481| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1482| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1483| 0| return result; 1484| | 1485| 575| case CONTAINER_PAIR(ARRAY, BITSET): ------------------ | | 65| 575| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 575|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 575|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1485:9): [True: 575, False: 5.78k] ------------------ 1486| 575| *result_type = 1487| 575| array_bitset_container_xor(const_CAST_array(c1), ------------------ | | 55| 575|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 575|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1487:17): [True: 520, False: 55] ------------------ 1488| 575| const_CAST_bitset(c2), &result) ------------------ | | 53| 575|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 575|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1489| 575| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 520|#define BITSET_CONTAINER_TYPE 1 ------------------ 1490| 575| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 630|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1491| 575| return result; 1492| | 1493| 0| case CONTAINER_PAIR(BITSET, RUN): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1493:9): [True: 0, False: 6.36k] ------------------ 1494| 0| *result_type = 1495| 0| run_bitset_container_xor(const_CAST_run(c2), ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1495:17): [True: 0, False: 0] ------------------ 1496| 0| const_CAST_bitset(c1), &result) ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1497| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1498| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1499| 0| return result; 1500| | 1501| 691| case CONTAINER_PAIR(RUN, BITSET): ------------------ | | 65| 691| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 691|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 691|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1501:9): [True: 691, False: 5.66k] ------------------ 1502| 691| *result_type = 1503| 691| run_bitset_container_xor(const_CAST_run(c1), ------------------ | | 78| 691|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1503:17): [True: 645, False: 46] ------------------ 1504| 691| const_CAST_bitset(c2), &result) ------------------ | | 53| 691|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1505| 691| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 645|#define BITSET_CONTAINER_TYPE 1 ------------------ 1506| 691| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 737|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1507| 691| return result; 1508| | 1509| 814| case CONTAINER_PAIR(ARRAY, RUN): ------------------ | | 65| 814| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 814|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 814|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1509:9): [True: 814, False: 5.54k] ------------------ 1510| 814| *result_type = (uint8_t)array_run_container_xor( 1511| 814| const_CAST_array(c1), const_CAST_run(c2), &result); ------------------ | | 55| 814|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 814|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c1), const_CAST_run(c2), &result); ------------------ | | 78| 814|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 814|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1512| 814| return result; 1513| | 1514| 1.27k| case CONTAINER_PAIR(RUN, ARRAY): ------------------ | | 65| 1.27k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 1.27k|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 1.27k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1514:9): [True: 1.27k, False: 5.08k] ------------------ 1515| 1.27k| *result_type = (uint8_t)array_run_container_xor( 1516| 1.27k| const_CAST_array(c2), const_CAST_run(c1), &result); ------------------ | | 55| 1.27k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c2), const_CAST_run(c1), &result); ------------------ | | 78| 1.27k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1517| 1.27k| return result; 1518| | 1519| 0| default: ------------------ | Branch (1519:9): [True: 0, False: 6.36k] ------------------ 1520| 0| assert(false); 1521| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 1522| 0| return NULL; // unreached 1523| 6.36k| } 1524| 6.36k|} roaring.c:container_ixor: 1667| 67.7k| uint8_t *result_type) { 1668| 67.7k| c1 = get_writable_copy_if_shared(c1, &type1); 1669| 67.7k| c2 = container_unwrap_shared(c2, &type2); 1670| 67.7k| container_t *result = NULL; 1671| 67.7k| switch (PAIR_CONTAINER_TYPES(type1, type2)) { ------------------ | | 62| 67.7k|#define PAIR_CONTAINER_TYPES(type1, type2) (4 * (type1) + (type2)) ------------------ 1672| 1.57k| case CONTAINER_PAIR(BITSET, BITSET): ------------------ | | 65| 1.57k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 1.57k|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 1.57k|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1672:9): [True: 1.57k, False: 66.1k] ------------------ 1673| 1.57k| *result_type = bitset_bitset_container_ixor( ------------------ | Branch (1673:28): [True: 0, False: 1.57k] ------------------ 1674| 1.57k| CAST_bitset(c1), const_CAST_bitset(c2), &result) ------------------ | | 52| 1.57k|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.57k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_bitset(c1), const_CAST_bitset(c2), &result) ------------------ | | 53| 1.57k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 1.57k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1675| 1.57k| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1676| 1.57k| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 3.15k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1677| 1.57k| return result; 1678| | 1679| 5.11k| case CONTAINER_PAIR(ARRAY, ARRAY): ------------------ | | 65| 5.11k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 5.11k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 5.11k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1679:9): [True: 5.11k, False: 62.6k] ------------------ 1680| 5.11k| *result_type = array_array_container_ixor( ------------------ | Branch (1680:28): [True: 0, False: 5.11k] ------------------ 1681| 5.11k| CAST_array(c1), const_CAST_array(c2), &result) ------------------ | | 54| 5.11k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 5.11k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_array(c1), const_CAST_array(c2), &result) ------------------ | | 55| 5.11k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 5.11k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1682| 5.11k| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1683| 5.11k| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 10.2k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1684| 5.11k| return result; 1685| | 1686| 60.5k| case CONTAINER_PAIR(RUN, RUN): ------------------ | | 65| 60.5k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 60.5k|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 60.5k|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1686:9): [True: 60.5k, False: 7.20k] ------------------ 1687| 60.5k| *result_type = (uint8_t)run_run_container_ixor( 1688| 60.5k| CAST_run(c1), const_CAST_run(c2), &result); ------------------ | | 77| 60.5k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 60.5k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_run(c1), const_CAST_run(c2), &result); ------------------ | | 78| 60.5k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 60.5k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1689| 60.5k| return result; 1690| | 1691| 0| case CONTAINER_PAIR(BITSET, ARRAY): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1691:9): [True: 0, False: 67.7k] ------------------ 1692| 0| *result_type = bitset_array_container_ixor( ------------------ | Branch (1692:28): [True: 0, False: 0] ------------------ 1693| 0| CAST_bitset(c1), const_CAST_array(c2), &result) ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_bitset(c1), const_CAST_array(c2), &result) ------------------ | | 55| 0|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1694| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1695| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1696| 0| return result; 1697| | 1698| 98| case CONTAINER_PAIR(ARRAY, BITSET): ------------------ | | 65| 98| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 98|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 98|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1698:9): [True: 98, False: 67.6k] ------------------ 1699| 98| *result_type = array_bitset_container_ixor( ------------------ | Branch (1699:28): [True: 0, False: 98] ------------------ 1700| 98| CAST_array(c1), const_CAST_bitset(c2), &result) ------------------ | | 54| 98|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 98|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_array(c1), const_CAST_bitset(c2), &result) ------------------ | | 53| 98|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 98|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1701| 98| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1702| 98| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 196|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1703| 98| return result; 1704| | 1705| 73| case CONTAINER_PAIR(BITSET, RUN): ------------------ | | 65| 73| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 73|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 73|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1705:9): [True: 73, False: 67.6k] ------------------ 1706| 73| *result_type = bitset_run_container_ixor( ------------------ | Branch (1706:28): [True: 0, False: 73] ------------------ 1707| 73| CAST_bitset(c1), const_CAST_run(c2), &result) ------------------ | | 52| 73|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 73|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_bitset(c1), const_CAST_run(c2), &result) ------------------ | | 78| 73|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 73|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1708| 73| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1709| 73| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 146|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1710| | 1711| 73| return result; 1712| | 1713| 18| case CONTAINER_PAIR(RUN, BITSET): ------------------ | | 65| 18| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 18|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 18|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1713:9): [True: 18, False: 67.7k] ------------------ 1714| 18| *result_type = run_bitset_container_ixor( ------------------ | Branch (1714:28): [True: 0, False: 18] ------------------ 1715| 18| CAST_run(c1), const_CAST_bitset(c2), &result) ------------------ | | 77| 18|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 18|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_run(c1), const_CAST_bitset(c2), &result) ------------------ | | 53| 18|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 18|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1716| 18| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1717| 18| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 36|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1718| 18| return result; 1719| | 1720| 269| case CONTAINER_PAIR(ARRAY, RUN): ------------------ | | 65| 269| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 269|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 269|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1720:9): [True: 269, False: 67.4k] ------------------ 1721| 269| *result_type = (uint8_t)array_run_container_ixor( 1722| 269| CAST_array(c1), const_CAST_run(c2), &result); ------------------ | | 54| 269|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 269|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_array(c1), const_CAST_run(c2), &result); ------------------ | | 78| 269|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 269|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1723| 269| return result; 1724| | 1725| 53| case CONTAINER_PAIR(RUN, ARRAY): ------------------ | | 65| 53| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 53|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 53|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1725:9): [True: 53, False: 67.6k] ------------------ 1726| 53| *result_type = (uint8_t)run_array_container_ixor( 1727| 53| CAST_run(c1), const_CAST_array(c2), &result); ------------------ | | 77| 53|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 53|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_run(c1), const_CAST_array(c2), &result); ------------------ | | 55| 53|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 53|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1728| 53| return result; 1729| | 1730| 0| default: ------------------ | Branch (1730:9): [True: 0, False: 67.7k] ------------------ 1731| 0| assert(false); 1732| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 1733| 0| return NULL; 1734| 67.7k| } 1735| 67.7k|} roaring.c:container_andnot: 1787| 6.36k| uint8_t *result_type) { 1788| 6.36k| c1 = container_unwrap_shared(c1, &type1); 1789| 6.36k| c2 = container_unwrap_shared(c2, &type2); 1790| 6.36k| container_t *result = NULL; 1791| 6.36k| switch (PAIR_CONTAINER_TYPES(type1, type2)) { ------------------ | | 62| 6.36k|#define PAIR_CONTAINER_TYPES(type1, type2) (4 * (type1) + (type2)) ------------------ 1792| 0| case CONTAINER_PAIR(BITSET, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1792:9): [True: 0, False: 6.36k] ------------------ 1793| 0| *result_type = 1794| 0| bitset_bitset_container_andnot(const_CAST_bitset(c1), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1794:17): [True: 0, False: 0] ------------------ 1795| 0| const_CAST_bitset(c2), &result) ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1796| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1797| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1798| 0| return result; 1799| | 1800| 2.32k| case CONTAINER_PAIR(ARRAY, ARRAY): ------------------ | | 65| 2.32k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.32k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.32k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1800:9): [True: 2.32k, False: 4.03k] ------------------ 1801| 2.32k| result = array_container_create(); 1802| 2.32k| array_array_container_andnot( 1803| 2.32k| const_CAST_array(c1), const_CAST_array(c2), CAST_array(result)); ------------------ | | 55| 2.32k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.32k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c1), const_CAST_array(c2), CAST_array(result)); ------------------ | | 55| 2.32k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.32k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c1), const_CAST_array(c2), CAST_array(result)); ------------------ | | 54| 2.32k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 2.32k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1804| 2.32k| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 2.32k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1805| 2.32k| return result; 1806| | 1807| 680| case CONTAINER_PAIR(RUN, RUN): ------------------ | | 65| 680| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 680|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 680|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1807:9): [True: 680, False: 5.68k] ------------------ 1808| 680| if (run_container_is_full(const_CAST_run(c2))) { ------------------ | | 78| 680|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 680|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1808:17): [True: 16, False: 664] ------------------ 1809| 16| result = array_container_create(); 1810| 16| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 16|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1811| 16| return result; 1812| 16| } 1813| 664| *result_type = (uint8_t)run_run_container_andnot( 1814| 664| const_CAST_run(c1), const_CAST_run(c2), &result); ------------------ | | 78| 664|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 664|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c1), const_CAST_run(c2), &result); ------------------ | | 78| 664|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 664|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1815| 664| return result; 1816| | 1817| 0| case CONTAINER_PAIR(BITSET, ARRAY): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1817:9): [True: 0, False: 6.36k] ------------------ 1818| 0| *result_type = 1819| 0| bitset_array_container_andnot(const_CAST_bitset(c1), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1819:17): [True: 0, False: 0] ------------------ 1820| 0| const_CAST_array(c2), &result) ------------------ | | 55| 0|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1821| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1822| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1823| 0| return result; 1824| | 1825| 575| case CONTAINER_PAIR(ARRAY, BITSET): ------------------ | | 65| 575| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 575|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 575|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1825:9): [True: 575, False: 5.78k] ------------------ 1826| 575| result = array_container_create(); 1827| 575| array_bitset_container_andnot(const_CAST_array(c1), ------------------ | | 55| 575|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 575|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1828| 575| const_CAST_bitset(c2), ------------------ | | 53| 575|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 575|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1829| 575| CAST_array(result)); ------------------ | | 54| 575|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 575|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1830| 575| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 575|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1831| 575| return result; 1832| | 1833| 0| case CONTAINER_PAIR(BITSET, RUN): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1833:9): [True: 0, False: 6.36k] ------------------ 1834| 0| if (run_container_is_full(const_CAST_run(c2))) { ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1834:17): [True: 0, False: 0] ------------------ 1835| 0| result = array_container_create(); 1836| 0| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1837| 0| return result; 1838| 0| } 1839| 0| *result_type = 1840| 0| bitset_run_container_andnot(const_CAST_bitset(c1), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1840:17): [True: 0, False: 0] ------------------ 1841| 0| const_CAST_run(c2), &result) ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1842| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1843| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1844| 0| return result; 1845| | 1846| 691| case CONTAINER_PAIR(RUN, BITSET): ------------------ | | 65| 691| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 691|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 691|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1846:9): [True: 691, False: 5.66k] ------------------ 1847| 691| *result_type = 1848| 691| run_bitset_container_andnot(const_CAST_run(c1), ------------------ | | 78| 691|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1848:17): [True: 0, False: 691] ------------------ 1849| 691| const_CAST_bitset(c2), &result) ------------------ | | 53| 691|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1850| 691| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1851| 691| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 1.38k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1852| 691| return result; 1853| | 1854| 814| case CONTAINER_PAIR(ARRAY, RUN): ------------------ | | 65| 814| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 814|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 814|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1854:9): [True: 814, False: 5.54k] ------------------ 1855| 814| if (run_container_is_full(const_CAST_run(c2))) { ------------------ | | 78| 814|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 814|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (1855:17): [True: 34, False: 780] ------------------ 1856| 34| result = array_container_create(); 1857| 34| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 34|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1858| 34| return result; 1859| 34| } 1860| 780| result = array_container_create(); 1861| 780| array_run_container_andnot(const_CAST_array(c1), const_CAST_run(c2), ------------------ | | 55| 780|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 780|#define CAST(type, value) ((type)value) | | ------------------ ------------------ array_run_container_andnot(const_CAST_array(c1), const_CAST_run(c2), ------------------ | | 78| 780|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 780|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1862| 780| CAST_array(result)); ------------------ | | 54| 780|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 780|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1863| 780| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 780|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1864| 780| return result; 1865| | 1866| 1.27k| case CONTAINER_PAIR(RUN, ARRAY): ------------------ | | 65| 1.27k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 1.27k|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 1.27k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1866:9): [True: 1.27k, False: 5.08k] ------------------ 1867| 1.27k| *result_type = (uint8_t)run_array_container_andnot( 1868| 1.27k| const_CAST_run(c1), const_CAST_array(c2), &result); ------------------ | | 78| 1.27k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c1), const_CAST_array(c2), &result); ------------------ | | 55| 1.27k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1869| 1.27k| return result; 1870| | 1871| 0| default: ------------------ | Branch (1871:9): [True: 0, False: 6.36k] ------------------ 1872| 0| assert(false); 1873| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 1874| 0| return NULL; // unreached 1875| 6.36k| } 1876| 6.36k|} roaring.c:container_iandnot: 1891| 6.05k| uint8_t *result_type) { 1892| 6.05k| c1 = get_writable_copy_if_shared(c1, &type1); 1893| 6.05k| c2 = container_unwrap_shared(c2, &type2); 1894| 6.05k| container_t *result = NULL; 1895| 6.05k| switch (PAIR_CONTAINER_TYPES(type1, type2)) { ------------------ | | 62| 6.05k|#define PAIR_CONTAINER_TYPES(type1, type2) (4 * (type1) + (type2)) ------------------ 1896| 0| case CONTAINER_PAIR(BITSET, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1896:9): [True: 0, False: 6.05k] ------------------ 1897| 0| *result_type = bitset_bitset_container_iandnot( ------------------ | Branch (1897:28): [True: 0, False: 0] ------------------ 1898| 0| CAST_bitset(c1), const_CAST_bitset(c2), &result) ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_bitset(c1), const_CAST_bitset(c2), &result) ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1899| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1900| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1901| 0| return result; 1902| | 1903| 3.54k| case CONTAINER_PAIR(ARRAY, ARRAY): ------------------ | | 65| 3.54k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 3.54k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 3.54k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1903:9): [True: 3.54k, False: 2.51k] ------------------ 1904| 3.54k| array_array_container_iandnot(CAST_array(c1), const_CAST_array(c2)); ------------------ | | 54| 3.54k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 3.54k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ array_array_container_iandnot(CAST_array(c1), const_CAST_array(c2)); ------------------ | | 55| 3.54k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 3.54k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1905| 3.54k| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 3.54k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1906| 3.54k| return c1; 1907| | 1908| 432| case CONTAINER_PAIR(RUN, RUN): ------------------ | | 65| 432| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 432|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 432|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1908:9): [True: 432, False: 5.62k] ------------------ 1909| 432| *result_type = (uint8_t)run_run_container_iandnot( 1910| 432| CAST_run(c1), const_CAST_run(c2), &result); ------------------ | | 77| 432|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 432|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_run(c1), const_CAST_run(c2), &result); ------------------ | | 78| 432|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 432|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1911| 432| return result; 1912| | 1913| 1.23k| case CONTAINER_PAIR(BITSET, ARRAY): ------------------ | | 65| 1.23k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 1.23k|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 1.23k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1913:9): [True: 1.23k, False: 4.81k] ------------------ 1914| 1.23k| *result_type = bitset_array_container_iandnot( ------------------ | Branch (1914:28): [True: 1.05k, False: 183] ------------------ 1915| 1.23k| CAST_bitset(c1), const_CAST_array(c2), &result) ------------------ | | 52| 1.23k|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.23k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_bitset(c1), const_CAST_array(c2), &result) ------------------ | | 55| 1.23k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 1.23k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1916| 1.23k| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 1.05k|#define BITSET_CONTAINER_TYPE 1 ------------------ 1917| 1.23k| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 1.42k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1918| 1.23k| return result; 1919| | 1920| 0| case CONTAINER_PAIR(ARRAY, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1920:9): [True: 0, False: 6.05k] ------------------ 1921| 0| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1922| 0| array_bitset_container_iandnot(CAST_array(c1), ------------------ | | 54| 0|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1923| 0| const_CAST_bitset(c2)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1924| 0| return c1; 1925| | 1926| 0| case CONTAINER_PAIR(BITSET, RUN): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1926:9): [True: 0, False: 6.05k] ------------------ 1927| 0| *result_type = bitset_run_container_iandnot( ------------------ | Branch (1927:28): [True: 0, False: 0] ------------------ 1928| 0| CAST_bitset(c1), const_CAST_run(c2), &result) ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_bitset(c1), const_CAST_run(c2), &result) ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1929| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1930| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1931| 0| return result; 1932| | 1933| 0| case CONTAINER_PAIR(RUN, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (1933:9): [True: 0, False: 6.05k] ------------------ 1934| 0| *result_type = run_bitset_container_iandnot( ------------------ | Branch (1934:28): [True: 0, False: 0] ------------------ 1935| 0| CAST_run(c1), const_CAST_bitset(c2), &result) ------------------ | | 77| 0|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_run(c1), const_CAST_bitset(c2), &result) ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1936| 0| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 1937| 0| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1938| 0| return result; 1939| | 1940| 45| case CONTAINER_PAIR(ARRAY, RUN): ------------------ | | 65| 45| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 45|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 45|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (1940:9): [True: 45, False: 6.01k] ------------------ 1941| 45| *result_type = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 45|#define ARRAY_CONTAINER_TYPE 2 ------------------ 1942| 45| array_run_container_iandnot(CAST_array(c1), const_CAST_run(c2)); ------------------ | | 54| 45|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 45|#define CAST(type, value) ((type)value) | | ------------------ ------------------ array_run_container_iandnot(CAST_array(c1), const_CAST_run(c2)); ------------------ | | 78| 45|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 45|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1943| 45| return c1; 1944| | 1945| 801| case CONTAINER_PAIR(RUN, ARRAY): ------------------ | | 65| 801| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 801|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 801|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (1945:9): [True: 801, False: 5.25k] ------------------ 1946| 801| *result_type = (uint8_t)run_array_container_iandnot( 1947| 801| CAST_run(c1), const_CAST_array(c2), &result); ------------------ | | 77| 801|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 801|#define CAST(type, value) ((type)value) | | ------------------ ------------------ CAST_run(c1), const_CAST_array(c2), &result); ------------------ | | 55| 801|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 801|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1948| 801| return result; 1949| | 1950| 0| default: ------------------ | Branch (1950:9): [True: 0, False: 6.05k] ------------------ 1951| 0| assert(false); 1952| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 1953| 0| return NULL; 1954| 6.05k| } 1955| 6.05k|} roaring.c:container_rank: 2209| 6.48k| uint16_t x) { 2210| 6.48k| c = container_unwrap_shared(c, &type); 2211| 6.48k| switch (type) { 2212| 0| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (2212:9): [True: 0, False: 6.48k] ------------------ 2213| 0| return bitset_container_rank(const_CAST_bitset(c), x); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2214| 3.80k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 3.80k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (2214:9): [True: 3.80k, False: 2.67k] ------------------ 2215| 3.80k| return array_container_rank(const_CAST_array(c), x); ------------------ | | 55| 3.80k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 3.80k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2216| 2.67k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 2.67k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (2216:9): [True: 2.67k, False: 3.80k] ------------------ 2217| 2.67k| return run_container_rank(const_CAST_run(c), x); ------------------ | | 78| 2.67k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 2.67k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2218| 0| default: ------------------ | Branch (2218:9): [True: 0, False: 6.48k] ------------------ 2219| 0| assert(false); 2220| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2221| 6.48k| } 2222| 6.48k| assert(false); 2223| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2224| 0| return false; 2225| 6.48k|} roaring.c:container_shrink_to_fit: 279| 6.48k|static inline int container_shrink_to_fit(container_t *c, uint8_t type) { 280| 6.48k| c = container_mutable_unwrap_shared(c, &type); 281| 6.48k| switch (type) { ------------------ | Branch (281:13): [True: 6.48k, False: 0] ------------------ 282| 0| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (282:9): [True: 0, False: 6.48k] ------------------ 283| 0| return 0; // no shrinking possible 284| 3.80k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 3.80k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (284:9): [True: 3.80k, False: 2.67k] ------------------ 285| 3.80k| return array_container_shrink_to_fit(CAST_array(c)); ------------------ | | 54| 3.80k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 3.80k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 286| 2.67k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 2.67k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (286:9): [True: 2.67k, False: 3.80k] ------------------ 287| 2.67k| return run_container_shrink_to_fit(CAST_run(c)); ------------------ | | 77| 2.67k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 2.67k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 288| 6.48k| } 289| 6.48k| assert(false); 290| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 291| 0| return 0; // unreached 292| 6.48k|} roaring.c:container_iterate: 1964| 69.3k| void *ptr) { 1965| 69.3k| c = container_unwrap_shared(c, &type); 1966| 69.3k| switch (type) { 1967| 1.69k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 1.69k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (1967:9): [True: 1.69k, False: 67.6k] ------------------ 1968| 1.69k| return bitset_container_iterate(const_CAST_bitset(c), base, ------------------ | | 53| 1.69k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 1.69k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1969| 1.69k| iterator, ptr); 1970| 6.58k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 6.58k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (1970:9): [True: 6.58k, False: 62.8k] ------------------ 1971| 6.58k| return array_container_iterate(const_CAST_array(c), base, iterator, ------------------ | | 55| 6.58k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 6.58k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1972| 6.58k| ptr); 1973| 61.1k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 61.1k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (1973:9): [True: 61.1k, False: 8.27k] ------------------ 1974| 61.1k| return run_container_iterate(const_CAST_run(c), base, iterator, ------------------ | | 78| 61.1k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 61.1k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1975| 61.1k| ptr); 1976| 0| default: ------------------ | Branch (1976:9): [True: 0, False: 69.3k] ------------------ 1977| 0| assert(false); 1978| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 1979| 69.3k| } 1980| 69.3k| assert(false); 1981| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 1982| 0| return false; 1983| 69.3k|} roaring.c:container_maximum: 2171| 12.8k|static inline uint16_t container_maximum(const container_t *c, uint8_t type) { 2172| 12.8k| c = container_unwrap_shared(c, &type); 2173| 12.8k| switch (type) { 2174| 2.58k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 2.58k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (2174:9): [True: 2.58k, False: 10.2k] ------------------ 2175| 2.58k| return bitset_container_maximum(const_CAST_bitset(c)); ------------------ | | 53| 2.58k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 2.58k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2176| 8.88k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 8.88k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (2176:9): [True: 8.88k, False: 3.95k] ------------------ 2177| 8.88k| return array_container_maximum(const_CAST_array(c)); ------------------ | | 55| 8.88k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 8.88k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2178| 1.37k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 1.37k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (2178:9): [True: 1.37k, False: 11.4k] ------------------ 2179| 1.37k| return run_container_maximum(const_CAST_run(c)); ------------------ | | 78| 1.37k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 1.37k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2180| 0| default: ------------------ | Branch (2180:9): [True: 0, False: 12.8k] ------------------ 2181| 0| assert(false); 2182| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2183| 12.8k| } 2184| 12.8k| assert(false); 2185| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2186| 0| return false; 2187| 12.8k|} roaring.c:container_equals: 620| 85.0k| const container_t *c2, uint8_t type2) { 621| 85.0k| c1 = container_unwrap_shared(c1, &type1); 622| 85.0k| c2 = container_unwrap_shared(c2, &type2); 623| 85.0k| switch (PAIR_CONTAINER_TYPES(type1, type2)) { ------------------ | | 62| 85.0k|#define PAIR_CONTAINER_TYPES(type1, type2) (4 * (type1) + (type2)) ------------------ 624| 73.9k| case CONTAINER_PAIR(BITSET, BITSET): ------------------ | | 65| 73.9k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 73.9k|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 73.9k|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (624:9): [True: 73.9k, False: 11.1k] ------------------ 625| 73.9k| return bitset_container_equals(const_CAST_bitset(c1), ------------------ | | 53| 73.9k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 73.9k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 626| 73.9k| const_CAST_bitset(c2)); ------------------ | | 53| 73.9k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 73.9k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 627| | 628| 0| case CONTAINER_PAIR(BITSET, RUN): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (628:9): [True: 0, False: 85.0k] ------------------ 629| 0| return run_container_equals_bitset(const_CAST_run(c2), ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 630| 0| const_CAST_bitset(c1)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 631| | 632| 0| case CONTAINER_PAIR(RUN, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (632:9): [True: 0, False: 85.0k] ------------------ 633| 0| return run_container_equals_bitset(const_CAST_run(c1), ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 634| 0| const_CAST_bitset(c2)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 635| | 636| 0| case CONTAINER_PAIR(BITSET, ARRAY): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (636:9): [True: 0, False: 85.0k] ------------------ 637| | // java would always return false? 638| 0| return array_container_equal_bitset(const_CAST_array(c2), ------------------ | | 55| 0|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 639| 0| const_CAST_bitset(c1)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 640| | 641| 469| case CONTAINER_PAIR(ARRAY, BITSET): ------------------ | | 65| 469| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 469|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 469|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (641:9): [True: 469, False: 84.6k] ------------------ 642| | // java would always return false? 643| 469| return array_container_equal_bitset(const_CAST_array(c1), ------------------ | | 55| 469|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 469|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 644| 469| const_CAST_bitset(c2)); ------------------ | | 53| 469|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 469|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 645| | 646| 609| case CONTAINER_PAIR(ARRAY, RUN): ------------------ | | 65| 609| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 609|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 609|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (646:9): [True: 609, False: 84.4k] ------------------ 647| 609| return run_container_equals_array(const_CAST_run(c2), ------------------ | | 78| 609|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 609|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 648| 609| const_CAST_array(c1)); ------------------ | | 55| 609|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 609|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 649| | 650| 277| case CONTAINER_PAIR(RUN, ARRAY): ------------------ | | 65| 277| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 277|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 277|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (650:9): [True: 277, False: 84.8k] ------------------ 651| 277| return run_container_equals_array(const_CAST_run(c1), ------------------ | | 78| 277|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 277|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 652| 277| const_CAST_array(c2)); ------------------ | | 55| 277|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 277|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 653| | 654| 9.61k| case CONTAINER_PAIR(ARRAY, ARRAY): ------------------ | | 65| 9.61k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 9.61k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 9.61k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (654:9): [True: 9.61k, False: 75.4k] ------------------ 655| 9.61k| return array_container_equals(const_CAST_array(c1), ------------------ | | 55| 9.61k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 9.61k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 656| 9.61k| const_CAST_array(c2)); ------------------ | | 55| 9.61k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 9.61k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 657| | 658| 133| case CONTAINER_PAIR(RUN, RUN): ------------------ | | 65| 133| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 133|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 133|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (658:9): [True: 133, False: 84.9k] ------------------ 659| 133| return run_container_equals(const_CAST_run(c1), const_CAST_run(c2)); ------------------ | | 78| 133|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 133|#define CAST(type, value) ((type)value) | | ------------------ ------------------ return run_container_equals(const_CAST_run(c1), const_CAST_run(c2)); ------------------ | | 78| 133|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 133|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 660| | 661| 0| default: ------------------ | Branch (661:9): [True: 0, False: 85.0k] ------------------ 662| 0| assert(false); 663| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 664| 0| return false; 665| 85.0k| } 666| 85.0k|} roaring.c:container_is_subset: 673| 11.1k| const container_t *c2, uint8_t type2) { 674| 11.1k| c1 = container_unwrap_shared(c1, &type1); 675| 11.1k| c2 = container_unwrap_shared(c2, &type2); 676| 11.1k| switch (PAIR_CONTAINER_TYPES(type1, type2)) { ------------------ | | 62| 11.1k|#define PAIR_CONTAINER_TYPES(type1, type2) (4 * (type1) + (type2)) ------------------ 677| 0| case CONTAINER_PAIR(BITSET, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (677:9): [True: 0, False: 11.1k] ------------------ 678| 0| return bitset_container_is_subset(const_CAST_bitset(c1), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 679| 0| const_CAST_bitset(c2)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 680| | 681| 0| case CONTAINER_PAIR(BITSET, RUN): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (681:9): [True: 0, False: 11.1k] ------------------ 682| 0| return bitset_container_is_subset_run(const_CAST_bitset(c1), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 683| 0| const_CAST_run(c2)); ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 684| | 685| 0| case CONTAINER_PAIR(RUN, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (685:9): [True: 0, False: 11.1k] ------------------ 686| 0| return run_container_is_subset_bitset(const_CAST_run(c1), ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 687| 0| const_CAST_bitset(c2)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 688| | 689| 0| case CONTAINER_PAIR(BITSET, ARRAY): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (689:9): [True: 0, False: 11.1k] ------------------ 690| 0| return false; // by construction, size(c1) > size(c2) 691| | 692| 2.45k| case CONTAINER_PAIR(ARRAY, BITSET): ------------------ | | 65| 2.45k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.45k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 2.45k|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (692:9): [True: 2.45k, False: 8.68k] ------------------ 693| 2.45k| return array_container_is_subset_bitset(const_CAST_array(c1), ------------------ | | 55| 2.45k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.45k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 694| 2.45k| const_CAST_bitset(c2)); ------------------ | | 53| 2.45k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 2.45k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 695| | 696| 2.25k| case CONTAINER_PAIR(ARRAY, RUN): ------------------ | | 65| 2.25k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.25k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 2.25k|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (696:9): [True: 2.25k, False: 8.87k] ------------------ 697| 2.25k| return array_container_is_subset_run(const_CAST_array(c1), ------------------ | | 55| 2.25k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.25k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 698| 2.25k| const_CAST_run(c2)); ------------------ | | 78| 2.25k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 2.25k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 699| | 700| 423| case CONTAINER_PAIR(RUN, ARRAY): ------------------ | | 65| 423| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 423|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 423|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (700:9): [True: 423, False: 10.7k] ------------------ 701| 423| return run_container_is_subset_array(const_CAST_run(c1), ------------------ | | 78| 423|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 423|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 702| 423| const_CAST_array(c2)); ------------------ | | 55| 423|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 423|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 703| | 704| 5.75k| case CONTAINER_PAIR(ARRAY, ARRAY): ------------------ | | 65| 5.75k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 5.75k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 5.75k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (704:9): [True: 5.75k, False: 5.37k] ------------------ 705| 5.75k| return array_container_is_subset(const_CAST_array(c1), ------------------ | | 55| 5.75k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 5.75k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 706| 5.75k| const_CAST_array(c2)); ------------------ | | 55| 5.75k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 5.75k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 707| | 708| 244| case CONTAINER_PAIR(RUN, RUN): ------------------ | | 65| 244| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 244|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 244|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (708:9): [True: 244, False: 10.8k] ------------------ 709| 244| return run_container_is_subset(const_CAST_run(c1), ------------------ | | 78| 244|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 244|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 710| 244| const_CAST_run(c2)); ------------------ | | 78| 244|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 244|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 711| | 712| 0| default: ------------------ | Branch (712:9): [True: 0, False: 11.1k] ------------------ 713| 0| assert(false); 714| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 715| 0| return false; 716| 11.1k| } 717| 11.1k|} roaring.c:container_inot_range: 2110| 6.62k| uint8_t *result_type) { 2111| 6.62k| c = get_writable_copy_if_shared(c, &type); 2112| 6.62k| container_t *result = NULL; 2113| 6.62k| switch (type) { 2114| 1.35k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 1.35k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (2114:9): [True: 1.35k, False: 5.27k] ------------------ 2115| 1.35k| *result_type = bitset_container_negation_range_inplace( ------------------ | Branch (2115:28): [True: 1.33k, False: 22] ------------------ 2116| 1.35k| CAST_bitset(c), range_start, range_end, &result) ------------------ | | 52| 1.35k|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.35k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2117| 1.35k| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 1.33k|#define BITSET_CONTAINER_TYPE 1 ------------------ 2118| 1.35k| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 1.37k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 2119| 1.35k| return result; 2120| 3.57k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 3.57k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (2120:9): [True: 3.57k, False: 3.05k] ------------------ 2121| 3.57k| *result_type = array_container_negation_range_inplace( ------------------ | Branch (2121:28): [True: 130, False: 3.44k] ------------------ 2122| 3.57k| CAST_array(c), range_start, range_end, &result) ------------------ | | 54| 3.57k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 3.57k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2123| 3.57k| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 130|#define BITSET_CONTAINER_TYPE 1 ------------------ 2124| 3.57k| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 7.02k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 2125| 3.57k| return result; 2126| 1.69k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 1.69k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (2126:9): [True: 1.69k, False: 4.93k] ------------------ 2127| 1.69k| *result_type = (uint8_t)run_container_negation_range_inplace( 2128| 1.69k| CAST_run(c), range_start, range_end, &result); ------------------ | | 77| 1.69k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.69k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2129| 1.69k| return result; 2130| | 2131| 0| default: ------------------ | Branch (2131:9): [True: 0, False: 6.62k] ------------------ 2132| 0| assert(false); 2133| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2134| 6.62k| } 2135| 6.62k| assert(false); 2136| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2137| 0| return NULL; 2138| 6.62k|} roaring.c:container_inot: 2076| 6.85k| uint8_t *result_type) { 2077| 6.85k| c = get_writable_copy_if_shared(c, &type); 2078| 6.85k| container_t *result = NULL; 2079| 6.85k| switch (type) { 2080| 87| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 87|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (2080:9): [True: 87, False: 6.76k] ------------------ 2081| 87| *result_type = 2082| 87| bitset_container_negation_inplace(CAST_bitset(c), &result) ------------------ | | 52| 87|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 87|#define CAST(type, value) ((type)value) | | ------------------ ------------------ | Branch (2082:17): [True: 34, False: 53] ------------------ 2083| 87| ? BITSET_CONTAINER_TYPE ------------------ | | 48| 34|#define BITSET_CONTAINER_TYPE 1 ------------------ 2084| 87| : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 140|#define ARRAY_CONTAINER_TYPE 2 ------------------ 2085| 87| return result; 2086| 185| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 185|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (2086:9): [True: 185, False: 6.67k] ------------------ 2087| | // will never be inplace 2088| 185| result = bitset_container_create(); 2089| 185| *result_type = BITSET_CONTAINER_TYPE; ------------------ | | 48| 185|#define BITSET_CONTAINER_TYPE 1 ------------------ 2090| 185| array_container_negation(CAST_array(c), CAST_bitset(result)); ------------------ | | 54| 185|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 185|#define CAST(type, value) ((type)value) | | ------------------ ------------------ array_container_negation(CAST_array(c), CAST_bitset(result)); ------------------ | | 52| 185|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 185|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2091| 185| array_container_free(CAST_array(c)); ------------------ | | 54| 185|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 185|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2092| 185| return result; 2093| 6.58k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 6.58k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (2093:9): [True: 6.58k, False: 272] ------------------ 2094| 6.58k| *result_type = 2095| 6.58k| (uint8_t)run_container_negation_inplace(CAST_run(c), &result); ------------------ | | 77| 6.58k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 6.58k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2096| 6.58k| return result; 2097| | 2098| 0| default: ------------------ | Branch (2098:9): [True: 0, False: 6.85k] ------------------ 2099| 0| assert(false); 2100| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2101| 6.85k| } 2102| 6.85k| assert(false); 2103| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2104| 0| return NULL; 2105| 6.85k|} roaring.c:container_unwrap_shared: 106| 1.16M| const container_t *candidate_shared_container, uint8_t *type) { 107| 1.16M| if (*type == SHARED_CONTAINER_TYPE) { ------------------ | | 51| 1.16M|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (107:9): [True: 0, False: 1.16M] ------------------ 108| 0| *type = const_CAST_shared(candidate_shared_container)->typecode; ------------------ | | 80| 0|#define const_CAST_shared(c) CAST(const shared_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 109| 0| assert(*type != SHARED_CONTAINER_TYPE); 110| 0| return const_CAST_shared(candidate_shared_container)->container; ------------------ | | 80| 0|#define const_CAST_shared(c) CAST(const shared_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 111| 1.16M| } else { 112| 1.16M| return candidate_shared_container; 113| 1.16M| } 114| 1.16M|} roaring.c:container_mutable_unwrap_shared: 118| 6.48k| uint8_t *type) { 119| 6.48k| if (*type == SHARED_CONTAINER_TYPE) { // the passed in container is shared ------------------ | | 51| 6.48k|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (119:9): [True: 0, False: 6.48k] ------------------ 120| 0| *type = CAST_shared(c)->typecode; ------------------ | | 79| 0|#define CAST_shared(c) CAST(shared_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 121| 0| assert(*type != SHARED_CONTAINER_TYPE); 122| 0| return CAST_shared(c)->container; // return the enclosed container ------------------ | | 79| 0|#define CAST_shared(c) CAST(shared_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 123| 6.48k| } else { 124| 6.48k| return c; // wasn't shared, so return as-is 125| 6.48k| } 126| 6.48k|} roaring.c:container_minimum: 2189| 6.43k|static inline uint16_t container_minimum(const container_t *c, uint8_t type) { 2190| 6.43k| c = container_unwrap_shared(c, &type); 2191| 6.43k| switch (type) { 2192| 1.27k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 1.27k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (2192:9): [True: 1.27k, False: 5.15k] ------------------ 2193| 1.27k| return bitset_container_minimum(const_CAST_bitset(c)); ------------------ | | 53| 1.27k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2194| 3.63k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 3.63k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (2194:9): [True: 3.63k, False: 2.79k] ------------------ 2195| 3.63k| return array_container_minimum(const_CAST_array(c)); ------------------ | | 55| 3.63k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 3.63k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2196| 1.51k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 1.51k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (2196:9): [True: 1.51k, False: 4.91k] ------------------ 2197| 1.51k| return run_container_minimum(const_CAST_run(c)); ------------------ | | 78| 1.51k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 1.51k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2198| 0| default: ------------------ | Branch (2198:9): [True: 0, False: 6.43k] ------------------ 2199| 0| assert(false); 2200| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2201| 6.43k| } 2202| 6.43k| assert(false); 2203| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2204| 0| return false; 2205| 6.43k|} roaring.c:container_select: 2150| 6.48k| uint32_t *element) { 2151| 6.48k| c = container_unwrap_shared(c, &type); 2152| 6.48k| switch (type) { 2153| 0| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (2153:9): [True: 0, False: 6.48k] ------------------ 2154| 0| return bitset_container_select(const_CAST_bitset(c), start_rank, ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2155| 0| rank, element); 2156| 3.80k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 3.80k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (2156:9): [True: 3.80k, False: 2.67k] ------------------ 2157| 3.80k| return array_container_select(const_CAST_array(c), start_rank, rank, ------------------ | | 55| 3.80k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 3.80k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2158| 3.80k| element); 2159| 2.67k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 2.67k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (2159:9): [True: 2.67k, False: 3.80k] ------------------ 2160| 2.67k| return run_container_select(const_CAST_run(c), start_rank, rank, ------------------ | | 78| 2.67k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 2.67k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2161| 2.67k| element); 2162| 0| default: ------------------ | Branch (2162:9): [True: 0, False: 6.48k] ------------------ 2163| 0| assert(false); 2164| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2165| 6.48k| } 2166| 6.48k| assert(false); 2167| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2168| 0| return false; 2169| 6.48k|} roaring.c:container_intersect: 865| 6.36k| const container_t *c2, uint8_t type2) { 866| 6.36k| c1 = container_unwrap_shared(c1, &type1); 867| 6.36k| c2 = container_unwrap_shared(c2, &type2); 868| 6.36k| switch (PAIR_CONTAINER_TYPES(type1, type2)) { ------------------ | | 62| 6.36k|#define PAIR_CONTAINER_TYPES(type1, type2) (4 * (type1) + (type2)) ------------------ 869| 0| case CONTAINER_PAIR(BITSET, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (869:9): [True: 0, False: 6.36k] ------------------ 870| 0| return bitset_container_intersect(const_CAST_bitset(c1), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 871| 0| const_CAST_bitset(c2)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 872| | 873| 2.32k| case CONTAINER_PAIR(ARRAY, ARRAY): ------------------ | | 65| 2.32k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.32k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.32k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (873:9): [True: 2.32k, False: 4.03k] ------------------ 874| 2.32k| return array_container_intersect(const_CAST_array(c1), ------------------ | | 55| 2.32k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.32k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 875| 2.32k| const_CAST_array(c2)); ------------------ | | 55| 2.32k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.32k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 876| | 877| 680| case CONTAINER_PAIR(RUN, RUN): ------------------ | | 65| 680| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 680|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 680|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (877:9): [True: 680, False: 5.68k] ------------------ 878| 680| return run_container_intersect(const_CAST_run(c1), ------------------ | | 78| 680|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 680|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 879| 680| const_CAST_run(c2)); ------------------ | | 78| 680|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 680|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 880| | 881| 0| case CONTAINER_PAIR(BITSET, ARRAY): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (881:9): [True: 0, False: 6.36k] ------------------ 882| 0| return array_bitset_container_intersect(const_CAST_array(c2), ------------------ | | 55| 0|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 883| 0| const_CAST_bitset(c1)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 884| | 885| 575| case CONTAINER_PAIR(ARRAY, BITSET): ------------------ | | 65| 575| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 575|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 575|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (885:9): [True: 575, False: 5.78k] ------------------ 886| 575| return array_bitset_container_intersect(const_CAST_array(c1), ------------------ | | 55| 575|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 575|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 887| 575| const_CAST_bitset(c2)); ------------------ | | 53| 575|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 575|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 888| | 889| 0| case CONTAINER_PAIR(BITSET, RUN): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (889:9): [True: 0, False: 6.36k] ------------------ 890| 0| return run_bitset_container_intersect(const_CAST_run(c2), ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 891| 0| const_CAST_bitset(c1)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 892| | 893| 691| case CONTAINER_PAIR(RUN, BITSET): ------------------ | | 65| 691| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 691|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 691|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (893:9): [True: 691, False: 5.66k] ------------------ 894| 691| return run_bitset_container_intersect(const_CAST_run(c1), ------------------ | | 78| 691|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 895| 691| const_CAST_bitset(c2)); ------------------ | | 53| 691|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 691|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 896| | 897| 814| case CONTAINER_PAIR(ARRAY, RUN): ------------------ | | 65| 814| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 814|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 814|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (897:9): [True: 814, False: 5.54k] ------------------ 898| 814| return array_run_container_intersect(const_CAST_array(c1), ------------------ | | 55| 814|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 814|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 899| 814| const_CAST_run(c2)); ------------------ | | 78| 814|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 814|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 900| | 901| 1.27k| case CONTAINER_PAIR(RUN, ARRAY): ------------------ | | 65| 1.27k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 1.27k|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 1.27k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (901:9): [True: 1.27k, False: 5.08k] ------------------ 902| 1.27k| return array_run_container_intersect(const_CAST_array(c2), ------------------ | | 55| 1.27k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 903| 1.27k| const_CAST_run(c1)); ------------------ | | 78| 1.27k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 1.27k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 904| | 905| 0| default: ------------------ | Branch (905:9): [True: 0, False: 6.36k] ------------------ 906| 0| assert(false); 907| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 908| 0| return 0; 909| 6.36k| } 910| 6.36k|} roaring.c:container_and_cardinality: 814| 25.4k| uint8_t type2) { 815| 25.4k| c1 = container_unwrap_shared(c1, &type1); 816| 25.4k| c2 = container_unwrap_shared(c2, &type2); 817| 25.4k| switch (PAIR_CONTAINER_TYPES(type1, type2)) { ------------------ | | 62| 25.4k|#define PAIR_CONTAINER_TYPES(type1, type2) (4 * (type1) + (type2)) ------------------ 818| 0| case CONTAINER_PAIR(BITSET, BITSET): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (818:9): [True: 0, False: 25.4k] ------------------ 819| 0| return bitset_container_and_justcard(const_CAST_bitset(c1), ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 820| 0| const_CAST_bitset(c2)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 821| | 822| 9.29k| case CONTAINER_PAIR(ARRAY, ARRAY): ------------------ | | 65| 9.29k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 9.29k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 9.29k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (822:9): [True: 9.29k, False: 16.1k] ------------------ 823| 9.29k| return array_container_intersection_cardinality( 824| 9.29k| const_CAST_array(c1), const_CAST_array(c2)); ------------------ | | 55| 9.29k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 9.29k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c1), const_CAST_array(c2)); ------------------ | | 55| 9.29k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 9.29k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 825| | 826| 2.72k| case CONTAINER_PAIR(RUN, RUN): ------------------ | | 65| 2.72k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 2.72k|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 2.72k|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (826:9): [True: 2.72k, False: 22.7k] ------------------ 827| 2.72k| return run_container_intersection_cardinality(const_CAST_run(c1), ------------------ | | 78| 2.72k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 2.72k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 828| 2.72k| const_CAST_run(c2)); ------------------ | | 78| 2.72k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 2.72k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 829| | 830| 0| case CONTAINER_PAIR(BITSET, ARRAY): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 0|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (830:9): [True: 0, False: 25.4k] ------------------ 831| 0| return array_bitset_container_intersection_cardinality( 832| 0| const_CAST_array(c2), const_CAST_bitset(c1)); ------------------ | | 55| 0|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c2), const_CAST_bitset(c1)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 833| | 834| 2.30k| case CONTAINER_PAIR(ARRAY, BITSET): ------------------ | | 65| 2.30k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 2.30k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 2.30k|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (834:9): [True: 2.30k, False: 23.1k] ------------------ 835| 2.30k| return array_bitset_container_intersection_cardinality( 836| 2.30k| const_CAST_array(c1), const_CAST_bitset(c2)); ------------------ | | 55| 2.30k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 2.30k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c1), const_CAST_bitset(c2)); ------------------ | | 53| 2.30k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 2.30k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 837| | 838| 0| case CONTAINER_PAIR(BITSET, RUN): ------------------ | | 65| 0| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 0|#define BITSET_CONTAINER_TYPE 1 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 0|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (838:9): [True: 0, False: 25.4k] ------------------ 839| 0| return run_bitset_container_intersection_cardinality( 840| 0| const_CAST_run(c2), const_CAST_bitset(c1)); ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c2), const_CAST_bitset(c1)); ------------------ | | 53| 0|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 841| | 842| 2.76k| case CONTAINER_PAIR(RUN, BITSET): ------------------ | | 65| 2.76k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 2.76k|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 48| 2.76k|#define BITSET_CONTAINER_TYPE 1 | | ------------------ ------------------ | Branch (842:9): [True: 2.76k, False: 22.6k] ------------------ 843| 2.76k| return run_bitset_container_intersection_cardinality( 844| 2.76k| const_CAST_run(c1), const_CAST_bitset(c2)); ------------------ | | 78| 2.76k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 2.76k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_run(c1), const_CAST_bitset(c2)); ------------------ | | 53| 2.76k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 2.76k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 845| | 846| 3.25k| case CONTAINER_PAIR(ARRAY, RUN): ------------------ | | 65| 3.25k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 3.25k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 3.25k|#define RUN_CONTAINER_TYPE 3 | | ------------------ ------------------ | Branch (846:9): [True: 3.25k, False: 22.1k] ------------------ 847| 3.25k| return array_run_container_intersection_cardinality( 848| 3.25k| const_CAST_array(c1), const_CAST_run(c2)); ------------------ | | 55| 3.25k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 3.25k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c1), const_CAST_run(c2)); ------------------ | | 78| 3.25k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 3.25k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 849| | 850| 5.10k| case CONTAINER_PAIR(RUN, ARRAY): ------------------ | | 65| 5.10k| (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 50| 5.10k|#define RUN_CONTAINER_TYPE 3 | | ------------------ | | (4 * (name1##_CONTAINER_TYPE) + (name2##_CONTAINER_TYPE)) | | ------------------ | | | | 49| 5.10k|#define ARRAY_CONTAINER_TYPE 2 | | ------------------ ------------------ | Branch (850:9): [True: 5.10k, False: 20.3k] ------------------ 851| 5.10k| return array_run_container_intersection_cardinality( 852| 5.10k| const_CAST_array(c2), const_CAST_run(c1)); ------------------ | | 55| 5.10k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 5.10k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ const_CAST_array(c2), const_CAST_run(c1)); ------------------ | | 78| 5.10k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 5.10k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 853| | 854| 0| default: ------------------ | Branch (854:9): [True: 0, False: 25.4k] ------------------ 855| 0| assert(false); 856| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 857| 0| return 0; 858| 25.4k| } 859| 25.4k|} roaring.c:container_contains_range: 596| 583|) { 597| 583| c = container_unwrap_shared(c, &typecode); 598| 583| switch (typecode) { 599| 263| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 263|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (599:9): [True: 263, False: 320] ------------------ 600| 263| return bitset_container_get_range(const_CAST_bitset(c), range_start, ------------------ | | 53| 263|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 263|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 601| 263| range_end); 602| 122| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 122|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (602:9): [True: 122, False: 461] ------------------ 603| 122| return array_container_contains_range(const_CAST_array(c), ------------------ | | 55| 122|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 122|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 604| 122| range_start, range_end); 605| 198| case RUN_CONTAINER_TYPE: ------------------ | | 50| 198|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (605:9): [True: 198, False: 385] ------------------ 606| 198| return run_container_contains_range(const_CAST_run(c), range_start, ------------------ | | 78| 198|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 198|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 607| 198| range_end); 608| 0| default: ------------------ | Branch (608:9): [True: 0, False: 583] ------------------ 609| 0| assert(false); 610| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 611| 0| return false; 612| 583| } 613| 583|} roaring_array.c:container_to_uint32_array: 465| 69.2k| uint8_t typecode, uint32_t base) { 466| 69.2k| c = container_unwrap_shared(c, &typecode); 467| 69.2k| switch (typecode) { ------------------ | Branch (467:13): [True: 69.2k, False: 0] ------------------ 468| 1.51k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 1.51k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (468:9): [True: 1.51k, False: 67.7k] ------------------ 469| 1.51k| return bitset_container_to_uint32_array(output, 470| 1.51k| const_CAST_bitset(c), base); ------------------ | | 53| 1.51k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 1.51k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 471| 5.98k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 5.98k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (471:9): [True: 5.98k, False: 63.2k] ------------------ 472| 5.98k| return array_container_to_uint32_array(output, const_CAST_array(c), ------------------ | | 55| 5.98k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 5.98k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 473| 5.98k| base); 474| 61.7k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 61.7k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (474:9): [True: 61.7k, False: 7.50k] ------------------ 475| 61.7k| return run_container_to_uint32_array(output, const_CAST_run(c), ------------------ | | 78| 61.7k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 61.7k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 476| 61.7k| base); 477| 69.2k| } 478| 69.2k| assert(false); 479| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 480| 0| return 0; // unreached 481| 69.2k|} roaring_array.c:container_unwrap_shared: 106| 400k| const container_t *candidate_shared_container, uint8_t *type) { 107| 400k| if (*type == SHARED_CONTAINER_TYPE) { ------------------ | | 51| 400k|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (107:9): [True: 0, False: 400k] ------------------ 108| 0| *type = const_CAST_shared(candidate_shared_container)->typecode; ------------------ | | 80| 0|#define const_CAST_shared(c) CAST(const shared_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 109| 0| assert(*type != SHARED_CONTAINER_TYPE); 110| 0| return const_CAST_shared(candidate_shared_container)->container; ------------------ | | 80| 0|#define const_CAST_shared(c) CAST(const shared_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 111| 400k| } else { 112| 400k| return candidate_shared_container; 113| 400k| } 114| 400k|} roaring_array.c:get_container_type: 129| 168k|static inline uint8_t get_container_type(const container_t *c, uint8_t type) { 130| 168k| if (type == SHARED_CONTAINER_TYPE) { ------------------ | | 51| 168k|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (130:9): [True: 0, False: 168k] ------------------ 131| 0| return const_CAST_shared(c)->typecode; ------------------ | | 80| 0|#define const_CAST_shared(c) CAST(const shared_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 132| 168k| } else { 133| 168k| return type; 134| 168k| } 135| 168k|} roaring_array.c:container_size_in_bytes: 403| 168k| uint8_t typecode) { 404| 168k| c = container_unwrap_shared(c, &typecode); 405| 168k| switch (typecode) { ------------------ | Branch (405:13): [True: 168k, False: 0] ------------------ 406| 147k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 147k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (406:9): [True: 147k, False: 20.8k] ------------------ 407| 147k| return bitset_container_size_in_bytes(const_CAST_bitset(c)); ------------------ | | 53| 147k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 147k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 408| 18.2k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 18.2k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (408:9): [True: 18.2k, False: 150k] ------------------ 409| 18.2k| return array_container_size_in_bytes(const_CAST_array(c)); ------------------ | | 55| 18.2k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 18.2k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 410| 2.67k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 2.67k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (410:9): [True: 2.67k, False: 166k] ------------------ 411| 2.67k| return run_container_size_in_bytes(const_CAST_run(c)); ------------------ | | 78| 2.67k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 2.67k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 412| 168k| } 413| 168k| assert(false); 414| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 415| 0| return 0; // unreached 416| 168k|} roaring_array.c:container_get_cardinality: 244| 81.1k| uint8_t typecode) { 245| 81.1k| c = container_unwrap_shared(c, &typecode); 246| 81.1k| switch (typecode) { ------------------ | Branch (246:13): [True: 81.1k, False: 0] ------------------ 247| 73.9k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 73.9k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (247:9): [True: 73.9k, False: 7.20k] ------------------ 248| 73.9k| return bitset_container_cardinality(const_CAST_bitset(c)); ------------------ | | 53| 73.9k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 73.9k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 249| 7.20k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 7.20k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (249:9): [True: 7.20k, False: 73.9k] ------------------ 250| 7.20k| return array_container_cardinality(const_CAST_array(c)); ------------------ | | 55| 7.20k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 7.20k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 251| 0| case RUN_CONTAINER_TYPE: ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (251:9): [True: 0, False: 81.1k] ------------------ 252| 0| return run_container_cardinality(const_CAST_run(c)); ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 253| 81.1k| } 254| 81.1k| assert(false); 255| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 256| 0| return 0; // unreached 257| 81.1k|} roaring_array.c:container_write: 382| 81.1k| char *buf) { 383| 81.1k| c = container_unwrap_shared(c, &typecode); 384| 81.1k| switch (typecode) { ------------------ | Branch (384:13): [True: 81.1k, False: 0] ------------------ 385| 73.9k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 73.9k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (385:9): [True: 73.9k, False: 7.20k] ------------------ 386| 73.9k| return bitset_container_write(const_CAST_bitset(c), buf); ------------------ | | 53| 73.9k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 73.9k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 387| 7.20k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 7.20k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (387:9): [True: 7.20k, False: 73.9k] ------------------ 388| 7.20k| return array_container_write(const_CAST_array(c), buf); ------------------ | | 55| 7.20k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 7.20k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 389| 0| case RUN_CONTAINER_TYPE: ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (389:9): [True: 0, False: 81.1k] ------------------ 390| 0| return run_container_write(const_CAST_run(c), buf); ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 391| 81.1k| } 392| 81.1k| assert(false); 393| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 394| 0| return 0; // unreached 395| 81.1k|} containers.c:container_unwrap_shared: 106| 6.64k| const container_t *candidate_shared_container, uint8_t *type) { 107| 6.64k| if (*type == SHARED_CONTAINER_TYPE) { ------------------ | | 51| 6.64k|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (107:9): [True: 0, False: 6.64k] ------------------ 108| 0| *type = const_CAST_shared(candidate_shared_container)->typecode; ------------------ | | 80| 0|#define const_CAST_shared(c) CAST(const shared_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 109| 0| assert(*type != SHARED_CONTAINER_TYPE); 110| 0| return const_CAST_shared(candidate_shared_container)->container; ------------------ | | 80| 0|#define const_CAST_shared(c) CAST(const shared_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 111| 6.64k| } else { 112| 6.64k| return candidate_shared_container; 113| 6.64k| } 114| 6.64k|} containers.c:container_maximum: 2171| 6.38k|static inline uint16_t container_maximum(const container_t *c, uint8_t type) { 2172| 6.38k| c = container_unwrap_shared(c, &type); 2173| 6.38k| switch (type) { 2174| 1.91k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 1.91k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (2174:9): [True: 1.91k, False: 4.46k] ------------------ 2175| 1.91k| return bitset_container_maximum(const_CAST_bitset(c)); ------------------ | | 53| 1.91k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 1.91k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2176| 4.46k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 4.46k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (2176:9): [True: 4.46k, False: 1.91k] ------------------ 2177| 4.46k| return array_container_maximum(const_CAST_array(c)); ------------------ | | 55| 4.46k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 4.46k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2178| 0| case RUN_CONTAINER_TYPE: ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (2178:9): [True: 0, False: 6.38k] ------------------ 2179| 0| return run_container_maximum(const_CAST_run(c)); ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 2180| 0| default: ------------------ | Branch (2180:9): [True: 0, False: 6.38k] ------------------ 2181| 0| assert(false); 2182| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2183| 6.38k| } 2184| 6.38k| assert(false); 2185| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 2186| 0| return false; 2187| 6.38k|} run_container_contains: 260| 9.49k|inline bool run_container_contains(const run_container_t *run, uint16_t pos) { 261| 9.49k| int32_t index = interleavedBinarySearch(run->runs, run->n_runs, pos); 262| 9.49k| if (index >= 0) return true; ------------------ | Branch (262:9): [True: 7.00k, False: 2.49k] ------------------ 263| 2.49k| index = -index - 2; // points to preceding value, possibly -1 264| 2.49k| if (index != -1) { // possible match ------------------ | Branch (264:9): [True: 861, False: 1.63k] ------------------ 265| 861| int32_t offset = pos - run->runs[index].value; 266| 861| int32_t le = run->runs[index].length; 267| 861| if (offset <= le) return true; ------------------ | Branch (267:13): [True: 573, False: 288] ------------------ 268| 861| } 269| 1.92k| return false; 270| 2.49k|} interleavedBinarySearch: 112| 426k| uint16_t ikey) { 113| 426k| int32_t low = 0; 114| 426k| int32_t high = lenarray - 1; 115| 1.95M| while (low <= high) { ------------------ | Branch (115:12): [True: 1.72M, False: 234k] ------------------ 116| 1.72M| int32_t middleIndex = (low + high) >> 1; 117| 1.72M| uint16_t middleValue = array[middleIndex].value; 118| 1.72M| if (middleValue < ikey) { ------------------ | Branch (118:13): [True: 682k, False: 1.03M] ------------------ 119| 682k| low = middleIndex + 1; 120| 1.03M| } else if (middleValue > ikey) { ------------------ | Branch (120:20): [True: 846k, False: 192k] ------------------ 121| 846k| high = middleIndex - 1; 122| 846k| } else { 123| 192k| return middleIndex; 124| 192k| } 125| 1.72M| } 126| 234k| return -(low + 1); 127| 426k|} run_container_minimum: 563| 16.6k|inline uint16_t run_container_minimum(const run_container_t *run) { 564| 16.6k| if (run->n_runs == 0) return 0; ------------------ | Branch (564:9): [True: 0, False: 16.6k] ------------------ 565| 16.6k| return run->runs[0].value; 566| 16.6k|} run_container_maximum: 569| 16.1k|inline uint16_t run_container_maximum(const run_container_t *run) { 570| 16.1k| if (run->n_runs == 0) return 0; ------------------ | Branch (570:9): [True: 0, False: 16.1k] ------------------ 571| 16.1k| return run->runs[run->n_runs - 1].value + run->runs[run->n_runs - 1].length; 572| 16.1k|} roaring.c:run_container_create_range: 532| 92.0k| uint32_t stop) { 533| 92.0k| run_container_t *rc = run_container_create_given_capacity(1); 534| 92.0k| if (rc) { ------------------ | Branch (534:9): [True: 92.0k, False: 0] ------------------ 535| 92.0k| rle16_t r; 536| 92.0k| r.value = (uint16_t)start; 537| 92.0k| r.length = (uint16_t)(stop - start - 1); 538| 92.0k| run_container_append_first(rc, r); 539| 92.0k| } 540| 92.0k| return rc; 541| 92.0k|} roaring.c:run_container_append_first: 351| 92.0k| rle16_t vl) { 352| 92.0k| run->runs[run->n_runs] = vl; 353| 92.0k| run->n_runs++; 354| 92.0k| return vl; 355| 92.0k|} roaring.c:rle16_count_greater: 178| 462| int32_t lenarray, uint16_t key) { 179| 462| if (lenarray == 0) return 0; ------------------ | Branch (179:9): [True: 0, False: 462] ------------------ 180| 462| int32_t low = 0; 181| 462| int32_t high = lenarray - 1; 182| 3.55k| while (low <= high) { ------------------ | Branch (182:12): [True: 3.12k, False: 439] ------------------ 183| 3.12k| int32_t middleIndex = (low + high) >> 1; 184| 3.12k| uint16_t min_value = array[middleIndex].value; 185| 3.12k| uint16_t max_value = 186| 3.12k| array[middleIndex].value + array[middleIndex].length; 187| 3.12k| if (max_value < key) { ------------------ | Branch (187:13): [True: 2.91k, False: 208] ------------------ 188| 2.91k| low = middleIndex + 1; 189| 2.91k| } else if (key + UINT32_C(1) < min_value) { // uint32 arithmetic ------------------ | Branch (189:20): [True: 185, False: 23] ------------------ 190| 185| high = middleIndex - 1; 191| 185| } else { 192| 23| return lenarray - (middleIndex + 1); 193| 23| } 194| 3.12k| } 195| 439| return lenarray - low; 196| 462|} roaring.c:rle16_count_less: 157| 462| uint16_t key) { 158| 462| if (lenarray == 0) return 0; ------------------ | Branch (158:9): [True: 1, False: 461] ------------------ 159| 461| int32_t low = 0; 160| 461| int32_t high = lenarray - 1; 161| 2.98k| while (low <= high) { ------------------ | Branch (161:12): [True: 2.77k, False: 207] ------------------ 162| 2.77k| int32_t middleIndex = (low + high) >> 1; 163| 2.77k| uint16_t min_value = array[middleIndex].value; 164| 2.77k| uint16_t max_value = 165| 2.77k| array[middleIndex].value + array[middleIndex].length; 166| 2.77k| if (max_value + UINT32_C(1) < key) { // uint32 arithmetic ------------------ | Branch (166:13): [True: 1.28k, False: 1.48k] ------------------ 167| 1.28k| low = middleIndex + 1; 168| 1.48k| } else if (key < min_value) { ------------------ | Branch (168:20): [True: 1.23k, False: 254] ------------------ 169| 1.23k| high = middleIndex - 1; 170| 1.23k| } else { 171| 254| return middleIndex; 172| 254| } 173| 2.77k| } 174| 207| return low; 175| 461|} roaring.c:run_container_add_range_nruns: 610| 462| int32_t nruns_greater) { 611| 462| int32_t nruns_common = run->n_runs - nruns_less - nruns_greater; 612| 462| if (nruns_common == 0) { ------------------ | Branch (612:9): [True: 139, False: 323] ------------------ 613| 139| makeRoomAtIndex(run, (uint16_t)nruns_less); 614| 139| run->runs[nruns_less].value = (uint16_t)min; 615| 139| run->runs[nruns_less].length = (uint16_t)(max - min); 616| 323| } else { 617| 323| uint32_t common_min = run->runs[nruns_less].value; 618| 323| uint32_t common_max = run->runs[nruns_less + nruns_common - 1].value + 619| 323| run->runs[nruns_less + nruns_common - 1].length; 620| 323| uint32_t result_min = (common_min < min) ? common_min : min; ------------------ | Branch (620:31): [True: 68, False: 255] ------------------ 621| 323| uint32_t result_max = (common_max > max) ? common_max : max; ------------------ | Branch (621:31): [True: 14, False: 309] ------------------ 622| | 623| 323| run->runs[nruns_less].value = (uint16_t)result_min; 624| 323| run->runs[nruns_less].length = (uint16_t)(result_max - result_min); 625| | 626| 323| memmove(&(run->runs[nruns_less + 1]), 627| 323| &(run->runs[run->n_runs - nruns_greater]), 628| 323| nruns_greater * sizeof(rle16_t)); 629| 323| run->n_runs = nruns_less + 1 + nruns_greater; 630| 323| } 631| 462|} roaring.c:makeRoomAtIndex: 208| 888|static inline void makeRoomAtIndex(run_container_t *run, uint16_t index) { 209| | /* This function calls realloc + memmove sequentially to move by one index. 210| | * Potentially copying twice the array. 211| | */ 212| 888| if (run->n_runs + 1 > run->capacity) ------------------ | Branch (212:9): [True: 736, False: 152] ------------------ 213| 736| run_container_grow(run, run->n_runs + 1, true); 214| 888| memmove(run->runs + 1 + index, run->runs + index, 215| 888| (run->n_runs - index) * sizeof(rle16_t)); 216| 888| run->n_runs++; 217| 888|} roaring.c:run_container_remove_range: 667| 1.75k| uint32_t min, uint32_t max) { 668| 1.75k| int32_t first = rle16_find_run(run->runs, run->n_runs, (uint16_t)min); 669| 1.75k| int32_t last = rle16_find_run(run->runs, run->n_runs, (uint16_t)max); 670| | 671| 1.75k| if (first >= 0 && min > run->runs[first].value && ------------------ | Branch (671:9): [True: 621, False: 1.13k] | Branch (671:23): [True: 276, False: 345] ------------------ 672| 276| max < ((uint32_t)run->runs[first].value + ------------------ | Branch (672:9): [True: 35, False: 241] ------------------ 673| 276| (uint32_t)run->runs[first].length)) { 674| | // split this run into two adjacent runs 675| | 676| | // right subinterval 677| 35| makeRoomAtIndex(run, (uint16_t)(first + 1)); 678| 35| run->runs[first + 1].value = (uint16_t)(max + 1); 679| 35| run->runs[first + 1].length = 680| 35| (uint16_t)((run->runs[first].value + run->runs[first].length) - 681| 35| (max + 1)); 682| | 683| | // left subinterval 684| 35| run->runs[first].length = 685| 35| (uint16_t)((min - 1) - run->runs[first].value); 686| | 687| 35| return; 688| 35| } 689| | 690| | // update left-most partial run 691| 1.72k| if (first >= 0) { ------------------ | Branch (691:9): [True: 586, False: 1.13k] ------------------ 692| 586| if (min > run->runs[first].value) { ------------------ | Branch (692:13): [True: 241, False: 345] ------------------ 693| 241| run->runs[first].length = 694| 241| (uint16_t)((min - 1) - run->runs[first].value); 695| 241| first++; 696| 241| } 697| 1.13k| } else { 698| 1.13k| first = -first - 1; 699| 1.13k| } 700| | 701| | // update right-most run 702| 1.72k| if (last >= 0) { ------------------ | Branch (702:9): [True: 557, False: 1.16k] ------------------ 703| 557| uint16_t run_max = run->runs[last].value + run->runs[last].length; 704| 557| if (run_max > max) { ------------------ | Branch (704:13): [True: 332, False: 225] ------------------ 705| 332| run->runs[last].value = (uint16_t)(max + 1); 706| 332| run->runs[last].length = (uint16_t)(run_max - (max + 1)); 707| 332| last--; 708| 332| } 709| 1.16k| } else { 710| 1.16k| last = (-last - 1) - 1; 711| 1.16k| } 712| | 713| | // remove intermediate runs 714| 1.72k| if (first <= last) { ------------------ | Branch (714:9): [True: 157, False: 1.56k] ------------------ 715| 157| run_container_shift_tail(run, run->n_runs - (last + 1), 716| 157| -(last - first + 1)); 717| 157| } 718| 1.72k|} roaring.c:rle16_find_run: 133| 3.51k| uint16_t ikey) { 134| 3.51k| int32_t low = 0; 135| 3.51k| int32_t high = lenarray - 1; 136| 14.9k| while (low <= high) { ------------------ | Branch (136:12): [True: 12.6k, False: 2.30k] ------------------ 137| 12.6k| int32_t middleIndex = (low + high) >> 1; 138| 12.6k| uint16_t min = array[middleIndex].value; 139| 12.6k| uint16_t max = array[middleIndex].value + array[middleIndex].length; 140| 12.6k| if (ikey > max) { ------------------ | Branch (140:13): [True: 1.52k, False: 11.1k] ------------------ 141| 1.52k| low = middleIndex + 1; 142| 11.1k| } else if (ikey < min) { ------------------ | Branch (142:20): [True: 9.88k, False: 1.21k] ------------------ 143| 9.88k| high = middleIndex - 1; 144| 9.88k| } else { 145| 1.21k| return middleIndex; 146| 1.21k| } 147| 12.6k| } 148| 2.30k| return -(low + 1); 149| 3.51k|} roaring.c:run_container_shift_tail: 650| 157| int32_t distance) { 651| 157| if (distance > 0) { ------------------ | Branch (651:9): [True: 0, False: 157] ------------------ 652| 0| if (run->capacity < count + distance) { ------------------ | Branch (652:13): [True: 0, False: 0] ------------------ 653| | run_container_grow(run, count + distance, true); 654| 0| } 655| 0| } 656| 157| int32_t srcpos = run->n_runs - count; 657| 157| int32_t dstpos = srcpos + distance; 658| 157| memmove(&(run->runs[dstpos]), &(run->runs[srcpos]), 659| 157| sizeof(rle16_t) * count); 660| 157| run->n_runs += distance; 661| 157|} roaring.c:run_container_remove: 223| 3.14k|static inline bool run_container_remove(run_container_t *run, uint16_t pos) { 224| 3.14k| int32_t index = interleavedBinarySearch(run->runs, run->n_runs, pos); 225| 3.14k| if (index >= 0) { ------------------ | Branch (225:9): [True: 1.19k, False: 1.94k] ------------------ 226| 1.19k| int32_t le = run->runs[index].length; 227| 1.19k| if (le == 0) { ------------------ | Branch (227:13): [True: 238, False: 958] ------------------ 228| 238| recoverRoomAtIndex(run, (uint16_t)index); 229| 958| } else { 230| 958| run->runs[index].value++; 231| 958| run->runs[index].length--; 232| 958| } 233| 1.19k| return true; 234| 1.19k| } 235| 1.94k| index = -index - 2; // points to preceding value, possibly -1 236| 1.94k| if (index >= 0) { // possible match ------------------ | Branch (236:9): [True: 960, False: 985] ------------------ 237| 960| int32_t offset = pos - run->runs[index].value; 238| 960| int32_t le = run->runs[index].length; 239| 960| if (offset < le) { ------------------ | Branch (239:13): [True: 714, False: 246] ------------------ 240| | // need to break in two 241| 714| run->runs[index].length = (uint16_t)(offset - 1); 242| | // need to insert 243| 714| uint16_t newvalue = pos + 1; 244| 714| int32_t newlength = le - offset - 1; 245| 714| makeRoomAtIndex(run, (uint16_t)(index + 1)); 246| 714| run->runs[index + 1].value = newvalue; 247| 714| run->runs[index + 1].length = (uint16_t)newlength; 248| 714| return true; 249| | 250| 714| } else if (offset == le) { ------------------ | Branch (250:20): [True: 38, False: 208] ------------------ 251| 38| run->runs[index].length--; 252| 38| return true; 253| 38| } 254| 960| } 255| | // no match 256| 1.19k| return false; 257| 1.94k|} roaring.c:recoverRoomAtIndex: 102| 238|static inline void recoverRoomAtIndex(run_container_t *run, uint16_t index) { 103| 238| memmove(run->runs + index, run->runs + (1 + index), 104| 238| (run->n_runs - index - 1) * sizeof(rle16_t)); 105| 238| run->n_runs--; 106| 238|} roaring.c:run_container_nonzero_cardinality: 309| 67.0k| const run_container_t *run) { 310| 67.0k| return run->n_runs > 0; // runs never empty 311| 67.0k|} roaring.c:run_container_is_full: 394| 6.41k|static inline bool run_container_is_full(const run_container_t *run) { 395| 6.41k| rle16_t vl = run->runs[0]; 396| 6.41k| return (run->n_runs == 1) && (vl.value == 0) && (vl.length == 0xFFFF); ------------------ | Branch (396:12): [True: 1.26k, False: 5.14k] | Branch (396:34): [True: 1.01k, False: 251] | Branch (396:53): [True: 896, False: 121] ------------------ 397| 6.41k|} roaring.c:run_container_equals: 502| 133| const run_container_t *container2) { 503| 133| if (container1->n_runs != container2->n_runs) { ------------------ | Branch (503:9): [True: 100, False: 33] ------------------ 504| 100| return false; 505| 100| } 506| 33| return memequals(container1->runs, container2->runs, 507| 33| container1->n_runs * sizeof(rle16_t)); 508| 133|} roaring.c:run_container_contains_range: 278| 198| uint32_t pos_end) { 279| 198| uint32_t count = 0; 280| 198| int32_t index = 281| 198| interleavedBinarySearch(run->runs, run->n_runs, (uint16_t)pos_start); 282| 198| if (index < 0) { ------------------ | Branch (282:9): [True: 127, False: 71] ------------------ 283| 127| index = -index - 2; 284| 127| if ((index == -1) || ------------------ | Branch (284:13): [True: 16, False: 111] ------------------ 285| 111| ((pos_start - run->runs[index].value) > run->runs[index].length)) { ------------------ | Branch (285:13): [True: 31, False: 80] ------------------ 286| 47| return false; 287| 47| } 288| 127| } 289| 3.27k| for (int32_t i = index; i < run->n_runs; ++i) { ------------------ | Branch (289:29): [True: 3.21k, False: 64] ------------------ 290| 3.21k| const uint32_t stop = run->runs[i].value + run->runs[i].length; 291| 3.21k| if (run->runs[i].value >= pos_end) break; ------------------ | Branch (291:13): [True: 18, False: 3.19k] ------------------ 292| 3.19k| if (stop >= pos_end) { ------------------ | Branch (292:13): [True: 69, False: 3.12k] ------------------ 293| 69| count += (((pos_end - run->runs[i].value) > 0) ------------------ | Branch (293:23): [True: 69, False: 0] ------------------ 294| 69| ? (pos_end - run->runs[i].value) 295| 69| : 0); 296| 69| break; 297| 69| } 298| 3.12k| const uint32_t min = (stop - pos_start) > 0 ? (stop - pos_start) : 0; ------------------ | Branch (298:30): [True: 3.11k, False: 13] ------------------ 299| 3.12k| count += (min < run->runs[i].length) ? min : run->runs[i].length; ------------------ | Branch (299:18): [True: 57, False: 3.07k] ------------------ 300| 3.12k| } 301| 151| return count >= (pos_end - pos_start - 1); 302| 198|} roaring_array.c:run_container_size_in_bytes: 493| 2.67k| const run_container_t *container) { 494| 2.67k| return run_container_serialized_size_in_bytes(container->n_runs); 495| 2.67k|} roaring_array.c:run_container_serialized_size_in_bytes: 456| 2.67k|static inline int32_t run_container_serialized_size_in_bytes(int32_t num_runs) { 457| 2.67k| return sizeof(uint16_t) + 458| 2.67k| sizeof(rle16_t) * num_runs; // each run requires 2 2-byte entries. 459| 2.67k|} convert.c:run_container_serialized_size_in_bytes: 456| 154k|static inline int32_t run_container_serialized_size_in_bytes(int32_t num_runs) { 457| 154k| return sizeof(uint16_t) + 458| 154k| sizeof(rle16_t) * num_runs; // each run requires 2 2-byte entries. 459| 154k|} mixed_intersection.c:run_container_is_full: 394| 17.9k|static inline bool run_container_is_full(const run_container_t *run) { 395| 17.9k| rle16_t vl = run->runs[0]; 396| 17.9k| return (run->n_runs == 1) && (vl.value == 0) && (vl.length == 0xFFFF); ------------------ | Branch (396:12): [True: 1.55k, False: 16.4k] | Branch (396:34): [True: 628, False: 929] | Branch (396:53): [True: 204, False: 424] ------------------ 397| 17.9k|} mixed_union.c:run_container_is_full: 394| 3.00k|static inline bool run_container_is_full(const run_container_t *run) { 395| 3.00k| rle16_t vl = run->runs[0]; 396| 3.00k| return (run->n_runs == 1) && (vl.value == 0) && (vl.length == 0xFFFF); ------------------ | Branch (396:12): [True: 389, False: 2.61k] | Branch (396:34): [True: 243, False: 146] | Branch (396:53): [True: 115, False: 128] ------------------ 397| 3.00k|} mixed_union.c:run_container_append_first: 351| 2.20k| rle16_t vl) { 352| 2.20k| run->runs[run->n_runs] = vl; 353| 2.20k| run->n_runs++; 354| 2.20k| return vl; 355| 2.20k|} mixed_union.c:run_container_append_value_first: 385| 684| uint16_t val) { 386| 684| rle16_t newrle = CROARING_MAKE_RLE16(val, 0); ------------------ | | 60| 684| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 387| 684| run->runs[run->n_runs] = newrle; 388| 684| run->n_runs++; 389| 684| return newrle; 390| 684|} mixed_union.c:run_container_append: 332| 173k| rle16_t *previousrl) { 333| 173k| const uint32_t previousend = previousrl->value + previousrl->length; 334| 173k| if (vl.value > previousend + 1) { // we add a new one ------------------ | Branch (334:9): [True: 160k, False: 13.1k] ------------------ 335| 160k| run->runs[run->n_runs] = vl; 336| 160k| run->n_runs++; 337| 160k| *previousrl = vl; 338| 160k| } else { 339| 13.1k| uint32_t newend = vl.value + vl.length + UINT32_C(1); 340| 13.1k| if (newend > previousend) { // we merge ------------------ | Branch (340:13): [True: 13.1k, False: 0] ------------------ 341| 13.1k| previousrl->length = (uint16_t)(newend - 1 - previousrl->value); 342| 13.1k| run->runs[run->n_runs - 1] = *previousrl; 343| 13.1k| } 344| 13.1k| } 345| 173k|} mixed_union.c:run_container_append_value: 368| 530k| rle16_t *previousrl) { 369| 530k| const uint32_t previousend = previousrl->value + previousrl->length; 370| 530k| if (val > previousend + 1) { // we add a new one ------------------ | Branch (370:9): [True: 55.9k, False: 474k] ------------------ 371| 55.9k| *previousrl = CROARING_MAKE_RLE16(val, 0); ------------------ | | 60| 55.9k| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 372| 55.9k| run->runs[run->n_runs] = *previousrl; 373| 55.9k| run->n_runs++; 374| 474k| } else if (val == previousend + 1) { // we merge ------------------ | Branch (374:16): [True: 166k, False: 307k] ------------------ 375| 166k| previousrl->length++; 376| 166k| run->runs[run->n_runs - 1] = *previousrl; 377| 166k| } 378| 530k|} run.c:run_container_is_full: 394| 136k|static inline bool run_container_is_full(const run_container_t *run) { 395| 136k| rle16_t vl = run->runs[0]; 396| 136k| return (run->n_runs == 1) && (vl.value == 0) && (vl.length == 0xFFFF); ------------------ | Branch (396:12): [True: 121k, False: 14.7k] | Branch (396:34): [True: 119k, False: 1.63k] | Branch (396:53): [True: 117k, False: 2.28k] ------------------ 397| 136k|} run.c:run_container_serialized_size_in_bytes: 456| 50|static inline int32_t run_container_serialized_size_in_bytes(int32_t num_runs) { 457| 50| return sizeof(uint16_t) + 458| 50| sizeof(rle16_t) * num_runs; // each run requires 2 2-byte entries. 459| 50|} run.c:recoverRoomAtIndex: 102| 6.57k|static inline void recoverRoomAtIndex(run_container_t *run, uint16_t index) { 103| 6.57k| memmove(run->runs + index, run->runs + (1 + index), 104| 6.57k| (run->n_runs - index - 1) * sizeof(rle16_t)); 105| 6.57k| run->n_runs--; 106| 6.57k|} run.c:makeRoomAtIndex: 208| 44.4k|static inline void makeRoomAtIndex(run_container_t *run, uint16_t index) { 209| | /* This function calls realloc + memmove sequentially to move by one index. 210| | * Potentially copying twice the array. 211| | */ 212| 44.4k| if (run->n_runs + 1 > run->capacity) ------------------ | Branch (212:9): [True: 2.53k, False: 41.9k] ------------------ 213| 2.53k| run_container_grow(run, run->n_runs + 1, true); 214| 44.4k| memmove(run->runs + 1 + index, run->runs + index, 215| 44.4k| (run->n_runs - index) * sizeof(rle16_t)); 216| 44.4k| run->n_runs++; 217| 44.4k|} run.c:run_container_append_first: 351| 2.37k| rle16_t vl) { 352| 2.37k| run->runs[run->n_runs] = vl; 353| 2.37k| run->n_runs++; 354| 2.37k| return vl; 355| 2.37k|} run.c:run_container_append: 332| 304k| rle16_t *previousrl) { 333| 304k| const uint32_t previousend = previousrl->value + previousrl->length; 334| 304k| if (vl.value > previousend + 1) { // we add a new one ------------------ | Branch (334:9): [True: 157k, False: 146k] ------------------ 335| 157k| run->runs[run->n_runs] = vl; 336| 157k| run->n_runs++; 337| 157k| *previousrl = vl; 338| 157k| } else { 339| 146k| uint32_t newend = vl.value + vl.length + UINT32_C(1); 340| 146k| if (newend > previousend) { // we merge ------------------ | Branch (340:13): [True: 115k, False: 31.3k] ------------------ 341| 115k| previousrl->length = (uint16_t)(newend - 1 - previousrl->value); 342| 115k| run->runs[run->n_runs - 1] = *previousrl; 343| 115k| } 344| 146k| } 345| 304k|} run.c:run_container_empty: 314| 16|static inline bool run_container_empty(const run_container_t *run) { 315| 16| return run->n_runs == 0; // runs never empty 316| 16|} run.c:run_container_size_in_bytes: 493| 50| const run_container_t *container) { 494| 50| return run_container_serialized_size_in_bytes(container->n_runs); 495| 50|} roaring_trailing_zeroes: 242| 73.1M|inline int roaring_trailing_zeroes(unsigned long long input_num) { 243| 73.1M| return __builtin_ctzll(input_num); 244| 73.1M|} roaring_leading_zeroes: 246| 4.49k|inline int roaring_leading_zeroes(unsigned long long input_num) { 247| 4.49k| return __builtin_clzll(input_num); 248| 4.49k|} roaring.c:roaring_hamming: 287| 153k|static inline int roaring_hamming(uint64_t x) { 288| |#if defined(_WIN64) && defined(CROARING_REGULAR_VISUAL_STUDIO) && \ 289| | CROARING_REGULAR_VISUAL_STUDIO 290| |#ifdef CROARING_USENEON 291| | return vaddv_u8(vcnt_u8(vcreate_u8(input_num))); 292| |#elif defined(_M_ARM64) 293| | return roaring_hamming_backup(x); 294| | // (int) _CountOneBits64(x); is unavailable 295| |#else // _M_ARM64 296| | return (int)__popcnt64(x); 297| |#endif // _M_ARM64 298| |#elif defined(_WIN32) && defined(CROARING_REGULAR_VISUAL_STUDIO) && \ 299| | CROARING_REGULAR_VISUAL_STUDIO 300| |#ifdef _M_ARM 301| | return roaring_hamming_backup(x); 302| | // _CountOneBits is unavailable 303| |#else // _M_ARM 304| | return (int)__popcnt((unsigned int)x) + 305| | (int)__popcnt((unsigned int)(x >> 32)); 306| |#endif // _M_ARM 307| |#else 308| 153k| return __builtin_popcountll(x); 309| 153k|#endif 310| 153k|} mixed_intersection.c:roaring_hamming: 287| 231k|static inline int roaring_hamming(uint64_t x) { 288| |#if defined(_WIN64) && defined(CROARING_REGULAR_VISUAL_STUDIO) && \ 289| | CROARING_REGULAR_VISUAL_STUDIO 290| |#ifdef CROARING_USENEON 291| | return vaddv_u8(vcnt_u8(vcreate_u8(input_num))); 292| |#elif defined(_M_ARM64) 293| | return roaring_hamming_backup(x); 294| | // (int) _CountOneBits64(x); is unavailable 295| |#else // _M_ARM64 296| | return (int)__popcnt64(x); 297| |#endif // _M_ARM64 298| |#elif defined(_WIN32) && defined(CROARING_REGULAR_VISUAL_STUDIO) && \ 299| | CROARING_REGULAR_VISUAL_STUDIO 300| |#ifdef _M_ARM 301| | return roaring_hamming_backup(x); 302| | // _CountOneBits is unavailable 303| |#else // _M_ARM 304| | return (int)__popcnt((unsigned int)x) + 305| | (int)__popcnt((unsigned int)(x >> 32)); 306| |#endif // _M_ARM 307| |#else 308| 231k| return __builtin_popcountll(x); 309| 231k|#endif 310| 231k|} roaring_bitmap_init_cleared: 72| 32.4k|inline void roaring_bitmap_init_cleared(roaring_bitmap_t *r) { 73| 32.4k| roaring_bitmap_init_with_capacity(r, 0); 74| 32.4k|} roaring_bitmap_add_range: 434| 6.48k| uint64_t max) { 435| 6.48k| if (max <= min || min > (uint64_t)UINT32_MAX + 1) { ------------------ | Branch (435:9): [True: 3.59k, False: 2.88k] | Branch (435:23): [True: 0, False: 2.88k] ------------------ 436| 3.59k| return; 437| 3.59k| } 438| 2.88k| roaring_bitmap_add_range_closed(r, (uint32_t)min, (uint32_t)(max - 1)); 439| 2.88k|} roaring_bitmap_remove_range: 456| 6.48k| uint64_t max) { 457| 6.48k| if (max <= min || min > (uint64_t)UINT32_MAX + 1) { ------------------ | Branch (457:9): [True: 5.52k, False: 951] | Branch (457:23): [True: 0, False: 951] ------------------ 458| 5.52k| return; 459| 5.52k| } 460| 951| roaring_bitmap_remove_range_closed(r, (uint32_t)min, (uint32_t)(max - 1)); 461| 951|} roaring_bitmap_contains: 478| 6.55k|inline bool roaring_bitmap_contains(const roaring_bitmap_t *r, uint32_t val) { 479| | // For performance reasons, this function is inline and uses internal 480| | // functions directly. 481| 6.55k|#ifdef __cplusplus 482| 6.55k| using namespace ::roaring::internal; 483| 6.55k|#endif 484| 6.55k| const uint16_t hb = val >> 16; 485| | /* 486| | * the next function call involves a binary search and lots of branching. 487| | */ 488| 6.55k| int32_t i = ra_get_index(&r->high_low_container, hb); 489| 6.55k| if (i < 0) return false; ------------------ | Branch (489:9): [True: 675, False: 5.87k] ------------------ 490| | 491| 5.87k| uint8_t typecode; 492| | // next call ought to be cheap 493| 5.87k| container_t *container = ra_get_container_at_index(&r->high_low_container, 494| 5.87k| (uint16_t)i, &typecode); 495| | // rest might be a tad expensive, possibly involving another round of binary 496| | // search 497| 5.87k| return container_contains(container, val & 0xFFFF, typecode); 498| 6.55k|} roaring_bitmap_set_copy_on_write: 117| 59.8k|inline void roaring_bitmap_set_copy_on_write(roaring_bitmap_t *r, bool cow) { 118| 59.8k| if (cow) { ------------------ | Branch (118:9): [True: 0, False: 59.8k] ------------------ 119| 0| r->high_low_container.flags |= ROARING_FLAG_COW; ------------------ | | 46| 0|#define ROARING_FLAG_COW UINT8_C(0x1) ------------------ 120| 59.8k| } else { 121| 59.8k| if (roaring_bitmap_get_copy_on_write(r)) { ------------------ | Branch (121:13): [True: 0, False: 59.8k] ------------------ 122| 0| roaring_unshare_all(r); 123| 0| } 124| | r->high_low_container.flags &= ~ROARING_FLAG_COW; ------------------ | | 46| 59.8k|#define ROARING_FLAG_COW UINT8_C(0x1) ------------------ 125| 59.8k| } 126| 59.8k|} roaring_bitmap_get_copy_on_write: 114| 72.8k|inline bool roaring_bitmap_get_copy_on_write(const roaring_bitmap_t *r) { 115| | return r->high_low_container.flags & ROARING_FLAG_COW; ------------------ | | 46| 72.8k|#define ROARING_FLAG_COW UINT8_C(0x1) ------------------ 116| 72.8k|} ra_get_index: 93| 91.3k|inline int32_t ra_get_index(const roaring_array_t *ra, uint16_t x) { 94| 91.3k| if ((ra->size == 0) || ra->keys[ra->size - 1] == x) return ra->size - 1; ------------------ | Branch (94:9): [True: 13.0k, False: 78.2k] | Branch (94:28): [True: 32.9k, False: 45.3k] ------------------ 95| 45.3k| return binarySearch(ra->keys, (int32_t)ra->size, x); 96| 91.3k|} ra_get_container_at_index: 102| 463k| uint16_t i, uint8_t *typecode) { 103| 463k| *typecode = ra->typecodes[i]; 104| 463k| return ra->containers[i]; 105| 463k|} ra_get_key_at_index: 110| 316k|inline uint16_t ra_get_key_at_index(const roaring_array_t *ra, uint16_t i) { 111| 316k| return ra->keys[i]; 112| 316k|} ra_set_container_at_index: 175| 116k| container_t *c, uint8_t typecode) { 176| | assert(i < ra->size); 177| 116k| ra->containers[i] = c; 178| 116k| ra->typecodes[i] = typecode; 179| 116k|} ra_get_size: 191| 13.7k|inline int32_t ra_get_size(const roaring_array_t *ra) { return ra->size; } ra_replace_key_and_container_at_index: 205| 156k| uint8_t typecode) { 206| 156k| assert(i < ra->size); 207| | 208| 156k| ra->keys[i] = key; 209| 156k| ra->containers[i] = c; 210| 156k| ra->typecodes[i] = typecode; 211| 156k|} roaring.c:ra_advance_until: 194| 518| int32_t pos) { 195| 518| return advanceUntil(ra->keys, pos, ra->size, x); 196| 518|} roaring.c:ra_unshare_container_at_index: 268| 64.0k| uint16_t i) { 269| | assert(i < ra->size); 270| 64.0k| ra->containers[i] = 271| 64.0k| get_writable_copy_if_shared(ra->containers[i], &ra->typecodes[i]); 272| 64.0k|} intersect_vector16: 386| 1.75k| size_t s_b, uint16_t *C) { 387| 1.75k| size_t count = 0; 388| 1.75k| size_t i_a = 0, i_b = 0; 389| 1.75k| const int vectorlength = sizeof(__m128i) / sizeof(uint16_t); 390| 1.75k| const size_t st_a = (s_a / vectorlength) * vectorlength; 391| 1.75k| const size_t st_b = (s_b / vectorlength) * vectorlength; 392| 1.75k| __m128i v_a, v_b; 393| 1.75k| if ((i_a < st_a) && (i_b < st_b)) { ------------------ | Branch (393:9): [True: 1.63k, False: 120] | Branch (393:25): [True: 1.47k, False: 158] ------------------ 394| 1.47k| v_a = _mm_lddqu_si128((__m128i *)&A[i_a]); 395| 1.47k| v_b = _mm_lddqu_si128((__m128i *)&B[i_b]); 396| 4.11k| while ((A[i_a] == 0) || (B[i_b] == 0)) { ------------------ | Branch (396:16): [True: 2.50k, False: 1.61k] | Branch (396:33): [True: 273, False: 1.34k] ------------------ 397| 2.77k| const __m128i res_v = _mm_cmpestrm( 398| 2.77k| v_b, vectorlength, v_a, vectorlength, 399| 2.77k| _SIDD_UWORD_OPS | _SIDD_CMP_EQUAL_ANY | _SIDD_BIT_MASK); 400| 2.77k| const int r = _mm_extract_epi32(res_v, 0); 401| 2.77k| __m128i sm16 = _mm_loadu_si128((const __m128i *)shuffle_mask16 + r); 402| 2.77k| __m128i p = _mm_shuffle_epi8(v_a, sm16); 403| 2.77k| _mm_storeu_si128((__m128i *)&C[count], p); // can overflow 404| 2.77k| count += _mm_popcnt_u32(r); 405| 2.77k| const uint16_t a_max = A[i_a + vectorlength - 1]; 406| 2.77k| const uint16_t b_max = B[i_b + vectorlength - 1]; 407| 2.77k| if (a_max <= b_max) { ------------------ | Branch (407:17): [True: 1.62k, False: 1.15k] ------------------ 408| 1.62k| i_a += vectorlength; 409| 1.62k| if (i_a == st_a) break; ------------------ | Branch (409:21): [True: 116, False: 1.50k] ------------------ 410| 1.50k| v_a = _mm_lddqu_si128((__m128i *)&A[i_a]); 411| 1.50k| } 412| 2.66k| if (b_max <= a_max) { ------------------ | Branch (412:17): [True: 1.24k, False: 1.41k] ------------------ 413| 1.24k| i_b += vectorlength; 414| 1.24k| if (i_b == st_b) break; ------------------ | Branch (414:21): [True: 15, False: 1.23k] ------------------ 415| 1.23k| v_b = _mm_lddqu_si128((__m128i *)&B[i_b]); 416| 1.23k| } 417| 2.66k| } 418| 1.47k| if ((i_a < st_a) && (i_b < st_b)) ------------------ | Branch (418:13): [True: 1.35k, False: 116] | Branch (418:29): [True: 1.34k, False: 15] ------------------ 419| 43.7k| while (true) { ------------------ | Branch (419:20): [True: 43.7k, Folded] ------------------ 420| 43.7k| const __m128i res_v = _mm_cmpistrm( 421| 43.7k| v_b, v_a, 422| 43.7k| _SIDD_UWORD_OPS | _SIDD_CMP_EQUAL_ANY | _SIDD_BIT_MASK); 423| 43.7k| const int r = _mm_extract_epi32(res_v, 0); 424| 43.7k| __m128i sm16 = 425| 43.7k| _mm_loadu_si128((const __m128i *)shuffle_mask16 + r); 426| 43.7k| __m128i p = _mm_shuffle_epi8(v_a, sm16); 427| 43.7k| _mm_storeu_si128((__m128i *)&C[count], p); // can overflow 428| 43.7k| count += _mm_popcnt_u32(r); 429| 43.7k| const uint16_t a_max = A[i_a + vectorlength - 1]; 430| 43.7k| const uint16_t b_max = B[i_b + vectorlength - 1]; 431| 43.7k| if (a_max <= b_max) { ------------------ | Branch (431:21): [True: 19.6k, False: 24.1k] ------------------ 432| 19.6k| i_a += vectorlength; 433| 19.6k| if (i_a == st_a) break; ------------------ | Branch (433:25): [True: 795, False: 18.8k] ------------------ 434| 18.8k| v_a = _mm_lddqu_si128((__m128i *)&A[i_a]); 435| 18.8k| } 436| 42.9k| if (b_max <= a_max) { ------------------ | Branch (436:21): [True: 25.6k, False: 17.3k] ------------------ 437| 25.6k| i_b += vectorlength; 438| 25.6k| if (i_b == st_b) break; ------------------ | Branch (438:25): [True: 546, False: 25.1k] ------------------ 439| 25.1k| v_b = _mm_lddqu_si128((__m128i *)&B[i_b]); 440| 25.1k| } 441| 42.9k| } 442| 1.47k| } 443| | // intersect the tail using scalar intersection 444| 28.2k| while (i_a < s_a && i_b < s_b) { ------------------ | Branch (444:12): [True: 26.8k, False: 1.38k] | Branch (444:25): [True: 26.4k, False: 364] ------------------ 445| 26.4k| uint16_t a = A[i_a]; 446| 26.4k| uint16_t b = B[i_b]; 447| 26.4k| if (a < b) { ------------------ | Branch (447:13): [True: 10.7k, False: 15.7k] ------------------ 448| 10.7k| i_a++; 449| 15.7k| } else if (b < a) { ------------------ | Branch (449:20): [True: 10.0k, False: 5.69k] ------------------ 450| 10.0k| i_b++; 451| 10.0k| } else { 452| 5.69k| C[count] = a; //==b; 453| 5.69k| count++; 454| 5.69k| i_a++; 455| 5.69k| i_b++; 456| 5.69k| } 457| 26.4k| } 458| 1.75k| return (int32_t)count; 459| 1.75k|} array_container_to_uint32_array_vector16: 464| 5.98k| uint32_t base) { 465| 5.98k| int outpos = 0; 466| 5.98k| uint32_t *out = (uint32_t *)vout; 467| 5.98k| size_t i = 0; 468| 203k| for (; i + sizeof(__m128i) / sizeof(uint16_t) <= cardinality; ------------------ | Branch (468:12): [True: 197k, False: 5.98k] ------------------ 469| 197k| i += sizeof(__m128i) / sizeof(uint16_t)) { 470| 197k| __m128i vinput = _mm_loadu_si128((const __m128i *)(array + i)); 471| 197k| __m256i voutput = _mm256_add_epi32(_mm256_cvtepu16_epi32(vinput), 472| 197k| _mm256_set1_epi32(base)); 473| 197k| _mm256_storeu_si256((__m256i *)(out + outpos), voutput); 474| 197k| outpos += sizeof(__m256i) / sizeof(uint32_t); 475| 197k| } 476| 20.6k| for (; i < cardinality; ++i) { ------------------ | Branch (476:12): [True: 14.6k, False: 5.98k] ------------------ 477| 14.6k| const uint32_t val = base + array[i]; 478| 14.6k| memcpy(out + outpos, &val, 479| 14.6k| sizeof(uint32_t)); // should be compiled as a MOV on x64 480| 14.6k| outpos++; 481| 14.6k| } 482| 5.98k| return outpos; 483| 5.98k|} intersect_vector16_inplace: 486| 4.64k| size_t s_b) { 487| 4.64k| size_t count = 0; 488| 4.64k| size_t i_a = 0, i_b = 0; 489| 4.64k| const int vectorlength = sizeof(__m128i) / sizeof(uint16_t); 490| 4.64k| const size_t st_a = (s_a / vectorlength) * vectorlength; 491| 4.64k| const size_t st_b = (s_b / vectorlength) * vectorlength; 492| 4.64k| __m128i v_a, v_b; 493| 4.64k| if ((i_a < st_a) && (i_b < st_b)) { ------------------ | Branch (493:9): [True: 2.29k, False: 2.35k] | Branch (493:25): [True: 2.09k, False: 195] ------------------ 494| 2.09k| v_a = _mm_lddqu_si128((__m128i *)&A[i_a]); 495| 2.09k| v_b = _mm_lddqu_si128((__m128i *)&B[i_b]); 496| 2.09k| __m128i tmp[2] = {_mm_setzero_si128()}; 497| 2.09k| size_t tmp_count = 0; 498| 4.20k| while ((A[i_a] == 0) || (B[i_b] == 0)) { ------------------ | Branch (498:16): [True: 1.82k, False: 2.38k] | Branch (498:33): [True: 376, False: 2.00k] ------------------ 499| 2.20k| const __m128i res_v = _mm_cmpestrm( 500| 2.20k| v_b, vectorlength, v_a, vectorlength, 501| 2.20k| _SIDD_UWORD_OPS | _SIDD_CMP_EQUAL_ANY | _SIDD_BIT_MASK); 502| 2.20k| const int r = _mm_extract_epi32(res_v, 0); 503| 2.20k| __m128i sm16 = _mm_loadu_si128((const __m128i *)shuffle_mask16 + r); 504| 2.20k| __m128i p = _mm_shuffle_epi8(v_a, sm16); 505| 2.20k| _mm_storeu_si128((__m128i *)&((uint16_t *)tmp)[tmp_count], p); 506| 2.20k| tmp_count += _mm_popcnt_u32(r); 507| 2.20k| const uint16_t a_max = A[i_a + vectorlength - 1]; 508| 2.20k| const uint16_t b_max = B[i_b + vectorlength - 1]; 509| 2.20k| if (a_max <= b_max) { ------------------ | Branch (509:17): [True: 2.05k, False: 142] ------------------ 510| 2.05k| _mm_storeu_si128((__m128i *)&A[count], tmp[0]); 511| 2.05k| _mm_storeu_si128(tmp, _mm_setzero_si128()); 512| 2.05k| count += tmp_count; 513| 2.05k| tmp_count = 0; 514| 2.05k| i_a += vectorlength; 515| 2.05k| if (i_a == st_a) break; ------------------ | Branch (515:21): [True: 86, False: 1.97k] ------------------ 516| 1.97k| v_a = _mm_lddqu_si128((__m128i *)&A[i_a]); 517| 1.97k| } 518| 2.11k| if (b_max <= a_max) { ------------------ | Branch (518:17): [True: 325, False: 1.78k] ------------------ 519| 325| i_b += vectorlength; 520| 325| if (i_b == st_b) break; ------------------ | Branch (520:21): [True: 6, False: 319] ------------------ 521| 319| v_b = _mm_lddqu_si128((__m128i *)&B[i_b]); 522| 319| } 523| 2.11k| } 524| 2.09k| if ((i_a < st_a) && (i_b < st_b)) { ------------------ | Branch (524:13): [True: 2.01k, False: 86] | Branch (524:29): [True: 2.00k, False: 6] ------------------ 525| 251k| while (true) { ------------------ | Branch (525:20): [True: 251k, Folded] ------------------ 526| 251k| const __m128i res_v = _mm_cmpistrm( 527| 251k| v_b, v_a, 528| 251k| _SIDD_UWORD_OPS | _SIDD_CMP_EQUAL_ANY | _SIDD_BIT_MASK); 529| 251k| const int r = _mm_extract_epi32(res_v, 0); 530| 251k| __m128i sm16 = 531| 251k| _mm_loadu_si128((const __m128i *)shuffle_mask16 + r); 532| 251k| __m128i p = _mm_shuffle_epi8(v_a, sm16); 533| 251k| _mm_storeu_si128((__m128i *)&((uint16_t *)tmp)[tmp_count], p); 534| 251k| tmp_count += _mm_popcnt_u32(r); 535| 251k| const uint16_t a_max = A[i_a + vectorlength - 1]; 536| 251k| const uint16_t b_max = B[i_b + vectorlength - 1]; 537| 251k| if (a_max <= b_max) { ------------------ | Branch (537:21): [True: 147k, False: 104k] ------------------ 538| 147k| _mm_storeu_si128((__m128i *)&A[count], tmp[0]); 539| 147k| _mm_storeu_si128(tmp, _mm_setzero_si128()); 540| 147k| count += tmp_count; 541| 147k| tmp_count = 0; 542| 147k| i_a += vectorlength; 543| 147k| if (i_a == st_a) break; ------------------ | Branch (543:25): [True: 1.12k, False: 146k] ------------------ 544| 146k| v_a = _mm_lddqu_si128((__m128i *)&A[i_a]); 545| 146k| } 546| 250k| if (b_max <= a_max) { ------------------ | Branch (546:21): [True: 141k, False: 109k] ------------------ 547| 141k| i_b += vectorlength; 548| 141k| if (i_b == st_b) break; ------------------ | Branch (548:25): [True: 875, False: 140k] ------------------ 549| 140k| v_b = _mm_lddqu_si128((__m128i *)&B[i_b]); 550| 140k| } 551| 250k| } 552| 2.00k| } 553| | // tmp_count <= 8, so this does not affect efficiency so much 554| 4.49k| for (size_t i = 0; i < tmp_count; i++) { ------------------ | Branch (554:28): [True: 2.40k, False: 2.09k] ------------------ 555| 2.40k| A[count] = ((uint16_t *)tmp)[i]; 556| 2.40k| count++; 557| 2.40k| } 558| 2.09k| i_a += tmp_count; // We can at least jump pass $tmp_count elements in A 559| 2.09k| } 560| | // intersect the tail using scalar intersection 561| 33.3k| while (i_a < s_a && i_b < s_b) { ------------------ | Branch (561:12): [True: 29.1k, False: 4.23k] | Branch (561:25): [True: 28.6k, False: 411] ------------------ 562| 28.6k| uint16_t a = A[i_a]; 563| 28.6k| uint16_t b = B[i_b]; 564| 28.6k| if (a < b) { ------------------ | Branch (564:13): [True: 9.39k, False: 19.2k] ------------------ 565| 9.39k| i_a++; 566| 19.2k| } else if (b < a) { ------------------ | Branch (566:20): [True: 7.33k, False: 11.9k] ------------------ 567| 7.33k| i_b++; 568| 11.9k| } else { 569| 11.9k| A[count] = a; //==b; 570| 11.9k| count++; 571| 11.9k| i_a++; 572| 11.9k| i_b++; 573| 11.9k| } 574| 28.6k| } 575| 4.64k| return (int32_t)count; 576| 4.64k|} intersect_vector16_cardinality: 581| 7.00k| const uint16_t *B, size_t s_b) { 582| 7.00k| size_t count = 0; 583| 7.00k| size_t i_a = 0, i_b = 0; 584| 7.00k| const int vectorlength = sizeof(__m128i) / sizeof(uint16_t); 585| 7.00k| const size_t st_a = (s_a / vectorlength) * vectorlength; 586| 7.00k| const size_t st_b = (s_b / vectorlength) * vectorlength; 587| 7.00k| __m128i v_a, v_b; 588| 7.00k| if ((i_a < st_a) && (i_b < st_b)) { ------------------ | Branch (588:9): [True: 6.52k, False: 480] | Branch (588:25): [True: 5.88k, False: 632] ------------------ 589| 5.88k| v_a = _mm_lddqu_si128((__m128i *)&A[i_a]); 590| 5.88k| v_b = _mm_lddqu_si128((__m128i *)&B[i_b]); 591| 16.4k| while ((A[i_a] == 0) || (B[i_b] == 0)) { ------------------ | Branch (591:16): [True: 10.0k, False: 6.45k] | Branch (591:33): [True: 1.09k, False: 5.36k] ------------------ 592| 11.1k| const __m128i res_v = _mm_cmpestrm( 593| 11.1k| v_b, vectorlength, v_a, vectorlength, 594| 11.1k| _SIDD_UWORD_OPS | _SIDD_CMP_EQUAL_ANY | _SIDD_BIT_MASK); 595| 11.1k| const int r = _mm_extract_epi32(res_v, 0); 596| 11.1k| count += _mm_popcnt_u32(r); 597| 11.1k| const uint16_t a_max = A[i_a + vectorlength - 1]; 598| 11.1k| const uint16_t b_max = B[i_b + vectorlength - 1]; 599| 11.1k| if (a_max <= b_max) { ------------------ | Branch (599:17): [True: 6.49k, False: 4.61k] ------------------ 600| 6.49k| i_a += vectorlength; 601| 6.49k| if (i_a == st_a) break; ------------------ | Branch (601:21): [True: 464, False: 6.02k] ------------------ 602| 6.02k| v_a = _mm_lddqu_si128((__m128i *)&A[i_a]); 603| 6.02k| } 604| 10.6k| if (b_max <= a_max) { ------------------ | Branch (604:17): [True: 4.98k, False: 5.66k] ------------------ 605| 4.98k| i_b += vectorlength; 606| 4.98k| if (i_b == st_b) break; ------------------ | Branch (606:21): [True: 60, False: 4.92k] ------------------ 607| 4.92k| v_b = _mm_lddqu_si128((__m128i *)&B[i_b]); 608| 4.92k| } 609| 10.6k| } 610| 5.88k| if ((i_a < st_a) && (i_b < st_b)) ------------------ | Branch (610:13): [True: 5.42k, False: 464] | Branch (610:29): [True: 5.36k, False: 60] ------------------ 611| 175k| while (true) { ------------------ | Branch (611:20): [True: 175k, Folded] ------------------ 612| 175k| const __m128i res_v = _mm_cmpistrm( 613| 175k| v_b, v_a, 614| 175k| _SIDD_UWORD_OPS | _SIDD_CMP_EQUAL_ANY | _SIDD_BIT_MASK); 615| 175k| const int r = _mm_extract_epi32(res_v, 0); 616| 175k| count += _mm_popcnt_u32(r); 617| 175k| const uint16_t a_max = A[i_a + vectorlength - 1]; 618| 175k| const uint16_t b_max = B[i_b + vectorlength - 1]; 619| 175k| if (a_max <= b_max) { ------------------ | Branch (619:21): [True: 78.5k, False: 96.6k] ------------------ 620| 78.5k| i_a += vectorlength; 621| 78.5k| if (i_a == st_a) break; ------------------ | Branch (621:25): [True: 3.18k, False: 75.3k] ------------------ 622| 75.3k| v_a = _mm_lddqu_si128((__m128i *)&A[i_a]); 623| 75.3k| } 624| 171k| if (b_max <= a_max) { ------------------ | Branch (624:21): [True: 102k, False: 69.2k] ------------------ 625| 102k| i_b += vectorlength; 626| 102k| if (i_b == st_b) break; ------------------ | Branch (626:25): [True: 2.18k, False: 100k] ------------------ 627| 100k| v_b = _mm_lddqu_si128((__m128i *)&B[i_b]); 628| 100k| } 629| 171k| } 630| 5.88k| } 631| | // intersect the tail using scalar intersection 632| 112k| while (i_a < s_a && i_b < s_b) { ------------------ | Branch (632:12): [True: 107k, False: 5.54k] | Branch (632:25): [True: 105k, False: 1.45k] ------------------ 633| 105k| uint16_t a = A[i_a]; 634| 105k| uint16_t b = B[i_b]; 635| 105k| if (a < b) { ------------------ | Branch (635:13): [True: 42.9k, False: 63.0k] ------------------ 636| 42.9k| i_a++; 637| 63.0k| } else if (b < a) { ------------------ | Branch (637:20): [True: 40.2k, False: 22.7k] ------------------ 638| 40.2k| i_b++; 639| 40.2k| } else { 640| 22.7k| count++; 641| 22.7k| i_a++; 642| 22.7k| i_b++; 643| 22.7k| } 644| 105k| } 645| 7.00k| return (int32_t)count; 646| 7.00k|} difference_vector16: 655| 2.32k| size_t s_b, uint16_t *C) { 656| | // we handle the degenerate case 657| 2.32k| if (s_a == 0) return 0; ------------------ | Branch (657:9): [True: 0, False: 2.32k] ------------------ 658| 2.32k| if (s_b == 0) { ------------------ | Branch (658:9): [True: 0, False: 2.32k] ------------------ 659| 0| if (A != C) memcpy(C, A, sizeof(uint16_t) * s_a); ------------------ | Branch (659:13): [True: 0, False: 0] ------------------ 660| 0| return (int32_t)s_a; 661| 0| } 662| | // handle the leading zeroes, it is messy but it allows us to use the fast 663| | // _mm_cmpistrm instrinsic safely 664| 2.32k| int32_t count = 0; 665| 2.32k| if ((A[0] == 0) || (B[0] == 0)) { ------------------ | Branch (665:9): [True: 2.12k, False: 195] | Branch (665:24): [True: 5, False: 190] ------------------ 666| 2.13k| if ((A[0] == 0) && (B[0] == 0)) { ------------------ | Branch (666:13): [True: 2.12k, False: 5] | Branch (666:28): [True: 307, False: 1.82k] ------------------ 667| 307| A++; 668| 307| s_a--; 669| 307| B++; 670| 307| s_b--; 671| 1.82k| } else if (A[0] == 0) { ------------------ | Branch (671:20): [True: 1.82k, False: 5] ------------------ 672| 1.82k| C[count++] = 0; 673| 1.82k| A++; 674| 1.82k| s_a--; 675| 1.82k| } else { 676| 5| B++; 677| 5| s_b--; 678| 5| } 679| 2.13k| } 680| | // at this point, we have two non-empty arrays, made of non-zero 681| | // increasing values. 682| 2.32k| size_t i_a = 0, i_b = 0; 683| 2.32k| const size_t vectorlength = sizeof(__m128i) / sizeof(uint16_t); 684| 2.32k| const size_t st_a = (s_a / vectorlength) * vectorlength; 685| 2.32k| const size_t st_b = (s_b / vectorlength) * vectorlength; 686| 2.32k| if ((i_a < st_a) && (i_b < st_b)) { // this is the vectorized code path ------------------ | Branch (686:9): [True: 1.89k, False: 425] | Branch (686:25): [True: 1.59k, False: 301] ------------------ 687| 1.59k| __m128i v_a, v_b; //, v_bmax; 688| | // we load a vector from A and a vector from B 689| 1.59k| v_a = _mm_lddqu_si128((__m128i *)&A[i_a]); 690| 1.59k| v_b = _mm_lddqu_si128((__m128i *)&B[i_b]); 691| | // we have a runningmask which indicates which values from A have been 692| | // spotted in B, these don't get written out. 693| 1.59k| __m128i runningmask_a_found_in_b = _mm_setzero_si128(); 694| | /**** 695| | * start of the main vectorized loop 696| | *****/ 697| 48.2k| while (true) { ------------------ | Branch (697:16): [True: 48.2k, Folded] ------------------ 698| | // afoundinb will contain a mask indicate for each entry in A 699| | // whether it is seen 700| | // in B 701| 48.2k| const __m128i a_found_in_b = _mm_cmpistrm( 702| 48.2k| v_b, v_a, 703| 48.2k| _SIDD_UWORD_OPS | _SIDD_CMP_EQUAL_ANY | _SIDD_BIT_MASK); 704| 48.2k| runningmask_a_found_in_b = 705| 48.2k| _mm_or_si128(runningmask_a_found_in_b, a_found_in_b); 706| | // we always compare the last values of A and B 707| 48.2k| const uint16_t a_max = A[i_a + vectorlength - 1]; 708| 48.2k| const uint16_t b_max = B[i_b + vectorlength - 1]; 709| 48.2k| if (a_max <= b_max) { ------------------ | Branch (709:17): [True: 21.6k, False: 26.5k] ------------------ 710| | // Ok. In this code path, we are ready to write our v_a 711| | // because there is no need to read more from B, they will 712| | // all be large values. 713| 21.6k| const int bitmask_belongs_to_difference = 714| 21.6k| _mm_extract_epi32(runningmask_a_found_in_b, 0) ^ 0xFF; 715| | /*** next few lines are probably expensive *****/ 716| 21.6k| __m128i sm16 = _mm_loadu_si128((const __m128i *)shuffle_mask16 + 717| 21.6k| bitmask_belongs_to_difference); 718| 21.6k| __m128i p = _mm_shuffle_epi8(v_a, sm16); 719| 21.6k| _mm_storeu_si128((__m128i *)&C[count], p); // can overflow 720| 21.6k| count += _mm_popcnt_u32(bitmask_belongs_to_difference); 721| | // we advance a 722| 21.6k| i_a += vectorlength; 723| 21.6k| if (i_a == st_a) // no more ------------------ | Branch (723:21): [True: 1.15k, False: 20.4k] ------------------ 724| 1.15k| break; 725| 20.4k| runningmask_a_found_in_b = _mm_setzero_si128(); 726| 20.4k| v_a = _mm_lddqu_si128((__m128i *)&A[i_a]); 727| 20.4k| } 728| 47.0k| if (b_max <= a_max) { ------------------ | Branch (728:17): [True: 29.5k, False: 17.4k] ------------------ 729| | // in this code path, the current v_b has become useless 730| 29.5k| i_b += vectorlength; 731| 29.5k| if (i_b == st_b) break; ------------------ | Branch (731:21): [True: 438, False: 29.1k] ------------------ 732| 29.1k| v_b = _mm_lddqu_si128((__m128i *)&B[i_b]); 733| 29.1k| } 734| 47.0k| } 735| | // at this point, either we have i_a == st_a, which is the end of the 736| | // vectorized processing, 737| | // or we have i_b == st_b, and we are not done processing the vector... 738| | // so we need to finish it off. 739| 1.59k| if (i_a < st_a) { // we have unfinished business... ------------------ | Branch (739:13): [True: 438, False: 1.15k] ------------------ 740| 438| uint16_t buffer[8]; // buffer to do a masked load 741| 438| memset(buffer, 0, 8 * sizeof(uint16_t)); 742| 438| memcpy(buffer, B + i_b, (s_b - i_b) * sizeof(uint16_t)); 743| 438| v_b = _mm_lddqu_si128((__m128i *)buffer); 744| 438| const __m128i a_found_in_b = _mm_cmpistrm( 745| 438| v_b, v_a, 746| 438| _SIDD_UWORD_OPS | _SIDD_CMP_EQUAL_ANY | _SIDD_BIT_MASK); 747| 438| runningmask_a_found_in_b = 748| 438| _mm_or_si128(runningmask_a_found_in_b, a_found_in_b); 749| 438| const int bitmask_belongs_to_difference = 750| 438| _mm_extract_epi32(runningmask_a_found_in_b, 0) ^ 0xFF; 751| 438| __m128i sm16 = _mm_loadu_si128((const __m128i *)shuffle_mask16 + 752| 438| bitmask_belongs_to_difference); 753| 438| __m128i p = _mm_shuffle_epi8(v_a, sm16); 754| 438| _mm_storeu_si128((__m128i *)&C[count], p); // can overflow 755| 438| count += _mm_popcnt_u32(bitmask_belongs_to_difference); 756| 438| i_a += vectorlength; 757| 438| } 758| | // at this point we should have i_a == st_a and i_b == st_b 759| 1.59k| } 760| | // do the tail using scalar code 761| 110k| while (i_a < s_a && i_b < s_b) { ------------------ | Branch (761:12): [True: 108k, False: 1.81k] | Branch (761:25): [True: 107k, False: 513] ------------------ 762| 107k| uint16_t a = A[i_a]; 763| 107k| uint16_t b = B[i_b]; 764| 107k| if (b < a) { ------------------ | Branch (764:13): [True: 76.8k, False: 31.0k] ------------------ 765| 76.8k| i_b++; 766| 76.8k| } else if (a < b) { ------------------ | Branch (766:20): [True: 25.4k, False: 5.68k] ------------------ 767| 25.4k| C[count] = a; 768| 25.4k| count++; 769| 25.4k| i_a++; 770| 25.4k| } else { //== 771| 5.68k| i_a++; 772| 5.68k| i_b++; 773| 5.68k| } 774| 107k| } 775| 2.32k| if (i_a < s_a) { ------------------ | Branch (775:9): [True: 513, False: 1.81k] ------------------ 776| 513| if (C == A) { ------------------ | Branch (776:13): [True: 0, False: 513] ------------------ 777| 0| assert((size_t)count <= i_a); 778| 0| if ((size_t)count < i_a) { ------------------ | Branch (778:17): [True: 0, False: 0] ------------------ 779| 0| memmove(C + count, A + i_a, sizeof(uint16_t) * (s_a - i_a)); 780| 0| } 781| 513| } else { 782| 26.5k| for (size_t i = 0; i < (s_a - i_a); i++) { ------------------ | Branch (782:32): [True: 26.0k, False: 513] ------------------ 783| 26.0k| C[count + i] = A[i + i_a]; 784| 26.0k| } 785| 513| } 786| 513| count += (int32_t)(s_a - i_a); 787| 513| } 788| 2.32k| return count; 789| 2.32k|} intersect_skewed_uint16: 855| 787| uint16_t *buffer) { 856| 787| size_t pos = 0, idx_l = 0, idx_s = 0; 857| | 858| 787| if (0 == size_s) { ------------------ | Branch (858:9): [True: 0, False: 787] ------------------ 859| 0| return 0; 860| 0| } 861| 787| int32_t index1 = 0, index2 = 0, index3 = 0, index4 = 0; 862| 2.17k| while ((idx_s + 4 <= size_s) && (idx_l < size_l)) { ------------------ | Branch (862:12): [True: 1.39k, False: 787] | Branch (862:37): [True: 1.39k, False: 0] ------------------ 863| 1.39k| uint16_t target1 = small[idx_s]; 864| 1.39k| uint16_t target2 = small[idx_s + 1]; 865| 1.39k| uint16_t target3 = small[idx_s + 2]; 866| 1.39k| uint16_t target4 = small[idx_s + 3]; 867| 1.39k| binarySearch4(large + idx_l, (int32_t)(size_l - idx_l), target1, 868| 1.39k| target2, target3, target4, &index1, &index2, &index3, 869| 1.39k| &index4); 870| 1.39k| if ((index1 + idx_l < size_l) && (large[idx_l + index1] == target1)) { ------------------ | Branch (870:13): [True: 1.39k, False: 1] | Branch (870:42): [True: 650, False: 741] ------------------ 871| 650| buffer[pos++] = target1; 872| 650| } 873| 1.39k| if ((index2 + idx_l < size_l) && (large[idx_l + index2] == target2)) { ------------------ | Branch (873:13): [True: 1.38k, False: 3] | Branch (873:42): [True: 814, False: 575] ------------------ 874| 814| buffer[pos++] = target2; 875| 814| } 876| 1.39k| if ((index3 + idx_l < size_l) && (large[idx_l + index3] == target3)) { ------------------ | Branch (876:13): [True: 1.38k, False: 4] | Branch (876:42): [True: 790, False: 598] ------------------ 877| 790| buffer[pos++] = target3; 878| 790| } 879| 1.39k| if ((index4 + idx_l < size_l) && (large[idx_l + index4] == target4)) { ------------------ | Branch (879:13): [True: 1.38k, False: 6] | Branch (879:42): [True: 758, False: 628] ------------------ 880| 758| buffer[pos++] = target4; 881| 758| } 882| 1.39k| idx_s += 4; 883| 1.39k| idx_l += index4; 884| 1.39k| } 885| 787| if ((idx_s + 2 <= size_s) && (idx_l < size_l)) { ------------------ | Branch (885:9): [True: 415, False: 372] | Branch (885:34): [True: 414, False: 1] ------------------ 886| 414| uint16_t target1 = small[idx_s]; 887| 414| uint16_t target2 = small[idx_s + 1]; 888| 414| binarySearch2(large + idx_l, (int32_t)(size_l - idx_l), target1, 889| 414| target2, &index1, &index2); 890| 414| if ((index1 + idx_l < size_l) && (large[idx_l + index1] == target1)) { ------------------ | Branch (890:13): [True: 409, False: 5] | Branch (890:42): [True: 280, False: 129] ------------------ 891| 280| buffer[pos++] = target1; 892| 280| } 893| 414| if ((index2 + idx_l < size_l) && (large[idx_l + index2] == target2)) { ------------------ | Branch (893:13): [True: 397, False: 17] | Branch (893:42): [True: 287, False: 110] ------------------ 894| 287| buffer[pos++] = target2; 895| 287| } 896| 414| idx_s += 2; 897| 414| idx_l += index2; 898| 414| } 899| 787| if ((idx_s < size_s) && (idx_l < size_l)) { ------------------ | Branch (899:9): [True: 353, False: 434] | Branch (899:29): [True: 350, False: 3] ------------------ 900| 350| uint16_t val_s = small[idx_s]; 901| 350| int32_t index = 902| 350| binarySearch(large + idx_l, (int32_t)(size_l - idx_l), val_s); 903| 350| if (index >= 0) buffer[pos++] = val_s; ------------------ | Branch (903:13): [True: 247, False: 103] ------------------ 904| 350| } 905| 787| return (int32_t)pos; 906| 787|} intersect_skewed_uint16_cardinality: 912| 2.29k| size_t size_l) { 913| 2.29k| size_t pos = 0, idx_l = 0, idx_s = 0; 914| | 915| 2.29k| if (0 == size_s) { ------------------ | Branch (915:9): [True: 0, False: 2.29k] ------------------ 916| 0| return 0; 917| 0| } 918| | 919| 2.29k| uint16_t val_l = large[idx_l], val_s = small[idx_s]; 920| | 921| 32.1k| while (true) { ------------------ | Branch (921:12): [True: 32.1k, Folded] ------------------ 922| 32.1k| if (val_l < val_s) { ------------------ | Branch (922:13): [True: 7.08k, False: 25.0k] ------------------ 923| 7.08k| idx_l = advanceUntil(large, (int32_t)idx_l, (int32_t)size_l, val_s); 924| 7.08k| if (idx_l == size_l) break; ------------------ | Branch (924:17): [True: 124, False: 6.96k] ------------------ 925| 6.96k| val_l = large[idx_l]; 926| 25.0k| } else if (val_s < val_l) { ------------------ | Branch (926:20): [True: 11.4k, False: 13.6k] ------------------ 927| 11.4k| idx_s++; 928| 11.4k| if (idx_s == size_s) break; ------------------ | Branch (928:17): [True: 880, False: 10.5k] ------------------ 929| 10.5k| val_s = small[idx_s]; 930| 13.6k| } else { 931| 13.6k| pos++; 932| 13.6k| idx_s++; 933| 13.6k| if (idx_s == size_s) break; ------------------ | Branch (933:17): [True: 1.20k, False: 12.4k] ------------------ 934| 12.4k| val_s = small[idx_s]; 935| 12.4k| idx_l = advanceUntil(large, (int32_t)idx_l, (int32_t)size_l, val_s); 936| 12.4k| if (idx_l == size_l) break; ------------------ | Branch (936:17): [True: 84, False: 12.3k] ------------------ 937| 12.3k| val_l = large[idx_l]; 938| 12.3k| } 939| 32.1k| } 940| | 941| 2.29k| return (int32_t)pos; 942| 2.29k|} intersect_skewed_uint16_nonempty: 945| 573| const uint16_t *large, size_t size_l) { 946| 573| size_t idx_l = 0, idx_s = 0; 947| | 948| 573| if (0 == size_s) { ------------------ | Branch (948:9): [True: 0, False: 573] ------------------ 949| 0| return false; 950| 0| } 951| | 952| 573| uint16_t val_l = large[idx_l], val_s = small[idx_s]; 953| | 954| 2.32k| while (true) { ------------------ | Branch (954:12): [True: 2.32k, Folded] ------------------ 955| 2.32k| if (val_l < val_s) { ------------------ | Branch (955:13): [True: 882, False: 1.44k] ------------------ 956| 882| idx_l = advanceUntil(large, (int32_t)idx_l, (int32_t)size_l, val_s); 957| 882| if (idx_l == size_l) break; ------------------ | Branch (957:17): [True: 30, False: 852] ------------------ 958| 852| val_l = large[idx_l]; 959| 1.44k| } else if (val_s < val_l) { ------------------ | Branch (959:20): [True: 973, False: 471] ------------------ 960| 973| idx_s++; 961| 973| if (idx_s == size_s) break; ------------------ | Branch (961:17): [True: 72, False: 901] ------------------ 962| 901| val_s = small[idx_s]; 963| 901| } else { 964| 471| return true; 965| 471| } 966| 2.32k| } 967| | 968| 102| return false; 969| 573|} intersect_uint16_nonempty: 1025| 1.75k| const uint16_t *B, const size_t lenB) { 1026| 1.75k| if (lenA == 0 || lenB == 0) return 0; ------------------ | Branch (1026:9): [True: 0, False: 1.75k] | Branch (1026:22): [True: 0, False: 1.75k] ------------------ 1027| 1.75k| const uint16_t *endA = A + lenA; 1028| 1.75k| const uint16_t *endB = B + lenB; 1029| | 1030| 1.75k| while (1) { ------------------ | Branch (1030:12): [True: 1.75k, Folded] ------------------ 1031| 8.78k| while (*A < *B) { ------------------ | Branch (1031:16): [True: 6.51k, False: 2.27k] ------------------ 1032| 7.07k| SKIP_FIRST_COMPARE: 1033| 7.07k| if (++A == endA) return false; ------------------ | Branch (1033:17): [True: 43, False: 7.03k] ------------------ 1034| 7.07k| } 1035| 4.28k| while (*A > *B) { ------------------ | Branch (1035:16): [True: 2.01k, False: 2.26k] ------------------ 1036| 2.01k| if (++B == endB) return false; ------------------ | Branch (1036:17): [True: 6, False: 2.01k] ------------------ 1037| 2.01k| } 1038| 2.26k| if (*A == *B) { ------------------ | Branch (1038:13): [True: 1.70k, False: 565] ------------------ 1039| 1.70k| return true; 1040| 1.70k| } else { 1041| 565| goto SKIP_FIRST_COMPARE; 1042| 565| } 1043| 2.26k| } 1044| 0| return false; // NOTREACHED 1045| 1.75k|} union_uint16: 1105| 4.61k| size_t size_2, uint16_t *buffer) { 1106| 4.61k| size_t pos = 0, idx_1 = 0, idx_2 = 0; 1107| | 1108| 4.61k| if (0 == size_2) { ------------------ | Branch (1108:9): [True: 63, False: 4.54k] ------------------ 1109| 63| memmove(buffer, set_1, size_1 * sizeof(uint16_t)); 1110| 63| return size_1; 1111| 63| } 1112| 4.54k| if (0 == size_1) { ------------------ | Branch (1112:9): [True: 0, False: 4.54k] ------------------ 1113| 0| memmove(buffer, set_2, size_2 * sizeof(uint16_t)); 1114| 0| return size_2; 1115| 0| } 1116| | 1117| 4.54k| uint16_t val_1 = set_1[idx_1], val_2 = set_2[idx_2]; 1118| | 1119| 219k| while (true) { ------------------ | Branch (1119:12): [True: 219k, Folded] ------------------ 1120| 219k| if (val_1 < val_2) { ------------------ | Branch (1120:13): [True: 7.65k, False: 211k] ------------------ 1121| 7.65k| buffer[pos++] = val_1; 1122| 7.65k| ++idx_1; 1123| 7.65k| if (idx_1 >= size_1) break; ------------------ | Branch (1123:17): [True: 376, False: 7.28k] ------------------ 1124| 7.28k| val_1 = set_1[idx_1]; 1125| 211k| } else if (val_2 < val_1) { ------------------ | Branch (1125:20): [True: 179k, False: 31.8k] ------------------ 1126| 179k| buffer[pos++] = val_2; 1127| 179k| ++idx_2; 1128| 179k| if (idx_2 >= size_2) break; ------------------ | Branch (1128:17): [True: 177, False: 179k] ------------------ 1129| 179k| val_2 = set_2[idx_2]; 1130| 179k| } else { 1131| 31.8k| buffer[pos++] = val_1; 1132| 31.8k| ++idx_1; 1133| 31.8k| ++idx_2; 1134| 31.8k| if (idx_1 >= size_1 || idx_2 >= size_2) break; ------------------ | Branch (1134:17): [True: 3.86k, False: 27.9k] | Branch (1134:36): [True: 131, False: 27.8k] ------------------ 1135| 27.8k| val_1 = set_1[idx_1]; 1136| 27.8k| val_2 = set_2[idx_2]; 1137| 27.8k| } 1138| 219k| } 1139| | 1140| 4.54k| if (idx_1 < size_1) { ------------------ | Branch (1140:9): [True: 308, False: 4.24k] ------------------ 1141| 308| const size_t n_elems = size_1 - idx_1; 1142| 308| memmove(buffer + pos, set_1 + idx_1, n_elems * sizeof(uint16_t)); 1143| 308| pos += n_elems; 1144| 4.24k| } else if (idx_2 < size_2) { ------------------ | Branch (1144:16): [True: 1.80k, False: 2.44k] ------------------ 1145| 1.80k| const size_t n_elems = size_2 - idx_2; 1146| 1.80k| memmove(buffer + pos, set_2 + idx_2, n_elems * sizeof(uint16_t)); 1147| 1.80k| pos += n_elems; 1148| 1.80k| } 1149| | 1150| 4.54k| return pos; 1151| 4.54k|} difference_uint16: 1154| 3.54k| int length2, uint16_t *a_out) { 1155| 3.54k| int out_card = 0; 1156| 3.54k| int k1 = 0, k2 = 0; 1157| 3.54k| if (length1 == 0) return 0; ------------------ | Branch (1157:9): [True: 0, False: 3.54k] ------------------ 1158| 3.54k| if (length2 == 0) { ------------------ | Branch (1158:9): [True: 0, False: 3.54k] ------------------ 1159| 0| if (a1 != a_out) memcpy(a_out, a1, sizeof(uint16_t) * length1); ------------------ | Branch (1159:13): [True: 0, False: 0] ------------------ 1160| 0| return length1; 1161| 0| } 1162| 3.54k| uint16_t s1 = a1[k1]; 1163| 3.54k| uint16_t s2 = a2[k2]; 1164| 449k| while (true) { ------------------ | Branch (1164:12): [True: 449k, Folded] ------------------ 1165| 449k| if (s1 < s2) { ------------------ | Branch (1165:13): [True: 222k, False: 227k] ------------------ 1166| 222k| a_out[out_card++] = s1; 1167| 222k| ++k1; 1168| 222k| if (k1 >= length1) { ------------------ | Branch (1168:17): [True: 0, False: 222k] ------------------ 1169| 0| break; 1170| 0| } 1171| 222k| s1 = a1[k1]; 1172| 227k| } else if (s1 == s2) { ------------------ | Branch (1172:20): [True: 227k, False: 0] ------------------ 1173| 227k| ++k1; 1174| 227k| ++k2; 1175| 227k| if (k1 >= length1) { ------------------ | Branch (1175:17): [True: 2.29k, False: 224k] ------------------ 1176| 2.29k| break; 1177| 2.29k| } 1178| 224k| if (k2 >= length2) { ------------------ | Branch (1178:17): [True: 1.24k, False: 223k] ------------------ 1179| 1.24k| memmove(a_out + out_card, a1 + k1, 1180| 1.24k| sizeof(uint16_t) * (length1 - k1)); 1181| 1.24k| return out_card + length1 - k1; 1182| 1.24k| } 1183| 223k| s1 = a1[k1]; 1184| 223k| s2 = a2[k2]; 1185| 223k| } else { // if (val1>val2) 1186| 0| ++k2; 1187| 0| if (k2 >= length2) { ------------------ | Branch (1187:17): [True: 0, False: 0] ------------------ 1188| 0| memmove(a_out + out_card, a1 + k1, 1189| 0| sizeof(uint16_t) * (length1 - k1)); 1190| 0| return out_card + length1 - k1; 1191| 0| } 1192| 0| s2 = a2[k2]; 1193| 0| } 1194| 449k| } 1195| 2.29k| return out_card; 1196| 3.54k|} xor_uint16: 1199| 8.04k| const uint16_t *array_2, int32_t card_2, uint16_t *out) { 1200| 8.04k| int32_t pos1 = 0, pos2 = 0, pos_out = 0; 1201| 221k| while (pos1 < card_1 && pos2 < card_2) { ------------------ | Branch (1201:12): [True: 214k, False: 7.44k] | Branch (1201:29): [True: 213k, False: 600] ------------------ 1202| 213k| const uint16_t v1 = array_1[pos1]; 1203| 213k| const uint16_t v2 = array_2[pos2]; 1204| 213k| if (v1 == v2) { ------------------ | Branch (1204:13): [True: 40.0k, False: 173k] ------------------ 1205| 40.0k| ++pos1; 1206| 40.0k| ++pos2; 1207| 40.0k| continue; 1208| 40.0k| } 1209| 173k| if (v1 < v2) { ------------------ | Branch (1209:13): [True: 29.6k, False: 143k] ------------------ 1210| 29.6k| out[pos_out++] = v1; 1211| 29.6k| ++pos1; 1212| 143k| } else { 1213| 143k| out[pos_out++] = v2; 1214| 143k| ++pos2; 1215| 143k| } 1216| 173k| } 1217| 8.04k| if (pos1 < card_1) { ------------------ | Branch (1217:9): [True: 600, False: 7.44k] ------------------ 1218| 600| const size_t n_elems = card_1 - pos1; 1219| 600| memcpy(out + pos_out, array_1 + pos1, n_elems * sizeof(uint16_t)); 1220| 600| pos_out += (int32_t)n_elems; 1221| 7.44k| } else if (pos2 < card_2) { ------------------ | Branch (1221:16): [True: 2.31k, False: 5.13k] ------------------ 1222| 2.31k| const size_t n_elems = card_2 - pos2; 1223| 2.31k| memcpy(out + pos_out, array_2 + pos2, n_elems * sizeof(uint16_t)); 1224| 2.31k| pos_out += (int32_t)n_elems; 1225| 2.31k| } 1226| 8.04k| return pos_out; 1227| 8.04k|} union_vector16: 1654| 4.61k| uint16_t *output) { 1655| 4.61k| if ((length1 < 8) || (length2 < 8)) { ------------------ | Branch (1655:9): [True: 1.06k, False: 3.54k] | Branch (1655:26): [True: 0, False: 3.54k] ------------------ 1656| 1.06k| return (uint32_t)union_uint16(array1, length1, array2, length2, output); 1657| 1.06k| } 1658| 3.54k| __m128i vA, vB, V, vecMin, vecMax; 1659| 3.54k| __m128i laststore; 1660| 3.54k| uint16_t *initoutput = output; 1661| 3.54k| uint32_t len1 = length1 / 8; 1662| 3.54k| uint32_t len2 = length2 / 8; 1663| 3.54k| uint32_t pos1 = 0; 1664| 3.54k| uint32_t pos2 = 0; 1665| | // we start the machine 1666| 3.54k| vA = _mm_lddqu_si128((const __m128i *)array1 + pos1); 1667| 3.54k| pos1++; 1668| 3.54k| vB = _mm_lddqu_si128((const __m128i *)array2 + pos2); 1669| 3.54k| pos2++; 1670| 3.54k| sse_merge(&vA, &vB, &vecMin, &vecMax); 1671| 3.54k| laststore = _mm_set1_epi16(-1); 1672| 3.54k| output += store_unique(laststore, vecMin, output); 1673| 3.54k| laststore = vecMin; 1674| 3.54k| if ((pos1 < len1) && (pos2 < len2)) { ------------------ | Branch (1674:9): [True: 3.04k, False: 500] | Branch (1674:26): [True: 3.04k, False: 0] ------------------ 1675| 3.04k| uint16_t curA, curB; 1676| 3.04k| curA = array1[8 * pos1]; 1677| 3.04k| curB = array2[8 * pos2]; 1678| 105k| while (true) { ------------------ | Branch (1678:16): [True: 105k, Folded] ------------------ 1679| 105k| if (curA <= curB) { ------------------ | Branch (1679:17): [True: 44.8k, False: 60.8k] ------------------ 1680| 44.8k| V = _mm_lddqu_si128((const __m128i *)array1 + pos1); 1681| 44.8k| pos1++; 1682| 44.8k| if (pos1 < len1) { ------------------ | Branch (1682:21): [True: 42.4k, False: 2.32k] ------------------ 1683| 42.4k| curA = array1[8 * pos1]; 1684| 42.4k| } else { 1685| 2.32k| break; 1686| 2.32k| } 1687| 60.8k| } else { 1688| 60.8k| V = _mm_lddqu_si128((const __m128i *)array2 + pos2); 1689| 60.8k| pos2++; 1690| 60.8k| if (pos2 < len2) { ------------------ | Branch (1690:21): [True: 60.1k, False: 719] ------------------ 1691| 60.1k| curB = array2[8 * pos2]; 1692| 60.1k| } else { 1693| 719| break; 1694| 719| } 1695| 60.8k| } 1696| 102k| sse_merge(&V, &vecMax, &vecMin, &vecMax); 1697| 102k| output += store_unique(laststore, vecMin, output); 1698| 102k| laststore = vecMin; 1699| 102k| } 1700| 3.04k| sse_merge(&V, &vecMax, &vecMin, &vecMax); 1701| 3.04k| output += store_unique(laststore, vecMin, output); 1702| 3.04k| laststore = vecMin; 1703| 3.04k| } 1704| | // we finish the rest off using a scalar algorithm 1705| | // could be improved? 1706| | // 1707| | // copy the small end on a tmp buffer 1708| 3.54k| uint32_t len = (uint32_t)(output - initoutput); 1709| 3.54k| uint16_t buffer[16]; 1710| 3.54k| uint32_t leftoversize = store_unique(laststore, vecMax, buffer); 1711| 3.54k| if (pos1 == len1) { ------------------ | Branch (1711:9): [True: 2.82k, False: 719] ------------------ 1712| 2.82k| memcpy(buffer + leftoversize, array1 + 8 * pos1, 1713| 2.82k| (length1 - 8 * len1) * sizeof(uint16_t)); 1714| 2.82k| leftoversize += length1 - 8 * len1; 1715| 2.82k| qsort(buffer, leftoversize, sizeof(uint16_t), uint16_compare); 1716| | 1717| 2.82k| leftoversize = unique(buffer, leftoversize); 1718| 2.82k| len += (uint32_t)union_uint16(buffer, leftoversize, array2 + 8 * pos2, 1719| 2.82k| length2 - 8 * pos2, output); 1720| 2.82k| } else { 1721| 719| memcpy(buffer + leftoversize, array2 + 8 * pos2, 1722| 719| (length2 - 8 * len2) * sizeof(uint16_t)); 1723| 719| leftoversize += length2 - 8 * len2; 1724| 719| qsort(buffer, leftoversize, sizeof(uint16_t), uint16_compare); 1725| 719| leftoversize = unique(buffer, leftoversize); 1726| 719| len += (uint32_t)union_uint16(buffer, leftoversize, array1 + 8 * pos1, 1727| 719| length1 - 8 * pos1, output); 1728| 719| } 1729| 3.54k| return len; 1730| 4.61k|} xor_vector16: 1778| 8.05k| uint16_t *output) { 1779| 8.05k| if ((length1 < 8) || (length2 < 8)) { ------------------ | Branch (1779:9): [True: 3.27k, False: 4.78k] | Branch (1779:26): [True: 363, False: 4.42k] ------------------ 1780| 3.63k| return xor_uint16(array1, length1, array2, length2, output); 1781| 3.63k| } 1782| 4.42k| __m128i vA, vB, V, vecMin, vecMax; 1783| 4.42k| __m128i laststore; 1784| 4.42k| uint16_t *initoutput = output; 1785| 4.42k| uint32_t len1 = length1 / 8; 1786| 4.42k| uint32_t len2 = length2 / 8; 1787| 4.42k| uint32_t pos1 = 0; 1788| 4.42k| uint32_t pos2 = 0; 1789| | // we start the machine 1790| 4.42k| vA = _mm_lddqu_si128((const __m128i *)array1 + pos1); 1791| 4.42k| pos1++; 1792| 4.42k| vB = _mm_lddqu_si128((const __m128i *)array2 + pos2); 1793| 4.42k| pos2++; 1794| 4.42k| sse_merge(&vA, &vB, &vecMin, &vecMax); 1795| 4.42k| laststore = _mm_set1_epi16(-1); 1796| 4.42k| uint16_t buffer[17]; 1797| 4.42k| output += store_unique_xor(laststore, vecMin, output); 1798| | 1799| 4.42k| laststore = vecMin; 1800| 4.42k| if ((pos1 < len1) && (pos2 < len2)) { ------------------ | Branch (1800:9): [True: 4.09k, False: 326] | Branch (1800:26): [True: 3.99k, False: 108] ------------------ 1801| 3.99k| uint16_t curA, curB; 1802| 3.99k| curA = array1[8 * pos1]; 1803| 3.99k| curB = array2[8 * pos2]; 1804| 193k| while (true) { ------------------ | Branch (1804:16): [True: 193k, Folded] ------------------ 1805| 193k| if (curA <= curB) { ------------------ | Branch (1805:17): [True: 92.3k, False: 101k] ------------------ 1806| 92.3k| V = _mm_lddqu_si128((const __m128i *)array1 + pos1); 1807| 92.3k| pos1++; 1808| 92.3k| if (pos1 < len1) { ------------------ | Branch (1808:21): [True: 89.7k, False: 2.62k] ------------------ 1809| 89.7k| curA = array1[8 * pos1]; 1810| 89.7k| } else { 1811| 2.62k| break; 1812| 2.62k| } 1813| 101k| } else { 1814| 101k| V = _mm_lddqu_si128((const __m128i *)array2 + pos2); 1815| 101k| pos2++; 1816| 101k| if (pos2 < len2) { ------------------ | Branch (1816:21): [True: 100k, False: 1.36k] ------------------ 1817| 100k| curB = array2[8 * pos2]; 1818| 100k| } else { 1819| 1.36k| break; 1820| 1.36k| } 1821| 101k| } 1822| 189k| sse_merge(&V, &vecMax, &vecMin, &vecMax); 1823| | // conditionally stores the last value of laststore as well as all 1824| | // but the 1825| | // last value of vecMin 1826| 189k| output += store_unique_xor(laststore, vecMin, output); 1827| 189k| laststore = vecMin; 1828| 189k| } 1829| 3.99k| sse_merge(&V, &vecMax, &vecMin, &vecMax); 1830| | // conditionally stores the last value of laststore as well as all but 1831| | // the 1832| | // last value of vecMin 1833| 3.99k| output += store_unique_xor(laststore, vecMin, output); 1834| 3.99k| laststore = vecMin; 1835| 3.99k| } 1836| 4.42k| uint32_t len = (uint32_t)(output - initoutput); 1837| | 1838| | // we finish the rest off using a scalar algorithm 1839| | // could be improved? 1840| | // conditionally stores the last value of laststore as well as all but the 1841| | // last value of vecMax, 1842| | // we store to "buffer" 1843| 4.42k| int leftoversize = store_unique_xor(laststore, vecMax, buffer); 1844| 4.42k| uint16_t vec7 = (uint16_t)_mm_extract_epi16(vecMax, 7); 1845| 4.42k| uint16_t vec6 = (uint16_t)_mm_extract_epi16(vecMax, 6); 1846| 4.42k| if (vec7 != vec6) buffer[leftoversize++] = vec7; ------------------ | Branch (1846:9): [True: 4.37k, False: 49] ------------------ 1847| 4.42k| if (pos1 == len1) { ------------------ | Branch (1847:9): [True: 2.94k, False: 1.47k] ------------------ 1848| 2.94k| memcpy(buffer + leftoversize, array1 + 8 * pos1, 1849| 2.94k| (length1 - 8 * len1) * sizeof(uint16_t)); 1850| 2.94k| leftoversize += length1 - 8 * len1; 1851| 2.94k| if (leftoversize == 0) { // trivial case ------------------ | Branch (1851:13): [True: 9, False: 2.94k] ------------------ 1852| 9| memcpy(output, array2 + 8 * pos2, 1853| 9| (length2 - 8 * pos2) * sizeof(uint16_t)); 1854| 9| len += (length2 - 8 * pos2); 1855| 2.94k| } else { 1856| 2.94k| qsort(buffer, leftoversize, sizeof(uint16_t), uint16_compare); 1857| 2.94k| leftoversize = unique_xor(buffer, leftoversize); 1858| 2.94k| len += xor_uint16(buffer, leftoversize, array2 + 8 * pos2, 1859| 2.94k| length2 - 8 * pos2, output); 1860| 2.94k| } 1861| 2.94k| } else { 1862| 1.47k| memcpy(buffer + leftoversize, array2 + 8 * pos2, 1863| 1.47k| (length2 - 8 * len2) * sizeof(uint16_t)); 1864| 1.47k| leftoversize += length2 - 8 * len2; 1865| 1.47k| if (leftoversize == 0) { // trivial case ------------------ | Branch (1865:13): [True: 4, False: 1.47k] ------------------ 1866| 4| memcpy(output, array1 + 8 * pos1, 1867| 4| (length1 - 8 * pos1) * sizeof(uint16_t)); 1868| 4| len += (length1 - 8 * pos1); 1869| 1.47k| } else { 1870| 1.47k| qsort(buffer, leftoversize, sizeof(uint16_t), uint16_compare); 1871| 1.47k| leftoversize = unique_xor(buffer, leftoversize); 1872| 1.47k| len += xor_uint16(buffer, leftoversize, array1 + 8 * pos1, 1873| 1.47k| length1 - 8 * pos1, output); 1874| 1.47k| } 1875| 1.47k| } 1876| 4.42k| return len; 1877| 8.05k|} fast_union_uint16: 1980| 4.61k| uint16_t *buffer) { 1981| 4.61k|#if CROARING_IS_X64 1982| 4.61k| if (croaring_hardware_support() & ROARING_SUPPORTS_AVX2) { ------------------ | Branch (1982:9): [True: 4.61k, False: 0] ------------------ 1983| | // compute union with smallest array first 1984| 4.61k| if (size_1 < size_2) { ------------------ | Branch (1984:13): [True: 2.36k, False: 2.24k] ------------------ 1985| 2.36k| return union_vector16(set_1, (uint32_t)size_1, set_2, 1986| 2.36k| (uint32_t)size_2, buffer); 1987| 2.36k| } else { 1988| 2.24k| return union_vector16(set_2, (uint32_t)size_2, set_1, 1989| 2.24k| (uint32_t)size_1, buffer); 1990| 2.24k| } 1991| 4.61k| } else { 1992| | // compute union with smallest array first 1993| 0| if (size_1 < size_2) { ------------------ | Branch (1993:13): [True: 0, False: 0] ------------------ 1994| 0| return union_uint16(set_1, size_1, set_2, size_2, buffer); 1995| 0| } else { 1996| 0| return union_uint16(set_2, size_2, set_1, size_1, buffer); 1997| 0| } 1998| 0| } 1999| |#else 2000| | // compute union with smallest array first 2001| | if (size_1 < size_2) { 2002| | return union_uint16(set_1, size_1, set_2, size_2, buffer); 2003| | } else { 2004| | return union_uint16(set_2, size_2, set_1, size_1, buffer); 2005| | } 2006| |#endif 2007| 4.61k|} memequals: 2111| 7.38k|bool memequals(const void *s1, const void *s2, size_t n) { 2112| 7.38k| if (n == 0) { ------------------ | Branch (2112:9): [True: 0, False: 7.38k] ------------------ 2113| 0| return true; 2114| 0| } 2115| 7.38k|#if CROARING_IS_X64 2116| 7.38k| int support = croaring_hardware_support(); 2117| 7.38k|#if CROARING_COMPILER_SUPPORTS_AVX512 2118| 7.38k| if (support & ROARING_SUPPORTS_AVX512) { ------------------ | Branch (2118:9): [True: 0, False: 7.38k] ------------------ 2119| 0| return _avx512_memequals(s1, s2, n); 2120| 0| } else 2121| 7.38k|#endif // CROARING_COMPILER_SUPPORTS_AVX512 2122| 7.38k| if (support & ROARING_SUPPORTS_AVX2) { ------------------ | Branch (2122:13): [True: 7.38k, False: 0] ------------------ 2123| 7.38k| return _avx2_memequals(s1, s2, n); 2124| 7.38k| } else { 2125| 0| return memcmp(s1, s2, n) == 0; 2126| 0| } 2127| |#else 2128| | return memcmp(s1, s2, n) == 0; 2129| |#endif 2130| 7.38k|} array_util.c:binarySearch4: 804| 1.39k| int32_t *index4) { 805| 1.39k| const uint16_t *base1 = array; 806| 1.39k| const uint16_t *base2 = array; 807| 1.39k| const uint16_t *base3 = array; 808| 1.39k| const uint16_t *base4 = array; 809| 1.39k| if (n == 0) return; ------------------ | Branch (809:9): [True: 0, False: 1.39k] ------------------ 810| 17.4k| while (n > 1) { ------------------ | Branch (810:12): [True: 16.0k, False: 1.39k] ------------------ 811| 16.0k| int32_t half = n >> 1; 812| 16.0k| base1 = (base1[half] < target1) ? &base1[half] : base1; ------------------ | Branch (812:17): [True: 1.21k, False: 14.8k] ------------------ 813| 16.0k| base2 = (base2[half] < target2) ? &base2[half] : base2; ------------------ | Branch (813:17): [True: 2.34k, False: 13.6k] ------------------ 814| 16.0k| base3 = (base3[half] < target3) ? &base3[half] : base3; ------------------ | Branch (814:17): [True: 3.06k, False: 12.9k] ------------------ 815| 16.0k| base4 = (base4[half] < target4) ? &base4[half] : base4; ------------------ | Branch (815:17): [True: 3.46k, False: 12.5k] ------------------ 816| 16.0k| n -= half; 817| 16.0k| } 818| 1.39k| *index1 = (int32_t)((*base1 < target1) + base1 - array); 819| 1.39k| *index2 = (int32_t)((*base2 < target2) + base2 - array); 820| 1.39k| *index3 = (int32_t)((*base3 < target3) + base3 - array); 821| 1.39k| *index4 = (int32_t)((*base4 < target4) + base4 - array); 822| 1.39k|} array_util.c:binarySearch2: 833| 414| uint16_t target2, int32_t *index1, int32_t *index2) { 834| 414| const uint16_t *base1 = array; 835| 414| const uint16_t *base2 = array; 836| 414| if (n == 0) return; ------------------ | Branch (836:9): [True: 0, False: 414] ------------------ 837| 3.93k| while (n > 1) { ------------------ | Branch (837:12): [True: 3.51k, False: 414] ------------------ 838| 3.51k| int32_t half = n >> 1; 839| 3.51k| base1 = (base1[half] < target1) ? &base1[half] : base1; ------------------ | Branch (839:17): [True: 881, False: 2.63k] ------------------ 840| 3.51k| base2 = (base2[half] < target2) ? &base2[half] : base2; ------------------ | Branch (840:17): [True: 1.39k, False: 2.12k] ------------------ 841| 3.51k| n -= half; 842| 3.51k| } 843| 414| *index1 = (int32_t)((*base1 < target1) + base1 - array); 844| 414| *index2 = (int32_t)((*base2 < target2) + base2 - array); 845| 414|} array_util.c:sse_merge: 1244| 307k| __m128i *vecMin, __m128i *vecMax) { // output 1245| 307k| __m128i vecTmp; 1246| 307k| vecTmp = _mm_min_epu16(*vInput1, *vInput2); 1247| 307k| *vecMax = _mm_max_epu16(*vInput1, *vInput2); 1248| 307k| vecTmp = _mm_alignr_epi8(vecTmp, vecTmp, 2); 1249| 307k| *vecMin = _mm_min_epu16(vecTmp, *vecMax); 1250| 307k| *vecMax = _mm_max_epu16(vecTmp, *vecMax); 1251| 307k| vecTmp = _mm_alignr_epi8(*vecMin, *vecMin, 2); 1252| 307k| *vecMin = _mm_min_epu16(vecTmp, *vecMax); 1253| 307k| *vecMax = _mm_max_epu16(vecTmp, *vecMax); 1254| 307k| vecTmp = _mm_alignr_epi8(*vecMin, *vecMin, 2); 1255| 307k| *vecMin = _mm_min_epu16(vecTmp, *vecMax); 1256| 307k| *vecMax = _mm_max_epu16(vecTmp, *vecMax); 1257| 307k| vecTmp = _mm_alignr_epi8(*vecMin, *vecMin, 2); 1258| 307k| *vecMin = _mm_min_epu16(vecTmp, *vecMax); 1259| 307k| *vecMax = _mm_max_epu16(vecTmp, *vecMax); 1260| 307k| vecTmp = _mm_alignr_epi8(*vecMin, *vecMin, 2); 1261| 307k| *vecMin = _mm_min_epu16(vecTmp, *vecMax); 1262| 307k| *vecMax = _mm_max_epu16(vecTmp, *vecMax); 1263| 307k| vecTmp = _mm_alignr_epi8(*vecMin, *vecMin, 2); 1264| 307k| *vecMin = _mm_min_epu16(vecTmp, *vecMax); 1265| 307k| *vecMax = _mm_max_epu16(vecTmp, *vecMax); 1266| 307k| vecTmp = _mm_alignr_epi8(*vecMin, *vecMin, 2); 1267| 307k| *vecMin = _mm_min_epu16(vecTmp, *vecMax); 1268| 307k| *vecMax = _mm_max_epu16(vecTmp, *vecMax); 1269| | *vecMin = _mm_alignr_epi8(*vecMin, *vecMin, 2); 1270| 307k|} array_util.c:store_unique: 1619| 112k|static inline int store_unique(__m128i old, __m128i newval, uint16_t *output) { 1620| 112k| __m128i vecTmp = _mm_alignr_epi8(newval, old, 16 - 2); 1621| | // lots of high latency instructions follow (optimize?) 1622| 112k| int M = _mm_movemask_epi8( 1623| 112k| _mm_packs_epi16(_mm_cmpeq_epi16(vecTmp, newval), _mm_setzero_si128())); 1624| 112k| int numberofnewvalues = 8 - _mm_popcnt_u32(M); 1625| 112k| __m128i key = _mm_lddqu_si128((const __m128i *)uniqshuf + M); 1626| 112k| __m128i val = _mm_shuffle_epi8(newval, key); 1627| 112k| _mm_storeu_si128((__m128i *)output, val); 1628| 112k| return numberofnewvalues; 1629| 112k|} array_util.c:uint16_compare: 1645| 126k|static int uint16_compare(const void *a, const void *b) { 1646| 126k| return (*(uint16_t *)a - *(uint16_t *)b); 1647| 126k|} array_util.c:unique: 1634| 3.54k|static inline uint32_t unique(uint16_t *out, uint32_t len) { 1635| 3.54k| uint32_t pos = 1; 1636| 36.3k| for (uint32_t i = 1; i < len; ++i) { ------------------ | Branch (1636:26): [True: 32.7k, False: 3.54k] ------------------ 1637| 32.7k| if (out[i] != out[i - 1]) { ------------------ | Branch (1637:13): [True: 32.6k, False: 95] ------------------ 1638| 32.6k| out[pos++] = out[i]; 1639| 32.6k| } 1640| 32.7k| } 1641| 3.54k| return pos; 1642| 3.54k|} array_util.c:store_unique_xor: 1746| 202k| uint16_t *output) { 1747| 202k| __m128i vecTmp1 = _mm_alignr_epi8(newval, old, 16 - 4); 1748| | __m128i vecTmp2 = _mm_alignr_epi8(newval, old, 16 - 2); 1749| 202k| __m128i equalleft = _mm_cmpeq_epi16(vecTmp2, vecTmp1); 1750| 202k| __m128i equalright = _mm_cmpeq_epi16(vecTmp2, newval); 1751| 202k| __m128i equalleftoright = _mm_or_si128(equalleft, equalright); 1752| 202k| int M = _mm_movemask_epi8( 1753| 202k| _mm_packs_epi16(equalleftoright, _mm_setzero_si128())); 1754| 202k| int numberofnewvalues = 8 - _mm_popcnt_u32(M); 1755| 202k| __m128i key = _mm_lddqu_si128((const __m128i *)uniqshuf + M); 1756| 202k| __m128i val = _mm_shuffle_epi8(vecTmp2, key); 1757| 202k| _mm_storeu_si128((__m128i *)output, val); 1758| 202k| return numberofnewvalues; 1759| 202k|} array_util.c:unique_xor: 1764| 4.41k|static inline uint32_t unique_xor(uint16_t *out, uint32_t len) { 1765| 4.41k| uint32_t pos = 1; 1766| 43.7k| for (uint32_t i = 1; i < len; ++i) { ------------------ | Branch (1766:26): [True: 39.3k, False: 4.41k] ------------------ 1767| 39.3k| if (out[i] != out[i - 1]) { ------------------ | Branch (1767:13): [True: 39.1k, False: 230] ------------------ 1768| 39.1k| out[pos++] = out[i]; 1769| 39.1k| } else 1770| 230| pos--; // if it is identical to previous, delete it 1771| 39.3k| } 1772| 4.41k| return pos; 1773| 4.41k|} array_util.c:_avx2_memequals: 2069| 7.38k|static inline bool _avx2_memequals(const void *s1, const void *s2, size_t n) { 2070| 7.38k| const uint8_t *ptr1 = (const uint8_t *)s1; 2071| 7.38k| const uint8_t *ptr2 = (const uint8_t *)s2; 2072| 7.38k| const uint8_t *end1 = ptr1 + n; 2073| 7.38k| const uint8_t *end8 = ptr1 + n / 8 * 8; 2074| 7.38k| const uint8_t *end32 = ptr1 + n / 32 * 32; 2075| | 2076| 145k| while (ptr1 < end32) { ------------------ | Branch (2076:12): [True: 138k, False: 7.31k] ------------------ 2077| 138k| __m256i r1 = _mm256_loadu_si256((const __m256i *)ptr1); 2078| 138k| __m256i r2 = _mm256_loadu_si256((const __m256i *)ptr2); 2079| 138k| int mask = _mm256_movemask_epi8(_mm256_cmpeq_epi8(r1, r2)); 2080| 138k| if ((uint32_t)mask != UINT32_MAX) { ------------------ | Branch (2080:13): [True: 63, False: 138k] ------------------ 2081| 63| return false; 2082| 63| } 2083| 138k| ptr1 += 32; 2084| 138k| ptr2 += 32; 2085| 138k| } 2086| | 2087| 13.1k| while (ptr1 < end8) { ------------------ | Branch (2087:12): [True: 5.95k, False: 7.23k] ------------------ 2088| 5.95k| uint64_t v1, v2; 2089| 5.95k| memcpy(&v1, ptr1, sizeof(uint64_t)); 2090| 5.95k| memcpy(&v2, ptr2, sizeof(uint64_t)); 2091| 5.95k| if (v1 != v2) { ------------------ | Branch (2091:13): [True: 84, False: 5.87k] ------------------ 2092| 84| return false; 2093| 84| } 2094| 5.87k| ptr1 += 8; 2095| 5.87k| ptr2 += 8; 2096| 5.87k| } 2097| | 2098| 25.6k| while (ptr1 < end1) { ------------------ | Branch (2098:12): [True: 18.4k, False: 7.20k] ------------------ 2099| 18.4k| if (*ptr1 != *ptr2) { ------------------ | Branch (2099:13): [True: 34, False: 18.4k] ------------------ 2100| 34| return false; 2101| 34| } 2102| 18.4k| ptr1++; 2103| 18.4k| ptr2++; 2104| 18.4k| } 2105| | 2106| 7.20k| return true; 2107| 7.23k|} bitset_extract_setbits_avx2: 687| 1.03k| uint32_t base) { 688| 1.03k| if (outcapacity == 0) return 0; ------------------ | Branch (688:9): [True: 0, False: 1.03k] ------------------ 689| 1.03k| uint32_t *initout = out; 690| 1.03k| __m256i baseVec = _mm256_set1_epi32(base - 1); 691| 1.03k| __m256i incVec = _mm256_set1_epi32(64); 692| 1.03k| __m256i add8 = _mm256_set1_epi32(8); 693| 1.03k| uint32_t *safeout = out + outcapacity; 694| 1.03k| size_t i = 0; 695| 873k| for (; (i < length) && (out + 64 <= safeout); ++i) { ------------------ | Branch (695:12): [True: 873k, False: 283] | Branch (695:28): [True: 872k, False: 755] ------------------ 696| 872k| uint64_t w = words[i]; 697| 872k| if (w == 0) { ------------------ | Branch (697:13): [True: 117k, False: 755k] ------------------ 698| 117k| baseVec = _mm256_add_epi32(baseVec, incVec); 699| 755k| } else { 700| 3.77M| for (int k = 0; k < 4; ++k) { ------------------ | Branch (700:29): [True: 3.02M, False: 755k] ------------------ 701| 3.02M| uint8_t byteA = (uint8_t)w; 702| 3.02M| uint8_t byteB = (uint8_t)(w >> 8); 703| 3.02M| w >>= 16; 704| 3.02M| __m256i vecA = 705| 3.02M| _mm256_loadu_si256((const __m256i *)vecDecodeTable[byteA]); 706| 3.02M| __m256i vecB = 707| 3.02M| _mm256_loadu_si256((const __m256i *)vecDecodeTable[byteB]); 708| 3.02M| uint8_t advanceA = lengthTable[byteA]; 709| 3.02M| uint8_t advanceB = lengthTable[byteB]; 710| 3.02M| vecA = _mm256_add_epi32(baseVec, vecA); 711| 3.02M| baseVec = _mm256_add_epi32(baseVec, add8); 712| 3.02M| vecB = _mm256_add_epi32(baseVec, vecB); 713| 3.02M| baseVec = _mm256_add_epi32(baseVec, add8); 714| 3.02M| _mm256_storeu_si256((__m256i *)out, vecA); 715| 3.02M| out += advanceA; 716| 3.02M| _mm256_storeu_si256((__m256i *)out, vecB); 717| 3.02M| out += advanceB; 718| 3.02M| } 719| 755k| } 720| 872k| } 721| 1.03k| base += i * 64; 722| 14.5k| for (; (i < length) && (out < safeout); ++i) { ------------------ | Branch (722:12): [True: 14.1k, False: 374] | Branch (722:28): [True: 13.5k, False: 664] ------------------ 723| 13.5k| uint64_t w = words[i]; 724| 32.5k| while ((w != 0) && (out < safeout)) { ------------------ | Branch (724:16): [True: 19.0k, False: 13.5k] | Branch (724:28): [True: 19.0k, False: 0] ------------------ 725| 19.0k| int r = 726| 19.0k| roaring_trailing_zeroes(w); // on x64, should compile to TZCNT 727| 19.0k| uint32_t val = r + base; 728| 19.0k| memcpy(out, &val, 729| 19.0k| sizeof(uint32_t)); // should be compiled as a MOV on x64 730| 19.0k| out++; 731| 19.0k| w &= (w - 1); 732| 19.0k| } 733| 13.5k| base += 64; 734| 13.5k| } 735| 1.03k| return out - initout; 736| 1.03k|} bitset_extract_setbits: 741| 480| uint32_t *out, uint32_t base) { 742| 480| int outpos = 0; 743| 492k| for (size_t i = 0; i < length; ++i) { ------------------ | Branch (743:24): [True: 491k, False: 480] ------------------ 744| 491k| uint64_t w = words[i]; 745| 2.79M| while (w != 0) { ------------------ | Branch (745:16): [True: 2.30M, False: 491k] ------------------ 746| 2.30M| int r = 747| 2.30M| roaring_trailing_zeroes(w); // on x64, should compile to TZCNT 748| 2.30M| uint32_t val = r + base; 749| 2.30M| memcpy(out + outpos, &val, 750| 2.30M| sizeof(uint32_t)); // should be compiled as a MOV on x64 751| 2.30M| outpos++; 752| 2.30M| w &= (w - 1); 753| 2.30M| } 754| 491k| base += 64; 755| 491k| } 756| 480| return outpos; 757| 480|} bitset_extract_setbits_uint16: 852| 3.01k| uint16_t *out, uint16_t base) { 853| 3.01k| int outpos = 0; 854| 3.08M| for (size_t i = 0; i < length; ++i) { ------------------ | Branch (854:24): [True: 3.08M, False: 3.01k] ------------------ 855| 3.08M| uint64_t w = words[i]; 856| 6.02M| while (w != 0) { ------------------ | Branch (856:16): [True: 2.93M, False: 3.08M] ------------------ 857| 2.93M| int r = roaring_trailing_zeroes(w); 858| 2.93M| out[outpos++] = (uint16_t)(r + base); 859| 2.93M| w &= (w - 1); 860| 2.93M| } 861| 3.08M| base += 64; 862| 3.08M| } 863| 3.01k| return outpos; 864| 3.01k|} bitset_clear_list: 1032| 1.75k| uint64_t length) { 1033| 1.75k| if (croaring_hardware_support() & ROARING_SUPPORTS_AVX2) { ------------------ | Branch (1033:9): [True: 1.75k, False: 0] ------------------ 1034| 1.75k| return _asm_bitset_clear_list(words, card, list, length); 1035| 1.75k| } else { 1036| 0| return _scalar_bitset_clear_list(words, card, list, length); 1037| 0| } 1038| 1.75k|} bitset_set_list_withcard: 1041| 1.42k| const uint16_t *list, uint64_t length) { 1042| 1.42k| if (croaring_hardware_support() & ROARING_SUPPORTS_AVX2) { ------------------ | Branch (1042:9): [True: 1.42k, False: 0] ------------------ 1043| 1.42k| return _asm_bitset_set_list_withcard(words, card, list, length); 1044| 1.42k| } else { 1045| 0| return _scalar_bitset_set_list_withcard(words, card, list, length); 1046| 0| } 1047| 1.42k|} bitset_set_list: 1049| 222|void bitset_set_list(uint64_t *words, const uint16_t *list, uint64_t length) { 1050| 222| if (croaring_hardware_support() & ROARING_SUPPORTS_AVX2) { ------------------ | Branch (1050:9): [True: 222, False: 0] ------------------ 1051| 222| _asm_bitset_set_list(words, list, length); 1052| 222| } else { 1053| 0| _scalar_bitset_set_list(words, list, length); 1054| 0| } 1055| 222|} bitset_flip_list_withcard: 1111| 1.52k| const uint16_t *list, uint64_t length) { 1112| 1.52k| uint64_t offset, load, newload, pos, index; 1113| 1.52k| const uint16_t *end = list + length; 1114| 2.73M| while (list != end) { ------------------ | Branch (1114:12): [True: 2.73M, False: 1.52k] ------------------ 1115| 2.73M| pos = *list; 1116| 2.73M| offset = pos >> 6; 1117| 2.73M| index = pos % 64; 1118| 2.73M| load = words[offset]; 1119| 2.73M| newload = load ^ (UINT64_C(1) << index); 1120| | // todo: is a branch here all that bad? 1121| 2.73M| card += 1122| | (1 - 2 * (((UINT64_C(1) << index) & load) >> index)); // +1 or -1 1123| 2.73M| words[offset] = newload; 1124| 2.73M| list++; 1125| 2.73M| } 1126| 1.52k| return card; 1127| 1.52k|} bitset_util.c:_asm_bitset_clear_list: 953| 1.75k| uint64_t length) { 954| 1.75k| uint64_t offset, load, pos; 955| 1.75k| uint64_t shift = 6; 956| 1.75k| const uint16_t *end = list + length; 957| 1.75k| if (!length) return card; ------------------ | Branch (957:9): [True: 0, False: 1.75k] ------------------ 958| | // btr is not available as an intrinsic in GCC 959| 1.75k| __asm volatile( 960| 1.75k| "1:\n" 961| 1.75k| "movzwq (%[list]), %[pos]\n" 962| 1.75k| "shrx %[shift], %[pos], %[offset]\n" 963| 1.75k| "mov (%[words],%[offset],8), %[load]\n" 964| 1.75k| "btr %[pos], %[load]\n" 965| 1.75k| "mov %[load], (%[words],%[offset],8)\n" 966| 1.75k| "sbb $0, %[card]\n" 967| 1.75k| "add $2, %[list]\n" 968| 1.75k| "cmp %[list], %[end]\n" 969| 1.75k| "jnz 1b" 970| 1.75k| : [card] "+&r"(card), [list] "+&r"(list), [load] "=&r"(load), 971| 1.75k| [pos] "=&r"(pos), [offset] "=&r"(offset) 972| 1.75k| : [end] "r"(end), [words] "r"(words), [shift] "r"(shift) 973| 1.75k| : 974| 1.75k| /* clobbers */ "memory"); 975| 1.75k| return card; 976| 1.75k|} bitset_util.c:_asm_bitset_set_list_withcard: 871| 1.42k| uint64_t length) { 872| 1.42k| uint64_t offset, load, pos; 873| 1.42k| uint64_t shift = 6; 874| 1.42k| const uint16_t *end = list + length; 875| 1.42k| if (!length) return card; ------------------ | Branch (875:9): [True: 0, False: 1.42k] ------------------ 876| | // TODO: could unroll for performance, see bitset_set_list 877| | // bts is not available as an intrinsic in GCC 878| 1.42k| __asm volatile( 879| 1.42k| "1:\n" 880| 1.42k| "movzwq (%[list]), %[pos]\n" 881| 1.42k| "shrx %[shift], %[pos], %[offset]\n" 882| 1.42k| "mov (%[words],%[offset],8), %[load]\n" 883| 1.42k| "bts %[pos], %[load]\n" 884| 1.42k| "mov %[load], (%[words],%[offset],8)\n" 885| 1.42k| "sbb $-1, %[card]\n" 886| 1.42k| "add $2, %[list]\n" 887| 1.42k| "cmp %[list], %[end]\n" 888| 1.42k| "jnz 1b" 889| 1.42k| : [card] "+&r"(card), [list] "+&r"(list), [load] "=&r"(load), 890| 1.42k| [pos] "=&r"(pos), [offset] "=&r"(offset) 891| 1.42k| : [end] "r"(end), [words] "r"(words), [shift] "r"(shift)); 892| 1.42k| return card; 893| 1.42k|} bitset_util.c:_asm_bitset_set_list: 896| 222| uint64_t length) { 897| 222| uint64_t pos; 898| 222| const uint16_t *end = list + length; 899| | 900| 222| uint64_t shift = 6; 901| 222| uint64_t offset; 902| 222| uint64_t load; 903| 6.19k| for (; list + 3 < end; list += 4) { ------------------ | Branch (903:12): [True: 5.97k, False: 222] ------------------ 904| 5.97k| pos = list[0]; 905| 5.97k| __asm volatile( 906| 5.97k| "shrx %[shift], %[pos], %[offset]\n" 907| 5.97k| "mov (%[words],%[offset],8), %[load]\n" 908| 5.97k| "bts %[pos], %[load]\n" 909| 5.97k| "mov %[load], (%[words],%[offset],8)" 910| 5.97k| : [load] "=&r"(load), [offset] "=&r"(offset) 911| 5.97k| : [words] "r"(words), [shift] "r"(shift), [pos] "r"(pos)); 912| 5.97k| pos = list[1]; 913| 5.97k| __asm volatile( 914| 5.97k| "shrx %[shift], %[pos], %[offset]\n" 915| 5.97k| "mov (%[words],%[offset],8), %[load]\n" 916| 5.97k| "bts %[pos], %[load]\n" 917| 5.97k| "mov %[load], (%[words],%[offset],8)" 918| 5.97k| : [load] "=&r"(load), [offset] "=&r"(offset) 919| 5.97k| : [words] "r"(words), [shift] "r"(shift), [pos] "r"(pos)); 920| 5.97k| pos = list[2]; 921| 5.97k| __asm volatile( 922| 5.97k| "shrx %[shift], %[pos], %[offset]\n" 923| 5.97k| "mov (%[words],%[offset],8), %[load]\n" 924| 5.97k| "bts %[pos], %[load]\n" 925| 5.97k| "mov %[load], (%[words],%[offset],8)" 926| 5.97k| : [load] "=&r"(load), [offset] "=&r"(offset) 927| 5.97k| : [words] "r"(words), [shift] "r"(shift), [pos] "r"(pos)); 928| 5.97k| pos = list[3]; 929| 5.97k| __asm volatile( 930| 5.97k| "shrx %[shift], %[pos], %[offset]\n" 931| 5.97k| "mov (%[words],%[offset],8), %[load]\n" 932| 5.97k| "bts %[pos], %[load]\n" 933| 5.97k| "mov %[load], (%[words],%[offset],8)" 934| 5.97k| : [load] "=&r"(load), [offset] "=&r"(offset) 935| 5.97k| : [words] "r"(words), [shift] "r"(shift), [pos] "r"(pos)); 936| 5.97k| } 937| | 938| 575| while (list != end) { ------------------ | Branch (938:12): [True: 353, False: 222] ------------------ 939| 353| pos = list[0]; 940| 353| __asm volatile( 941| 353| "shrx %[shift], %[pos], %[offset]\n" 942| 353| "mov (%[words],%[offset],8), %[load]\n" 943| 353| "bts %[pos], %[load]\n" 944| 353| "mov %[load], (%[words],%[offset],8)" 945| 353| : [load] "=&r"(load), [offset] "=&r"(offset) 946| 353| : [words] "r"(words), [shift] "r"(shift), [pos] "r"(pos)); 947| 353| list++; 948| 353| } 949| 222|} array_container_create_given_capacity: 49| 150k|array_container_t *array_container_create_given_capacity(int32_t size) { 50| 150k| array_container_t *container; 51| | 52| 150k| if ((container = (array_container_t *)roaring_malloc( ------------------ | Branch (52:9): [True: 0, False: 150k] ------------------ 53| 150k| sizeof(array_container_t))) == NULL) { 54| 0| return NULL; 55| 0| } 56| | 57| 150k| if (size <= 0) { // we don't want to rely on malloc(0) ------------------ | Branch (57:9): [True: 93.6k, False: 57.1k] ------------------ 58| 93.6k| container->array = NULL; 59| 93.6k| } else if ((container->array = (uint16_t *)roaring_malloc(sizeof(uint16_t) * ------------------ | Branch (59:16): [True: 0, False: 57.1k] ------------------ 60| 57.1k| size)) == NULL) { 61| 0| roaring_free(container); 62| 0| return NULL; 63| 0| } 64| | 65| 150k| container->capacity = size; 66| 150k| container->cardinality = 0; 67| | 68| 150k| return container; 69| 150k|} array_container_create: 72| 26.4k|array_container_t *array_container_create(void) { 73| 26.4k| return array_container_create_given_capacity(ARRAY_DEFAULT_INIT_SIZE); 74| 26.4k|} array_container_create_range: 77| 144|array_container_t *array_container_create_range(uint32_t min, uint32_t max) { 78| 144| array_container_t *answer = 79| 144| array_container_create_given_capacity(max - min + 1); 80| 144| if (answer == NULL) return answer; ------------------ | Branch (80:9): [True: 0, False: 144] ------------------ 81| 144| answer->cardinality = 0; 82| 288| for (uint32_t k = min; k < max; k++) { ------------------ | Branch (82:28): [True: 144, False: 144] ------------------ 83| 144| answer->array[answer->cardinality++] = k; 84| 144| } 85| 144| return answer; 86| 144|} array_container_clone: 90| 23.2k|array_container_t *array_container_clone(const array_container_t *src) { 91| 23.2k| array_container_t *newcontainer = 92| 23.2k| array_container_create_given_capacity(src->capacity); 93| 23.2k| if (newcontainer == NULL) return NULL; ------------------ | Branch (93:9): [True: 0, False: 23.2k] ------------------ 94| | 95| 23.2k| newcontainer->cardinality = src->cardinality; 96| | 97| 23.2k| memcpy(newcontainer->array, src->array, 98| 23.2k| src->cardinality * sizeof(uint16_t)); 99| | 100| 23.2k| return newcontainer; 101| 23.2k|} array_container_shrink_to_fit: 131| 3.80k|int array_container_shrink_to_fit(array_container_t *src) { 132| 3.80k| if (src->cardinality == src->capacity) return 0; // nothing to do ------------------ | Branch (132:9): [True: 497, False: 3.30k] ------------------ 133| 3.30k| int savings = src->capacity - src->cardinality; 134| 3.30k| src->capacity = src->cardinality; 135| 3.30k| if (src->capacity == ------------------ | Branch (135:9): [True: 0, False: 3.30k] ------------------ 136| 3.30k| 0) { // we do not want to rely on realloc for zero allocs 137| 0| roaring_free(src->array); 138| 0| src->array = NULL; 139| 3.30k| } else { 140| 3.30k| uint16_t *oldarray = src->array; 141| 3.30k| src->array = (uint16_t *)roaring_realloc( 142| 3.30k| oldarray, src->capacity * sizeof(uint16_t)); 143| 3.30k| if (src->array == NULL) roaring_free(oldarray); // should never happen? ------------------ | Branch (143:13): [True: 0, False: 3.30k] ------------------ 144| 3.30k| } 145| 3.30k| return savings; 146| 3.80k|} array_container_free: 149| 150k|void array_container_free(array_container_t *arr) { 150| 150k| if (arr == NULL) return; ------------------ | Branch (150:9): [True: 0, False: 150k] ------------------ 151| 150k| roaring_free(arr->array); 152| 150k| roaring_free(arr); 153| 150k|} array_container_grow: 167| 121k| bool preserve) { 168| 121k| int32_t max = (min <= DEFAULT_MAX_SIZE ? DEFAULT_MAX_SIZE : 65536); ------------------ | Branch (168:20): [True: 121k, False: 0] ------------------ 169| 121k| int32_t new_capacity = clamp(grow_capacity(container->capacity), min, max); 170| | 171| 121k| container->capacity = new_capacity; 172| 121k| uint16_t *array = container->array; 173| | 174| 121k| if (preserve) { ------------------ | Branch (174:9): [True: 111k, False: 10.0k] ------------------ 175| 111k| container->array = 176| 111k| (uint16_t *)roaring_realloc(array, new_capacity * sizeof(uint16_t)); 177| 111k| if (container->array == NULL) roaring_free(array); ------------------ | Branch (177:13): [True: 0, False: 111k] ------------------ 178| 111k| } else { 179| 10.0k| roaring_free(array); 180| 10.0k| container->array = 181| 10.0k| (uint16_t *)roaring_malloc(new_capacity * sizeof(uint16_t)); 182| 10.0k| } 183| | 184| | // if realloc fails, we have container->array == NULL. 185| 121k|} array_container_copy: 189| 34| array_container_t *dst) { 190| 34| const int32_t cardinality = src->cardinality; 191| 34| if (cardinality > dst->capacity) { ------------------ | Branch (191:9): [True: 34, False: 0] ------------------ 192| 34| array_container_grow(dst, cardinality, false); 193| 34| } 194| | 195| 34| dst->cardinality = cardinality; 196| 34| memcpy(dst->array, src->array, cardinality * sizeof(uint16_t)); 197| 34|} array_container_union: 211| 4.61k| array_container_t *out) { 212| 4.61k| const int32_t card_1 = array_1->cardinality, card_2 = array_2->cardinality; 213| 4.61k| const int32_t max_cardinality = card_1 + card_2; 214| | 215| 4.61k| if (out->capacity < max_cardinality) { ------------------ | Branch (215:9): [True: 0, False: 4.61k] ------------------ 216| | array_container_grow(out, max_cardinality, false); 217| 0| } 218| 4.61k| out->cardinality = (int32_t)fast_union_uint16( 219| 4.61k| array_1->array, card_1, array_2->array, card_2, out->array); 220| 4.61k|} array_container_andnot: 228| 5.86k| array_container_t *out) { 229| 5.86k| if (out->capacity < array_1->cardinality) ------------------ | Branch (229:9): [True: 2.32k, False: 3.54k] ------------------ 230| 2.32k| array_container_grow(out, array_1->cardinality, false); 231| 5.86k|#if CROARING_IS_X64 232| 5.86k| if ((croaring_hardware_support() & ROARING_SUPPORTS_AVX2) && ------------------ | Branch (232:9): [True: 5.86k, False: 0] ------------------ 233| 5.86k| (out != array_1) && (out != array_2)) { ------------------ | Branch (233:9): [True: 2.32k, False: 3.54k] | Branch (233:29): [True: 2.32k, False: 0] ------------------ 234| 2.32k| out->cardinality = difference_vector16( 235| 2.32k| array_1->array, array_1->cardinality, array_2->array, 236| 2.32k| array_2->cardinality, out->array); 237| 3.54k| } else { 238| 3.54k| out->cardinality = 239| 3.54k| difference_uint16(array_1->array, array_1->cardinality, 240| 3.54k| array_2->array, array_2->cardinality, out->array); 241| 3.54k| } 242| |#else 243| | out->cardinality = 244| | difference_uint16(array_1->array, array_1->cardinality, array_2->array, 245| | array_2->cardinality, out->array); 246| |#endif 247| 5.86k|} array_container_xor: 256| 8.05k| array_container_t *out) { 257| 8.05k| const int32_t card_1 = array_1->cardinality, card_2 = array_2->cardinality; 258| 8.05k| const int32_t max_cardinality = card_1 + card_2; 259| 8.05k| if (out->capacity < max_cardinality) { ------------------ | Branch (259:9): [True: 0, False: 8.05k] ------------------ 260| 0| array_container_grow(out, max_cardinality, false); 261| 0| } 262| | 263| 8.05k|#if CROARING_IS_X64 264| 8.05k| if (croaring_hardware_support() & ROARING_SUPPORTS_AVX2) { ------------------ | Branch (264:9): [True: 8.05k, False: 0] ------------------ 265| 8.05k| out->cardinality = 266| 8.05k| xor_vector16(array_1->array, array_1->cardinality, array_2->array, 267| 8.05k| array_2->cardinality, out->array); 268| 8.05k| } else { 269| 0| out->cardinality = 270| 0| xor_uint16(array_1->array, array_1->cardinality, array_2->array, 271| 0| array_2->cardinality, out->array); 272| 0| } 273| |#else 274| | out->cardinality = 275| | xor_uint16(array_1->array, array_1->cardinality, array_2->array, 276| | array_2->cardinality, out->array); 277| |#endif 278| 8.05k|} array_container_intersection: 290| 2.32k| array_container_t *out) { 291| 2.32k| int32_t card_1 = array1->cardinality, card_2 = array2->cardinality, 292| 2.32k| min_card = minimum_int32(card_1, card_2); 293| 2.32k| const int threshold = 64; // subject to tuning 294| 2.32k|#if CROARING_IS_X64 295| 2.32k| if (out->capacity < min_card) { ------------------ | Branch (295:9): [True: 2.32k, False: 0] ------------------ 296| 2.32k| array_container_grow(out, min_card + sizeof(__m128i) / sizeof(uint16_t), 297| 2.32k| false); 298| 2.32k| } 299| |#else 300| | if (out->capacity < min_card) { 301| | array_container_grow(out, min_card, false); 302| | } 303| |#endif 304| | 305| 2.32k| if (card_1 * threshold < card_2) { ------------------ | Branch (305:9): [True: 368, False: 1.95k] ------------------ 306| 368| out->cardinality = intersect_skewed_uint16( 307| 368| array1->array, card_1, array2->array, card_2, out->array); 308| 1.95k| } else if (card_2 * threshold < card_1) { ------------------ | Branch (308:16): [True: 205, False: 1.75k] ------------------ 309| 205| out->cardinality = intersect_skewed_uint16( 310| 205| array2->array, card_2, array1->array, card_1, out->array); 311| 1.75k| } else { 312| 1.75k|#if CROARING_IS_X64 313| 1.75k| if (croaring_hardware_support() & ROARING_SUPPORTS_AVX2) { ------------------ | Branch (313:13): [True: 1.75k, False: 0] ------------------ 314| 1.75k| out->cardinality = intersect_vector16( 315| 1.75k| array1->array, card_1, array2->array, card_2, out->array); 316| 1.75k| } else { 317| 0| out->cardinality = intersect_uint16( 318| 0| array1->array, card_1, array2->array, card_2, out->array); 319| 0| } 320| |#else 321| | out->cardinality = intersect_uint16(array1->array, card_1, 322| | array2->array, card_2, out->array); 323| |#endif 324| 1.75k| } 325| 2.32k|} array_container_intersection_cardinality: 330| 9.29k| const array_container_t *array2) { 331| 9.29k| int32_t card_1 = array1->cardinality, card_2 = array2->cardinality; 332| 9.29k| const int threshold = 64; // subject to tuning 333| 9.29k| if (card_1 * threshold < card_2) { ------------------ | Branch (333:9): [True: 1.47k, False: 7.82k] ------------------ 334| 1.47k| return intersect_skewed_uint16_cardinality(array1->array, card_1, 335| 1.47k| array2->array, card_2); 336| 7.82k| } else if (card_2 * threshold < card_1) { ------------------ | Branch (336:16): [True: 820, False: 7.00k] ------------------ 337| 820| return intersect_skewed_uint16_cardinality(array2->array, card_2, 338| 820| array1->array, card_1); 339| 7.00k| } else { 340| 7.00k|#if CROARING_IS_X64 341| 7.00k| if (croaring_hardware_support() & ROARING_SUPPORTS_AVX2) { ------------------ | Branch (341:13): [True: 7.00k, False: 0] ------------------ 342| 7.00k| return intersect_vector16_cardinality(array1->array, card_1, 343| 7.00k| array2->array, card_2); 344| 7.00k| } else { 345| 0| return intersect_uint16_cardinality(array1->array, card_1, 346| 0| array2->array, card_2); 347| 0| } 348| |#else 349| | return intersect_uint16_cardinality(array1->array, card_1, 350| | array2->array, card_2); 351| |#endif 352| 7.00k| } 353| 9.29k|} array_container_intersect: 356| 2.32k| const array_container_t *array2) { 357| 2.32k| int32_t card_1 = array1->cardinality, card_2 = array2->cardinality; 358| 2.32k| const int threshold = 64; // subject to tuning 359| 2.32k| if (card_1 * threshold < card_2) { ------------------ | Branch (359:9): [True: 368, False: 1.95k] ------------------ 360| 368| return intersect_skewed_uint16_nonempty(array1->array, card_1, 361| 368| array2->array, card_2); 362| 1.95k| } else if (card_2 * threshold < card_1) { ------------------ | Branch (362:16): [True: 205, False: 1.75k] ------------------ 363| 205| return intersect_skewed_uint16_nonempty(array2->array, card_2, 364| 205| array1->array, card_1); 365| 1.75k| } else { 366| | // we do not bother vectorizing 367| 1.75k| return intersect_uint16_nonempty(array1->array, card_1, array2->array, 368| 1.75k| card_2); 369| 1.75k| } 370| 2.32k|} array_container_intersection_inplace: 376| 4.86k| const array_container_t *src_2) { 377| 4.86k| int32_t card_1 = src_1->cardinality, card_2 = src_2->cardinality; 378| 4.86k| const int threshold = 64; // subject to tuning 379| 4.86k| if (card_1 * threshold < card_2) { ------------------ | Branch (379:9): [True: 0, False: 4.86k] ------------------ 380| 0| src_1->cardinality = intersect_skewed_uint16( 381| 0| src_1->array, card_1, src_2->array, card_2, src_1->array); 382| 4.86k| } else if (card_2 * threshold < card_1) { ------------------ | Branch (382:16): [True: 214, False: 4.64k] ------------------ 383| 214| src_1->cardinality = intersect_skewed_uint16( 384| 214| src_2->array, card_2, src_1->array, card_1, src_1->array); 385| 4.64k| } else { 386| 4.64k|#if CROARING_IS_X64 387| 4.64k| if (croaring_hardware_support() & ROARING_SUPPORTS_AVX2) { ------------------ | Branch (387:13): [True: 4.64k, False: 0] ------------------ 388| 4.64k| src_1->cardinality = intersect_vector16_inplace( 389| 4.64k| src_1->array, card_1, src_2->array, card_2); 390| 4.64k| } else { 391| 0| src_1->cardinality = intersect_uint16( 392| 0| src_1->array, card_1, src_2->array, card_2, src_1->array); 393| 0| } 394| |#else 395| | src_1->cardinality = intersect_uint16( 396| | src_1->array, card_1, src_2->array, card_2, src_1->array); 397| |#endif 398| 4.64k| } 399| 4.86k|} array_container_to_uint32_array: 403| 5.98k| uint32_t base) { 404| 5.98k|#if CROARING_IS_X64 405| 5.98k| int support = croaring_hardware_support(); 406| 5.98k|#if CROARING_COMPILER_SUPPORTS_AVX512 407| 5.98k| if (support & ROARING_SUPPORTS_AVX512) { ------------------ | Branch (407:9): [True: 0, False: 5.98k] ------------------ 408| 0| return avx512_array_container_to_uint32_array(vout, cont->array, 409| 0| cont->cardinality, base); 410| 0| } 411| 5.98k|#endif 412| 5.98k| if (support & ROARING_SUPPORTS_AVX2) { ------------------ | Branch (412:9): [True: 5.98k, False: 0] ------------------ 413| 5.98k| return array_container_to_uint32_array_vector16( 414| 5.98k| vout, cont->array, cont->cardinality, base); 415| 5.98k| } 416| 0|#endif // CROARING_IS_X64 417| 0| int outpos = 0; 418| 0| uint32_t *out = (uint32_t *)vout; 419| 0| size_t i = 0; 420| 0| for (; i < (size_t)cont->cardinality; ++i) { ------------------ | Branch (420:12): [True: 0, False: 0] ------------------ 421| 0| const uint32_t val = base + cont->array[i]; 422| 0| memcpy(out + outpos, &val, 423| 0| sizeof(uint32_t)); // should be compiled as a MOV on x64 424| 0| outpos++; 425| 0| } 426| 0| return outpos; 427| 5.98k|} array_container_number_of_runs: 495| 12.9k|int32_t array_container_number_of_runs(const array_container_t *ac) { 496| | // Can SIMD work here? 497| 12.9k| int32_t nr_runs = 0; 498| 12.9k| int32_t prev = -2; 499| 1.67M| for (const uint16_t *p = ac->array; p != ac->array + ac->cardinality; ++p) { ------------------ | Branch (499:41): [True: 1.65M, False: 12.9k] ------------------ 500| 1.65M| if (*p != prev + 1) nr_runs++; ------------------ | Branch (500:13): [True: 905k, False: 754k] ------------------ 501| 1.65M| prev = *p; 502| 1.65M| } 503| 12.9k| return nr_runs; 504| 12.9k|} array_container_write: 512| 7.20k|int32_t array_container_write(const array_container_t *container, char *buf) { 513| |#if CROARING_IS_BIG_ENDIAN 514| | for (int32_t i = 0; i < container->cardinality; ++i) { 515| | uint16_t v_le = croaring_htole16(container->array[i]); 516| | memcpy(buf + i * sizeof(uint16_t), &v_le, sizeof(uint16_t)); 517| | } 518| |#else 519| 7.20k| memcpy(buf, container->array, container->cardinality * sizeof(uint16_t)); 520| 7.20k|#endif 521| 7.20k| return array_container_size_in_bytes(container); 522| 7.20k|} array_container_is_subset: 525| 5.75k| const array_container_t *container2) { 526| 5.75k| if (container1->cardinality > container2->cardinality) { ------------------ | Branch (526:9): [True: 592, False: 5.16k] ------------------ 527| 592| return false; 528| 592| } 529| 5.16k| int i1 = 0, i2 = 0; 530| 55.0k| while (i1 < container1->cardinality && i2 < container2->cardinality) { ------------------ | Branch (530:12): [True: 55.0k, False: 0] | Branch (530:44): [True: 55.0k, False: 23] ------------------ 531| 55.0k| if (container1->array[i1] == container2->array[i2]) { ------------------ | Branch (531:13): [True: 0, False: 55.0k] ------------------ 532| 0| i1++; 533| 0| i2++; 534| 55.0k| } else if (container1->array[i1] > container2->array[i2]) { ------------------ | Branch (534:20): [True: 49.9k, False: 5.14k] ------------------ 535| 49.9k| i2++; 536| 49.9k| } else { // container1->array[i1] < container2->array[i2] 537| 5.14k| return false; 538| 5.14k| } 539| 55.0k| } 540| 23| if (i1 == container1->cardinality) { ------------------ | Branch (540:9): [True: 0, False: 23] ------------------ 541| 0| return true; 542| 23| } else { 543| | return false; 544| 23| } 545| 23|} array_container_read: 548| 7.25k| const char *buf) { 549| 7.25k| if (container->capacity < cardinality) { ------------------ | Branch (549:9): [True: 0, False: 7.25k] ------------------ 550| 0| array_container_grow(container, cardinality, false); 551| 0| } 552| 7.25k| container->cardinality = cardinality; 553| |#if CROARING_IS_BIG_ENDIAN 554| | for (int32_t i = 0; i < cardinality; ++i) { 555| | uint16_t v_le; 556| | memcpy(&v_le, buf + i * sizeof(uint16_t), sizeof(uint16_t)); 557| | container->array[i] = croaring_letoh16(v_le); 558| | } 559| |#else 560| 7.25k| memcpy(container->array, buf, container->cardinality * sizeof(uint16_t)); 561| 7.25k|#endif 562| | 563| 7.25k| return array_container_size_in_bytes(container); 564| 7.25k|} array_container_iterate: 567| 6.58k| roaring_iterator iterator, void *ptr) { 568| 2.18M| for (int i = 0; i < cont->cardinality; i++) ------------------ | Branch (568:21): [True: 2.17M, False: 6.58k] ------------------ 569| 2.17M| if (!iterator(cont->array[i] + base, ptr)) return false; ------------------ | Branch (569:13): [True: 0, False: 2.17M] ------------------ 570| 6.58k| return true; 571| 6.58k|} array.c:clamp: 162| 121k|static inline int32_t clamp(int32_t val, int32_t min, int32_t max) { 163| 121k| return ((val < min) ? min : (val > max) ? max : val); ------------------ | Branch (163:13): [True: 27.0k, False: 94.6k] | Branch (163:33): [True: 268, False: 94.4k] ------------------ 164| 121k|} array.c:grow_capacity: 155| 121k|static inline int32_t grow_capacity(int32_t capacity) { 156| 121k| return (capacity <= 0) ? ARRAY_DEFAULT_INIT_SIZE ------------------ | Branch (156:12): [True: 26.3k, False: 95.3k] ------------------ 157| 121k| : capacity < 64 ? capacity * 2 ------------------ | Branch (157:14): [True: 69.1k, False: 26.2k] ------------------ 158| 95.3k| : capacity < 1024 ? capacity * 3 / 2 ------------------ | Branch (158:14): [True: 25.9k, False: 305] ------------------ 159| 26.2k| : capacity * 5 / 4; 160| 121k|} array.c:minimum_int32: 280| 2.32k|static inline int32_t minimum_int32(int32_t a, int32_t b) { 281| 2.32k| return (a < b) ? a : b; ------------------ | Branch (281:12): [True: 1.02k, False: 1.30k] ------------------ 282| 2.32k|} bitset_container_clear: 51| 153k|void bitset_container_clear(bitset_container_t *bitset) { 52| 153k| memset(bitset->words, 0, sizeof(uint64_t) * BITSET_CONTAINER_SIZE_IN_WORDS); 53| 153k| bitset->cardinality = 0; 54| 153k|} bitset_container_set_all: 56| 185|void bitset_container_set_all(bitset_container_t *bitset) { 57| | memset(bitset->words, INT64_C(-1), 58| 185| sizeof(uint64_t) * BITSET_CONTAINER_SIZE_IN_WORDS); 59| 185| bitset->cardinality = (1 << 16); 60| 185|} bitset_container_create: 63| 153k|bitset_container_t *bitset_container_create(void) { 64| 153k| bitset_container_t *bitset = 65| 153k| (bitset_container_t *)roaring_malloc(sizeof(bitset_container_t)); 66| | 67| 153k| if (!bitset) { ------------------ | Branch (67:9): [True: 0, False: 153k] ------------------ 68| 0| return NULL; 69| 0| } 70| | 71| 153k| size_t align_size = 32; 72| 153k|#if CROARING_IS_X64 73| 153k| int support = croaring_hardware_support(); 74| 153k| if (support & ROARING_SUPPORTS_AVX512) { ------------------ | Branch (74:9): [True: 0, False: 153k] ------------------ 75| | // sizeof(__m512i) == 64 76| 0| align_size = 64; 77| 153k| } else { 78| | // sizeof(__m256i) == 32 79| 153k| align_size = 32; 80| 153k| } 81| 153k|#endif 82| 153k| bitset->words = (uint64_t *)roaring_aligned_malloc( 83| 153k| align_size, sizeof(uint64_t) * BITSET_CONTAINER_SIZE_IN_WORDS); 84| 153k| if (!bitset->words) { ------------------ | Branch (84:9): [True: 0, False: 153k] ------------------ 85| 0| roaring_free(bitset); 86| 0| return NULL; 87| 0| } 88| 153k| bitset_container_clear(bitset); 89| 153k| return bitset; 90| 153k|} bitset_container_copy: 94| 4.03k| bitset_container_t *dest) { 95| 4.03k| dest->cardinality = source->cardinality; 96| 4.03k| memcpy(dest->words, source->words, 97| 4.03k| sizeof(uint64_t) * BITSET_CONTAINER_SIZE_IN_WORDS); 98| 4.03k|} bitset_container_free: 129| 302k|void bitset_container_free(bitset_container_t *bitset) { 130| 302k| if (bitset == NULL) return; ------------------ | Branch (130:9): [True: 0, False: 302k] ------------------ 131| 302k| roaring_aligned_free(bitset->words); 132| 302k| roaring_free(bitset); 133| 302k|} bitset_container_clone: 137| 148k|bitset_container_t *bitset_container_clone(const bitset_container_t *src) { 138| 148k| bitset_container_t *bitset = 139| 148k| (bitset_container_t *)roaring_malloc(sizeof(bitset_container_t)); 140| | 141| 148k| if (!bitset) { ------------------ | Branch (141:9): [True: 0, False: 148k] ------------------ 142| 0| return NULL; 143| 0| } 144| | 145| 148k| size_t align_size = 32; 146| 148k|#if CROARING_IS_X64 147| 148k| if (croaring_hardware_support() & ROARING_SUPPORTS_AVX512) { ------------------ | Branch (147:9): [True: 0, False: 148k] ------------------ 148| | // sizeof(__m512i) == 64 149| 0| align_size = 64; 150| 148k| } else { 151| | // sizeof(__m256i) == 32 152| 148k| align_size = 32; 153| 148k| } 154| 148k|#endif 155| 148k| bitset->words = (uint64_t *)roaring_aligned_malloc( 156| 148k| align_size, sizeof(uint64_t) * BITSET_CONTAINER_SIZE_IN_WORDS); 157| 148k| if (!bitset->words) { ------------------ | Branch (157:9): [True: 0, False: 148k] ------------------ 158| 0| roaring_free(bitset); 159| 0| return NULL; 160| 0| } 161| 148k| bitset->cardinality = src->cardinality; 162| 148k| memcpy(bitset->words, src->words, 163| 148k| sizeof(uint64_t) * BITSET_CONTAINER_SIZE_IN_WORDS); 164| 148k| return bitset; 165| 148k|} bitset_container_compute_cardinality: 269| 3.60k|int bitset_container_compute_cardinality(const bitset_container_t *bitset) { 270| 3.60k| int support = croaring_hardware_support(); 271| 3.60k|#if CROARING_COMPILER_SUPPORTS_AVX512 272| 3.60k| if (support & ROARING_SUPPORTS_AVX512) { ------------------ | Branch (272:9): [True: 0, False: 3.60k] ------------------ 273| 0| return (int)avx512_vpopcount( 274| 0| (const __m512i *)bitset->words, 275| 0| BITSET_CONTAINER_SIZE_IN_WORDS / (WORDS_IN_AVX512_REG)); ------------------ | | 253| 0|#define WORDS_IN_AVX512_REG sizeof(__m512i) / sizeof(uint64_t) ------------------ 276| 0| } else 277| 3.60k|#endif // CROARING_COMPILER_SUPPORTS_AVX512 278| 3.60k| if (support & ROARING_SUPPORTS_AVX2) { ------------------ | Branch (278:13): [True: 3.60k, False: 0] ------------------ 279| 3.60k| return (int)avx2_harley_seal_popcount256( 280| 3.60k| (const __m256i *)bitset->words, 281| 3.60k| BITSET_CONTAINER_SIZE_IN_WORDS / (CROARING_WORDS_IN_AVX2_REG)); ------------------ | | 250| 3.60k|#define CROARING_WORDS_IN_AVX2_REG sizeof(__m256i) / sizeof(uint64_t) ------------------ 282| 3.60k| } else { 283| 0| return _scalar_bitset_container_compute_cardinality(bitset); 284| 0| } 285| 3.60k|} bitset_container_and_nocard: 724| 1.51k| bitset_container_t *dst) { \ 725| 1.51k| int support = croaring_hardware_support(); \ 726| 1.51k| if ( support & ROARING_SUPPORTS_AVX512 ) { \ ------------------ | Branch (726:10): [True: 0, False: 1.51k] ------------------ 727| 0| return _avx512_bitset_container_##opname##_nocard(src_1, src_2, dst); \ 728| 0| } \ 729| 1.51k| else if ( support & ROARING_SUPPORTS_AVX2 ) { \ ------------------ | Branch (729:15): [True: 1.51k, False: 0] ------------------ 730| 1.51k| return _avx2_bitset_container_##opname##_nocard(src_1, src_2, dst); \ 731| 1.51k| } else { \ 732| 0| return _scalar_bitset_container_##opname##_nocard(src_1, src_2, dst); \ 733| 0| } \ 734| 1.51k| } \ bitset_container_and_justcard: 736| 1.51k| const bitset_container_t *src_2) { \ 737| 1.51k| int support = croaring_hardware_support(); \ 738| 1.51k| if ( support & ROARING_SUPPORTS_AVX512 ) { \ ------------------ | Branch (738:10): [True: 0, False: 1.51k] ------------------ 739| 0| return _avx512_bitset_container_##opname##_justcard(src_1, src_2); \ 740| 0| } \ 741| 1.51k| else if ( support & ROARING_SUPPORTS_AVX2 ) { \ ------------------ | Branch (741:15): [True: 1.51k, False: 0] ------------------ 742| 1.51k| return _avx2_bitset_container_##opname##_justcard(src_1, src_2); \ 743| 1.51k| } else { \ 744| 0| return _scalar_bitset_container_##opname##_justcard(src_1, src_2); \ 745| 0| } \ 746| 1.51k| } bitset_container_xor: 711| 1.57k| bitset_container_t *dst) { \ 712| 1.57k| int support = croaring_hardware_support(); \ 713| 1.57k| if ( support & ROARING_SUPPORTS_AVX512 ) { \ ------------------ | Branch (713:10): [True: 0, False: 1.57k] ------------------ 714| 0| return _avx512_bitset_container_##opname(src_1, src_2, dst); \ 715| 0| } \ 716| 1.57k| else if ( support & ROARING_SUPPORTS_AVX2 ) { \ ------------------ | Branch (716:15): [True: 1.57k, False: 0] ------------------ 717| 1.57k| return _avx2_bitset_container_##opname(src_1, src_2, dst); \ 718| 1.57k| } else { \ 719| 0| return _scalar_bitset_container_##opname(src_1, src_2, dst); \ 720| 0| } \ 721| 1.57k| } \ bitset_container_to_uint32_array: 936| 1.51k|){ 937| 1.51k|#if CROARING_IS_X64 938| 1.51k| int support = croaring_hardware_support(); 939| 1.51k|#if CROARING_COMPILER_SUPPORTS_AVX512 940| 1.51k| if(( support & ROARING_SUPPORTS_AVX512 ) && (bc->cardinality >= 8192)) // heuristic ------------------ | Branch (940:7): [True: 0, False: 1.51k] | Branch (940:49): [True: 0, False: 0] ------------------ 941| 0| return (int) bitset_extract_setbits_avx512(bc->words, 942| 0| BITSET_CONTAINER_SIZE_IN_WORDS, out, bc->cardinality, base); 943| 1.51k| else 944| 1.51k|#endif 945| 1.51k| if(( support & ROARING_SUPPORTS_AVX2 ) && (bc->cardinality >= 8192)) // heuristic ------------------ | Branch (945:7): [True: 1.51k, False: 0] | Branch (945:47): [True: 1.03k, False: 480] ------------------ 946| 1.03k| return (int) bitset_extract_setbits_avx2(bc->words, 947| 1.03k| BITSET_CONTAINER_SIZE_IN_WORDS, out, bc->cardinality, base); 948| 480| else 949| 480| return (int) bitset_extract_setbits(bc->words, 950| 480| BITSET_CONTAINER_SIZE_IN_WORDS, out, base); 951| |#else 952| | return (int) bitset_extract_setbits(bc->words, 953| | BITSET_CONTAINER_SIZE_IN_WORDS, out, base); 954| |#endif 955| 1.51k|} bitset_container_write: 1050| 73.9k| char *buf) { 1051| |#if CROARING_IS_BIG_ENDIAN 1052| | for (int32_t i = 0; i < BITSET_CONTAINER_SIZE_IN_WORDS; ++i) { 1053| | uint64_t w_le = croaring_htole64(container->words[i]); 1054| | memcpy(buf + i * sizeof(uint64_t), &w_le, sizeof(uint64_t)); 1055| | } 1056| |#else 1057| 73.9k| memcpy(buf, container->words, BITSET_CONTAINER_SIZE_IN_WORDS * sizeof(uint64_t)); 1058| 73.9k|#endif 1059| 73.9k| return bitset_container_size_in_bytes(container); 1060| 73.9k|} bitset_container_read: 1064| 73.9k| const char *buf) { 1065| 73.9k| container->cardinality = cardinality; 1066| |#if CROARING_IS_BIG_ENDIAN 1067| | for (int32_t i = 0; i < BITSET_CONTAINER_SIZE_IN_WORDS; ++i) { 1068| | uint64_t w_le; 1069| | memcpy(&w_le, buf + i * sizeof(uint64_t), sizeof(uint64_t)); 1070| | container->words[i] = croaring_letoh64(w_le); 1071| | } 1072| |#else 1073| 73.9k| memcpy(container->words, buf, BITSET_CONTAINER_SIZE_IN_WORDS * sizeof(uint64_t)); 1074| 73.9k|#endif 1075| 73.9k| return bitset_container_size_in_bytes(container); 1076| 73.9k|} bitset_container_iterate: 1078| 1.69k|bool bitset_container_iterate(const bitset_container_t *cont, uint32_t base, roaring_iterator iterator, void *ptr) { 1079| 1.73M| for (int32_t i = 0; i < BITSET_CONTAINER_SIZE_IN_WORDS; ++i ) { ------------------ | Branch (1079:23): [True: 1.73M, False: 1.69k] ------------------ 1080| 1.73M| uint64_t w = cont->words[i]; 1081| 65.3M| while (w != 0) { ------------------ | Branch (1081:12): [True: 63.5M, False: 1.73M] ------------------ 1082| 63.5M| uint64_t t = w & (~w + 1); 1083| 63.5M| int r = roaring_trailing_zeroes(w); 1084| 63.5M| if(!iterator(r + base, ptr)) return false; ------------------ | Branch (1084:10): [True: 0, False: 63.5M] ------------------ 1085| 63.5M| w ^= t; 1086| 63.5M| } 1087| 1.73M| base += 64; 1088| 1.73M| } 1089| 1.69k| return true; 1090| 1.69k|} bitset_container_equals: 1144| 73.9k|bool bitset_container_equals(const bitset_container_t *container1, const bitset_container_t *container2) { 1145| 73.9k| if((container1->cardinality != BITSET_UNKNOWN_CARDINALITY) && (container2->cardinality != BITSET_UNKNOWN_CARDINALITY)) { ------------------ | Branch (1145:6): [True: 73.9k, False: 0] | Branch (1145:65): [True: 73.9k, False: 0] ------------------ 1146| 73.9k| if(container1->cardinality != container2->cardinality) { ------------------ | Branch (1146:8): [True: 0, False: 73.9k] ------------------ 1147| 0| return false; 1148| 0| } 1149| 73.9k| if (container1->cardinality == INT32_C(0x10000)) { ------------------ | Branch (1149:9): [True: 69.9k, False: 4.07k] ------------------ 1150| 69.9k| return true; 1151| 69.9k| } 1152| 73.9k| } 1153| 4.07k|#if CROARING_IS_X64 1154| 4.07k| int support = croaring_hardware_support(); 1155| 4.07k|#if CROARING_COMPILER_SUPPORTS_AVX512 1156| 4.07k| if( support & ROARING_SUPPORTS_AVX512 ) { ------------------ | Branch (1156:7): [True: 0, False: 4.07k] ------------------ 1157| 0| return _avx512_bitset_container_equals(container1, container2); 1158| 0| } 1159| 4.07k| else 1160| 4.07k|#endif 1161| 4.07k| if( support & ROARING_SUPPORTS_AVX2 ) { ------------------ | Branch (1161:7): [True: 4.07k, False: 0] ------------------ 1162| 4.07k| return _avx2_bitset_container_equals(container1, container2); 1163| 4.07k| } 1164| 0|#endif 1165| 0| return memcmp(container1->words, 1166| 0| container2->words, 1167| 0| BITSET_CONTAINER_SIZE_IN_WORDS*sizeof(uint64_t)) == 0; 1168| 4.07k|} bitset_container_minimum: 1218| 1.27k|uint16_t bitset_container_minimum(const bitset_container_t *container) { 1219| 13.5k| for (int32_t i = 0; i < BITSET_CONTAINER_SIZE_IN_WORDS; ++i ) { ------------------ | Branch (1219:23): [True: 13.5k, False: 0] ------------------ 1220| 13.5k| uint64_t w = container->words[i]; 1221| 13.5k| if (w != 0) { ------------------ | Branch (1221:9): [True: 1.27k, False: 12.2k] ------------------ 1222| 1.27k| int r = roaring_trailing_zeroes(w); 1223| 1.27k| return r + i * 64; 1224| 1.27k| } 1225| 13.5k| } 1226| 0| return UINT16_MAX; 1227| 1.27k|} bitset_container_maximum: 1230| 4.49k|uint16_t bitset_container_maximum(const bitset_container_t *container) { 1231| 1.12M| for (int32_t i = BITSET_CONTAINER_SIZE_IN_WORDS - 1; i > 0; --i ) { ------------------ | Branch (1231:56): [True: 1.12M, False: 0] ------------------ 1232| 1.12M| uint64_t w = container->words[i]; 1233| 1.12M| if (w != 0) { ------------------ | Branch (1233:9): [True: 4.49k, False: 1.12M] ------------------ 1234| 4.49k| int r = roaring_leading_zeroes(w); 1235| 4.49k| return i * 64 + 63 - r; 1236| 4.49k| } 1237| 1.12M| } 1238| 0| return 0; 1239| 4.49k|} bitset_container_index_equalorlarger: 1299| 1.91k|int bitset_container_index_equalorlarger(const bitset_container_t *container, uint16_t x) { 1300| 1.91k| uint32_t x32 = x; 1301| 1.91k| uint32_t k = x32 / 64; 1302| 1.91k| uint64_t word = container->words[k]; 1303| 1.91k| const int diff = x32 - k * 64; // in [0,64) 1304| 1.91k| word = (word >> diff) << diff; // a mask is faster, but we don't care 1305| 5.42k| while(word == 0) { ------------------ | Branch (1305:9): [True: 3.51k, False: 1.91k] ------------------ 1306| 3.51k| k++; 1307| 3.51k| if(k == BITSET_CONTAINER_SIZE_IN_WORDS) return -1; ------------------ | Branch (1307:8): [True: 0, False: 3.51k] ------------------ 1308| 3.51k| word = container->words[k]; 1309| 3.51k| } 1310| 1.91k| return k * 64 + roaring_trailing_zeroes(word); 1311| 1.91k|} bitset.c:_avx2_bitset_container_and_nocard: 503| 1.51k| bitset_container_t *dst) { \ 504| 1.51k| const uint8_t *__restrict__ words_1 = (const uint8_t *)src_1->words; \ 505| 1.51k| const uint8_t *__restrict__ words_2 = (const uint8_t *)src_2->words; \ 506| 1.51k| /* not using the blocking optimization for some reason*/ \ 507| 1.51k| uint8_t *out = (uint8_t *)dst->words; \ 508| 1.51k| const int innerloop = 8; \ 509| 1.51k| for (size_t i = 0; \ 510| 50.0k| i < BITSET_CONTAINER_SIZE_IN_WORDS / (CROARING_WORDS_IN_AVX2_REG); \ ------------------ | | 250| 50.0k|#define CROARING_WORDS_IN_AVX2_REG sizeof(__m256i) / sizeof(uint64_t) ------------------ | Branch (510:10): [True: 48.5k, False: 1.51k] ------------------ 511| 48.5k| i += innerloop) { \ 512| 48.5k| __m256i A1, A2, AO; \ 513| 48.5k| A1 = _mm256_lddqu_si256((const __m256i *)(words_1)); \ 514| 48.5k| A2 = _mm256_lddqu_si256((const __m256i *)(words_2)); \ 515| 48.5k| AO = avx_intrinsic(A2, A1); \ 516| 48.5k| _mm256_storeu_si256((__m256i *)out, AO); \ 517| 48.5k| A1 = _mm256_lddqu_si256((const __m256i *)(words_1 + 32)); \ 518| 48.5k| A2 = _mm256_lddqu_si256((const __m256i *)(words_2 + 32)); \ 519| 48.5k| AO = avx_intrinsic(A2, A1); \ 520| 48.5k| _mm256_storeu_si256((__m256i *)(out + 32), AO); \ 521| 48.5k| A1 = _mm256_lddqu_si256((const __m256i *)(words_1 + 64)); \ 522| 48.5k| A2 = _mm256_lddqu_si256((const __m256i *)(words_2 + 64)); \ 523| 48.5k| AO = avx_intrinsic(A2, A1); \ 524| 48.5k| _mm256_storeu_si256((__m256i *)(out + 64), AO); \ 525| 48.5k| A1 = _mm256_lddqu_si256((const __m256i *)(words_1 + 96)); \ 526| 48.5k| A2 = _mm256_lddqu_si256((const __m256i *)(words_2 + 96)); \ 527| 48.5k| AO = avx_intrinsic(A2, A1); \ 528| 48.5k| _mm256_storeu_si256((__m256i *)(out + 96), AO); \ 529| 48.5k| A1 = _mm256_lddqu_si256((const __m256i *)(words_1 + 128)); \ 530| 48.5k| A2 = _mm256_lddqu_si256((const __m256i *)(words_2 + 128)); \ 531| 48.5k| AO = avx_intrinsic(A2, A1); \ 532| 48.5k| _mm256_storeu_si256((__m256i *)(out + 128), AO); \ 533| 48.5k| A1 = _mm256_lddqu_si256((const __m256i *)(words_1 + 160)); \ 534| 48.5k| A2 = _mm256_lddqu_si256((const __m256i *)(words_2 + 160)); \ 535| 48.5k| AO = avx_intrinsic(A2, A1); \ 536| 48.5k| _mm256_storeu_si256((__m256i *)(out + 160), AO); \ 537| 48.5k| A1 = _mm256_lddqu_si256((const __m256i *)(words_1 + 192)); \ 538| 48.5k| A2 = _mm256_lddqu_si256((const __m256i *)(words_2 + 192)); \ 539| 48.5k| AO = avx_intrinsic(A2, A1); \ 540| 48.5k| _mm256_storeu_si256((__m256i *)(out + 192), AO); \ 541| 48.5k| A1 = _mm256_lddqu_si256((const __m256i *)(words_1 + 224)); \ 542| 48.5k| A2 = _mm256_lddqu_si256((const __m256i *)(words_2 + 224)); \ 543| 48.5k| AO = avx_intrinsic(A2, A1); \ 544| 48.5k| _mm256_storeu_si256((__m256i *)(out + 224), AO); \ 545| 48.5k| out += 256; \ 546| 48.5k| words_1 += 256; \ 547| 48.5k| words_2 += 256; \ 548| 48.5k| } \ 549| 1.51k| dst->cardinality = BITSET_UNKNOWN_CARDINALITY; \ 550| 1.51k| return dst->cardinality; \ 551| 1.51k| } bitset.c:_avx2_bitset_container_and_justcard: 572| 1.51k| const bitset_container_t *src_1, const bitset_container_t *src_2) { \ 573| 1.51k| const __m256i *__restrict__ data1 = (const __m256i *)src_1->words; \ 574| 1.51k| const __m256i *__restrict__ data2 = (const __m256i *)src_2->words; \ 575| 1.51k| return (int)avx2_harley_seal_popcount256_##opname( \ 576| 1.51k| data2, data1, BITSET_CONTAINER_SIZE_IN_WORDS / (CROARING_WORDS_IN_AVX2_REG)); \ ------------------ | | 250| 1.51k|#define CROARING_WORDS_IN_AVX2_REG sizeof(__m256i) / sizeof(uint64_t) ------------------ 577| 1.51k| } bitset.c:_avx2_bitset_container_xor: 558| 1.57k| bitset_container_t *dst) { \ 559| 1.57k| const __m256i *__restrict__ words_1 = (const __m256i *)src_1->words; \ 560| 1.57k| const __m256i *__restrict__ words_2 = (const __m256i *)src_2->words; \ 561| 1.57k| __m256i *out = (__m256i *)dst->words; \ 562| 1.57k| dst->cardinality = (int32_t)avx2_harley_seal_popcount256andstore_##opname( \ 563| 1.57k| words_2, words_1, out, \ 564| 1.57k| BITSET_CONTAINER_SIZE_IN_WORDS / (CROARING_WORDS_IN_AVX2_REG)); \ ------------------ | | 250| 1.57k|#define CROARING_WORDS_IN_AVX2_REG sizeof(__m256i) / sizeof(uint64_t) ------------------ 565| 1.57k| return dst->cardinality; \ 566| 1.57k| } \ bitset.c:_avx2_bitset_container_equals: 1127| 4.07k|static inline bool _avx2_bitset_container_equals(const bitset_container_t *container1, const bitset_container_t *container2) { 1128| 4.07k| const __m256i *ptr1 = (const __m256i*)container1->words; 1129| 4.07k| const __m256i *ptr2 = (const __m256i*)container2->words; 1130| 1.04M| for (size_t i = 0; i < BITSET_CONTAINER_SIZE_IN_WORDS*sizeof(uint64_t)/32; i++) { ------------------ | Branch (1130:24): [True: 1.04M, False: 4.07k] ------------------ 1131| 1.04M| __m256i r1 = _mm256_loadu_si256(ptr1+i); 1132| 1.04M| __m256i r2 = _mm256_loadu_si256(ptr2+i); 1133| 1.04M| int mask = _mm256_movemask_epi8(_mm256_cmpeq_epi8(r1, r2)); 1134| 1.04M| if ((uint32_t)mask != UINT32_MAX) { ------------------ | Branch (1134:11): [True: 0, False: 1.04M] ------------------ 1135| 0| return false; 1136| 0| } 1137| 1.04M| } 1138| 4.07k| return true; 1139| 4.07k|} container_free: 58| 637k|void container_free(container_t *c, uint8_t type) { 59| 637k| switch (type) { 60| 299k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 299k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (60:9): [True: 299k, False: 338k] ------------------ 61| 299k| bitset_container_free(CAST_bitset(c)); ------------------ | | 52| 299k|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 299k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 62| 299k| break; 63| 136k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 136k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (63:9): [True: 136k, False: 501k] ------------------ 64| 136k| array_container_free(CAST_array(c)); ------------------ | | 54| 136k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 136k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 65| 136k| break; 66| 202k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 202k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (66:9): [True: 202k, False: 435k] ------------------ 67| 202k| run_container_free(CAST_run(c)); ------------------ | | 77| 202k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 202k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 68| 202k| break; 69| 0| case SHARED_CONTAINER_TYPE: ------------------ | | 51| 0|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (69:9): [True: 0, False: 637k] ------------------ 70| 0| shared_container_free(CAST_shared(c)); ------------------ | | 79| 0|#define CAST_shared(c) CAST(shared_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 71| 0| break; 72| 0| default: ------------------ | Branch (72:9): [True: 0, False: 637k] ------------------ 73| 0| assert(false); 74| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 75| 637k| } 76| 637k|} get_copy_of_container: 181| 259| bool copy_on_write) { 182| 259| if (copy_on_write) { ------------------ | Branch (182:9): [True: 0, False: 259] ------------------ 183| 0| shared_container_t *shared_container; 184| 0| if (*typecode == SHARED_CONTAINER_TYPE) { ------------------ | | 51| 0|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (184:13): [True: 0, False: 0] ------------------ 185| 0| shared_container = CAST_shared(c); ------------------ | | 79| 0|#define CAST_shared(c) CAST(shared_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 186| 0| croaring_refcount_inc(&shared_container->counter); 187| 0| return shared_container; 188| 0| } 189| 0| assert(*typecode != SHARED_CONTAINER_TYPE); 190| | 191| 0| if ((shared_container = (shared_container_t *)roaring_malloc( ------------------ | Branch (191:13): [True: 0, False: 0] ------------------ 192| 0| sizeof(shared_container_t))) == NULL) { 193| 0| return NULL; 194| 0| } 195| | 196| 0| shared_container->container = c; 197| 0| shared_container->typecode = *typecode; 198| | // At this point, we are creating new shared container 199| | // so there should be no other references, and setting 200| | // the counter to 2 - even non-atomically - is safe as 201| | // long as the value is set before the return statement. 202| 0| shared_container->counter = 2; 203| 0| *typecode = SHARED_CONTAINER_TYPE; ------------------ | | 51| 0|#define SHARED_CONTAINER_TYPE 4 ------------------ 204| | 205| 0| return shared_container; 206| 0| } // copy_on_write 207| | // otherwise, no copy on write... 208| 259| const container_t *actual_container = container_unwrap_shared(c, typecode); 209| | assert(*typecode != SHARED_CONTAINER_TYPE); 210| 259| return container_clone(actual_container, *typecode); 211| 259|} container_clone: 217| 353k|container_t *container_clone(const container_t *c, uint8_t typecode) { 218| | // We do not want to allow cloning of shared containers. 219| | // c = container_unwrap_shared(c, &typecode); 220| 353k| switch (typecode) { 221| 148k| case BITSET_CONTAINER_TYPE: ------------------ | | 48| 148k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (221:9): [True: 148k, False: 204k] ------------------ 222| 148k| return bitset_container_clone(const_CAST_bitset(c)); ------------------ | | 53| 148k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 148k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 223| 23.2k| case ARRAY_CONTAINER_TYPE: ------------------ | | 49| 23.2k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (223:9): [True: 23.2k, False: 329k] ------------------ 224| 23.2k| return array_container_clone(const_CAST_array(c)); ------------------ | | 55| 23.2k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 23.2k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 225| 181k| case RUN_CONTAINER_TYPE: ------------------ | | 50| 181k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (225:9): [True: 181k, False: 172k] ------------------ 226| 181k| return run_container_clone(const_CAST_run(c)); ------------------ | | 78| 181k|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 181k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 227| 0| case SHARED_CONTAINER_TYPE: ------------------ | | 51| 0|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (227:9): [True: 0, False: 353k] ------------------ 228| | // Shared containers are not cloneable. Are you mixing COW and 229| | // non-COW bitmaps? 230| 0| return NULL; 231| 0| default: ------------------ | Branch (231:9): [True: 0, False: 353k] ------------------ 232| 0| assert(false); 233| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 234| 0| return NULL; 235| 353k| } 236| 353k|} container_init_iterator: 304| 19.1k| uint16_t *value) { 305| 19.1k| switch (typecode) { 306| 3.87k| case BITSET_CONTAINER_TYPE: { ------------------ | | 48| 3.87k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (306:9): [True: 3.87k, False: 15.2k] ------------------ 307| 3.87k| const bitset_container_t *bc = const_CAST_bitset(c); ------------------ | | 53| 3.87k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 3.87k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 308| 3.87k| uint32_t wordindex = 0; 309| 3.87k| uint64_t word; 310| 20.8k| while ((word = bc->words[wordindex]) == 0) { ------------------ | Branch (310:20): [True: 17.0k, False: 3.87k] ------------------ 311| 17.0k| wordindex++; 312| 17.0k| } 313| | // word is non-zero 314| 3.87k| int32_t index = wordindex * 64 + roaring_trailing_zeroes(word); 315| 3.87k| *value = index; 316| 3.87k| return ROARING_INIT_ROARING_CONTAINER_ITERATOR_T{ ------------------ | | 13| 3.87k|#define ROARING_INIT_ROARING_CONTAINER_ITERATOR_T (roaring_container_iterator_t) ------------------ 317| 3.87k| .index = index, 318| 3.87k| }; 319| 0| } 320| 14.7k| case ARRAY_CONTAINER_TYPE: { ------------------ | | 49| 14.7k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (320:9): [True: 14.7k, False: 4.37k] ------------------ 321| 14.7k| const array_container_t *ac = const_CAST_array(c); ------------------ | | 55| 14.7k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 14.7k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 322| 14.7k| *value = ac->array[0]; 323| 14.7k| return ROARING_INIT_ROARING_CONTAINER_ITERATOR_T{ ------------------ | | 13| 14.7k|#define ROARING_INIT_ROARING_CONTAINER_ITERATOR_T (roaring_container_iterator_t) ------------------ 324| 14.7k| .index = 0, 325| 14.7k| }; 326| 0| } 327| 503| case RUN_CONTAINER_TYPE: { ------------------ | | 50| 503|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (327:9): [True: 503, False: 18.6k] ------------------ 328| 503| const run_container_t *rc = const_CAST_run(c); ------------------ | | 78| 503|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 503|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 329| 503| *value = rc->runs[0].value; 330| 503| return ROARING_INIT_ROARING_CONTAINER_ITERATOR_T{ ------------------ | | 13| 503|#define ROARING_INIT_ROARING_CONTAINER_ITERATOR_T (roaring_container_iterator_t) ------------------ 331| 503| .index = 0, 332| 503| }; 333| 0| } 334| 0| default: ------------------ | Branch (334:9): [True: 0, False: 19.1k] ------------------ 335| 0| assert(false); 336| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 337| 0| return ROARING_INIT_ROARING_CONTAINER_ITERATOR_T{0}; ------------------ | | 13| 0|#define ROARING_INIT_ROARING_CONTAINER_ITERATOR_T (roaring_container_iterator_t) ------------------ 338| 19.1k| } 339| 19.1k|} container_iterator_lower_bound: 386| 6.38k| uint16_t *value_out, uint16_t val) { 387| 6.38k| if (val > container_maximum(c, typecode)) { ------------------ | Branch (387:9): [True: 0, False: 6.38k] ------------------ 388| 0| return false; 389| 0| } 390| 6.38k| switch (typecode) { 391| 1.91k| case BITSET_CONTAINER_TYPE: { ------------------ | | 48| 1.91k|#define BITSET_CONTAINER_TYPE 1 ------------------ | Branch (391:9): [True: 1.91k, False: 4.46k] ------------------ 392| 1.91k| const bitset_container_t *bc = const_CAST_bitset(c); ------------------ | | 53| 1.91k|#define const_CAST_bitset(c) CAST(const bitset_container_t *, c) | | ------------------ | | | | 86| 1.91k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 393| 1.91k| it->index = bitset_container_index_equalorlarger(bc, val); 394| 1.91k| *value_out = it->index; 395| 1.91k| return true; 396| 0| } 397| 4.46k| case ARRAY_CONTAINER_TYPE: { ------------------ | | 49| 4.46k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (397:9): [True: 4.46k, False: 1.91k] ------------------ 398| 4.46k| const array_container_t *ac = const_CAST_array(c); ------------------ | | 55| 4.46k|#define const_CAST_array(c) CAST(const array_container_t *, c) | | ------------------ | | | | 86| 4.46k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 399| 4.46k| it->index = array_container_index_equalorlarger(ac, val); 400| 4.46k| *value_out = ac->array[it->index]; 401| 4.46k| return true; 402| 0| } 403| 0| case RUN_CONTAINER_TYPE: { ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (403:9): [True: 0, False: 6.38k] ------------------ 404| 0| const run_container_t *rc = const_CAST_run(c); ------------------ | | 78| 0|#define const_CAST_run(c) CAST(const run_container_t *, c) | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 405| 0| it->index = run_container_index_equalorlarger(rc, val); 406| 0| if (rc->runs[it->index].value <= val) { ------------------ | Branch (406:17): [True: 0, False: 0] ------------------ 407| 0| *value_out = val; 408| 0| } else { 409| 0| *value_out = rc->runs[it->index].value; 410| 0| } 411| 0| return true; 412| 0| } 413| 0| default: ------------------ | Branch (413:9): [True: 0, False: 6.38k] ------------------ 414| 0| assert(false); 415| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 416| 0| return false; 417| 6.38k| } 418| 6.38k|} bitset_container_from_array: 22| 2.37k|bitset_container_t *bitset_container_from_array(const array_container_t *ac) { 23| 2.37k| bitset_container_t *ans = bitset_container_create(); 24| 2.37k| int limit = array_container_cardinality(ac); 25| 2.46M| for (int i = 0; i < limit; ++i) bitset_container_set(ans, ac->array[i]); ------------------ | Branch (25:21): [True: 2.46M, False: 2.37k] ------------------ 26| 2.37k| return ans; 27| 2.37k|} bitset_container_from_run: 29| 612|bitset_container_t *bitset_container_from_run(const run_container_t *arr) { 30| 612| int card = run_container_cardinality(arr); 31| 612| bitset_container_t *answer = bitset_container_create(); 32| 20.7k| for (int rlepos = 0; rlepos < arr->n_runs; ++rlepos) { ------------------ | Branch (32:26): [True: 20.1k, False: 612] ------------------ 33| 20.1k| rle16_t vl = arr->runs[rlepos]; 34| 20.1k| bitset_set_lenrange(answer->words, vl.value, vl.length); 35| 20.1k| } 36| 612| answer->cardinality = card; 37| 612| return answer; 38| 612|} array_container_from_run: 40| 1.18k|array_container_t *array_container_from_run(const run_container_t *arr) { 41| 1.18k| array_container_t *answer = 42| 1.18k| array_container_create_given_capacity(run_container_cardinality(arr)); 43| 1.18k| answer->cardinality = 0; 44| 88.1k| for (int rlepos = 0; rlepos < arr->n_runs; ++rlepos) { ------------------ | Branch (44:26): [True: 86.9k, False: 1.18k] ------------------ 45| 86.9k| int run_start = arr->runs[rlepos].value; 46| 86.9k| int run_end = run_start + arr->runs[rlepos].length; 47| | 48| 450k| for (int run_value = run_start; run_value <= run_end; ++run_value) { ------------------ | Branch (48:41): [True: 363k, False: 86.9k] ------------------ 49| 363k| answer->array[answer->cardinality++] = (uint16_t)run_value; 50| 363k| } 51| 86.9k| } 52| 1.18k| return answer; 53| 1.18k|} array_container_from_bitset: 55| 2.91k|array_container_t *array_container_from_bitset(const bitset_container_t *bits) { 56| 2.91k| array_container_t *result = 57| 2.91k| array_container_create_given_capacity(bits->cardinality); 58| 2.91k| result->cardinality = bits->cardinality; 59| 2.91k|#if CROARING_IS_X64 60| 2.91k|#if CROARING_COMPILER_SUPPORTS_AVX512 61| 2.91k| if (croaring_hardware_support() & ROARING_SUPPORTS_AVX512) { ------------------ | Branch (61:9): [True: 0, False: 2.91k] ------------------ 62| 0| bitset_extract_setbits_avx512_uint16( 63| 0| bits->words, BITSET_CONTAINER_SIZE_IN_WORDS, result->array, 64| 0| bits->cardinality, 0); 65| 0| } else 66| 2.91k|#endif 67| 2.91k| { 68| | // sse version ends up being slower here 69| | // (bitset_extract_setbits_sse_uint16) 70| | // because of the sparsity of the data 71| 2.91k| bitset_extract_setbits_uint16( 72| 2.91k| bits->words, BITSET_CONTAINER_SIZE_IN_WORDS, result->array, 0); 73| 2.91k| } 74| |#else 75| | // If the system is not x64, then we have no accelerated function. 76| | bitset_extract_setbits_uint16(bits->words, BITSET_CONTAINER_SIZE_IN_WORDS, 77| | result->array, 0); 78| |#endif 79| | 80| 2.91k| return result; 81| 2.91k|} convert_to_bitset_or_array_container: 120| 73.4k| uint8_t *resulttype) { 121| 73.4k| if (card <= DEFAULT_MAX_SIZE) { ------------------ | Branch (121:9): [True: 1.21k, False: 72.2k] ------------------ 122| 1.21k| array_container_t *answer = array_container_create_given_capacity(card); 123| 1.21k| answer->cardinality = 0; 124| 65.8k| for (int rlepos = 0; rlepos < rc->n_runs; ++rlepos) { ------------------ | Branch (124:30): [True: 64.6k, False: 1.21k] ------------------ 125| 64.6k| uint16_t run_start = rc->runs[rlepos].value; 126| 64.6k| uint16_t run_end = run_start + rc->runs[rlepos].length; 127| 645k| for (uint16_t run_value = run_start; run_value < run_end; ------------------ | Branch (127:50): [True: 580k, False: 64.6k] ------------------ 128| 580k| ++run_value) { 129| 580k| answer->array[answer->cardinality++] = run_value; 130| 580k| } 131| 64.6k| answer->array[answer->cardinality++] = run_end; 132| 64.6k| } 133| 1.21k| assert(card == answer->cardinality); 134| 1.21k| *resulttype = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 1.21k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 135| | // run_container_free(r); 136| 1.21k| return answer; 137| 1.21k| } 138| 72.2k| bitset_container_t *answer = bitset_container_create(); 139| 164k| for (int rlepos = 0; rlepos < rc->n_runs; ++rlepos) { ------------------ | Branch (139:26): [True: 92.1k, False: 72.2k] ------------------ 140| 92.1k| uint16_t run_start = rc->runs[rlepos].value; 141| 92.1k| bitset_set_lenrange(answer->words, run_start, rc->runs[rlepos].length); 142| 92.1k| } 143| 72.2k| answer->cardinality = card; 144| 72.2k| *resulttype = BITSET_CONTAINER_TYPE; ------------------ | | 48| 72.2k|#define BITSET_CONTAINER_TYPE 1 ------------------ 145| | // run_container_free(r); 146| 72.2k| return answer; 147| 73.4k|} convert_run_to_efficient_container: 155| 141k| uint8_t *typecode_after) { 156| 141k| int32_t size_as_run_container = 157| 141k| run_container_serialized_size_in_bytes(c->n_runs); 158| | 159| 141k| int32_t size_as_bitset_container = 160| 141k| bitset_container_serialized_size_in_bytes(); 161| 141k| int32_t card = run_container_cardinality(c); 162| 141k| int32_t size_as_array_container = 163| 141k| array_container_serialized_size_in_bytes(card); 164| | 165| 141k| int32_t min_size_non_run = 166| 141k| size_as_bitset_container < size_as_array_container ------------------ | Branch (166:9): [True: 63.4k, False: 77.8k] ------------------ 167| 141k| ? size_as_bitset_container 168| 141k| : size_as_array_container; 169| 141k| if (size_as_run_container <= min_size_non_run) { // no conversion ------------------ | Branch (169:9): [True: 71.8k, False: 69.4k] ------------------ 170| 71.8k| *typecode_after = RUN_CONTAINER_TYPE; ------------------ | | 50| 71.8k|#define RUN_CONTAINER_TYPE 3 ------------------ 171| 71.8k| return c; 172| 71.8k| } 173| 69.4k| if (card <= DEFAULT_MAX_SIZE) { ------------------ | Branch (173:9): [True: 69.4k, False: 0] ------------------ 174| | // to array 175| 69.4k| array_container_t *answer = array_container_create_given_capacity(card); 176| 69.4k| answer->cardinality = 0; 177| 212k| for (int rlepos = 0; rlepos < c->n_runs; ++rlepos) { ------------------ | Branch (177:30): [True: 142k, False: 69.4k] ------------------ 178| 142k| int run_start = c->runs[rlepos].value; 179| 142k| int run_end = run_start + c->runs[rlepos].length; 180| | 181| 381k| for (int run_value = run_start; run_value <= run_end; ++run_value) { ------------------ | Branch (181:45): [True: 238k, False: 142k] ------------------ 182| 238k| answer->array[answer->cardinality++] = (uint16_t)run_value; 183| 238k| } 184| 142k| } 185| 69.4k| *typecode_after = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 69.4k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 186| 69.4k| return answer; 187| 69.4k| } 188| | 189| | // else to bitset 190| 0| bitset_container_t *answer = bitset_container_create(); 191| | 192| 0| for (int rlepos = 0; rlepos < c->n_runs; ++rlepos) { ------------------ | Branch (192:26): [True: 0, False: 0] ------------------ 193| 0| int start = c->runs[rlepos].value; 194| 0| int end = start + c->runs[rlepos].length; 195| 0| bitset_set_range(answer->words, start, end + 1); 196| 0| } 197| 0| answer->cardinality = card; 198| 0| *typecode_after = BITSET_CONTAINER_TYPE; ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 199| 0| return answer; 200| 69.4k|} convert_run_to_efficient_container_and_free: 204| 128k| run_container_t *c, uint8_t *typecode_after) { 205| 128k| container_t *answer = convert_run_to_efficient_container(c, typecode_after); 206| 128k| if (answer != c) run_container_free(c); ------------------ | Branch (206:9): [True: 62.3k, False: 66.5k] ------------------ 207| 128k| return answer; 208| 128k|} convert_run_optimize: 218| 12.9k| uint8_t *typecode_after) { 219| 12.9k| if (typecode_original == RUN_CONTAINER_TYPE) { ------------------ | | 50| 12.9k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (219:9): [True: 0, False: 12.9k] ------------------ 220| 0| container_t *newc = 221| 0| convert_run_to_efficient_container(CAST_run(c), typecode_after); ------------------ | | 77| 0|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 222| 0| if (newc != c) { ------------------ | Branch (222:13): [True: 0, False: 0] ------------------ 223| 0| container_free(c, typecode_original); 224| 0| } 225| 0| return newc; 226| 12.9k| } else if (typecode_original == ARRAY_CONTAINER_TYPE) { ------------------ | | 49| 12.9k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (226:16): [True: 12.9k, False: 0] ------------------ 227| | // it might need to be converted to a run container. 228| 12.9k| array_container_t *c_qua_array = CAST_array(c); ------------------ | | 54| 12.9k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 12.9k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 229| 12.9k| int32_t n_runs = array_container_number_of_runs(c_qua_array); 230| 12.9k| int32_t size_as_run_container = 231| 12.9k| run_container_serialized_size_in_bytes(n_runs); 232| 12.9k| int32_t card = array_container_cardinality(c_qua_array); 233| 12.9k| int32_t size_as_array_container = 234| 12.9k| array_container_serialized_size_in_bytes(card); 235| | 236| 12.9k| if (size_as_run_container >= size_as_array_container) { ------------------ | Branch (236:13): [True: 8.85k, False: 4.10k] ------------------ 237| 8.85k| *typecode_after = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 8.85k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 238| 8.85k| return c; 239| 8.85k| } 240| | // else convert array to run container 241| 4.10k| run_container_t *answer = run_container_create_given_capacity(n_runs); 242| 4.10k| int prev = -2; 243| 4.10k| int run_start = -1; 244| | 245| 4.10k| assert(card > 0); 246| 663k| for (int i = 0; i < card; ++i) { ------------------ | Branch (246:25): [True: 659k, False: 4.10k] ------------------ 247| 659k| uint16_t cur_val = c_qua_array->array[i]; 248| 659k| if (cur_val != prev + 1) { ------------------ | Branch (248:17): [True: 269k, False: 389k] ------------------ 249| | // new run starts; flush old one, if any 250| 269k| if (run_start != -1) add_run(answer, run_start, prev); ------------------ | Branch (250:21): [True: 265k, False: 4.10k] ------------------ 251| 269k| run_start = cur_val; 252| 269k| } 253| 659k| prev = c_qua_array->array[i]; 254| 659k| } 255| 4.10k| assert(run_start >= 0); 256| | // now prev is the last seen value 257| 4.10k| add_run(answer, run_start, prev); 258| 4.10k| *typecode_after = RUN_CONTAINER_TYPE; ------------------ | | 50| 4.10k|#define RUN_CONTAINER_TYPE 3 ------------------ 259| 4.10k| array_container_free(c_qua_array); 260| 4.10k| return answer; 261| 12.9k| } else if (typecode_original == ------------------ | Branch (261:16): [True: 0, False: 0] ------------------ 262| 0| BITSET_CONTAINER_TYPE) { // run conversions on bitset ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 263| | // does bitset need conversion to run? 264| 0| bitset_container_t *c_qua_bitset = CAST_bitset(c); ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 265| 0| int32_t n_runs = bitset_container_number_of_runs(c_qua_bitset); 266| 0| int32_t size_as_run_container = 267| 0| run_container_serialized_size_in_bytes(n_runs); 268| 0| int32_t size_as_bitset_container = 269| 0| bitset_container_serialized_size_in_bytes(); 270| | 271| 0| if (size_as_bitset_container <= size_as_run_container) { ------------------ | Branch (271:13): [True: 0, False: 0] ------------------ 272| | // no conversion needed. 273| 0| *typecode_after = BITSET_CONTAINER_TYPE; ------------------ | | 48| 0|#define BITSET_CONTAINER_TYPE 1 ------------------ 274| 0| return c; 275| 0| } 276| | // bitset to runcontainer (ported from Java RunContainer( 277| | // BitmapContainer bc, int nbrRuns)) 278| 0| assert(n_runs > 0); // no empty bitmaps 279| 0| run_container_t *answer = run_container_create_given_capacity(n_runs); 280| | 281| 0| int long_ctr = 0; 282| 0| uint64_t cur_word = c_qua_bitset->words[0]; 283| 0| while (true) { ------------------ | Branch (283:16): [True: 0, Folded] ------------------ 284| 0| while (cur_word == UINT64_C(0) && ------------------ | Branch (284:20): [True: 0, False: 0] ------------------ 285| 0| long_ctr < BITSET_CONTAINER_SIZE_IN_WORDS - 1) ------------------ | Branch (285:20): [True: 0, False: 0] ------------------ 286| 0| cur_word = c_qua_bitset->words[++long_ctr]; 287| | 288| 0| if (cur_word == UINT64_C(0)) { ------------------ | Branch (288:17): [True: 0, False: 0] ------------------ 289| 0| bitset_container_free(c_qua_bitset); 290| 0| *typecode_after = RUN_CONTAINER_TYPE; ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ 291| 0| return answer; 292| 0| } 293| | 294| 0| int local_run_start = roaring_trailing_zeroes(cur_word); 295| 0| int run_start = local_run_start + 64 * long_ctr; 296| 0| uint64_t cur_word_with_1s = cur_word | (cur_word - 1); 297| | 298| 0| int run_end = 0; 299| 0| while (cur_word_with_1s == UINT64_C(0xFFFFFFFFFFFFFFFF) && ------------------ | Branch (299:20): [True: 0, False: 0] ------------------ 300| 0| long_ctr < BITSET_CONTAINER_SIZE_IN_WORDS - 1) ------------------ | Branch (300:20): [True: 0, False: 0] ------------------ 301| 0| cur_word_with_1s = c_qua_bitset->words[++long_ctr]; 302| | 303| 0| if (cur_word_with_1s == UINT64_C(0xFFFFFFFFFFFFFFFF)) { ------------------ | Branch (303:17): [True: 0, False: 0] ------------------ 304| 0| run_end = 64 + long_ctr * 64; // exclusive, I guess 305| 0| add_run(answer, run_start, run_end - 1); 306| 0| bitset_container_free(c_qua_bitset); 307| 0| *typecode_after = RUN_CONTAINER_TYPE; ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ 308| 0| return answer; 309| 0| } 310| 0| int local_run_end = roaring_trailing_zeroes(~cur_word_with_1s); 311| 0| run_end = local_run_end + long_ctr * 64; 312| 0| add_run(answer, run_start, run_end - 1); 313| 0| cur_word = cur_word_with_1s & (cur_word_with_1s + 1); 314| 0| } 315| 0| return answer; 316| 0| } else { 317| 0| assert(false); 318| 0| roaring_unreachable; ------------------ | | 240| 0|#define roaring_unreachable __builtin_unreachable() ------------------ 319| 0| return NULL; 320| 0| } 321| 12.9k|} convert.c:add_run: 84| 269k|static void add_run(run_container_t *rc, int s, int e) { 85| 269k| rc->runs[rc->n_runs].value = s; 86| 269k| rc->runs[rc->n_runs].length = e - s; 87| 269k| rc->n_runs++; 88| 269k|} array_bitset_container_andnot: 26| 575| array_container_t *dst) { 27| | // follows Java implementation as of June 2016 28| 575| if (dst->capacity < src_1->cardinality) { ------------------ | Branch (28:9): [True: 575, False: 0] ------------------ 29| 575| array_container_grow(dst, src_1->cardinality, false); 30| 575| } 31| 575| int32_t newcard = 0; 32| 575| const int32_t origcard = src_1->cardinality; 33| 102k| for (int i = 0; i < origcard; ++i) { ------------------ | Branch (33:21): [True: 101k, False: 575] ------------------ 34| 101k| uint16_t key = src_1->array[i]; 35| 101k| dst->array[newcard] = key; 36| 101k| newcard += 1 - bitset_container_contains(src_2, key); 37| 101k| } 38| 575| dst->cardinality = newcard; 39| 575|} bitset_array_container_iandnot: 83| 1.57k| container_t **dst) { 84| 1.57k| *dst = src_1; 85| 1.57k| src_1->cardinality = 86| 1.57k| (int32_t)bitset_clear_list(src_1->words, (uint64_t)src_1->cardinality, 87| 1.57k| src_2->array, (uint64_t)src_2->cardinality); 88| | 89| 1.57k| if (src_1->cardinality <= DEFAULT_MAX_SIZE) { ------------------ | Branch (89:9): [True: 201, False: 1.36k] ------------------ 90| 201| *dst = array_container_from_bitset(src_1); 91| 201| bitset_container_free(src_1); 92| 201| return false; // not bitset 93| 201| } else 94| 1.36k| return true; 95| 1.57k|} run_bitset_container_andnot: 106| 691| container_t **dst) { 107| | // follows the Java implementation as of June 2016 108| 691| int card = run_container_cardinality(src_1); 109| 691| if (card <= DEFAULT_MAX_SIZE) { ------------------ | Branch (109:9): [True: 691, False: 0] ------------------ 110| | // must be an array 111| 691| array_container_t *answer = array_container_create_given_capacity(card); 112| 691| answer->cardinality = 0; 113| 57.0k| for (int32_t rlepos = 0; rlepos < src_1->n_runs; ++rlepos) { ------------------ | Branch (113:34): [True: 56.3k, False: 691] ------------------ 114| 56.3k| rle16_t rle = src_1->runs[rlepos]; 115| 202k| for (int run_value = rle.value; run_value <= rle.value + rle.length; ------------------ | Branch (115:45): [True: 146k, False: 56.3k] ------------------ 116| 146k| ++run_value) { 117| 146k| if (!bitset_container_get(src_2, (uint16_t)run_value)) { ------------------ | Branch (117:21): [True: 42.8k, False: 103k] ------------------ 118| 42.8k| answer->array[answer->cardinality++] = (uint16_t)run_value; 119| 42.8k| } 120| 146k| } 121| 56.3k| } 122| 691| *dst = answer; 123| 691| return false; 124| 691| } else { // we guess it will be a bitset, though have to check guess when 125| | // done 126| 0| bitset_container_t *answer = bitset_container_clone(src_2); 127| | 128| 0| uint32_t last_pos = 0; 129| 0| for (int32_t rlepos = 0; rlepos < src_1->n_runs; ++rlepos) { ------------------ | Branch (129:34): [True: 0, False: 0] ------------------ 130| 0| rle16_t rle = src_1->runs[rlepos]; 131| | 132| 0| uint32_t start = rle.value; 133| 0| uint32_t end = start + rle.length + 1; 134| 0| bitset_reset_range(answer->words, last_pos, start); 135| 0| bitset_flip_range(answer->words, start, end); 136| 0| last_pos = end; 137| 0| } 138| 0| bitset_reset_range(answer->words, last_pos, (uint32_t)(1 << 16)); 139| | 140| 0| answer->cardinality = bitset_container_compute_cardinality(answer); 141| | 142| 0| if (answer->cardinality <= DEFAULT_MAX_SIZE) { ------------------ | Branch (142:13): [True: 0, False: 0] ------------------ 143| 0| *dst = array_container_from_bitset(answer); 144| 0| bitset_container_free(answer); 145| 0| return false; // not bitset 146| 0| } 147| 0| *dst = answer; 148| | return true; // bitset 149| 0| } 150| 691|} run_array_container_andnot: 279| 2.07k| container_t **dst) { 280| | // follows the Java impl as of June 2016 281| | 282| 2.07k| int card = run_container_cardinality(src_1); 283| 2.07k| const int arbitrary_threshold = 32; 284| | 285| 2.07k| if (card <= arbitrary_threshold) { ------------------ | Branch (285:9): [True: 381, False: 1.69k] ------------------ 286| 381| if (src_2->cardinality == 0) { ------------------ | Branch (286:13): [True: 0, False: 381] ------------------ 287| 0| *dst = run_container_clone(src_1); 288| 0| return RUN_CONTAINER_TYPE; ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ 289| 0| } 290| | // Java's "lazyandNot.toEfficientContainer" thing 291| 381| run_container_t *answer = run_container_create_given_capacity( 292| 381| card + array_container_cardinality(src_2)); 293| | 294| 381| int rlepos = 0; 295| 381| int xrlepos = 0; // "x" is src_2 296| 381| rle16_t rle = src_1->runs[rlepos]; 297| 381| int32_t start = rle.value; 298| 381| int32_t end = start + rle.length + 1; 299| 381| int32_t xstart = src_2->array[xrlepos]; 300| | 301| 13.7k| while ((rlepos < src_1->n_runs) && (xrlepos < src_2->cardinality)) { ------------------ | Branch (301:16): [True: 13.4k, False: 302] | Branch (301:44): [True: 13.3k, False: 79] ------------------ 302| 13.3k| if (end <= xstart) { ------------------ | Branch (302:17): [True: 639, False: 12.7k] ------------------ 303| | // output the first run 304| 639| answer->runs[answer->n_runs++] = 305| 639| CROARING_MAKE_RLE16(start, end - start - 1); ------------------ | | 60| 639| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 306| 639| rlepos++; 307| 639| if (rlepos < src_1->n_runs) { ------------------ | Branch (307:21): [True: 584, False: 55] ------------------ 308| 584| start = src_1->runs[rlepos].value; 309| 584| end = start + src_1->runs[rlepos].length + 1; 310| 584| } 311| 12.7k| } else if (xstart + 1 <= start) { ------------------ | Branch (311:24): [True: 8.14k, False: 4.57k] ------------------ 312| | // exit the second run 313| 8.14k| xrlepos++; 314| 8.14k| if (xrlepos < src_2->cardinality) { ------------------ | Branch (314:21): [True: 8.06k, False: 79] ------------------ 315| 8.06k| xstart = src_2->array[xrlepos]; 316| 8.06k| } 317| 8.14k| } else { 318| 4.57k| if (start < xstart) { ------------------ | Branch (318:21): [True: 323, False: 4.25k] ------------------ 319| 323| answer->runs[answer->n_runs++] = 320| 323| CROARING_MAKE_RLE16(start, xstart - start - 1); ------------------ | | 60| 323| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 321| 323| } 322| 4.57k| if (xstart + 1 < end) { ------------------ | Branch (322:21): [True: 2.77k, False: 1.80k] ------------------ 323| 2.77k| start = xstart + 1; 324| 2.77k| } else { 325| 1.80k| rlepos++; 326| 1.80k| if (rlepos < src_1->n_runs) { ------------------ | Branch (326:25): [True: 1.55k, False: 247] ------------------ 327| 1.55k| start = src_1->runs[rlepos].value; 328| 1.55k| end = start + src_1->runs[rlepos].length + 1; 329| 1.55k| } 330| 1.80k| } 331| 4.57k| } 332| 13.3k| } 333| 381| if (rlepos < src_1->n_runs) { ------------------ | Branch (333:13): [True: 79, False: 302] ------------------ 334| 79| answer->runs[answer->n_runs++] = 335| 79| CROARING_MAKE_RLE16(start, end - start - 1); ------------------ | | 60| 79| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 336| 79| rlepos++; 337| 79| if (rlepos < src_1->n_runs) { ------------------ | Branch (337:17): [True: 25, False: 54] ------------------ 338| 25| memcpy(answer->runs + answer->n_runs, src_1->runs + rlepos, 339| 25| (src_1->n_runs - rlepos) * sizeof(rle16_t)); 340| 25| answer->n_runs += (src_1->n_runs - rlepos); 341| 25| } 342| 79| } 343| 381| uint8_t return_type; 344| 381| *dst = convert_run_to_efficient_container(answer, &return_type); 345| 381| if (answer != *dst) run_container_free(answer); ------------------ | Branch (345:13): [True: 242, False: 139] ------------------ 346| 381| return return_type; 347| 381| } 348| | // else it's a bitmap or array 349| | 350| 1.69k| if (card <= DEFAULT_MAX_SIZE) { ------------------ | Branch (350:9): [True: 1.36k, False: 333] ------------------ 351| 1.36k| array_container_t *ac = array_container_create_given_capacity(card); 352| | // nb Java code used a generic iterator-based merge to compute 353| | // difference 354| 1.36k| ac->cardinality = run_array_array_subtract(src_1, src_2, ac); 355| 1.36k| *dst = ac; 356| 1.36k| return ARRAY_CONTAINER_TYPE; ------------------ | | 49| 1.36k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 357| 1.36k| } 358| 333| bitset_container_t *ans = bitset_container_from_run(src_1); 359| 333| bool result_is_bitset = bitset_array_container_iandnot(ans, src_2, dst); 360| 333| return (result_is_bitset ? BITSET_CONTAINER_TYPE : ARRAY_CONTAINER_TYPE); ------------------ | | 48| 315|#define BITSET_CONTAINER_TYPE 1 ------------------ return (result_is_bitset ? BITSET_CONTAINER_TYPE : ARRAY_CONTAINER_TYPE); ------------------ | | 49| 18|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (360:13): [True: 315, False: 18] ------------------ 361| 1.69k|} run_array_container_iandnot: 372| 801| container_t **dst) { 373| | // dummy implementation same as June 2016 Java 374| 801| int ans = run_array_container_andnot(src_1, src_2, dst); 375| 801| run_container_free(src_1); 376| 801| return ans; 377| 801|} array_run_container_andnot: 383| 825| array_container_t *dst) { 384| | // basically following Java impl as of June 2016 385| 825| if (src_1->cardinality > dst->capacity) { ------------------ | Branch (385:9): [True: 780, False: 45] ------------------ 386| 780| array_container_grow(dst, src_1->cardinality, false); 387| 780| } 388| | 389| 825| if (src_2->n_runs == 0) { ------------------ | Branch (389:9): [True: 0, False: 825] ------------------ 390| 0| memmove(dst->array, src_1->array, 391| 0| sizeof(uint16_t) * src_1->cardinality); 392| 0| dst->cardinality = src_1->cardinality; 393| 0| return; 394| 0| } 395| 825| int32_t run_start = src_2->runs[0].value; 396| 825| int32_t run_end = run_start + src_2->runs[0].length; 397| 825| int which_run = 0; 398| | 399| 825| uint16_t val = 0; 400| 825| int dest_card = 0; 401| 176k| for (int i = 0; i < src_1->cardinality; ++i) { ------------------ | Branch (401:21): [True: 175k, False: 825] ------------------ 402| 175k| val = src_1->array[i]; 403| 175k| if (val < run_start) ------------------ | Branch (403:13): [True: 44.3k, False: 130k] ------------------ 404| 44.3k| dst->array[dest_card++] = val; 405| 130k| else if (val <= run_end) { ------------------ | Branch (405:18): [True: 95.3k, False: 35.6k] ------------------ 406| 95.3k| ; // omitted item 407| 95.3k| } else { 408| 49.8k| do { 409| 49.8k| if (which_run + 1 < src_2->n_runs) { ------------------ | Branch (409:21): [True: 49.6k, False: 197] ------------------ 410| 49.6k| ++which_run; 411| 49.6k| run_start = src_2->runs[which_run].value; 412| 49.6k| run_end = run_start + src_2->runs[which_run].length; 413| | 414| 49.6k| } else 415| 197| run_start = run_end = (1 << 16) + 1; 416| 49.8k| } while (val > run_end); ------------------ | Branch (416:22): [True: 14.2k, False: 35.6k] ------------------ 417| 35.6k| --i; 418| 35.6k| } 419| 175k| } 420| 825| dst->cardinality = dest_card; 421| 825|} array_run_container_iandnot: 428| 45| const run_container_t *src_2) { 429| 45| array_run_container_andnot(src_1, src_2, src_1); 430| 45|} run_run_container_andnot: 437| 1.09k| const run_container_t *src_2, container_t **dst) { 438| 1.09k| run_container_t *ans = run_container_create(); 439| 1.09k| run_container_andnot(src_1, src_2, ans); 440| 1.09k| uint8_t typecode_after; 441| 1.09k| *dst = convert_run_to_efficient_container_and_free(ans, &typecode_after); 442| 1.09k| return typecode_after; 443| 1.09k|} run_run_container_iandnot: 453| 432| const run_container_t *src_2, container_t **dst) { 454| | // following Java impl as of June 2016 (dummy) 455| 432| int ans = run_run_container_andnot(src_1, src_2, dst); 456| 432| run_container_free(src_1); 457| 432| return ans; 458| 432|} array_array_container_andnot: 466| 2.32k| array_container_t *dst) { 467| 2.32k| array_container_andnot(src_1, src_2, dst); 468| 2.32k|} array_array_container_iandnot: 473| 3.54k| const array_container_t *src_2) { 474| 3.54k| array_container_andnot(src_1, src_2, src_1); 475| 3.54k|} mixed_andnot.c:run_array_array_subtract: 235| 1.36k| array_container_t *a_out) { 236| 1.36k| int out_card = 0; 237| 1.36k| int32_t in_array_pos = 238| 1.36k| -1; // since advanceUntil always assumes we start the search AFTER this 239| | 240| 103k| for (int rlepos = 0; rlepos < rc->n_runs; rlepos++) { ------------------ | Branch (240:26): [True: 101k, False: 1.36k] ------------------ 241| 101k| int32_t start = rc->runs[rlepos].value; 242| 101k| int32_t end = start + rc->runs[rlepos].length + 1; 243| | 244| 101k| in_array_pos = advanceUntil(a_in->array, in_array_pos, 245| 101k| a_in->cardinality, (uint16_t)start); 246| | 247| 101k| if (in_array_pos >= a_in->cardinality) { // run has no items subtracted ------------------ | Branch (247:13): [True: 5.00k, False: 96.7k] ------------------ 248| 79.1k| for (int32_t i = start; i < end; ++i) ------------------ | Branch (248:37): [True: 74.1k, False: 5.00k] ------------------ 249| 74.1k| a_out->array[out_card++] = (uint16_t)i; 250| 96.7k| } else { 251| 96.7k| uint16_t next_nonincluded = a_in->array[in_array_pos]; 252| 96.7k| if (next_nonincluded >= end) { ------------------ | Branch (252:17): [True: 33.1k, False: 63.5k] ------------------ 253| | // another case when run goes unaltered 254| 89.1k| for (int32_t i = start; i < end; ++i) ------------------ | Branch (254:41): [True: 55.9k, False: 33.1k] ------------------ 255| 55.9k| a_out->array[out_card++] = (uint16_t)i; 256| 33.1k| in_array_pos--; // ensure we see this item again if necessary 257| 63.5k| } else { 258| 318k| for (int32_t i = start; i < end; ++i) ------------------ | Branch (258:41): [True: 254k, False: 63.5k] ------------------ 259| 254k| if (i != next_nonincluded) ------------------ | Branch (259:25): [True: 113k, False: 141k] ------------------ 260| 113k| a_out->array[out_card++] = (uint16_t)i; 261| 141k| else // 0 should ensure we don't match 262| 141k| next_nonincluded = 263| 141k| (in_array_pos + 1 >= a_in->cardinality) ------------------ | Branch (263:29): [True: 1.08k, False: 140k] ------------------ 264| 141k| ? 0 265| 141k| : a_in->array[++in_array_pos]; 266| 63.5k| in_array_pos--; // see again 267| 63.5k| } 268| 96.7k| } 269| 101k| } 270| 1.36k| return out_card; 271| 1.36k|} array_container_equal_bitset: 10| 469| const bitset_container_t* container2) { 11| 469| if (container2->cardinality != BITSET_UNKNOWN_CARDINALITY) { ------------------ | Branch (11:9): [True: 469, False: 0] ------------------ 12| 469| if (container2->cardinality != container1->cardinality) { ------------------ | Branch (12:13): [True: 469, False: 0] ------------------ 13| 469| return false; 14| 469| } 15| 469| } 16| 0| int32_t pos = 0; 17| 0| for (int32_t i = 0; i < BITSET_CONTAINER_SIZE_IN_WORDS; ++i) { ------------------ | Branch (17:25): [True: 0, False: 0] ------------------ 18| 0| uint64_t w = container2->words[i]; 19| 0| while (w != 0) { ------------------ | Branch (19:16): [True: 0, False: 0] ------------------ 20| 0| uint64_t t = w & (~w + 1); 21| 0| uint16_t r = i * 64 + roaring_trailing_zeroes(w); 22| 0| if (pos >= container1->cardinality) { ------------------ | Branch (22:17): [True: 0, False: 0] ------------------ 23| 0| return false; 24| 0| } 25| 0| if (container1->array[pos] != r) { ------------------ | Branch (25:17): [True: 0, False: 0] ------------------ 26| 0| return false; 27| 0| } 28| 0| ++pos; 29| 0| w ^= t; 30| 0| } 31| 0| } 32| 0| return (pos == container1->cardinality); 33| 0|} run_container_equals_array: 36| 886| const array_container_t* container2) { 37| 886| if (run_container_cardinality(container1) != container2->cardinality) ------------------ | Branch (37:9): [True: 832, False: 54] ------------------ 38| 832| return false; 39| 54| int32_t pos = 0; 40| 54| for (int i = 0; i < container1->n_runs; ++i) { ------------------ | Branch (40:21): [True: 54, False: 0] ------------------ 41| 54| const uint32_t run_start = container1->runs[i].value; 42| 54| const uint32_t le = container1->runs[i].length; 43| | 44| 54| if (container2->array[pos] != run_start) { ------------------ | Branch (44:13): [True: 54, False: 0] ------------------ 45| 54| return false; 46| 54| } 47| | 48| 0| if (container2->array[pos + le] != run_start + le) { ------------------ | Branch (48:13): [True: 0, False: 0] ------------------ 49| 0| return false; 50| 0| } 51| | 52| 0| pos += le + 1; 53| 0| } 54| 0| return true; 55| 54|} array_bitset_container_intersection: 21| 627| array_container_t *dst) { 22| 627| if (dst->capacity < src_1->cardinality) { ------------------ | Branch (22:9): [True: 627, False: 0] ------------------ 23| 627| array_container_grow(dst, src_1->cardinality, false); 24| 627| } 25| 627| int32_t newcard = 0; // dst could be src_1 26| 627| const int32_t origcard = src_1->cardinality; 27| 314k| for (int i = 0; i < origcard; ++i) { ------------------ | Branch (27:21): [True: 313k, False: 627] ------------------ 28| 313k| uint16_t key = src_1->array[i]; 29| | // this branchless approach is much faster... 30| 313k| dst->array[newcard] = key; 31| 313k| newcard += bitset_container_contains(src_2, key); 32| | /** 33| | * we could do it this way instead... 34| | * if (bitset_container_contains(src_2, key)) { 35| | * dst->array[newcard++] = key; 36| | * } 37| | * but if the result is unpredictible, the processor generates 38| | * many mispredicted branches. 39| | * Difference can be huge (from 3 cycles when predictible all the way 40| | * to 16 cycles when unpredictible. 41| | * See 42| | * https://github.com/lemire/Code-used-on-Daniel-Lemire-s-blog/blob/master/extra/bitset/c/arraybitsetintersection.c 43| | */ 44| 313k| } 45| 627| dst->cardinality = newcard; 46| 627|} array_bitset_container_intersection_cardinality: 50| 2.30k| const array_container_t *src_1, const bitset_container_t *src_2) { 51| 2.30k| int32_t newcard = 0; 52| 2.30k| const int32_t origcard = src_1->cardinality; 53| 408k| for (int i = 0; i < origcard; ++i) { ------------------ | Branch (53:21): [True: 406k, False: 2.30k] ------------------ 54| 406k| uint16_t key = src_1->array[i]; 55| 406k| newcard += bitset_container_contains(src_2, key); 56| 406k| } 57| 2.30k| return newcard; 58| 2.30k|} array_bitset_container_intersect: 61| 575| const bitset_container_t *src_2) { 62| 575| const int32_t origcard = src_1->cardinality; 63| 3.93k| for (int i = 0; i < origcard; ++i) { ------------------ | Branch (63:21): [True: 3.91k, False: 14] ------------------ 64| 3.91k| uint16_t key = src_1->array[i]; 65| 3.91k| if (bitset_container_contains(src_2, key)) return true; ------------------ | Branch (65:13): [True: 561, False: 3.35k] ------------------ 66| 3.91k| } 67| 14| return false; 68| 575|} array_run_container_intersection: 75| 3.35k| array_container_t *dst) { 76| 3.35k| if (run_container_is_full(src_2)) { ------------------ | Branch (76:9): [True: 34, False: 3.32k] ------------------ 77| 34| if (dst != src_1) array_container_copy(src_1, dst); ------------------ | Branch (77:13): [True: 34, False: 0] ------------------ 78| 34| return; 79| 34| } 80| 3.32k| if (dst->capacity < src_1->cardinality) { ------------------ | Branch (80:9): [True: 3.32k, False: 0] ------------------ 81| 3.32k| array_container_grow(dst, src_1->cardinality, false); 82| 3.32k| } 83| 3.32k| if (src_2->n_runs == 0) { ------------------ | Branch (83:9): [True: 0, False: 3.32k] ------------------ 84| 0| return; 85| 0| } 86| 3.32k| int32_t rlepos = 0; 87| 3.32k| int32_t arraypos = 0; 88| 3.32k| rle16_t rle = src_2->runs[rlepos]; 89| 3.32k| int32_t newcard = 0; 90| 457k| while (arraypos < src_1->cardinality) { ------------------ | Branch (90:12): [True: 454k, False: 2.75k] ------------------ 91| 454k| const uint16_t arrayval = src_1->array[arraypos]; 92| 610k| while (rle.value + rle.length < ------------------ | Branch (92:16): [True: 156k, False: 453k] ------------------ 93| 610k| arrayval) { // this will frequently be false 94| 156k| ++rlepos; 95| 156k| if (rlepos == src_2->n_runs) { ------------------ | Branch (95:17): [True: 570, False: 156k] ------------------ 96| 570| dst->cardinality = newcard; 97| 570| return; // we are done 98| 570| } 99| 156k| rle = src_2->runs[rlepos]; 100| 156k| } 101| 453k| if (rle.value > arrayval) { ------------------ | Branch (101:13): [True: 30.6k, False: 422k] ------------------ 102| 30.6k| arraypos = advanceUntil(src_1->array, arraypos, src_1->cardinality, 103| 30.6k| rle.value); 104| 422k| } else { 105| 422k| dst->array[newcard] = arrayval; 106| 422k| newcard++; 107| 422k| arraypos++; 108| 422k| } 109| 453k| } 110| 2.75k| dst->cardinality = newcard; 111| 2.75k|} run_bitset_container_intersection: 119| 691| container_t **dst) { 120| 691| if (run_container_is_full(src_1)) { ------------------ | Branch (120:9): [True: 0, False: 691] ------------------ 121| 0| if (*dst != src_2) *dst = bitset_container_clone(src_2); ------------------ | Branch (121:13): [True: 0, False: 0] ------------------ 122| 0| return true; 123| 0| } 124| 691| int32_t card = run_container_cardinality(src_1); 125| 691| if (card <= DEFAULT_MAX_SIZE) { ------------------ | Branch (125:9): [True: 691, False: 0] ------------------ 126| | // result can only be an array (assuming that we never make a 127| | // RunContainer) 128| 691| if (card > src_2->cardinality) { ------------------ | Branch (128:13): [True: 0, False: 691] ------------------ 129| 0| card = src_2->cardinality; 130| 0| } 131| 691| array_container_t *answer = array_container_create_given_capacity(card); 132| 691| *dst = answer; 133| 691| if (*dst == NULL) { ------------------ | Branch (133:13): [True: 0, False: 691] ------------------ 134| 0| return false; 135| 0| } 136| 57.0k| for (int32_t rlepos = 0; rlepos < src_1->n_runs; ++rlepos) { ------------------ | Branch (136:34): [True: 56.3k, False: 691] ------------------ 137| 56.3k| rle16_t rle = src_1->runs[rlepos]; 138| 56.3k| uint32_t endofrun = (uint32_t)rle.value + rle.length; 139| 202k| for (uint32_t runValue = rle.value; runValue <= endofrun; ------------------ | Branch (139:49): [True: 146k, False: 56.3k] ------------------ 140| 146k| ++runValue) { 141| 146k| answer->array[answer->cardinality] = (uint16_t)runValue; 142| 146k| answer->cardinality += 143| 146k| bitset_container_contains(src_2, runValue); 144| 146k| } 145| 56.3k| } 146| 691| return false; 147| 691| } 148| 0| if (*dst == src_2) { // we attempt in-place ------------------ | Branch (148:9): [True: 0, False: 0] ------------------ 149| 0| bitset_container_t *answer = CAST_bitset(*dst); ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 150| 0| uint32_t start = 0; 151| 0| for (int32_t rlepos = 0; rlepos < src_1->n_runs; ++rlepos) { ------------------ | Branch (151:34): [True: 0, False: 0] ------------------ 152| 0| const rle16_t rle = src_1->runs[rlepos]; 153| 0| uint32_t end = rle.value; 154| 0| bitset_reset_range(src_2->words, start, end); 155| | 156| 0| start = end + rle.length + 1; 157| 0| } 158| 0| bitset_reset_range(src_2->words, start, UINT32_C(1) << 16); 159| 0| answer->cardinality = bitset_container_compute_cardinality(answer); 160| 0| if (src_2->cardinality > DEFAULT_MAX_SIZE) { ------------------ | Branch (160:13): [True: 0, False: 0] ------------------ 161| 0| return true; 162| 0| } else { 163| 0| array_container_t *newanswer = array_container_from_bitset(src_2); 164| 0| if (newanswer == NULL) { ------------------ | Branch (164:17): [True: 0, False: 0] ------------------ 165| 0| *dst = NULL; 166| 0| return false; 167| 0| } 168| 0| *dst = newanswer; 169| 0| return false; 170| 0| } 171| 0| } else { // no inplace 172| | // we expect the answer to be a bitmap (if we are lucky) 173| 0| bitset_container_t *answer = bitset_container_clone(src_2); 174| | 175| 0| *dst = answer; 176| 0| if (answer == NULL) { ------------------ | Branch (176:13): [True: 0, False: 0] ------------------ 177| 0| return true; 178| 0| } 179| 0| uint32_t start = 0; 180| 0| for (int32_t rlepos = 0; rlepos < src_1->n_runs; ++rlepos) { ------------------ | Branch (180:34): [True: 0, False: 0] ------------------ 181| 0| const rle16_t rle = src_1->runs[rlepos]; 182| 0| uint32_t end = rle.value; 183| 0| bitset_reset_range(answer->words, start, end); 184| 0| start = end + rle.length + 1; 185| 0| } 186| 0| bitset_reset_range(answer->words, start, UINT32_C(1) << 16); 187| 0| answer->cardinality = bitset_container_compute_cardinality(answer); 188| | 189| 0| if (answer->cardinality > DEFAULT_MAX_SIZE) { ------------------ | Branch (189:13): [True: 0, False: 0] ------------------ 190| 0| return true; 191| 0| } else { 192| 0| array_container_t *newanswer = array_container_from_bitset(answer); 193| 0| bitset_container_free(CAST_bitset(*dst)); ------------------ | | 52| 0|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 194| 0| if (newanswer == NULL) { ------------------ | Branch (194:17): [True: 0, False: 0] ------------------ 195| 0| *dst = NULL; 196| 0| return false; 197| 0| } 198| 0| *dst = newanswer; 199| | return false; 200| 0| } 201| 0| } 202| 0|} array_run_container_intersection_cardinality: 206| 8.36k| const run_container_t *src_2) { 207| 8.36k| if (run_container_is_full(src_2)) { ------------------ | Branch (207:9): [True: 136, False: 8.22k] ------------------ 208| 136| return src_1->cardinality; 209| 136| } 210| 8.22k| if (src_2->n_runs == 0) { ------------------ | Branch (210:9): [True: 0, False: 8.22k] ------------------ 211| 0| return 0; 212| 0| } 213| 8.22k| int32_t rlepos = 0; 214| 8.22k| int32_t arraypos = 0; 215| 8.22k| rle16_t rle = src_2->runs[rlepos]; 216| 8.22k| int32_t newcard = 0; 217| 819k| while (arraypos < src_1->cardinality) { ------------------ | Branch (217:12): [True: 812k, False: 6.36k] ------------------ 218| 812k| const uint16_t arrayval = src_1->array[arraypos]; 219| 1.21M| while (rle.value + rle.length < ------------------ | Branch (219:16): [True: 402k, False: 811k] ------------------ 220| 1.21M| arrayval) { // this will frequently be false 221| 402k| ++rlepos; 222| 402k| if (rlepos == src_2->n_runs) { ------------------ | Branch (222:17): [True: 1.86k, False: 400k] ------------------ 223| 1.86k| return newcard; // we are done 224| 1.86k| } 225| 400k| rle = src_2->runs[rlepos]; 226| 400k| } 227| 811k| if (rle.value > arrayval) { ------------------ | Branch (227:13): [True: 110k, False: 700k] ------------------ 228| 110k| arraypos = advanceUntil(src_1->array, arraypos, src_1->cardinality, 229| 110k| rle.value); 230| 700k| } else { 231| 700k| newcard++; 232| 700k| arraypos++; 233| 700k| } 234| 811k| } 235| 6.36k| return newcard; 236| 8.22k|} run_bitset_container_intersection_cardinality: 241| 2.76k| const run_container_t *src_1, const bitset_container_t *src_2) { 242| 2.76k| if (run_container_is_full(src_1)) { ------------------ | Branch (242:9): [True: 0, False: 2.76k] ------------------ 243| 0| return bitset_container_cardinality(src_2); 244| 0| } 245| 2.76k| int answer = 0; 246| 228k| for (int32_t rlepos = 0; rlepos < src_1->n_runs; ++rlepos) { ------------------ | Branch (246:30): [True: 225k, False: 2.76k] ------------------ 247| 225k| rle16_t rle = src_1->runs[rlepos]; 248| 225k| answer += 249| 225k| bitset_lenrange_cardinality(src_2->words, rle.value, rle.length); 250| 225k| } 251| 2.76k| return answer; 252| 2.76k|} array_run_container_intersect: 255| 2.09k| const run_container_t *src_2) { 256| 2.09k| if (run_container_is_full(src_2)) { ------------------ | Branch (256:9): [True: 34, False: 2.05k] ------------------ 257| 34| return !array_container_empty(src_1); 258| 34| } 259| 2.05k| if (src_2->n_runs == 0) { ------------------ | Branch (259:9): [True: 0, False: 2.05k] ------------------ 260| 0| return false; 261| 0| } 262| 2.05k| int32_t rlepos = 0; 263| 2.05k| int32_t arraypos = 0; 264| 2.05k| rle16_t rle = src_2->runs[rlepos]; 265| 3.54k| while (arraypos < src_1->cardinality) { ------------------ | Branch (265:12): [True: 3.48k, False: 56] ------------------ 266| 3.48k| const uint16_t arrayval = src_1->array[arraypos]; 267| 9.41k| while (rle.value + rle.length < ------------------ | Branch (267:16): [True: 5.98k, False: 3.43k] ------------------ 268| 9.41k| arrayval) { // this will frequently be false 269| 5.98k| ++rlepos; 270| 5.98k| if (rlepos == src_2->n_runs) { ------------------ | Branch (270:17): [True: 50, False: 5.93k] ------------------ 271| 50| return false; // we are done 272| 50| } 273| 5.93k| rle = src_2->runs[rlepos]; 274| 5.93k| } 275| 3.43k| if (rle.value > arrayval) { ------------------ | Branch (275:13): [True: 1.48k, False: 1.95k] ------------------ 276| 1.48k| arraypos = advanceUntil(src_1->array, arraypos, src_1->cardinality, 277| 1.48k| rle.value); 278| 1.95k| } else { 279| 1.95k| return true; 280| 1.95k| } 281| 3.43k| } 282| 56| return false; 283| 2.05k|} run_bitset_container_intersect: 288| 691| const bitset_container_t *src_2) { 289| 691| if (run_container_is_full(src_1)) { ------------------ | Branch (289:9): [True: 0, False: 691] ------------------ 290| 0| return !bitset_container_empty(src_2); 291| 0| } 292| 4.14k| for (int32_t rlepos = 0; rlepos < src_1->n_runs; ++rlepos) { ------------------ | Branch (292:30): [True: 4.12k, False: 15] ------------------ 293| 4.12k| rle16_t rle = src_1->runs[rlepos]; 294| 4.12k| if (!bitset_lenrange_empty(src_2->words, rle.value, rle.length)) ------------------ | Branch (294:13): [True: 676, False: 3.44k] ------------------ 295| 676| return true; 296| 4.12k| } 297| 15| return false; 298| 691|} bitset_bitset_container_intersection_inplace: 329| 1.51k| container_t **dst) { 330| 1.51k| const int newCardinality = bitset_container_and_justcard(src_1, src_2); 331| 1.51k| if (newCardinality > DEFAULT_MAX_SIZE) { ------------------ | Branch (331:9): [True: 1.51k, False: 0] ------------------ 332| 1.51k| *dst = src_1; 333| 1.51k| bitset_container_and_nocard(src_1, src_2, src_1); 334| 1.51k| CAST_bitset(*dst)->cardinality = newCardinality; ------------------ | | 52| 1.51k|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 1.51k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 335| 1.51k| return true; // it is a bitset 336| 1.51k| } 337| 0| *dst = array_container_create_given_capacity(newCardinality); 338| 0| if (*dst != NULL) { ------------------ | Branch (338:9): [True: 0, False: 0] ------------------ 339| 0| CAST_array(*dst)->cardinality = newCardinality; ------------------ | | 54| 0|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 340| 0| bitset_extract_intersection_setbits_uint16( 341| 0| src_1->words, src_2->words, BITSET_CONTAINER_SIZE_IN_WORDS, 342| 0| CAST_array(*dst)->array, 0); ------------------ | | 54| 0|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 343| 0| } 344| | return false; // not a bitset 345| 1.51k|} array_container_negation: 33| 185| bitset_container_t *dst) { 34| 185| uint64_t card = UINT64_C(1 << 16); 35| 185| bitset_container_set_all(dst); 36| | 37| 185| if (src->cardinality == 0) { ------------------ | Branch (37:9): [True: 0, False: 185] ------------------ 38| 0| return; 39| 0| } 40| | 41| 185| dst->cardinality = (int32_t)bitset_clear_list(dst->words, card, src->array, 42| 185| (uint64_t)src->cardinality); 43| 185|} bitset_container_negation_inplace: 67| 87| container_t **dst) { 68| 87| return bitset_container_negation_range_inplace(src, 0, (1 << 16), dst); 69| 87|} run_container_negation_inplace: 88| 6.58k|int run_container_negation_inplace(run_container_t *src, container_t **dst) { 89| 6.58k| return run_container_negation_range_inplace(src, 0, (1 << 16), dst); 90| 6.58k|} array_container_negation_range: 99| 3.57k| container_t **dst) { 100| | /* close port of the Java implementation */ 101| 3.57k| if (range_start >= range_end) { ------------------ | Branch (101:9): [True: 0, False: 3.57k] ------------------ 102| 0| *dst = array_container_clone(src); 103| 0| return false; 104| 0| } 105| | 106| 3.57k| int32_t start_index = 107| 3.57k| binarySearch(src->array, src->cardinality, (uint16_t)range_start); 108| 3.57k| if (start_index < 0) start_index = -start_index - 1; ------------------ | Branch (108:9): [True: 3.24k, False: 327] ------------------ 109| | 110| 3.57k| int32_t last_index = 111| 3.57k| binarySearch(src->array, src->cardinality, (uint16_t)(range_end - 1)); 112| 3.57k| if (last_index < 0) last_index = -last_index - 2; ------------------ | Branch (112:9): [True: 3.30k, False: 273] ------------------ 113| | 114| 3.57k| const int32_t current_values_in_range = last_index - start_index + 1; 115| 3.57k| const int32_t span_to_be_flipped = range_end - range_start; 116| 3.57k| const int32_t new_values_in_range = 117| 3.57k| span_to_be_flipped - current_values_in_range; 118| 3.57k| const int32_t cardinality_change = 119| 3.57k| new_values_in_range - current_values_in_range; 120| 3.57k| const int32_t new_cardinality = src->cardinality + cardinality_change; 121| | 122| 3.57k| if (new_cardinality > DEFAULT_MAX_SIZE) { ------------------ | Branch (122:9): [True: 130, False: 3.44k] ------------------ 123| 130| bitset_container_t *temp = bitset_container_from_array(src); 124| 130| bitset_flip_range(temp->words, (uint32_t)range_start, 125| 130| (uint32_t)range_end); 126| 130| temp->cardinality = new_cardinality; 127| 130| *dst = temp; 128| 130| return true; 129| 130| } 130| | 131| 3.44k| array_container_t *arr = 132| 3.44k| array_container_create_given_capacity(new_cardinality); 133| 3.44k| *dst = (container_t *)arr; 134| 3.44k| if (new_cardinality == 0) { ------------------ | Branch (134:9): [True: 12, False: 3.43k] ------------------ 135| 12| arr->cardinality = new_cardinality; 136| 12| return false; // we are done. 137| 12| } 138| | // copy stuff before the active area 139| 3.43k| memcpy(arr->array, src->array, start_index * sizeof(uint16_t)); 140| | 141| | // work on the range 142| 3.43k| int32_t out_pos = start_index, in_pos = start_index; 143| 3.43k| int32_t val_in_range = range_start; 144| 35.0k| for (; val_in_range < range_end && in_pos <= last_index; ++val_in_range) { ------------------ | Branch (144:12): [True: 34.7k, False: 260] | Branch (144:40): [True: 31.5k, False: 3.17k] ------------------ 145| 31.5k| if ((uint16_t)val_in_range != src->array[in_pos]) { ------------------ | Branch (145:13): [True: 15.7k, False: 15.8k] ------------------ 146| 15.7k| arr->array[out_pos++] = (uint16_t)val_in_range; 147| 15.8k| } else { 148| 15.8k| ++in_pos; 149| 15.8k| } 150| 31.5k| } 151| 69.1k| for (; val_in_range < range_end; ++val_in_range) ------------------ | Branch (151:12): [True: 65.6k, False: 3.43k] ------------------ 152| 65.6k| arr->array[out_pos++] = (uint16_t)val_in_range; 153| | 154| | // content after the active range 155| 3.43k| memcpy(arr->array + out_pos, src->array + (last_index + 1), 156| 3.43k| (src->cardinality - (last_index + 1)) * sizeof(uint16_t)); 157| 3.43k| arr->cardinality = new_cardinality; 158| | return false; 159| 3.44k|} array_container_negation_range_inplace: 168| 3.57k| container_t **dst) { 169| 3.57k| bool ans = array_container_negation_range(src, range_start, range_end, dst); 170| | // TODO : try a real inplace version 171| 3.57k| array_container_free(src); 172| 3.57k| return ans; 173| 3.57k|} bitset_container_negation_range_inplace: 217| 1.44k| container_t **dst) { 218| 1.44k| bitset_flip_range(src->words, (uint32_t)range_start, (uint32_t)range_end); 219| 1.44k| src->cardinality = bitset_container_compute_cardinality(src); 220| 1.44k| if (src->cardinality > DEFAULT_MAX_SIZE) { ------------------ | Branch (220:9): [True: 1.36k, False: 75] ------------------ 221| 1.36k| *dst = src; 222| 1.36k| return true; 223| 1.36k| } 224| 75| *dst = array_container_from_bitset(src); 225| 75| bitset_container_free(src); 226| | return false; 227| 1.44k|} run_container_negation_range: 237| 146| container_t **dst) { 238| 146| uint8_t return_typecode; 239| | 240| | // follows the Java implementation 241| 146| if (range_end <= range_start) { ------------------ | Branch (241:9): [True: 0, False: 146] ------------------ 242| 0| *dst = run_container_clone(src); 243| 0| return RUN_CONTAINER_TYPE; ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ 244| 0| } 245| | 246| 146| run_container_t *ans = run_container_create_given_capacity( 247| 146| src->n_runs + 1); // src->n_runs + 1); 248| 146| int k = 0; 249| 3.05k| for (; k < src->n_runs && src->runs[k].value < range_start; ++k) { ------------------ | Branch (249:12): [True: 3.00k, False: 47] | Branch (249:31): [True: 2.90k, False: 99] ------------------ 250| 2.90k| ans->runs[k] = src->runs[k]; 251| 2.90k| ans->n_runs++; 252| 2.90k| } 253| | 254| 146| run_container_smart_append_exclusive( 255| 146| ans, (uint16_t)range_start, (uint16_t)(range_end - range_start - 1)); 256| | 257| 3.73k| for (; k < src->n_runs; ++k) { ------------------ | Branch (257:12): [True: 3.58k, False: 146] ------------------ 258| 3.58k| run_container_smart_append_exclusive(ans, src->runs[k].value, 259| 3.58k| src->runs[k].length); 260| 3.58k| } 261| | 262| 146| *dst = convert_run_to_efficient_container(ans, &return_typecode); 263| 146| if (return_typecode != RUN_CONTAINER_TYPE) run_container_free(ans); ------------------ | | 50| 146|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (263:9): [True: 8, False: 138] ------------------ 264| | 265| 146| return return_typecode; 266| 146|} run_container_negation_range_inplace: 278| 8.28k| container_t **dst) { 279| 8.28k| uint8_t return_typecode; 280| | 281| 8.28k| if (range_end <= range_start) { ------------------ | Branch (281:9): [True: 0, False: 8.28k] ------------------ 282| 0| *dst = src; 283| 0| return RUN_CONTAINER_TYPE; ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ 284| 0| } 285| | 286| | // TODO: efficient special case when range is 0 to 65535 inclusive 287| | 288| 8.28k| if (src->capacity == src->n_runs) { ------------------ | Branch (288:9): [True: 7.12k, False: 1.15k] ------------------ 289| | // no excess room. More checking to see if result can fit 290| 7.12k| bool last_val_before_range = false; 291| 7.12k| bool first_val_in_range = false; 292| 7.12k| bool last_val_in_range = false; 293| 7.12k| bool first_val_past_range = false; 294| | 295| 7.12k| if (range_start > 0) ------------------ | Branch (295:13): [True: 166, False: 6.95k] ------------------ 296| 166| last_val_before_range = 297| 166| run_container_contains(src, (uint16_t)(range_start - 1)); 298| 7.12k| first_val_in_range = run_container_contains(src, (uint16_t)range_start); 299| | 300| 7.12k| if (last_val_before_range == first_val_in_range) { ------------------ | Branch (300:13): [True: 459, False: 6.66k] ------------------ 301| 459| last_val_in_range = 302| 459| run_container_contains(src, (uint16_t)(range_end - 1)); 303| 459| if (range_end != 0x10000) ------------------ | Branch (303:17): [True: 294, False: 165] ------------------ 304| 294| first_val_past_range = 305| 294| run_container_contains(src, (uint16_t)range_end); 306| | 307| 459| if (last_val_in_range == ------------------ | Branch (307:17): [True: 146, False: 313] ------------------ 308| 459| first_val_past_range) { // no space for inplace 309| 146| int ans = run_container_negation_range(src, range_start, 310| 146| range_end, dst); 311| 146| run_container_free(src); 312| 146| return ans; 313| 146| } 314| 459| } 315| 7.12k| } 316| | // all other cases: result will fit 317| | 318| 8.13k| run_container_t *ans = src; 319| 8.13k| int my_nbr_runs = src->n_runs; 320| | 321| 8.13k| ans->n_runs = 0; 322| 8.13k| int k = 0; 323| 10.6k| for (; (k < my_nbr_runs) && (src->runs[k].value < range_start); ++k) { ------------------ | Branch (323:12): [True: 10.6k, False: 69] | Branch (323:33): [True: 2.54k, False: 8.06k] ------------------ 324| | // ans->runs[k] = src->runs[k]; (would be self-copy) 325| 2.54k| ans->n_runs++; 326| 2.54k| } 327| | 328| | // as with Java implementation, use locals to give self a buffer of depth 1 329| 8.13k| rle16_t buffered = CROARING_MAKE_RLE16(0, 0); ------------------ | | 60| 8.13k| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 330| 8.13k| rle16_t next = buffered; 331| 8.13k| if (k < my_nbr_runs) buffered = src->runs[k]; ------------------ | Branch (331:9): [True: 8.06k, False: 69] ------------------ 332| | 333| 8.13k| run_container_smart_append_exclusive( 334| 8.13k| ans, (uint16_t)range_start, (uint16_t)(range_end - range_start - 1)); 335| | 336| 97.7k| for (; k < my_nbr_runs; ++k) { ------------------ | Branch (336:12): [True: 89.6k, False: 8.13k] ------------------ 337| 89.6k| if (k + 1 < my_nbr_runs) next = src->runs[k + 1]; ------------------ | Branch (337:13): [True: 81.5k, False: 8.06k] ------------------ 338| | 339| 89.6k| run_container_smart_append_exclusive(ans, buffered.value, 340| 89.6k| buffered.length); 341| 89.6k| buffered = next; 342| 89.6k| } 343| | 344| 8.13k| *dst = convert_run_to_efficient_container(ans, &return_typecode); 345| 8.13k| if (return_typecode != RUN_CONTAINER_TYPE) run_container_free(ans); ------------------ | | 50| 8.13k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (345:9): [True: 6.44k, False: 1.68k] ------------------ 346| | 347| 8.13k| return return_typecode; 348| 8.28k|} array_container_is_subset_bitset: 11| 2.45k| const bitset_container_t* container2) { 12| 2.45k| if (container2->cardinality != BITSET_UNKNOWN_CARDINALITY) { ------------------ | Branch (12:9): [True: 2.45k, False: 0] ------------------ 13| 2.45k| if (container2->cardinality < container1->cardinality) { ------------------ | Branch (13:13): [True: 0, False: 2.45k] ------------------ 14| 0| return false; 15| 0| } 16| 2.45k| } 17| 2.45k| for (int i = 0; i < container1->cardinality; ++i) { ------------------ | Branch (17:21): [True: 2.45k, False: 0] ------------------ 18| 2.45k| if (!bitset_container_contains(container2, container1->array[i])) { ------------------ | Branch (18:13): [True: 2.45k, False: 0] ------------------ 19| 2.45k| return false; 20| 2.45k| } 21| 2.45k| } 22| 0| return true; 23| 2.45k|} run_container_is_subset_array: 26| 423| const array_container_t* container2) { 27| 423| if (run_container_cardinality(container1) > container2->cardinality) ------------------ | Branch (27:9): [True: 171, False: 252] ------------------ 28| 171| return false; 29| 252| int32_t start_pos = -1, stop_pos = -1; 30| 252| for (int i = 0; i < container1->n_runs; ++i) { ------------------ | Branch (30:21): [True: 252, False: 0] ------------------ 31| 252| int32_t start = container1->runs[i].value; 32| 252| int32_t stop = start + container1->runs[i].length; 33| 252| start_pos = advanceUntil(container2->array, stop_pos, 34| 252| container2->cardinality, start); 35| 252| stop_pos = advanceUntil(container2->array, stop_pos, 36| 252| container2->cardinality, stop); 37| 252| if (stop_pos == container2->cardinality) { ------------------ | Branch (37:13): [True: 5, False: 247] ------------------ 38| 5| return false; 39| 247| } else if (stop_pos - start_pos != stop - start || ------------------ | Branch (39:20): [True: 87, False: 160] ------------------ 40| 160| container2->array[start_pos] != start || ------------------ | Branch (40:20): [True: 160, False: 0] ------------------ 41| 247| container2->array[stop_pos] != stop) { ------------------ | Branch (41:20): [True: 0, False: 0] ------------------ 42| 247| return false; 43| 247| } 44| 252| } 45| 0| return true; 46| 252|} array_container_is_subset_run: 49| 2.25k| const run_container_t* container2) { 50| 2.25k| if (container1->cardinality > run_container_cardinality(container2)) ------------------ | Branch (50:9): [True: 161, False: 2.09k] ------------------ 51| 161| return false; 52| 2.09k| int i_array = 0, i_run = 0; 53| 7.03k| while (i_array < container1->cardinality && i_run < container2->n_runs) { ------------------ | Branch (53:12): [True: 7.03k, False: 0] | Branch (53:49): [True: 7.02k, False: 14] ------------------ 54| 7.02k| uint32_t start = container2->runs[i_run].value; 55| 7.02k| uint32_t stop = start + container2->runs[i_run].length; 56| 7.02k| if (container1->array[i_array] < start) { ------------------ | Branch (56:13): [True: 2.08k, False: 4.94k] ------------------ 57| 2.08k| return false; 58| 4.94k| } else if (container1->array[i_array] > stop) { ------------------ | Branch (58:20): [True: 4.94k, False: 0] ------------------ 59| 4.94k| i_run++; 60| 4.94k| } else { // the value of the array is in the run 61| 0| i_array++; 62| 0| } 63| 7.02k| } 64| 14| if (i_array == container1->cardinality) { ------------------ | Branch (64:9): [True: 0, False: 14] ------------------ 65| 0| return true; 66| 14| } else { 67| | return false; 68| 14| } 69| 14|} array_bitset_container_union: 24| 1.20k| bitset_container_t *dst) { 25| 1.20k| if (src_2 != dst) bitset_container_copy(src_2, dst); ------------------ | Branch (25:9): [True: 1.20k, False: 0] ------------------ 26| 1.20k| dst->cardinality = (int32_t)bitset_set_list_withcard( 27| 1.20k| dst->words, dst->cardinality, src_1->array, src_1->cardinality); 28| 1.20k|} run_bitset_container_union: 43| 1.38k| bitset_container_t *dst) { 44| 1.38k| assert(!run_container_is_full(src_1)); // catch this case upstream 45| 1.38k| if (src_2 != dst) bitset_container_copy(src_2, dst); ------------------ | Branch (45:9): [True: 1.38k, False: 0] ------------------ 46| 114k| for (int32_t rlepos = 0; rlepos < src_1->n_runs; ++rlepos) { ------------------ | Branch (46:30): [True: 112k, False: 1.38k] ------------------ 47| 112k| rle16_t rle = src_1->runs[rlepos]; 48| 112k| bitset_set_lenrange(dst->words, rle.value, rle.length); 49| 112k| } 50| 1.38k| dst->cardinality = bitset_container_compute_cardinality(dst); 51| 1.38k|} array_run_container_union: 68| 2.75k| run_container_t *dst) { 69| 2.75k| if (run_container_is_full(src_2)) { ------------------ | Branch (69:9): [True: 115, False: 2.64k] ------------------ 70| 115| run_container_copy(src_2, dst); 71| 115| return; 72| 115| } 73| | // TODO: see whether the "2*" is spurious 74| 2.64k| run_container_grow(dst, 2 * (src_1->cardinality + src_2->n_runs), false); 75| 2.64k| int32_t rlepos = 0; 76| 2.64k| int32_t arraypos = 0; 77| 2.64k| rle16_t previousrle; 78| 2.64k| if (src_2->runs[rlepos].value <= src_1->array[arraypos]) { ------------------ | Branch (78:9): [True: 1.99k, False: 642] ------------------ 79| 1.99k| previousrle = run_container_append_first(dst, src_2->runs[rlepos]); 80| 1.99k| rlepos++; 81| 1.99k| } else { 82| 642| previousrle = 83| 642| run_container_append_value_first(dst, src_1->array[arraypos]); 84| 642| arraypos++; 85| 642| } 86| 477k| while ((rlepos < src_2->n_runs) && (arraypos < src_1->cardinality)) { ------------------ | Branch (86:12): [True: 475k, False: 1.73k] | Branch (86:40): [True: 474k, False: 910] ------------------ 87| 474k| if (src_2->runs[rlepos].value <= src_1->array[arraypos]) { ------------------ | Branch (87:13): [True: 155k, False: 318k] ------------------ 88| 155k| run_container_append(dst, src_2->runs[rlepos], &previousrle); 89| 155k| rlepos++; 90| 318k| } else { 91| 318k| run_container_append_value(dst, src_1->array[arraypos], 92| 318k| &previousrle); 93| 318k| arraypos++; 94| 318k| } 95| 474k| } 96| 2.64k| if (arraypos < src_1->cardinality) { ------------------ | Branch (96:9): [True: 1.73k, False: 910] ------------------ 97| 174k| while (arraypos < src_1->cardinality) { ------------------ | Branch (97:16): [True: 172k, False: 1.73k] ------------------ 98| 172k| run_container_append_value(dst, src_1->array[arraypos], 99| 172k| &previousrle); 100| 172k| arraypos++; 101| 172k| } 102| 1.73k| } else { 103| 6.21k| while (rlepos < src_2->n_runs) { ------------------ | Branch (103:16): [True: 5.30k, False: 910] ------------------ 104| 5.30k| run_container_append(dst, src_2->runs[rlepos], &previousrle); 105| 5.30k| rlepos++; 106| 5.30k| } 107| 910| } 108| 2.64k|} array_run_container_inplace_union: 111| 248| run_container_t *src_2) { 112| 248| if (run_container_is_full(src_2)) { ------------------ | Branch (112:9): [True: 0, False: 248] ------------------ 113| 0| return; 114| 0| } 115| 248| const int32_t maxoutput = src_1->cardinality + src_2->n_runs; 116| 248| const int32_t neededcapacity = maxoutput + src_2->n_runs; 117| 248| if (src_2->capacity < neededcapacity) ------------------ | Branch (117:9): [True: 248, False: 0] ------------------ 118| 248| run_container_grow(src_2, neededcapacity, true); 119| 248| memmove(src_2->runs + maxoutput, src_2->runs, 120| 248| src_2->n_runs * sizeof(rle16_t)); 121| 248| rle16_t *inputsrc2 = src_2->runs + maxoutput; 122| 248| int32_t rlepos = 0; 123| 248| int32_t arraypos = 0; 124| 248| int src2nruns = src_2->n_runs; 125| 248| src_2->n_runs = 0; 126| | 127| 248| rle16_t previousrle; 128| | 129| 248| if (inputsrc2[rlepos].value <= src_1->array[arraypos]) { ------------------ | Branch (129:9): [True: 206, False: 42] ------------------ 130| 206| previousrle = run_container_append_first(src_2, inputsrc2[rlepos]); 131| 206| rlepos++; 132| 206| } else { 133| 42| previousrle = 134| 42| run_container_append_value_first(src_2, src_1->array[arraypos]); 135| 42| arraypos++; 136| 42| } 137| | 138| 49.9k| while ((rlepos < src2nruns) && (arraypos < src_1->cardinality)) { ------------------ | Branch (138:12): [True: 49.7k, False: 248] | Branch (138:36): [True: 49.7k, False: 0] ------------------ 139| 49.7k| if (inputsrc2[rlepos].value <= src_1->array[arraypos]) { ------------------ | Branch (139:13): [True: 12.6k, False: 37.0k] ------------------ 140| 12.6k| run_container_append(src_2, inputsrc2[rlepos], &previousrle); 141| 12.6k| rlepos++; 142| 37.0k| } else { 143| 37.0k| run_container_append_value(src_2, src_1->array[arraypos], 144| 37.0k| &previousrle); 145| 37.0k| arraypos++; 146| 37.0k| } 147| 49.7k| } 148| 248| if (arraypos < src_1->cardinality) { ------------------ | Branch (148:9): [True: 248, False: 0] ------------------ 149| 2.55k| while (arraypos < src_1->cardinality) { ------------------ | Branch (149:16): [True: 2.30k, False: 248] ------------------ 150| 2.30k| run_container_append_value(src_2, src_1->array[arraypos], 151| 2.30k| &previousrle); 152| 2.30k| arraypos++; 153| 2.30k| } 154| 248| } else { 155| 0| while (rlepos < src2nruns) { ------------------ | Branch (155:16): [True: 0, False: 0] ------------------ 156| 0| run_container_append(src_2, inputsrc2[rlepos], &previousrle); 157| 0| rlepos++; 158| 0| } 159| 0| } 160| 248|} array_array_container_union: 164| 2.32k| container_t **dst) { 165| 2.32k| int totalCardinality = src_1->cardinality + src_2->cardinality; 166| 2.32k| if (totalCardinality <= DEFAULT_MAX_SIZE) { ------------------ | Branch (166:9): [True: 2.19k, False: 126] ------------------ 167| 2.19k| *dst = array_container_create_given_capacity(totalCardinality); 168| 2.19k| if (*dst != NULL) { ------------------ | Branch (168:13): [True: 2.19k, False: 0] ------------------ 169| 2.19k| array_container_union(src_1, src_2, CAST_array(*dst)); ------------------ | | 54| 2.19k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 2.19k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 170| 2.19k| } else { 171| 0| return true; // otherwise failure won't be caught 172| 0| } 173| 2.19k| return false; // not a bitset 174| 2.19k| } 175| 126| *dst = bitset_container_create(); 176| 126| bool returnval = true; // expect a bitset 177| 126| if (*dst != NULL) { ------------------ | Branch (177:9): [True: 126, False: 0] ------------------ 178| 126| bitset_container_t *ourbitset = CAST_bitset(*dst); ------------------ | | 52| 126|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 126|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 179| 126| bitset_set_list(ourbitset->words, src_1->array, src_1->cardinality); 180| 126| ourbitset->cardinality = (int32_t)bitset_set_list_withcard( 181| 126| ourbitset->words, src_1->cardinality, src_2->array, 182| 126| src_2->cardinality); 183| 126| if (ourbitset->cardinality <= DEFAULT_MAX_SIZE) { ------------------ | Branch (183:13): [True: 74, False: 52] ------------------ 184| | // need to convert! 185| 74| *dst = array_container_from_bitset(ourbitset); 186| 74| bitset_container_free(ourbitset); 187| | returnval = false; // not going to be a bitset 188| 74| } 189| 126| } 190| 126| return returnval; 191| 2.32k|} array_array_container_inplace_union: 195| 2.51k| container_t **dst) { 196| 2.51k| int totalCardinality = src_1->cardinality + src_2->cardinality; 197| 2.51k| *dst = NULL; 198| 2.51k| if (totalCardinality <= DEFAULT_MAX_SIZE) { ------------------ | Branch (198:9): [True: 2.41k, False: 96] ------------------ 199| 2.41k| if (src_1->capacity < totalCardinality) { ------------------ | Branch (199:13): [True: 2.41k, False: 0] ------------------ 200| 2.41k| *dst = array_container_create_given_capacity( 201| 2.41k| 2 * totalCardinality); // be purposefully generous 202| 2.41k| if (*dst != NULL) { ------------------ | Branch (202:17): [True: 2.41k, False: 0] ------------------ 203| 2.41k| array_container_union(src_1, src_2, CAST_array(*dst)); ------------------ | | 54| 2.41k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 2.41k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 204| 2.41k| } else { 205| 0| return true; // otherwise failure won't be caught 206| 0| } 207| 2.41k| return false; // not a bitset 208| 2.41k| } else { 209| 0| memmove(src_1->array + src_2->cardinality, src_1->array, 210| 0| src_1->cardinality * sizeof(uint16_t)); 211| | // In theory, we could use fast_union_uint16, but it is unsafe. It 212| | // fails with Intel compilers in particular. 213| | // https://github.com/RoaringBitmap/CRoaring/pull/452 214| | // See report https://github.com/RoaringBitmap/CRoaring/issues/476 215| 0| src_1->cardinality = (int32_t)union_uint16( 216| 0| src_1->array + src_2->cardinality, src_1->cardinality, 217| 0| src_2->array, src_2->cardinality, src_1->array); 218| 0| return false; // not a bitset 219| 0| } 220| 2.41k| } 221| 96| *dst = bitset_container_create(); 222| 96| bool returnval = true; // expect a bitset 223| 96| if (*dst != NULL) { ------------------ | Branch (223:9): [True: 96, False: 0] ------------------ 224| 96| bitset_container_t *ourbitset = CAST_bitset(*dst); ------------------ | | 52| 96|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 96|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 225| 96| bitset_set_list(ourbitset->words, src_1->array, src_1->cardinality); 226| 96| ourbitset->cardinality = (int32_t)bitset_set_list_withcard( 227| 96| ourbitset->words, src_1->cardinality, src_2->array, 228| 96| src_2->cardinality); 229| 96| if (ourbitset->cardinality <= DEFAULT_MAX_SIZE) { ------------------ | Branch (229:13): [True: 96, False: 0] ------------------ 230| | // need to convert! 231| 96| if (src_1->capacity < ourbitset->cardinality) { ------------------ | Branch (231:17): [True: 96, False: 0] ------------------ 232| 96| array_container_grow(src_1, ourbitset->cardinality, false); 233| 96| } 234| | 235| 96| bitset_extract_setbits_uint16(ourbitset->words, 236| 96| BITSET_CONTAINER_SIZE_IN_WORDS, 237| 96| src_1->array, 0); 238| 96| src_1->cardinality = ourbitset->cardinality; 239| 96| *dst = src_1; 240| 96| bitset_container_free(ourbitset); 241| | returnval = false; // not going to be a bitset 242| 96| } 243| 96| } 244| 96| return returnval; 245| 2.51k|} array_bitset_container_xor: 25| 673| container_t **dst) { 26| 673| bitset_container_t *result = bitset_container_create(); 27| 673| bitset_container_copy(src_2, result); 28| 673| result->cardinality = (int32_t)bitset_flip_list_withcard( 29| 673| result->words, result->cardinality, src_1->array, src_1->cardinality); 30| | 31| | // do required type conversions. 32| 673| if (result->cardinality <= DEFAULT_MAX_SIZE) { ------------------ | Branch (32:9): [True: 153, False: 520] ------------------ 33| 153| *dst = array_container_from_bitset(result); 34| 153| bitset_container_free(result); 35| 153| return false; // not bitset 36| 153| } 37| 520| *dst = result; 38| | return true; // bitset 39| 673|} run_bitset_container_xor: 63| 782| container_t **dst) { 64| 782| bitset_container_t *result = bitset_container_create(); 65| | 66| 782| bitset_container_copy(src_2, result); 67| 60.6k| for (int32_t rlepos = 0; rlepos < src_1->n_runs; ++rlepos) { ------------------ | Branch (67:30): [True: 59.8k, False: 782] ------------------ 68| 59.8k| rle16_t rle = src_1->runs[rlepos]; 69| 59.8k| bitset_flip_range(result->words, rle.value, 70| 59.8k| rle.value + rle.length + UINT32_C(1)); 71| 59.8k| } 72| 782| result->cardinality = bitset_container_compute_cardinality(result); 73| | 74| 782| if (result->cardinality <= DEFAULT_MAX_SIZE) { ------------------ | Branch (74:9): [True: 137, False: 645] ------------------ 75| 137| *dst = array_container_from_bitset(result); 76| 137| bitset_container_free(result); 77| 137| return false; // not bitset 78| 137| } 79| 645| *dst = result; 80| | return true; // bitset 81| 782|} array_run_container_xor: 105| 2.41k| const run_container_t *src_2, container_t **dst) { 106| | // semi following Java XOR implementation as of May 2016 107| | // the C OR implementation works quite differently and can return a run 108| | // container 109| | // TODO could optimize for full run containers. 110| | 111| | // use of lazy following Java impl. 112| 2.41k| const int arbitrary_threshold = 32; 113| 2.41k| if (src_1->cardinality < arbitrary_threshold) { ------------------ | Branch (113:9): [True: 945, False: 1.46k] ------------------ 114| 945| run_container_t *ans = run_container_create(); 115| 945| array_run_container_lazy_xor(src_1, src_2, ans); // keeps runs. 116| 945| uint8_t typecode_after; 117| 945| *dst = 118| 945| convert_run_to_efficient_container_and_free(ans, &typecode_after); 119| 945| return typecode_after; 120| 945| } 121| | 122| 1.46k| int card = run_container_cardinality(src_2); 123| 1.46k| if (card <= DEFAULT_MAX_SIZE) { ------------------ | Branch (123:9): [True: 1.18k, False: 279] ------------------ 124| | // Java implementation works with the array, xoring the run elements via 125| | // iterator 126| 1.18k| array_container_t *temp = array_container_from_run(src_2); 127| 1.18k| bool ret_is_bitset = array_array_container_xor(temp, src_1, dst); 128| 1.18k| array_container_free(temp); 129| 1.18k| return ret_is_bitset ? BITSET_CONTAINER_TYPE : ARRAY_CONTAINER_TYPE; ------------------ | | 48| 6|#define BITSET_CONTAINER_TYPE 1 ------------------ return ret_is_bitset ? BITSET_CONTAINER_TYPE : ARRAY_CONTAINER_TYPE; ------------------ | | 49| 1.18k|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (129:16): [True: 6, False: 1.18k] ------------------ 130| | 131| 1.18k| } else { // guess that it will end up as a bitset 132| 279| bitset_container_t *result = bitset_container_from_run(src_2); 133| 279| bool is_bitset = bitset_array_container_ixor(result, src_1, dst); 134| | // any necessary type conversion has been done by the ixor 135| 279| int retval = (is_bitset ? BITSET_CONTAINER_TYPE : ARRAY_CONTAINER_TYPE); ------------------ | | 48| 262|#define BITSET_CONTAINER_TYPE 1 ------------------ int retval = (is_bitset ? BITSET_CONTAINER_TYPE : ARRAY_CONTAINER_TYPE); ------------------ | | 49| 17|#define ARRAY_CONTAINER_TYPE 2 ------------------ | Branch (135:23): [True: 262, False: 17] ------------------ 136| 279| return retval; 137| 279| } 138| 1.46k|} array_run_container_lazy_xor: 147| 945| run_container_t *dst) { 148| 945| run_container_grow(dst, src_1->cardinality + src_2->n_runs, false); 149| 945| int32_t rlepos = 0; 150| 945| int32_t arraypos = 0; 151| 945| dst->n_runs = 0; 152| | 153| 29.3k| while ((rlepos < src_2->n_runs) && (arraypos < src_1->cardinality)) { ------------------ | Branch (153:12): [True: 28.7k, False: 593] | Branch (153:40): [True: 28.4k, False: 352] ------------------ 154| 28.4k| if (src_2->runs[rlepos].value <= src_1->array[arraypos]) { ------------------ | Branch (154:13): [True: 17.3k, False: 11.1k] ------------------ 155| 17.3k| run_container_smart_append_exclusive(dst, src_2->runs[rlepos].value, 156| 17.3k| src_2->runs[rlepos].length); 157| 17.3k| rlepos++; 158| 17.3k| } else { 159| 11.1k| run_container_smart_append_exclusive(dst, src_1->array[arraypos], 160| 11.1k| 0); 161| 11.1k| arraypos++; 162| 11.1k| } 163| 28.4k| } 164| 3.51k| while (arraypos < src_1->cardinality) { ------------------ | Branch (164:12): [True: 2.56k, False: 945] ------------------ 165| 2.56k| run_container_smart_append_exclusive(dst, src_1->array[arraypos], 0); 166| 2.56k| arraypos++; 167| 2.56k| } 168| 8.19k| while (rlepos < src_2->n_runs) { ------------------ | Branch (168:12): [True: 7.25k, False: 945] ------------------ 169| 7.25k| run_container_smart_append_exclusive(dst, src_2->runs[rlepos].value, 170| 7.25k| src_2->runs[rlepos].length); 171| 7.25k| rlepos++; 172| 7.25k| } 173| 945|} run_run_container_xor: 180| 61.2k| const run_container_t *src_2, container_t **dst) { 181| 61.2k| run_container_t *ans = run_container_create(); 182| 61.2k| run_container_xor(src_1, src_2, ans); 183| 61.2k| uint8_t typecode_after; 184| 61.2k| *dst = convert_run_to_efficient_container_and_free(ans, &typecode_after); 185| 61.2k| return typecode_after; 186| 61.2k|} array_array_container_xor: 198| 8.62k| container_t **dst) { 199| 8.62k| int totalCardinality = 200| 8.62k| src_1->cardinality + src_2->cardinality; // upper bound 201| 8.62k| if (totalCardinality <= DEFAULT_MAX_SIZE) { ------------------ | Branch (201:9): [True: 8.05k, False: 572] ------------------ 202| 8.05k| *dst = array_container_create_given_capacity(totalCardinality); 203| 8.05k| array_container_xor(src_1, src_2, CAST_array(*dst)); ------------------ | | 54| 8.05k|#define CAST_array(c) CAST(array_container_t *, c) // safer downcast | | ------------------ | | | | 86| 8.05k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 204| 8.05k| return false; // not a bitset 205| 8.05k| } 206| 572| *dst = bitset_container_from_array(src_1); 207| 572| bool returnval = true; // expect a bitset 208| 572| bitset_container_t *ourbitset = CAST_bitset(*dst); ------------------ | | 52| 572|#define CAST_bitset(c) CAST(bitset_container_t *, c) // safer downcast | | ------------------ | | | | 86| 572|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 209| 572| ourbitset->cardinality = (uint32_t)bitset_flip_list_withcard( 210| 572| ourbitset->words, src_1->cardinality, src_2->array, src_2->cardinality); 211| 572| if (ourbitset->cardinality <= DEFAULT_MAX_SIZE) { ------------------ | Branch (211:9): [True: 558, False: 14] ------------------ 212| | // need to convert! 213| 558| *dst = array_container_from_bitset(ourbitset); 214| 558| bitset_container_free(ourbitset); 215| 558| returnval = false; // not going to be a bitset 216| 558| } 217| | 218| 572| return returnval; 219| 8.62k|} bitset_array_container_ixor: 284| 279| container_t **dst) { 285| 279| *dst = src_1; 286| 279| src_1->cardinality = (uint32_t)bitset_flip_list_withcard( 287| 279| src_1->words, src_1->cardinality, src_2->array, src_2->cardinality); 288| | 289| 279| if (src_1->cardinality <= DEFAULT_MAX_SIZE) { ------------------ | Branch (289:9): [True: 17, False: 262] ------------------ 290| 17| *dst = array_container_from_bitset(src_1); 291| 17| bitset_container_free(src_1); 292| 17| return false; // not bitset 293| 17| } else 294| 262| return true; 295| 279|} bitset_bitset_container_ixor: 304| 1.57k| container_t **dst) { 305| 1.57k| int card = bitset_container_xor(src_1, src_2, src_1); 306| 1.57k| if (card <= DEFAULT_MAX_SIZE) { ------------------ | Branch (306:9): [True: 1.57k, False: 0] ------------------ 307| 1.57k| *dst = array_container_from_bitset(src_1); 308| 1.57k| bitset_container_free(src_1); 309| 1.57k| return false; // not bitset 310| 1.57k| } else { 311| 0| *dst = src_1; 312| | return true; 313| 0| } 314| 1.57k|} array_bitset_container_ixor: 318| 98| container_t **dst) { 319| | bool ans = array_bitset_container_xor(src_1, src_2, dst); 320| 98| array_container_free(src_1); 321| 98| return ans; 322| 98|} run_bitset_container_ixor: 333| 18| container_t **dst) { 334| | bool ans = run_bitset_container_xor(src_1, src_2, dst); 335| 18| run_container_free(src_1); 336| 18| return ans; 337| 18|} bitset_run_container_ixor: 341| 73| container_t **dst) { 342| | bool ans = run_bitset_container_xor(src_2, src_1, dst); 343| 73| bitset_container_free(src_1); 344| 73| return ans; 345| 73|} array_run_container_ixor: 352| 269| const run_container_t *src_2, container_t **dst) { 353| 269| int ans = array_run_container_xor(src_1, src_2, dst); 354| 269| array_container_free(src_1); 355| 269| return ans; 356| 269|} run_array_container_ixor: 360| 53| container_t **dst) { 361| 53| int ans = array_run_container_xor(src_2, src_1, dst); 362| 53| run_container_free(src_1); 363| 53| return ans; 364| 53|} array_array_container_ixor: 368| 5.11k| container_t **dst) { 369| | bool ans = array_array_container_xor(src_1, src_2, dst); 370| 5.11k| array_container_free(src_1); 371| 5.11k| return ans; 372| 5.11k|} run_run_container_ixor: 375| 60.5k| container_t **dst) { 376| 60.5k| int ans = run_run_container_xor(src_1, src_2, dst); 377| 60.5k| run_container_free(src_1); 378| 60.5k| return ans; 379| 60.5k|} run_container_add: 39| 413k|bool run_container_add(run_container_t *run, uint16_t pos) { 40| 413k| int32_t index = interleavedBinarySearch(run->runs, run->n_runs, pos); 41| 413k| if (index >= 0) return false; // already there ------------------ | Branch (41:9): [True: 183k, False: 229k] ------------------ 42| 229k| index = -index - 2; // points to preceding value, possibly -1 43| 229k| if (index >= 0) { // possible match ------------------ | Branch (43:9): [True: 229k, False: 380] ------------------ 44| 229k| int32_t offset = pos - run->runs[index].value; 45| 229k| int32_t le = run->runs[index].length; 46| 229k| if (offset <= le) return false; // already there ------------------ | Branch (46:13): [True: 137k, False: 91.8k] ------------------ 47| 91.8k| if (offset == le + 1) { ------------------ | Branch (47:13): [True: 14.5k, False: 77.3k] ------------------ 48| | // we may need to fuse 49| 14.5k| if (index + 1 < run->n_runs) { ------------------ | Branch (49:17): [True: 14.2k, False: 287] ------------------ 50| 14.2k| if (run->runs[index + 1].value == pos + 1) { ------------------ | Branch (50:21): [True: 6.57k, False: 7.70k] ------------------ 51| | // indeed fusion is needed 52| 6.57k| run->runs[index].length = run->runs[index + 1].value + 53| 6.57k| run->runs[index + 1].length - 54| 6.57k| run->runs[index].value; 55| 6.57k| recoverRoomAtIndex(run, (uint16_t)(index + 1)); 56| 6.57k| return true; 57| 6.57k| } 58| 14.2k| } 59| 7.99k| run->runs[index].length++; 60| 7.99k| return true; 61| 14.5k| } 62| 77.3k| if (index + 1 < run->n_runs) { ------------------ | Branch (62:13): [True: 75.4k, False: 1.83k] ------------------ 63| | // we may need to fuse 64| 75.4k| if (run->runs[index + 1].value == pos + 1) { ------------------ | Branch (64:17): [True: 32.9k, False: 42.5k] ------------------ 65| | // indeed fusion is needed 66| 32.9k| run->runs[index + 1].value = pos; 67| 32.9k| run->runs[index + 1].length = run->runs[index + 1].length + 1; 68| 32.9k| return true; 69| 32.9k| } 70| 75.4k| } 71| 77.3k| } 72| 44.7k| if (index == -1) { ------------------ | Branch (72:9): [True: 380, False: 44.3k] ------------------ 73| | // we may need to extend the first run 74| 380| if (0 < run->n_runs) { ------------------ | Branch (74:13): [True: 380, False: 0] ------------------ 75| 380| if (run->runs[0].value == pos + 1) { ------------------ | Branch (75:17): [True: 260, False: 120] ------------------ 76| 260| run->runs[0].length++; 77| 260| run->runs[0].value--; 78| 260| return true; 79| 260| } 80| 380| } 81| 380| } 82| 44.4k| makeRoomAtIndex(run, (uint16_t)(index + 1)); 83| 44.4k| run->runs[index + 1].value = pos; 84| 44.4k| run->runs[index + 1].length = 0; 85| | return true; 86| 44.7k|} run_container_create_given_capacity: 89| 406k|run_container_t *run_container_create_given_capacity(int32_t size) { 90| 406k| run_container_t *run; 91| | /* Allocate the run container itself. */ 92| 406k| if ((run = (run_container_t *)roaring_malloc(sizeof(run_container_t))) == ------------------ | Branch (92:9): [True: 0, False: 406k] ------------------ 93| 406k| NULL) { 94| 0| return NULL; 95| 0| } 96| 406k| if (size <= 0) { // we don't want to rely on malloc(0) ------------------ | Branch (96:9): [True: 128k, False: 277k] ------------------ 97| 128k| run->runs = NULL; 98| 277k| } else if ((run->runs = (rle16_t *)roaring_malloc(sizeof(rle16_t) * ------------------ | Branch (98:16): [True: 0, False: 277k] ------------------ 99| 277k| size)) == NULL) { 100| 0| roaring_free(run); 101| 0| return NULL; 102| 0| } 103| 406k| run->capacity = size; 104| 406k| run->n_runs = 0; 105| 406k| return run; 106| 406k|} run_container_shrink_to_fit: 108| 2.67k|int run_container_shrink_to_fit(run_container_t *src) { 109| 2.67k| if (src->n_runs == src->capacity) return 0; // nothing to do ------------------ | Branch (109:9): [True: 2.67k, False: 0] ------------------ 110| 0| int savings = src->capacity - src->n_runs; 111| 0| src->capacity = src->n_runs; 112| 0| rle16_t *oldruns = src->runs; 113| 0| src->runs = 114| 0| (rle16_t *)roaring_realloc(oldruns, src->capacity * sizeof(rle16_t)); 115| 0| if (src->runs == NULL) roaring_free(oldruns); // should never happen? ------------------ | Branch (115:9): [True: 0, False: 0] ------------------ 116| 0| return savings; 117| 2.67k|} run_container_create: 119| 128k|run_container_t *run_container_create(void) { 120| 128k| return run_container_create_given_capacity(RUN_DEFAULT_INIT_SIZE); 121| 128k|} run_container_clone: 124| 181k|run_container_t *run_container_clone(const run_container_t *src) { 125| 181k| run_container_t *run = run_container_create_given_capacity(src->capacity); 126| 181k| if (run == NULL) return NULL; ------------------ | Branch (126:9): [True: 0, False: 181k] ------------------ 127| 181k| run->capacity = src->capacity; 128| 181k| run->n_runs = src->n_runs; 129| 181k| memcpy(run->runs, src->runs, src->n_runs * sizeof(rle16_t)); 130| 181k| return run; 131| 181k|} run_container_free: 193| 406k|void run_container_free(run_container_t *run) { 194| 406k| if (run == NULL) return; ------------------ | Branch (194:9): [True: 0, False: 406k] ------------------ 195| 406k| roaring_free(run->runs); 196| 406k| roaring_free(run); 197| 406k|} run_container_grow: 199| 134k|void run_container_grow(run_container_t *run, int32_t min, bool copy) { 200| 134k| int32_t newCapacity = (run->capacity == 0) ? RUN_DEFAULT_INIT_SIZE ------------------ | Branch (200:27): [True: 128k, False: 5.22k] ------------------ 201| 134k| : run->capacity < 64 ? run->capacity * 2 ------------------ | Branch (201:29): [True: 3.64k, False: 1.58k] ------------------ 202| 5.22k| : run->capacity < 1024 ? run->capacity * 3 / 2 ------------------ | Branch (202:29): [True: 1.58k, False: 0] ------------------ 203| 1.58k| : run->capacity * 5 / 4; 204| 134k| if (newCapacity < min) newCapacity = min; ------------------ | Branch (204:9): [True: 130k, False: 3.27k] ------------------ 205| 134k| run->capacity = newCapacity; 206| 134k| assert(run->capacity >= min); 207| 134k| if (copy) { ------------------ | Branch (207:9): [True: 5.22k, False: 128k] ------------------ 208| 5.22k| rle16_t *oldruns = run->runs; 209| 5.22k| run->runs = (rle16_t *)roaring_realloc(oldruns, 210| 5.22k| run->capacity * sizeof(rle16_t)); 211| 5.22k| if (run->runs == NULL) roaring_free(oldruns); ------------------ | Branch (211:13): [True: 0, False: 5.22k] ------------------ 212| 128k| } else { 213| 128k| roaring_free(run->runs); 214| 128k| run->runs = (rle16_t *)roaring_malloc(run->capacity * sizeof(rle16_t)); 215| 128k| } 216| | // We may have run->runs == NULL. 217| 134k|} run_container_copy: 220| 58.8k|void run_container_copy(const run_container_t *src, run_container_t *dst) { 221| 58.8k| const int32_t n_runs = src->n_runs; 222| 58.8k| if (src->n_runs > dst->capacity) { ------------------ | Branch (222:9): [True: 58.8k, False: 32] ------------------ 223| | run_container_grow(dst, n_runs, false); 224| 58.8k| } 225| 58.8k| dst->n_runs = n_runs; 226| 58.8k| memcpy(dst->runs, src->runs, sizeof(rle16_t) * n_runs); 227| 58.8k|} run_container_union: 232| 680| const run_container_t *src_2, run_container_t *dst) { 233| | // TODO: this could be a lot more efficient 234| | 235| | // we start out with inexpensive checks 236| 680| const bool if1 = run_container_is_full(src_1); 237| 680| const bool if2 = run_container_is_full(src_2); 238| 680| if (if1 || if2) { ------------------ | Branch (238:9): [True: 0, False: 680] | Branch (238:16): [True: 16, False: 664] ------------------ 239| 16| if (if1) { ------------------ | Branch (239:13): [True: 0, False: 16] ------------------ 240| 0| run_container_copy(src_1, dst); 241| 0| return; 242| 0| } 243| 16| if (if2) { ------------------ | Branch (243:13): [True: 16, False: 0] ------------------ 244| 16| run_container_copy(src_2, dst); 245| 16| return; 246| 16| } 247| 16| } 248| 664| const int32_t neededcapacity = src_1->n_runs + src_2->n_runs; 249| 664| if (dst->capacity < neededcapacity) ------------------ | Branch (249:9): [True: 664, False: 0] ------------------ 250| 664| run_container_grow(dst, neededcapacity, false); 251| 664| dst->n_runs = 0; 252| 664| int32_t rlepos = 0; 253| 664| int32_t xrlepos = 0; 254| | 255| 664| rle16_t previousrle; 256| 664| if (src_1->runs[rlepos].value <= src_2->runs[xrlepos].value) { ------------------ | Branch (256:9): [True: 604, False: 60] ------------------ 257| 604| previousrle = run_container_append_first(dst, src_1->runs[rlepos]); 258| 604| rlepos++; 259| 604| } else { 260| 60| previousrle = run_container_append_first(dst, src_2->runs[xrlepos]); 261| 60| xrlepos++; 262| 60| } 263| | 264| 85.0k| while ((xrlepos < src_2->n_runs) && (rlepos < src_1->n_runs)) { ------------------ | Branch (264:12): [True: 84.7k, False: 283] | Branch (264:41): [True: 84.4k, False: 381] ------------------ 265| 84.4k| rle16_t newrl; 266| 84.4k| if (src_1->runs[rlepos].value <= src_2->runs[xrlepos].value) { ------------------ | Branch (266:13): [True: 46.0k, False: 38.3k] ------------------ 267| 46.0k| newrl = src_1->runs[rlepos]; 268| 46.0k| rlepos++; 269| 46.0k| } else { 270| 38.3k| newrl = src_2->runs[xrlepos]; 271| 38.3k| xrlepos++; 272| 38.3k| } 273| 84.4k| run_container_append(dst, newrl, &previousrle); 274| 84.4k| } 275| 1.61k| while (xrlepos < src_2->n_runs) { ------------------ | Branch (275:12): [True: 950, False: 664] ------------------ 276| 950| run_container_append(dst, src_2->runs[xrlepos], &previousrle); 277| 950| xrlepos++; 278| 950| } 279| 11.1k| while (rlepos < src_1->n_runs) { ------------------ | Branch (279:12): [True: 10.5k, False: 664] ------------------ 280| 10.5k| run_container_append(dst, src_1->runs[rlepos], &previousrle); 281| 10.5k| rlepos++; 282| 10.5k| } 283| 664|} run_container_union_inplace: 288| 1.73k| const run_container_t *src_2) { 289| | // TODO: this could be a lot more efficient 290| | 291| | // we start out with inexpensive checks 292| 1.73k| const bool if1 = run_container_is_full(src_1); 293| 1.73k| const bool if2 = run_container_is_full(src_2); 294| 1.73k| if (if1 || if2) { ------------------ | Branch (294:9): [True: 0, False: 1.73k] | Branch (294:16): [True: 32, False: 1.70k] ------------------ 295| 32| if (if1) { ------------------ | Branch (295:13): [True: 0, False: 32] ------------------ 296| 0| return; 297| 0| } 298| 32| if (if2) { ------------------ | Branch (298:13): [True: 32, False: 0] ------------------ 299| 32| run_container_copy(src_2, src_1); 300| 32| return; 301| 32| } 302| 32| } 303| | // we move the data to the end of the current array 304| 1.70k| const int32_t maxoutput = src_1->n_runs + src_2->n_runs; 305| 1.70k| const int32_t neededcapacity = maxoutput + src_1->n_runs; 306| 1.70k| if (src_1->capacity < neededcapacity) ------------------ | Branch (306:9): [True: 1.70k, False: 0] ------------------ 307| 1.70k| run_container_grow(src_1, neededcapacity, true); 308| 1.70k| memmove(src_1->runs + maxoutput, src_1->runs, 309| 1.70k| src_1->n_runs * sizeof(rle16_t)); 310| 1.70k| rle16_t *inputsrc1 = src_1->runs + maxoutput; 311| 1.70k| const int32_t input1nruns = src_1->n_runs; 312| 1.70k| src_1->n_runs = 0; 313| 1.70k| int32_t rlepos = 0; 314| 1.70k| int32_t xrlepos = 0; 315| | 316| 1.70k| rle16_t previousrle; 317| 1.70k| if (inputsrc1[rlepos].value <= src_2->runs[xrlepos].value) { ------------------ | Branch (317:9): [True: 1.61k, False: 90] ------------------ 318| 1.61k| previousrle = run_container_append_first(src_1, inputsrc1[rlepos]); 319| 1.61k| rlepos++; 320| 1.61k| } else { 321| 90| previousrle = run_container_append_first(src_1, src_2->runs[xrlepos]); 322| 90| xrlepos++; 323| 90| } 324| 200k| while ((xrlepos < src_2->n_runs) && (rlepos < input1nruns)) { ------------------ | Branch (324:12): [True: 200k, False: 142] | Branch (324:41): [True: 198k, False: 1.56k] ------------------ 325| 198k| rle16_t newrl; 326| 198k| if (inputsrc1[rlepos].value <= src_2->runs[xrlepos].value) { ------------------ | Branch (326:13): [True: 95.0k, False: 103k] ------------------ 327| 95.0k| newrl = inputsrc1[rlepos]; 328| 95.0k| rlepos++; 329| 103k| } else { 330| 103k| newrl = src_2->runs[xrlepos]; 331| 103k| xrlepos++; 332| 103k| } 333| 198k| run_container_append(src_1, newrl, &previousrle); 334| 198k| } 335| 4.28k| while (xrlepos < src_2->n_runs) { ------------------ | Branch (335:12): [True: 2.58k, False: 1.70k] ------------------ 336| 2.58k| run_container_append(src_1, src_2->runs[xrlepos], &previousrle); 337| 2.58k| xrlepos++; 338| 2.58k| } 339| 8.73k| while (rlepos < input1nruns) { ------------------ | Branch (339:12): [True: 7.02k, False: 1.70k] ------------------ 340| 7.02k| run_container_append(src_1, inputsrc1[rlepos], &previousrle); 341| 7.02k| rlepos++; 342| 7.02k| } 343| 1.70k|} run_container_xor: 349| 61.2k| const run_container_t *src_2, run_container_t *dst) { 350| | // don't bother to convert xor with full range into negation 351| | // since negation is implemented similarly 352| | 353| 61.2k| const int32_t neededcapacity = src_1->n_runs + src_2->n_runs; 354| 61.2k| if (dst->capacity < neededcapacity) ------------------ | Branch (354:9): [True: 61.2k, False: 0] ------------------ 355| 61.2k| run_container_grow(dst, neededcapacity, false); 356| | 357| 61.2k| int32_t pos1 = 0; 358| 61.2k| int32_t pos2 = 0; 359| 61.2k| dst->n_runs = 0; 360| | 361| 239k| while ((pos1 < src_1->n_runs) && (pos2 < src_2->n_runs)) { ------------------ | Branch (361:12): [True: 178k, False: 60.9k] | Branch (361:38): [True: 178k, False: 304] ------------------ 362| 178k| if (src_1->runs[pos1].value <= src_2->runs[pos2].value) { ------------------ | Branch (362:13): [True: 121k, False: 56.4k] ------------------ 363| 121k| run_container_smart_append_exclusive(dst, src_1->runs[pos1].value, 364| 121k| src_1->runs[pos1].length); 365| 121k| pos1++; 366| 121k| } else { 367| 56.4k| run_container_smart_append_exclusive(dst, src_2->runs[pos2].value, 368| 56.4k| src_2->runs[pos2].length); 369| 56.4k| pos2++; 370| 56.4k| } 371| 178k| } 372| 73.1k| while (pos1 < src_1->n_runs) { ------------------ | Branch (372:12): [True: 11.9k, False: 61.2k] ------------------ 373| 11.9k| run_container_smart_append_exclusive(dst, src_1->runs[pos1].value, 374| 11.9k| src_1->runs[pos1].length); 375| 11.9k| pos1++; 376| 11.9k| } 377| | 378| 123k| while (pos2 < src_2->n_runs) { ------------------ | Branch (378:12): [True: 62.6k, False: 61.2k] ------------------ 379| 62.6k| run_container_smart_append_exclusive(dst, src_2->runs[pos2].value, 380| 62.6k| src_2->runs[pos2].length); 381| 62.6k| pos2++; 382| 62.6k| } 383| 61.2k|} run_container_intersection: 389| 62.2k| run_container_t *dst) { 390| 62.2k| const bool if1 = run_container_is_full(src_1); 391| 62.2k| const bool if2 = run_container_is_full(src_2); 392| 62.2k| if (if1 || if2) { ------------------ | Branch (392:9): [True: 58.6k, False: 3.57k] | Branch (392:16): [True: 16, False: 3.55k] ------------------ 393| 58.7k| if (if1) { ------------------ | Branch (393:13): [True: 58.6k, False: 16] ------------------ 394| 58.6k| run_container_copy(src_2, dst); 395| 58.6k| return; 396| 58.6k| } 397| 16| if (if2) { ------------------ | Branch (397:13): [True: 16, False: 0] ------------------ 398| 16| run_container_copy(src_1, dst); 399| 16| return; 400| 16| } 401| 16| } 402| | // TODO: this could be a lot more efficient, could use SIMD optimizations 403| 3.55k| const int32_t neededcapacity = src_1->n_runs + src_2->n_runs; 404| 3.55k| if (dst->capacity < neededcapacity) ------------------ | Branch (404:9): [True: 3.55k, False: 0] ------------------ 405| 3.55k| run_container_grow(dst, neededcapacity, false); 406| 3.55k| dst->n_runs = 0; 407| 3.55k| int32_t rlepos = 0; 408| 3.55k| int32_t xrlepos = 0; 409| 3.55k| int32_t start = src_1->runs[rlepos].value; 410| 3.55k| int32_t end = start + src_1->runs[rlepos].length + 1; 411| 3.55k| int32_t xstart = src_2->runs[xrlepos].value; 412| 3.55k| int32_t xend = xstart + src_2->runs[xrlepos].length + 1; 413| 187k| while ((rlepos < src_1->n_runs) && (xrlepos < src_2->n_runs)) { ------------------ | Branch (413:12): [True: 184k, False: 3.07k] | Branch (413:40): [True: 183k, False: 483] ------------------ 414| 183k| if (end <= xstart) { ------------------ | Branch (414:13): [True: 47.0k, False: 136k] ------------------ 415| 47.0k| ++rlepos; 416| 47.0k| if (rlepos < src_1->n_runs) { ------------------ | Branch (416:17): [True: 46.9k, False: 88] ------------------ 417| 46.9k| start = src_1->runs[rlepos].value; 418| 46.9k| end = start + src_1->runs[rlepos].length + 1; 419| 46.9k| } 420| 136k| } else if (xend <= start) { ------------------ | Branch (420:20): [True: 12.4k, False: 124k] ------------------ 421| 12.4k| ++xrlepos; 422| 12.4k| if (xrlepos < src_2->n_runs) { ------------------ | Branch (422:17): [True: 12.4k, False: 52] ------------------ 423| 12.4k| xstart = src_2->runs[xrlepos].value; 424| 12.4k| xend = xstart + src_2->runs[xrlepos].length + 1; 425| 12.4k| } 426| 124k| } else { // they overlap 427| 124k| const int32_t lateststart = start > xstart ? start : xstart; ------------------ | Branch (427:41): [True: 14.4k, False: 109k] ------------------ 428| 124k| int32_t earliestend; 429| 124k| if (end == xend) { // improbable ------------------ | Branch (429:17): [True: 95.1k, False: 29.2k] ------------------ 430| 95.1k| earliestend = end; 431| 95.1k| rlepos++; 432| 95.1k| xrlepos++; 433| 95.1k| if (rlepos < src_1->n_runs) { ------------------ | Branch (433:21): [True: 92.2k, False: 2.84k] ------------------ 434| 92.2k| start = src_1->runs[rlepos].value; 435| 92.2k| end = start + src_1->runs[rlepos].length + 1; 436| 92.2k| } 437| 95.1k| if (xrlepos < src_2->n_runs) { ------------------ | Branch (437:21): [True: 92.0k, False: 3.04k] ------------------ 438| 92.0k| xstart = src_2->runs[xrlepos].value; 439| 92.0k| xend = xstart + src_2->runs[xrlepos].length + 1; 440| 92.0k| } 441| 95.1k| } else if (end < xend) { ------------------ | Branch (441:24): [True: 14.7k, False: 14.5k] ------------------ 442| 14.7k| earliestend = end; 443| 14.7k| rlepos++; 444| 14.7k| if (rlepos < src_1->n_runs) { ------------------ | Branch (444:21): [True: 14.5k, False: 145] ------------------ 445| 14.5k| start = src_1->runs[rlepos].value; 446| 14.5k| end = start + src_1->runs[rlepos].length + 1; 447| 14.5k| } 448| | 449| 14.7k| } else { // end > xend 450| 14.5k| earliestend = xend; 451| 14.5k| xrlepos++; 452| 14.5k| if (xrlepos < src_2->n_runs) { ------------------ | Branch (452:21): [True: 14.3k, False: 161] ------------------ 453| 14.3k| xstart = src_2->runs[xrlepos].value; 454| 14.3k| xend = xstart + src_2->runs[xrlepos].length + 1; 455| 14.3k| } 456| 14.5k| } 457| 124k| dst->runs[dst->n_runs].value = (uint16_t)lateststart; 458| 124k| dst->runs[dst->n_runs].length = 459| 124k| (uint16_t)(earliestend - lateststart - 1); 460| 124k| dst->n_runs++; 461| 124k| } 462| 183k| } 463| 3.55k|} run_container_intersection_cardinality: 467| 2.72k| const run_container_t *src_2) { 468| 2.72k| const bool if1 = run_container_is_full(src_1); 469| 2.72k| const bool if2 = run_container_is_full(src_2); 470| 2.72k| if (if1 || if2) { ------------------ | Branch (470:9): [True: 0, False: 2.72k] | Branch (470:16): [True: 64, False: 2.65k] ------------------ 471| 64| if (if1) { ------------------ | Branch (471:13): [True: 0, False: 64] ------------------ 472| 0| return run_container_cardinality(src_2); 473| 0| } 474| 64| if (if2) { ------------------ | Branch (474:13): [True: 64, False: 0] ------------------ 475| 64| return run_container_cardinality(src_1); 476| 64| } 477| 64| } 478| 2.65k| int answer = 0; 479| 2.65k| int32_t rlepos = 0; 480| 2.65k| int32_t xrlepos = 0; 481| 2.65k| int32_t start = src_1->runs[rlepos].value; 482| 2.65k| int32_t end = start + src_1->runs[rlepos].length + 1; 483| 2.65k| int32_t xstart = src_2->runs[xrlepos].value; 484| 2.65k| int32_t xend = xstart + src_2->runs[xrlepos].length + 1; 485| 285k| while ((rlepos < src_1->n_runs) && (xrlepos < src_2->n_runs)) { ------------------ | Branch (485:12): [True: 283k, False: 1.82k] | Branch (485:40): [True: 282k, False: 836] ------------------ 486| 282k| if (end <= xstart) { ------------------ | Branch (486:13): [True: 68.9k, False: 213k] ------------------ 487| 68.9k| ++rlepos; 488| 68.9k| if (rlepos < src_1->n_runs) { ------------------ | Branch (488:17): [True: 68.5k, False: 352] ------------------ 489| 68.5k| start = src_1->runs[rlepos].value; 490| 68.5k| end = start + src_1->runs[rlepos].length + 1; 491| 68.5k| } 492| 213k| } else if (xend <= start) { ------------------ | Branch (492:20): [True: 49.8k, False: 163k] ------------------ 493| 49.8k| ++xrlepos; 494| 49.8k| if (xrlepos < src_2->n_runs) { ------------------ | Branch (494:17): [True: 49.6k, False: 208] ------------------ 495| 49.6k| xstart = src_2->runs[xrlepos].value; 496| 49.6k| xend = xstart + src_2->runs[xrlepos].length + 1; 497| 49.6k| } 498| 163k| } else { // they overlap 499| 163k| const int32_t lateststart = start > xstart ? start : xstart; ------------------ | Branch (499:41): [True: 57.6k, False: 106k] ------------------ 500| 163k| int32_t earliestend; 501| 163k| if (end == xend) { // improbable ------------------ | Branch (501:17): [True: 86.5k, False: 77.3k] ------------------ 502| 86.5k| earliestend = end; 503| 86.5k| rlepos++; 504| 86.5k| xrlepos++; 505| 86.5k| if (rlepos < src_1->n_runs) { ------------------ | Branch (505:21): [True: 85.6k, False: 888] ------------------ 506| 85.6k| start = src_1->runs[rlepos].value; 507| 85.6k| end = start + src_1->runs[rlepos].length + 1; 508| 85.6k| } 509| 86.5k| if (xrlepos < src_2->n_runs) { ------------------ | Branch (509:21): [True: 85.5k, False: 1.04k] ------------------ 510| 85.5k| xstart = src_2->runs[xrlepos].value; 511| 85.5k| xend = xstart + src_2->runs[xrlepos].length + 1; 512| 85.5k| } 513| 86.5k| } else if (end < xend) { ------------------ | Branch (513:24): [True: 58.9k, False: 18.4k] ------------------ 514| 58.9k| earliestend = end; 515| 58.9k| rlepos++; 516| 58.9k| if (rlepos < src_1->n_runs) { ------------------ | Branch (516:21): [True: 58.3k, False: 580] ------------------ 517| 58.3k| start = src_1->runs[rlepos].value; 518| 58.3k| end = start + src_1->runs[rlepos].length + 1; 519| 58.3k| } 520| | 521| 58.9k| } else { // end > xend 522| 18.4k| earliestend = xend; 523| 18.4k| xrlepos++; 524| 18.4k| if (xrlepos < src_2->n_runs) { ------------------ | Branch (524:21): [True: 18.2k, False: 204] ------------------ 525| 18.2k| xstart = src_2->runs[xrlepos].value; 526| 18.2k| xend = xstart + src_2->runs[xrlepos].length + 1; 527| 18.2k| } 528| 18.4k| } 529| 163k| answer += earliestend - lateststart; 530| 163k| } 531| 282k| } 532| 2.65k| return answer; 533| 2.72k|} run_container_intersect: 536| 680| const run_container_t *src_2) { 537| 680| const bool if1 = run_container_is_full(src_1); 538| 680| const bool if2 = run_container_is_full(src_2); 539| 680| if (if1 || if2) { ------------------ | Branch (539:9): [True: 0, False: 680] | Branch (539:16): [True: 16, False: 664] ------------------ 540| 16| if (if1) { ------------------ | Branch (540:13): [True: 0, False: 16] ------------------ 541| 0| return !run_container_empty(src_2); 542| 0| } 543| 16| if (if2) { ------------------ | Branch (543:13): [True: 16, False: 0] ------------------ 544| 16| return !run_container_empty(src_1); 545| 16| } 546| 16| } 547| 664| int32_t rlepos = 0; 548| 664| int32_t xrlepos = 0; 549| 664| int32_t start = src_1->runs[rlepos].value; 550| 664| int32_t end = start + src_1->runs[rlepos].length + 1; 551| 664| int32_t xstart = src_2->runs[xrlepos].value; 552| 664| int32_t xend = xstart + src_2->runs[xrlepos].length + 1; 553| 3.27k| while ((rlepos < src_1->n_runs) && (xrlepos < src_2->n_runs)) { ------------------ | Branch (553:12): [True: 3.25k, False: 15] | Branch (553:40): [True: 3.25k, False: 4] ------------------ 554| 3.25k| if (end <= xstart) { ------------------ | Branch (554:13): [True: 2.21k, False: 1.03k] ------------------ 555| 2.21k| ++rlepos; 556| 2.21k| if (rlepos < src_1->n_runs) { ------------------ | Branch (556:17): [True: 2.20k, False: 15] ------------------ 557| 2.20k| start = src_1->runs[rlepos].value; 558| 2.20k| end = start + src_1->runs[rlepos].length + 1; 559| 2.20k| } 560| 2.21k| } else if (xend <= start) { ------------------ | Branch (560:20): [True: 393, False: 645] ------------------ 561| 393| ++xrlepos; 562| 393| if (xrlepos < src_2->n_runs) { ------------------ | Branch (562:17): [True: 389, False: 4] ------------------ 563| 389| xstart = src_2->runs[xrlepos].value; 564| 389| xend = xstart + src_2->runs[xrlepos].length + 1; 565| 389| } 566| 645| } else { // they overlap 567| 645| return true; 568| 645| } 569| 3.25k| } 570| 19| return false; 571| 664|} run_container_andnot: 576| 1.09k| const run_container_t *src_2, run_container_t *dst) { 577| | // following Java implementation as of June 2016 578| | 579| 1.09k| if (dst->capacity < src_1->n_runs + src_2->n_runs) ------------------ | Branch (579:9): [True: 1.09k, False: 0] ------------------ 580| 1.09k| run_container_grow(dst, src_1->n_runs + src_2->n_runs, false); 581| | 582| 1.09k| dst->n_runs = 0; 583| | 584| 1.09k| int rlepos1 = 0; 585| 1.09k| int rlepos2 = 0; 586| 1.09k| int32_t start = src_1->runs[rlepos1].value; 587| 1.09k| int32_t end = start + src_1->runs[rlepos1].length + 1; 588| 1.09k| int32_t start2 = src_2->runs[rlepos2].value; 589| 1.09k| int32_t end2 = start2 + src_2->runs[rlepos2].length + 1; 590| | 591| 156k| while ((rlepos1 < src_1->n_runs) && (rlepos2 < src_2->n_runs)) { ------------------ | Branch (591:12): [True: 155k, False: 670] | Branch (591:41): [True: 155k, False: 426] ------------------ 592| 155k| if (end <= start2) { ------------------ | Branch (592:13): [True: 23.1k, False: 132k] ------------------ 593| | // output the first run 594| 23.1k| dst->runs[dst->n_runs++] = 595| 23.1k| CROARING_MAKE_RLE16(start, end - start - 1); ------------------ | | 60| 23.1k| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 596| 23.1k| rlepos1++; 597| 23.1k| if (rlepos1 < src_1->n_runs) { ------------------ | Branch (597:17): [True: 23.0k, False: 88] ------------------ 598| 23.0k| start = src_1->runs[rlepos1].value; 599| 23.0k| end = start + src_1->runs[rlepos1].length + 1; 600| 23.0k| } 601| 132k| } else if (end2 <= start) { ------------------ | Branch (601:20): [True: 64.7k, False: 67.4k] ------------------ 602| | // exit the second run 603| 64.7k| rlepos2++; 604| 64.7k| if (rlepos2 < src_2->n_runs) { ------------------ | Branch (604:17): [True: 64.2k, False: 426] ------------------ 605| 64.2k| start2 = src_2->runs[rlepos2].value; 606| 64.2k| end2 = start2 + src_2->runs[rlepos2].length + 1; 607| 64.2k| } 608| 67.4k| } else { 609| 67.4k| if (start < start2) { ------------------ | Branch (609:17): [True: 17.9k, False: 49.4k] ------------------ 610| 17.9k| dst->runs[dst->n_runs++] = 611| 17.9k| CROARING_MAKE_RLE16(start, start2 - start - 1); ------------------ | | 60| 17.9k| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 612| 17.9k| } 613| 67.4k| if (end2 < end) { ------------------ | Branch (613:17): [True: 17.4k, False: 49.9k] ------------------ 614| 17.4k| start = end2; 615| 49.9k| } else { 616| 49.9k| rlepos1++; 617| 49.9k| if (rlepos1 < src_1->n_runs) { ------------------ | Branch (617:21): [True: 49.3k, False: 582] ------------------ 618| 49.3k| start = src_1->runs[rlepos1].value; 619| 49.3k| end = start + src_1->runs[rlepos1].length + 1; 620| 49.3k| } 621| 49.9k| } 622| 67.4k| } 623| 155k| } 624| 1.09k| if (rlepos1 < src_1->n_runs) { ------------------ | Branch (624:9): [True: 426, False: 670] ------------------ 625| 426| dst->runs[dst->n_runs++] = CROARING_MAKE_RLE16(start, end - start - 1); ------------------ | | 60| 426| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 626| 426| rlepos1++; 627| 426| if (rlepos1 < src_1->n_runs) { ------------------ | Branch (627:13): [True: 182, False: 244] ------------------ 628| 182| memcpy(dst->runs + dst->n_runs, src_1->runs + rlepos1, 629| 182| sizeof(rle16_t) * (src_1->n_runs - rlepos1)); 630| 182| dst->n_runs += src_1->n_runs - rlepos1; 631| 182| } 632| 426| } 633| 1.09k|} run_container_read: 739| 50| const char *buf) { 740| 50| (void)cardinality; 741| 50| uint16_t cast_16; 742| 50| memcpy(&cast_16, buf, sizeof(uint16_t)); 743| 50| container->n_runs = croaring_letoh16(cast_16); ------------------ | | 523| 50|#define croaring_letoh16(x) croaring_htole16(x) | | ------------------ | | | | 517| 50|#define croaring_htole16(x) (x) | | ------------------ ------------------ 744| 50| if (container->n_runs > container->capacity) ------------------ | Branch (744:9): [True: 7, False: 43] ------------------ 745| 7| run_container_grow(container, container->n_runs, false); 746| 50| if (container->n_runs > 0) { ------------------ | Branch (746:9): [True: 7, False: 43] ------------------ 747| |#if CROARING_IS_BIG_ENDIAN 748| | const char *in = buf + sizeof(uint16_t); 749| | for (int32_t i = 0; i < container->n_runs; ++i) { 750| | uint16_t v_le, l_le; 751| | memcpy(&v_le, in, sizeof(uint16_t)); 752| | memcpy(&l_le, in + sizeof(uint16_t), sizeof(uint16_t)); 753| | container->runs[i].value = croaring_letoh16(v_le); 754| | container->runs[i].length = croaring_letoh16(l_le); 755| | in += sizeof(rle16_t); 756| | } 757| |#else 758| 7| memcpy(container->runs, buf + sizeof(uint16_t), 759| 7| container->n_runs * sizeof(rle16_t)); 760| 7|#endif 761| 7| } 762| 50| return run_container_size_in_bytes(container); 763| 50|} run_container_iterate: 766| 61.1k| roaring_iterator iterator, void *ptr) { 767| 162k| for (int i = 0; i < cont->n_runs; ++i) { ------------------ | Branch (767:21): [True: 101k, False: 61.1k] ------------------ 768| 101k| uint32_t run_start = base + cont->runs[i].value; 769| 101k| uint16_t le = cont->runs[i].length; 770| | 771| 3.91G| for (int j = 0; j <= le; ++j) ------------------ | Branch (771:25): [True: 3.91G, False: 101k] ------------------ 772| 3.91G| if (!iterator(run_start + j, ptr)) return false; ------------------ | Branch (772:17): [True: 0, False: 3.91G] ------------------ 773| 101k| } 774| 61.1k| return true; 775| 61.1k|} run_container_is_subset: 792| 244| const run_container_t *container2) { 793| 244| int i1 = 0, i2 = 0; 794| 643| while (i1 < container1->n_runs && i2 < container2->n_runs) { ------------------ | Branch (794:12): [True: 643, False: 0] | Branch (794:39): [True: 631, False: 12] ------------------ 795| 631| int start1 = container1->runs[i1].value; 796| 631| int stop1 = start1 + container1->runs[i1].length; 797| 631| int start2 = container2->runs[i2].value; 798| 631| int stop2 = start2 + container2->runs[i2].length; 799| 631| if (start1 < start2) { ------------------ | Branch (799:13): [True: 232, False: 399] ------------------ 800| 232| return false; 801| 399| } else { // start1 >= start2 802| 399| if (stop1 < stop2) { ------------------ | Branch (802:17): [True: 0, False: 399] ------------------ 803| 0| i1++; 804| 399| } else if (stop1 == stop2) { ------------------ | Branch (804:24): [True: 0, False: 399] ------------------ 805| 0| i1++; 806| 0| i2++; 807| 399| } else { // stop1 > stop2 808| 399| i2++; 809| 399| } 810| 399| } 811| 631| } 812| 12| if (i1 == container1->n_runs) { ------------------ | Branch (812:9): [True: 0, False: 12] ------------------ 813| 0| return true; 814| 12| } else { 815| | return false; 816| 12| } 817| 12|} run_container_smart_append_exclusive: 826| 392k| const uint16_t length) { 827| 392k| int old_end; 828| 392k| rle16_t *last_run = src->n_runs ? src->runs + (src->n_runs - 1) : NULL; ------------------ | Branch (828:25): [True: 315k, False: 76.9k] ------------------ 829| 392k| rle16_t *appended_last_run = src->runs + src->n_runs; 830| | 831| 392k| if (!src->n_runs || ------------------ | Branch (831:9): [True: 76.9k, False: 315k] ------------------ 832| 315k| (start > (old_end = last_run->value + last_run->length + 1))) { ------------------ | Branch (832:9): [True: 170k, False: 145k] ------------------ 833| 247k| *appended_last_run = CROARING_MAKE_RLE16(start, length); ------------------ | | 60| 247k| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 834| 247k| src->n_runs++; 835| 247k| return; 836| 247k| } 837| 145k| if (old_end == start) { ------------------ | Branch (837:9): [True: 3.75k, False: 141k] ------------------ 838| | // we merge 839| 3.75k| last_run->length += (length + 1); 840| 3.75k| return; 841| 3.75k| } 842| 141k| int new_end = start + length + 1; 843| | 844| 141k| if (start == last_run->value) { ------------------ | Branch (844:9): [True: 113k, False: 28.6k] ------------------ 845| | // wipe out previous 846| 113k| if (new_end < old_end) { ------------------ | Branch (846:13): [True: 7.20k, False: 105k] ------------------ 847| 7.20k| *last_run = CROARING_MAKE_RLE16(new_end, old_end - new_end - 1); ------------------ | | 60| 7.20k| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 848| 7.20k| return; 849| 105k| } else if (new_end > old_end) { ------------------ | Branch (849:20): [True: 2.33k, False: 103k] ------------------ 850| 2.33k| *last_run = CROARING_MAKE_RLE16(old_end, new_end - old_end - 1); ------------------ | | 60| 2.33k| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 851| 2.33k| return; 852| 103k| } else { 853| 103k| src->n_runs--; 854| 103k| return; 855| 103k| } 856| 113k| } 857| 28.6k| last_run->length = start - last_run->value - 1; 858| 28.6k| if (new_end < old_end) { ------------------ | Branch (858:9): [True: 22.6k, False: 5.93k] ------------------ 859| 22.6k| *appended_last_run = 860| 22.6k| CROARING_MAKE_RLE16(new_end, old_end - new_end - 1); ------------------ | | 60| 22.6k| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 861| 22.6k| src->n_runs++; 862| 22.6k| } else if (new_end > old_end) { ------------------ | Branch (862:16): [True: 1.07k, False: 4.85k] ------------------ 863| 1.07k| *appended_last_run = 864| 1.07k| CROARING_MAKE_RLE16(old_end, new_end - old_end - 1); ------------------ | | 60| 1.07k| (rle16_t) { .value = (uint16_t)(val), .length = (uint16_t)(len) } ------------------ 865| 1.07k| src->n_runs++; 866| 1.07k| } 867| 28.6k|} run_container_select: 871| 2.67k| uint32_t *element) { 872| 15.7k| for (int i = 0; i < container->n_runs; i++) { ------------------ | Branch (872:21): [True: 15.6k, False: 104] ------------------ 873| 15.6k| uint16_t length = container->runs[i].length; 874| 15.6k| if (rank <= *start_rank + length) { ------------------ | Branch (874:13): [True: 2.57k, False: 13.0k] ------------------ 875| 2.57k| uint16_t value = container->runs[i].value; 876| 2.57k| *element = value + rank - (*start_rank); 877| 2.57k| return true; 878| 2.57k| } else 879| 13.0k| *start_rank += length + 1; 880| 15.6k| } 881| 104| return false; 882| 2.67k|} run_container_rank: 884| 2.67k|int run_container_rank(const run_container_t *container, uint16_t x) { 885| 2.67k| int sum = 0; 886| 2.67k| uint32_t x32 = x; 887| 12.9k| for (int i = 0; i < container->n_runs; i++) { ------------------ | Branch (887:21): [True: 12.9k, False: 74] ------------------ 888| 12.9k| uint32_t startpoint = container->runs[i].value; 889| 12.9k| uint32_t length = container->runs[i].length; 890| 12.9k| uint32_t endpoint = length + startpoint; 891| 12.9k| if (x <= endpoint) { ------------------ | Branch (891:13): [True: 2.60k, False: 10.3k] ------------------ 892| 2.60k| if (x < startpoint) break; ------------------ | Branch (892:17): [True: 293, False: 2.31k] ------------------ 893| 2.31k| return sum + (x32 - startpoint) + 1; 894| 10.3k| } else { 895| 10.3k| sum += length + 1; 896| 10.3k| } 897| 12.9k| } 898| 367| return sum; 899| 2.67k|} _avx2_run_container_to_uint32_array: 1021| 61.7k| uint32_t base) { 1022| 61.7k| int outpos = 0; 1023| 61.7k| uint32_t *out = (uint32_t *)vout; 1024| | 1025| 210k| for (int i = 0; i < cont->n_runs; ++i) { ------------------ | Branch (1025:21): [True: 149k, False: 61.7k] ------------------ 1026| 149k| uint32_t run_start = base + cont->runs[i].value; 1027| 149k| uint16_t le = cont->runs[i].length; 1028| 149k| if (le < 8) { ------------------ | Branch (1028:13): [True: 83.6k, False: 65.5k] ------------------ 1029| 216k| for (int j = 0; j <= le; ++j) { ------------------ | Branch (1029:29): [True: 133k, False: 83.6k] ------------------ 1030| 133k| uint32_t val = run_start + j; 1031| 133k| memcpy(out + outpos, &val, 1032| 133k| sizeof(uint32_t)); // should be compiled as a MOV on x64 1033| 133k| outpos++; 1034| 133k| } 1035| 83.6k| } else { 1036| 65.5k| int j = 0; 1037| 65.5k| __m256i run_start_v = _mm256_set1_epi32(run_start); 1038| | // [8,8,8,8....] 1039| 65.5k| __m256i inc = _mm256_set1_epi32(8); 1040| | // used for generate sequence: 1041| | // [0, 1, 2, 3...], [8, 9, 10,...] 1042| 65.5k| __m256i delta = _mm256_setr_epi32(0, 1, 2, 3, 4, 5, 6, 7); 1043| 491M| for (j = 0; j + 8 <= le; j += 8) { ------------------ | Branch (1043:25): [True: 491M, False: 65.5k] ------------------ 1044| 491M| __m256i val_v = _mm256_add_epi32(run_start_v, delta); 1045| 491M| _mm256_storeu_si256((__m256i *)(out + outpos), val_v); 1046| 491M| delta = _mm256_add_epi32(inc, delta); 1047| 491M| outpos += 8; 1048| 491M| } 1049| 564k| for (; j <= le; ++j) { ------------------ | Branch (1049:20): [True: 499k, False: 65.5k] ------------------ 1050| 499k| uint32_t val = run_start + j; 1051| 499k| memcpy(out + outpos, &val, 1052| 499k| sizeof(uint32_t)); // should be compiled as a MOV on x64 1053| 499k| outpos++; 1054| 499k| } 1055| 65.5k| } 1056| 149k| } 1057| 61.7k| return outpos; 1058| 61.7k|} run_container_cardinality: 1077| 550k|int run_container_cardinality(const run_container_t *run) { 1078| | // Empirically AVX-512 is not always faster than AVX2 1079| |#if CROARING_COMPILER_SUPPORTS_AVX512 && \ 1080| | CROARING_ENABLE_AVX512_RUN_CONTAINER_CARDINALITY 1081| | if (croaring_hardware_support() & ROARING_SUPPORTS_AVX512) { 1082| | return _avx512_run_container_cardinality(run); 1083| | } else 1084| |#endif 1085| 550k| if (croaring_hardware_support() & ROARING_SUPPORTS_AVX2) { ------------------ | Branch (1085:13): [True: 550k, False: 0] ------------------ 1086| 550k| return _avx2_run_container_cardinality(run); 1087| 550k| } else { 1088| 0| return _scalar_run_container_cardinality(run); 1089| 0| } 1090| 550k|} run_container_to_uint32_array: 1111| 61.7k| uint32_t base) { 1112| 61.7k| if (croaring_hardware_support() & ROARING_SUPPORTS_AVX2) { ------------------ | Branch (1112:9): [True: 61.7k, False: 0] ------------------ 1113| 61.7k| return _avx2_run_container_to_uint32_array(vout, cont, base); 1114| 61.7k| } else { 1115| 0| return _scalar_run_container_to_uint32_array(vout, cont, base); 1116| 0| } 1117| 61.7k|} run.c:_avx2_run_container_cardinality: 991| 550k|static inline int _avx2_run_container_cardinality(const run_container_t *run) { 992| 550k| const int32_t n_runs = run->n_runs; 993| 550k| const rle16_t *runs = run->runs; 994| | 995| | /* by initializing with n_runs, we omit counting the +1 for each pair. */ 996| 550k| int sum = n_runs; 997| 550k| int32_t k = 0; 998| 550k| const int32_t step = sizeof(__m256i) / sizeof(rle16_t); 999| 550k| if (n_runs > step) { ------------------ | Branch (999:9): [True: 33.7k, False: 516k] ------------------ 1000| 33.7k| __m256i total = _mm256_setzero_si256(); 1001| 382k| for (; k + step <= n_runs; k += step) { ------------------ | Branch (1001:16): [True: 348k, False: 33.7k] ------------------ 1002| 348k| __m256i ymm1 = _mm256_lddqu_si256((const __m256i *)(runs + k)); 1003| 348k| __m256i justlengths = _mm256_srli_epi32(ymm1, 16); 1004| 348k| total = _mm256_add_epi32(total, justlengths); 1005| 348k| } 1006| | // a store might be faster than extract? 1007| 33.7k| uint32_t buffer[sizeof(__m256i) / sizeof(rle16_t)]; 1008| 33.7k| _mm256_storeu_si256((__m256i *)buffer, total); 1009| 33.7k| sum += (buffer[0] + buffer[1]) + (buffer[2] + buffer[3]) + 1010| 33.7k| (buffer[4] + buffer[5]) + (buffer[6] + buffer[7]); 1011| 33.7k| } 1012| 1.14M| for (; k < n_runs; ++k) { ------------------ | Branch (1012:12): [True: 593k, False: 550k] ------------------ 1013| 593k| sum += runs[k].length; 1014| 593k| } 1015| | 1016| 550k| return sum; 1017| 550k|} croaring_hardware_support: 321| 980k|int croaring_hardware_support(void) { 322| 980k| static 323| 980k|#if CROARING_ATOMIC_IMPL == CROARING_ATOMIC_IMPL_C 324| 980k| _Atomic 325| 980k|#endif 326| 980k| int support = 0xFFFFFFF; 327| 980k| if (support == 0xFFFFFFF) { ------------------ | Branch (327:9): [True: 1, False: 980k] ------------------ 328| 1| bool has_avx2 = (croaring_detect_supported_architectures() & 329| 1| CROARING_AVX2) == CROARING_AVX2; 330| 1| bool has_avx512 = false; 331| 1|#if CROARING_COMPILER_SUPPORTS_AVX512 332| 1| has_avx512 = (croaring_detect_supported_architectures() & 333| 1| CROARING_AVX512_REQUIRED) == CROARING_AVX512_REQUIRED; 334| 1|#endif // CROARING_COMPILER_SUPPORTS_AVX512 335| 1| support = (has_avx2 ? ROARING_SUPPORTS_AVX2 : 0) | ------------------ | Branch (335:20): [True: 1, False: 0] ------------------ 336| 1| (has_avx512 ? ROARING_SUPPORTS_AVX512 : 0); ------------------ | Branch (336:20): [True: 0, False: 1] ------------------ 337| 1| } 338| 980k| return support; 339| 980k|} isadetection.c:croaring_detect_supported_architectures: 269| 2|static uint32_t croaring_detect_supported_architectures(void) { 270| | // we use an atomic for thread safety 271| 2| static _Atomic uint32_t buffer = CROARING_UNINITIALIZED; 272| 2| if (buffer == CROARING_UNINITIALIZED) { ------------------ | Branch (272:9): [True: 1, False: 1] ------------------ 273| | // atomicity is sufficient 274| 1| buffer = dynamic_croaring_detect_supported_architectures(); 275| 1| } 276| 2| return buffer; 277| 2|} isadetection.c:dynamic_croaring_detect_supported_architectures: 154| 1|static inline uint32_t dynamic_croaring_detect_supported_architectures(void) { 155| 1| uint32_t eax, ebx, ecx, edx; 156| 1| uint32_t host_isa = 0x0; 157| | // Can be found on Intel ISA Reference for CPUID 158| 1| static uint32_t cpuid_avx2_bit = 159| 1| 1 << 5; ///< @private Bit 5 of EBX for EAX=0x7 160| 1| static uint32_t cpuid_bmi1_bit = 161| 1| 1 << 3; ///< @private bit 3 of EBX for EAX=0x7 162| 1| static uint32_t cpuid_bmi2_bit = 163| 1| 1 << 8; ///< @private bit 8 of EBX for EAX=0x7 164| 1| static uint32_t cpuid_avx512f_bit = 165| 1| 1 << 16; ///< @private bit 16 of EBX for EAX=0x7 166| 1| static uint32_t cpuid_avx512dq_bit = 167| 1| 1 << 17; ///< @private bit 17 of EBX for EAX=0x7 168| 1| static uint32_t cpuid_avx512bw_bit = 169| 1| 1 << 30; ///< @private bit 30 of EBX for EAX=0x7 170| 1| static uint32_t cpuid_avx512vbmi2_bit = 171| 1| 1 << 6; ///< @private bit 6 of ECX for EAX=0x7 172| 1| static uint32_t cpuid_avx512bitalg_bit = 173| 1| 1 << 12; ///< @private bit 12 of ECX for EAX=0x7 174| 1| static uint32_t cpuid_avx512vpopcntdq_bit = 175| 1| 1 << 14; ///< @private bit 14 of ECX for EAX=0x7 176| 1| static uint64_t cpuid_avx256_saved = 1 << 2; ///< @private bit 2 = AVX 177| 1| static uint64_t cpuid_avx512_saved = 178| 1| 7 << 5; ///< @private bits 5,6,7 = opmask, ZMM_hi256, hi16_ZMM 179| 1| static uint32_t cpuid_sse42_bit = 180| 1| 1 << 20; ///< @private bit 20 of ECX for EAX=0x1 181| 1| static uint32_t cpuid_osxsave = 182| 1| (1 << 26) | (1 << 27); ///< @private bits 26+27 of ECX for EAX=0x1 183| 1| static uint32_t cpuid_pclmulqdq_bit = 184| 1| 1 << 1; ///< @private bit 1 of ECX for EAX=0x1 185| | 186| | // EBX for EAX=0x1 187| 1| eax = 0x1; 188| 1| ecx = 0x0; 189| 1| cpuid(&eax, &ebx, &ecx, &edx); 190| | 191| 1| if (ecx & cpuid_sse42_bit) { ------------------ | Branch (191:9): [True: 1, False: 0] ------------------ 192| 1| host_isa |= CROARING_SSE42; 193| 1| } else { 194| 0| return host_isa; // everything after is redundant 195| 0| } 196| | 197| 1| if (ecx & cpuid_pclmulqdq_bit) { ------------------ | Branch (197:9): [True: 1, False: 0] ------------------ 198| 1| host_isa |= CROARING_PCLMULQDQ; 199| 1| } 200| | 201| 1| if ((ecx & cpuid_osxsave) != cpuid_osxsave) { ------------------ | Branch (201:9): [True: 0, False: 1] ------------------ 202| 0| return host_isa; 203| 0| } 204| | 205| | // xgetbv for checking if the OS saves registers 206| 1| uint64_t xcr0 = xgetbv(); 207| | 208| 1| if ((xcr0 & cpuid_avx256_saved) == 0) { ------------------ | Branch (208:9): [True: 0, False: 1] ------------------ 209| 0| return host_isa; 210| 0| } 211| | 212| | // ECX for EAX=0x7 213| 1| eax = 0x7; 214| 1| ecx = 0x0; 215| 1| cpuid(&eax, &ebx, &ecx, &edx); 216| 1| if (ebx & cpuid_avx2_bit) { ------------------ | Branch (216:9): [True: 1, False: 0] ------------------ 217| 1| host_isa |= CROARING_AVX2; 218| 1| } 219| 1| if (ebx & cpuid_bmi1_bit) { ------------------ | Branch (219:9): [True: 1, False: 0] ------------------ 220| 1| host_isa |= CROARING_BMI1; 221| 1| } 222| | 223| 1| if (ebx & cpuid_bmi2_bit) { ------------------ | Branch (223:9): [True: 1, False: 0] ------------------ 224| 1| host_isa |= CROARING_BMI2; 225| 1| } 226| | 227| 1| if (!((xcr0 & cpuid_avx512_saved) == cpuid_avx512_saved)) { ------------------ | Branch (227:9): [True: 1, False: 0] ------------------ 228| 1| return host_isa; 229| 1| } 230| | 231| 0| if (ebx & cpuid_avx512f_bit) { ------------------ | Branch (231:9): [True: 0, False: 0] ------------------ 232| 0| host_isa |= CROARING_AVX512F; 233| 0| } 234| | 235| 0| if (ebx & cpuid_avx512bw_bit) { ------------------ | Branch (235:9): [True: 0, False: 0] ------------------ 236| 0| host_isa |= CROARING_AVX512BW; 237| 0| } 238| | 239| 0| if (ebx & cpuid_avx512dq_bit) { ------------------ | Branch (239:9): [True: 0, False: 0] ------------------ 240| 0| host_isa |= CROARING_AVX512DQ; 241| 0| } 242| | 243| 0| if (ecx & cpuid_avx512vbmi2_bit) { ------------------ | Branch (243:9): [True: 0, False: 0] ------------------ 244| 0| host_isa |= CROARING_AVX512VBMI2; 245| 0| } 246| | 247| 0| if (ecx & cpuid_avx512bitalg_bit) { ------------------ | Branch (247:9): [True: 0, False: 0] ------------------ 248| 0| host_isa |= CROARING_AVX512BITALG; 249| 0| } 250| | 251| 0| if (ecx & cpuid_avx512vpopcntdq_bit) { ------------------ | Branch (251:9): [True: 0, False: 0] ------------------ 252| 0| host_isa |= CROARING_AVX512VPOPCNTDQ; 253| 0| } 254| | 255| 0| return host_isa; 256| 1|} isadetection.c:cpuid: 115| 2| uint32_t *edx) { 116| |#if CROARING_REGULAR_VISUAL_STUDIO 117| | int cpu_info[4]; 118| | __cpuidex(cpu_info, *eax, *ecx); 119| | *eax = cpu_info[0]; 120| | *ebx = cpu_info[1]; 121| | *ecx = cpu_info[2]; 122| | *edx = cpu_info[3]; 123| |#elif (defined(HAVE_GCC_GET_CPUID) && defined(USE_GCC_GET_CPUID)) || \ 124| | defined(__FILC__) 125| | uint32_t level = *eax; 126| | __get_cpuid(level, eax, ebx, ecx, edx); 127| |#else 128| 2| uint32_t a = *eax, b, c = *ecx, d; 129| 2| __asm__("cpuid\n\t" : "+a"(a), "=b"(b), "+c"(c), "=d"(d)); 130| 2| *eax = a; 131| 2| *ebx = b; 132| 2| *ecx = c; 133| 2| *edx = d; 134| 2|#endif 135| 2|} isadetection.c:xgetbv: 137| 1|static inline uint64_t xgetbv(void) { 138| |#if defined(_MSC_VER) 139| | return _xgetbv(0); 140| |#elif defined(__FILC__) 141| | return zxgetbv(); 142| |#else 143| 1| uint32_t xcr0_lo, xcr0_hi; 144| 1| __asm__("xgetbv\n\t" : "=a"(xcr0_lo), "=d"(xcr0_hi) : "c"(0)); 145| 1| return xcr0_lo | ((uint64_t)xcr0_hi << 32); 146| 1|#endif 147| 1|} roaring_malloc: 48| 1.44M|void* roaring_malloc(size_t n) { return global_memory_hook.malloc(n); } roaring_realloc: 50| 120k|void* roaring_realloc(void* p, size_t new_sz) { 51| 120k| return global_memory_hook.realloc(p, new_sz); 52| 120k|} roaring_free: 58| 1.70M|void roaring_free(void* p) { global_memory_hook.free(p); } roaring_aligned_malloc: 60| 302k|void* roaring_aligned_malloc(size_t alignment, size_t size) { 61| 302k| return global_memory_hook.aligned_malloc(alignment, size); 62| 302k|} roaring_aligned_free: 64| 302k|void roaring_aligned_free(void* p) { global_memory_hook.aligned_free(p); } memory.c:roaring_bitmap_aligned_malloc: 11| 302k|static void* roaring_bitmap_aligned_malloc(size_t alignment, size_t size) { 12| 302k| void* p; 13| |#ifdef _MSC_VER 14| | p = _aligned_malloc(size, alignment); 15| |#elif defined(__MINGW32__) || defined(__MINGW64__) 16| | p = __mingw_aligned_malloc(size, alignment); 17| |#else 18| | // somehow, if this is used before including "x86intrin.h", it creates an 19| | // implicit defined warning. 20| 302k| if (posix_memalign(&p, alignment, size) != 0) return NULL; ------------------ | Branch (20:9): [True: 0, False: 302k] ------------------ 21| 302k|#endif 22| 302k| return p; 23| 302k|} memory.c:roaring_bitmap_aligned_free: 25| 302k|static void roaring_bitmap_aligned_free(void* memblock) { 26| |#ifdef _MSC_VER 27| | _aligned_free(memblock); 28| |#elif defined(__MINGW32__) || defined(__MINGW64__) 29| | __mingw_aligned_free(memblock); 30| |#else 31| 302k| free(memblock); 32| 302k|#endif 33| 302k|} roaring_bitmap_create_with_capacity: 86| 25.7k|roaring_bitmap_t *roaring_bitmap_create_with_capacity(uint32_t cap) { 87| 25.7k| roaring_bitmap_t *ans = 88| 25.7k| (roaring_bitmap_t *)roaring_malloc(sizeof(roaring_bitmap_t)); 89| 25.7k| if (!ans) { ------------------ | Branch (89:9): [True: 0, False: 25.7k] ------------------ 90| 0| return NULL; 91| 0| } 92| 25.7k| bool is_ok = ra_init_with_capacity(&ans->high_low_container, cap); 93| 25.7k| if (!is_ok) { ------------------ | Branch (93:9): [True: 0, False: 25.7k] ------------------ 94| 0| roaring_free(ans); 95| 0| return NULL; 96| 0| } 97| 25.7k| return ans; 98| 25.7k|} roaring_bitmap_init_with_capacity: 100| 32.4k|bool roaring_bitmap_init_with_capacity(roaring_bitmap_t *r, uint32_t cap) { 101| 32.4k| return ra_init_with_capacity(&r->high_low_container, cap); 102| 32.4k|} roaring_bitmap_add_many: 135| 19.4k| const uint32_t *vals) { 136| 19.4k| uint32_t val; 137| 19.4k| const uint32_t *start = vals; 138| 19.4k| const uint32_t *end = vals + n_args; 139| 19.4k| const uint32_t *current_val = start; 140| | 141| 19.4k| if (n_args == 0) { ------------------ | Branch (141:9): [True: 0, False: 19.4k] ------------------ 142| 0| return; 143| 0| } 144| | 145| 19.4k| uint8_t typecode; 146| 19.4k| int idx; 147| 19.4k| container_t *container; 148| 19.4k| val = *current_val; 149| 19.4k| container = containerptr_roaring_bitmap_add(r, val, &typecode, &idx); 150| 19.4k| roaring_bulk_context_t context = {container, idx, (uint16_t)(val >> 16), 151| 19.4k| typecode}; 152| | 153| 8.11M| for (; current_val != end; current_val++) { ------------------ | Branch (153:12): [True: 8.10M, False: 19.4k] ------------------ 154| 8.10M| memcpy(&val, current_val, sizeof(val)); 155| 8.10M| add_bulk_impl(r, &context, val); 156| 8.10M| } 157| 19.4k|} roaring_bitmap_add_range_closed: 252| 2.88k| uint32_t max) { 253| 2.88k| if (min > max) { ------------------ | Branch (253:9): [True: 0, False: 2.88k] ------------------ 254| 0| return; 255| 0| } 256| | 257| 2.88k| roaring_array_t *ra = &r->high_low_container; 258| | 259| 2.88k| uint32_t min_key = min >> 16; 260| 2.88k| uint32_t max_key = max >> 16; 261| | 262| 2.88k| int32_t num_required_containers = max_key - min_key + 1; 263| 2.88k| int32_t suffix_length = 264| 2.88k| count_greater(ra->keys, ra->size, (uint16_t)max_key); 265| 2.88k| int32_t prefix_length = 266| 2.88k| count_less(ra->keys, ra->size - suffix_length, (uint16_t)min_key); 267| 2.88k| int32_t common_length = ra->size - prefix_length - suffix_length; 268| | 269| 2.88k| if (num_required_containers > common_length) { ------------------ | Branch (269:9): [True: 1.33k, False: 1.54k] ------------------ 270| 1.33k| ra_shift_tail(ra, suffix_length, 271| 1.33k| num_required_containers - common_length); 272| 1.33k| } 273| | 274| 2.88k| int32_t src = prefix_length + common_length - 1; 275| 2.88k| int32_t dst = ra->size - suffix_length - 1; 276| 79.4k| for (uint32_t key = max_key; key != min_key - 1; ------------------ | Branch (276:34): [True: 76.5k, False: 2.88k] ------------------ 277| 76.5k| key--) { // beware of min_key==0 278| 76.5k| uint32_t container_min = (min_key == key) ? (min & 0xffff) : 0; ------------------ | Branch (278:34): [True: 2.88k, False: 73.7k] ------------------ 279| 76.5k| uint32_t container_max = (max_key == key) ? (max & 0xffff) : 0xffff; ------------------ | Branch (279:34): [True: 2.88k, False: 73.7k] ------------------ 280| 76.5k| container_t *new_container; 281| 76.5k| uint8_t new_type; 282| | 283| 76.5k| if (src >= 0 && ra->keys[src] == key) { ------------------ | Branch (283:13): [True: 76.5k, False: 0] | Branch (283:25): [True: 3.34k, False: 73.2k] ------------------ 284| 3.34k| ra_unshare_container_at_index(ra, (uint16_t)src); 285| 3.34k| new_container = 286| 3.34k| container_add_range(ra->containers[src], ra->typecodes[src], 287| 3.34k| container_min, container_max, &new_type); 288| 3.34k| if (new_container != ra->containers[src]) { ------------------ | Branch (288:17): [True: 2.23k, False: 1.10k] ------------------ 289| 2.23k| container_free(ra->containers[src], ra->typecodes[src]); 290| 2.23k| } 291| 3.34k| src--; 292| 73.2k| } else { 293| 73.2k| new_container = container_from_range(&new_type, container_min, 294| 73.2k| container_max + 1, 1); 295| 73.2k| } 296| 76.5k| ra_replace_key_and_container_at_index(ra, dst, (uint16_t)key, 297| 76.5k| new_container, new_type); 298| 76.5k| dst--; 299| 76.5k| } 300| 2.88k|} roaring_bitmap_remove_range_closed: 303| 7.43k| uint32_t max) { 304| 7.43k| if (min > max) { ------------------ | Branch (304:9): [True: 866, False: 6.56k] ------------------ 305| 866| return; 306| 866| } 307| | 308| 6.56k| roaring_array_t *ra = &r->high_low_container; 309| | 310| 6.56k| uint32_t min_key = min >> 16; 311| 6.56k| uint32_t max_key = max >> 16; 312| | 313| 6.56k| int32_t src = count_less(ra->keys, ra->size, (uint16_t)min_key); 314| 6.56k| int32_t dst = src; 315| 26.6k| while (src < ra->size && ra->keys[src] <= max_key) { ------------------ | Branch (315:12): [True: 22.2k, False: 4.37k] | Branch (315:30): [True: 20.0k, False: 2.18k] ------------------ 316| 20.0k| uint32_t container_min = 317| 20.0k| (min_key == ra->keys[src]) ? (min & 0xffff) : 0; ------------------ | Branch (317:13): [True: 6.13k, False: 13.9k] ------------------ 318| 20.0k| uint32_t container_max = 319| 20.0k| (max_key == ra->keys[src]) ? (max & 0xffff) : 0xffff; ------------------ | Branch (319:13): [True: 5.80k, False: 14.2k] ------------------ 320| 20.0k| ra_unshare_container_at_index(ra, (uint16_t)src); 321| 20.0k| container_t *new_container; 322| 20.0k| uint8_t new_type; 323| 20.0k| new_container = 324| 20.0k| container_remove_range(ra->containers[src], ra->typecodes[src], 325| 20.0k| container_min, container_max, &new_type); 326| 20.0k| if (new_container != ra->containers[src]) { ------------------ | Branch (326:13): [True: 14.0k, False: 6.02k] ------------------ 327| 14.0k| container_free(ra->containers[src], ra->typecodes[src]); 328| 14.0k| } 329| 20.0k| if (new_container) { ------------------ | Branch (329:13): [True: 6.27k, False: 13.8k] ------------------ 330| 6.27k| ra_replace_key_and_container_at_index(ra, dst, ra->keys[src], 331| 6.27k| new_container, new_type); 332| 6.27k| dst++; 333| 6.27k| } 334| 20.0k| src++; 335| 20.0k| } 336| 6.56k| if (src > dst) { ------------------ | Branch (336:9): [True: 634, False: 5.93k] ------------------ 337| 634| ra_shift_tail(ra, ra->size - src, dst - src); 338| 634| } 339| 6.56k|} roaring_bitmap_copy: 525| 138|roaring_bitmap_t *roaring_bitmap_copy(const roaring_bitmap_t *r) { 526| 138| roaring_bitmap_t *ans = 527| 138| (roaring_bitmap_t *)roaring_malloc(sizeof(roaring_bitmap_t)); 528| 138| if (!ans) { ------------------ | Branch (528:9): [True: 0, False: 138] ------------------ 529| 0| return NULL; 530| 0| } 531| 138| if (!ra_init_with_capacity( // allocation of list of containers can fail ------------------ | Branch (531:9): [True: 0, False: 138] ------------------ 532| 138| &ans->high_low_container, r->high_low_container.size)) { 533| 0| roaring_free(ans); 534| 0| return NULL; 535| 0| } 536| 138| if (!ra_overwrite( // memory allocation of individual containers may fail ------------------ | Branch (536:9): [True: 0, False: 138] ------------------ 537| 138| &r->high_low_container, &ans->high_low_container, is_cow(r))) { 538| 0| roaring_bitmap_free(ans); // overwrite should leave in freeable state 539| 0| return NULL; 540| 0| } 541| 138| roaring_bitmap_set_copy_on_write(ans, is_cow(r)); 542| 138| return ans; 543| 138|} roaring_bitmap_overwrite: 546| 14.5k| const roaring_bitmap_t *src) { 547| 14.5k| roaring_bitmap_set_copy_on_write(dest, is_cow(src)); 548| 14.5k| return ra_overwrite(&src->high_low_container, &dest->high_low_container, 549| 14.5k| is_cow(src)); 550| 14.5k|} roaring_bitmap_clear: 562| 64.8k|void roaring_bitmap_clear(roaring_bitmap_t *r) { 563| 64.8k| ra_reset(&r->high_low_container); 564| 64.8k|} roaring_bitmap_add: 566| 6.48k|void roaring_bitmap_add(roaring_bitmap_t *r, uint32_t val) { 567| 6.48k| roaring_array_t *ra = &r->high_low_container; 568| | 569| 6.48k| const uint16_t hb = val >> 16; 570| 6.48k| const int i = ra_get_index(ra, hb); 571| 6.48k| uint8_t typecode; 572| 6.48k| if (i >= 0) { ------------------ | Branch (572:9): [True: 4.64k, False: 1.83k] ------------------ 573| 4.64k| ra_unshare_container_at_index(ra, (uint16_t)i); 574| 4.64k| container_t *container = 575| 4.64k| ra_get_container_at_index(ra, (uint16_t)i, &typecode); 576| 4.64k| uint8_t newtypecode = typecode; 577| 4.64k| container_t *container2 = 578| 4.64k| container_add(container, val & 0xFFFF, typecode, &newtypecode); 579| 4.64k| if (container2 != container) { ------------------ | Branch (579:13): [True: 0, False: 4.64k] ------------------ 580| 0| container_free(container, typecode); 581| 0| ra_set_container_at_index(&r->high_low_container, i, container2, 582| 0| newtypecode); 583| 0| } 584| 4.64k| } else { 585| 1.83k| array_container_t *newac = array_container_create(); 586| 1.83k| container_t *container = 587| 1.83k| container_add(newac, val & 0xFFFF, ARRAY_CONTAINER_TYPE, &typecode); ------------------ | | 49| 1.83k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 588| | // we could just assume that it stays an array container 589| 1.83k| ra_insert_new_key_value_at(&r->high_low_container, -i - 1, hb, 590| 1.83k| container, typecode); 591| 1.83k| } 592| 6.48k|} roaring_bitmap_add_checked: 594| 6.48k|bool roaring_bitmap_add_checked(roaring_bitmap_t *r, uint32_t val) { 595| 6.48k| const uint16_t hb = val >> 16; 596| 6.48k| const int i = ra_get_index(&r->high_low_container, hb); 597| 6.48k| uint8_t typecode; 598| 6.48k| bool result = false; 599| 6.48k| if (i >= 0) { ------------------ | Branch (599:9): [True: 4.87k, False: 1.60k] ------------------ 600| 4.87k| ra_unshare_container_at_index(&r->high_low_container, (uint16_t)i); 601| 4.87k| container_t *container = ra_get_container_at_index( 602| 4.87k| &r->high_low_container, (uint16_t)i, &typecode); 603| | 604| 4.87k| const int oldCardinality = 605| 4.87k| container_get_cardinality(container, typecode); 606| | 607| 4.87k| uint8_t newtypecode = typecode; 608| 4.87k| container_t *container2 = 609| 4.87k| container_add(container, val & 0xFFFF, typecode, &newtypecode); 610| 4.87k| if (container2 != container) { ------------------ | Branch (610:13): [True: 0, False: 4.87k] ------------------ 611| 0| container_free(container, typecode); 612| 0| ra_set_container_at_index(&r->high_low_container, i, container2, 613| 0| newtypecode); 614| 0| result = true; 615| 4.87k| } else { 616| 4.87k| const int newCardinality = 617| 4.87k| container_get_cardinality(container, newtypecode); 618| | 619| 4.87k| result = oldCardinality != newCardinality; 620| 4.87k| } 621| 4.87k| } else { 622| 1.60k| array_container_t *newac = array_container_create(); 623| 1.60k| container_t *container = 624| 1.60k| container_add(newac, val & 0xFFFF, ARRAY_CONTAINER_TYPE, &typecode); ------------------ | | 49| 1.60k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 625| | // we could just assume that it stays an array container 626| 1.60k| ra_insert_new_key_value_at(&r->high_low_container, -i - 1, hb, 627| 1.60k| container, typecode); 628| 1.60k| result = true; 629| 1.60k| } 630| | 631| 6.48k| return result; 632| 6.48k|} roaring_bitmap_remove: 634| 6.48k|void roaring_bitmap_remove(roaring_bitmap_t *r, uint32_t val) { 635| 6.48k| const uint16_t hb = val >> 16; 636| 6.48k| const int i = ra_get_index(&r->high_low_container, hb); 637| 6.48k| uint8_t typecode; 638| 6.48k| if (i >= 0) { ------------------ | Branch (638:9): [True: 5.83k, False: 644] ------------------ 639| 5.83k| ra_unshare_container_at_index(&r->high_low_container, (uint16_t)i); 640| 5.83k| container_t *container = ra_get_container_at_index( 641| 5.83k| &r->high_low_container, (uint16_t)i, &typecode); 642| 5.83k| uint8_t newtypecode = typecode; 643| 5.83k| container_t *container2 = 644| 5.83k| container_remove(container, val & 0xFFFF, typecode, &newtypecode); 645| 5.83k| if (container2 != container) { ------------------ | Branch (645:13): [True: 12, False: 5.82k] ------------------ 646| 12| container_free(container, typecode); 647| 12| ra_set_container_at_index(&r->high_low_container, i, container2, 648| 12| newtypecode); 649| 12| } 650| 5.83k| if (container_nonzero_cardinality(container2, newtypecode)) { ------------------ | Branch (650:13): [True: 5.77k, False: 61] ------------------ 651| 5.77k| ra_set_container_at_index(&r->high_low_container, i, container2, 652| 5.77k| newtypecode); 653| 5.77k| } else { 654| 61| ra_remove_at_index_and_free(&r->high_low_container, i); 655| 61| } 656| 5.83k| } 657| 6.48k|} roaring_bitmap_remove_checked: 659| 6.48k|bool roaring_bitmap_remove_checked(roaring_bitmap_t *r, uint32_t val) { 660| 6.48k| const uint16_t hb = val >> 16; 661| 6.48k| const int i = ra_get_index(&r->high_low_container, hb); 662| 6.48k| uint8_t typecode; 663| 6.48k| bool result = false; 664| 6.48k| if (i >= 0) { ------------------ | Branch (664:9): [True: 5.84k, False: 637] ------------------ 665| 5.84k| ra_unshare_container_at_index(&r->high_low_container, (uint16_t)i); 666| 5.84k| container_t *container = ra_get_container_at_index( 667| 5.84k| &r->high_low_container, (uint16_t)i, &typecode); 668| | 669| 5.84k| const int oldCardinality = 670| 5.84k| container_get_cardinality(container, typecode); 671| | 672| 5.84k| uint8_t newtypecode = typecode; 673| 5.84k| container_t *container2 = 674| 5.84k| container_remove(container, val & 0xFFFF, typecode, &newtypecode); 675| 5.84k| if (container2 != container) { ------------------ | Branch (675:13): [True: 2, False: 5.84k] ------------------ 676| 2| container_free(container, typecode); 677| 2| ra_set_container_at_index(&r->high_low_container, i, container2, 678| 2| newtypecode); 679| 2| } 680| | 681| 5.84k| const int newCardinality = 682| 5.84k| container_get_cardinality(container2, newtypecode); 683| | 684| 5.84k| if (newCardinality != 0) { ------------------ | Branch (684:13): [True: 5.83k, False: 8] ------------------ 685| 5.83k| ra_set_container_at_index(&r->high_low_container, i, container2, 686| 5.83k| newtypecode); 687| 5.83k| } else { 688| 8| ra_remove_at_index_and_free(&r->high_low_container, i); 689| 8| } 690| | 691| 5.84k| result = oldCardinality != newCardinality; 692| 5.84k| } 693| 6.48k| return result; 694| 6.48k|} roaring_bitmap_and: 732| 6.48k| const roaring_bitmap_t *x2) { 733| 6.48k| uint8_t result_type = 0; 734| 6.48k| const int length1 = x1->high_low_container.size, 735| 6.48k| length2 = x2->high_low_container.size; 736| 6.48k| uint32_t neededcap = length1 > length2 ? length2 : length1; ------------------ | Branch (736:26): [True: 46, False: 6.43k] ------------------ 737| 6.48k| roaring_bitmap_t *answer = roaring_bitmap_create_with_capacity(neededcap); 738| 6.48k| roaring_bitmap_set_copy_on_write(answer, is_cow(x1) || is_cow(x2)); ------------------ | Branch (738:46): [True: 0, False: 6.48k] | Branch (738:60): [True: 0, False: 6.48k] ------------------ 739| | 740| 6.48k| int pos1 = 0, pos2 = 0; 741| | 742| 12.9k| while (pos1 < length1 && pos2 < length2) { ------------------ | Branch (742:12): [True: 6.48k, False: 6.43k] | Branch (742:30): [True: 6.43k, False: 46] ------------------ 743| 6.43k| const uint16_t s1 = 744| 6.43k| ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 745| 6.43k| const uint16_t s2 = 746| 6.43k| ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 747| | 748| 6.43k| if (s1 == s2) { ------------------ | Branch (748:13): [True: 6.36k, False: 74] ------------------ 749| 6.36k| uint8_t type1, type2; 750| 6.36k| container_t *c1 = ra_get_container_at_index(&x1->high_low_container, 751| 6.36k| (uint16_t)pos1, &type1); 752| 6.36k| container_t *c2 = ra_get_container_at_index(&x2->high_low_container, 753| 6.36k| (uint16_t)pos2, &type2); 754| 6.36k| container_t *c = container_and(c1, type1, c2, type2, &result_type); 755| | 756| 6.36k| if (container_nonzero_cardinality(c, result_type)) { ------------------ | Branch (756:17): [True: 6.05k, False: 305] ------------------ 757| 6.05k| ra_append(&answer->high_low_container, s1, c, result_type); 758| 6.05k| } else { 759| 305| container_free(c, result_type); // otherwise: memory leak! 760| 305| } 761| 6.36k| ++pos1; 762| 6.36k| ++pos2; 763| 6.36k| } else if (s1 < s2) { // s1 < s2 ------------------ | Branch (763:20): [True: 74, False: 0] ------------------ 764| 74| pos1 = ra_advance_until(&x1->high_low_container, s2, pos1); 765| 74| } else { // s1 > s2 766| 0| pos2 = ra_advance_until(&x2->high_low_container, s1, pos2); 767| 0| } 768| 6.43k| } 769| 6.48k| return answer; 770| 6.48k|} roaring_bitmap_and_inplace: 813| 6.48k| const roaring_bitmap_t *x2) { 814| 6.48k| if (x1 == x2) return; ------------------ | Branch (814:9): [True: 0, False: 6.48k] ------------------ 815| 6.48k| int pos1 = 0, pos2 = 0, intersection_size = 0; 816| 6.48k| const int length1 = ra_get_size(&x1->high_low_container); 817| 6.48k| const int length2 = ra_get_size(&x2->high_low_container); 818| | 819| | // any skipped-over or newly emptied containers in x1 820| | // have to be freed. 821| 75.8k| while (pos1 < length1 && pos2 < length2) { ------------------ | Branch (821:12): [True: 69.4k, False: 6.43k] | Branch (821:30): [True: 69.3k, False: 46] ------------------ 822| 69.3k| const uint16_t s1 = 823| 69.3k| ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 824| 69.3k| const uint16_t s2 = 825| 69.3k| ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 826| | 827| 69.3k| if (s1 == s2) { ------------------ | Branch (827:13): [True: 69.2k, False: 74] ------------------ 828| 69.2k| uint8_t type1, type2, result_type; 829| 69.2k| container_t *c1 = ra_get_container_at_index(&x1->high_low_container, 830| 69.2k| (uint16_t)pos1, &type1); 831| 69.2k| container_t *c2 = ra_get_container_at_index(&x2->high_low_container, 832| 69.2k| (uint16_t)pos2, &type2); 833| | 834| | // We do the computation "in place" only when c1 is not a shared 835| | // container. Rationale: using a shared container safely with in 836| | // place computation would require making a copy and then doing the 837| | // computation in place which is likely less efficient than avoiding 838| | // in place entirely and always generating a new container. 839| 69.2k| container_t *c = 840| 69.2k| (type1 == SHARED_CONTAINER_TYPE) ------------------ | | 51| 69.2k|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (840:17): [True: 0, False: 69.2k] ------------------ 841| 69.2k| ? container_and(c1, type1, c2, type2, &result_type) 842| 69.2k| : container_iand(c1, type1, c2, type2, &result_type); 843| | 844| 69.2k| if (c != c1) { // in this instance a new container was created, and ------------------ | Branch (844:17): [True: 62.9k, False: 6.38k] ------------------ 845| | // we need to free the old one 846| 62.9k| container_free(c1, type1); 847| 62.9k| } 848| 69.2k| if (container_nonzero_cardinality(c, result_type)) { ------------------ | Branch (848:17): [True: 69.2k, False: 0] ------------------ 849| 69.2k| ra_replace_key_and_container_at_index(&x1->high_low_container, 850| 69.2k| intersection_size, s1, c, 851| 69.2k| result_type); 852| 69.2k| intersection_size++; 853| 69.2k| } else { 854| 0| container_free(c, result_type); 855| 0| } 856| 69.2k| ++pos1; 857| 69.2k| ++pos2; 858| 69.2k| } else if (s1 < s2) { ------------------ | Branch (858:20): [True: 74, False: 0] ------------------ 859| 74| pos1 = ra_advance_until_freeing(&x1->high_low_container, s2, pos1); 860| 74| } else { // s1 > s2 861| 0| pos2 = ra_advance_until(&x2->high_low_container, s1, pos2); 862| 0| } 863| 69.3k| } 864| | 865| | // if we ended early because x2 ran out, then all remaining in x1 should be 866| | // freed 867| 6.52k| while (pos1 < length1) { ------------------ | Branch (867:12): [True: 46, False: 6.48k] ------------------ 868| 46| container_free(x1->high_low_container.containers[pos1], 869| 46| x1->high_low_container.typecodes[pos1]); 870| 46| ++pos1; 871| 46| } 872| | 873| | // all containers after this have either been copied or freed 874| 6.48k| ra_downsize(&x1->high_low_container, intersection_size); 875| 6.48k|} roaring_bitmap_or: 878| 6.48k| const roaring_bitmap_t *x2) { 879| 6.48k| uint8_t result_type = 0; 880| 6.48k| const int length1 = x1->high_low_container.size, 881| 6.48k| length2 = x2->high_low_container.size; 882| 6.48k| if (0 == length1) { ------------------ | Branch (882:9): [True: 0, False: 6.48k] ------------------ 883| 0| return roaring_bitmap_copy(x2); 884| 0| } 885| 6.48k| if (0 == length2) { ------------------ | Branch (885:9): [True: 46, False: 6.43k] ------------------ 886| 46| return roaring_bitmap_copy(x1); 887| 46| } 888| 6.43k| roaring_bitmap_t *answer = 889| 6.43k| roaring_bitmap_create_with_capacity(length1 + length2); 890| 6.43k| roaring_bitmap_set_copy_on_write(answer, is_cow(x1) || is_cow(x2)); ------------------ | Branch (890:46): [True: 0, False: 6.43k] | Branch (890:60): [True: 0, False: 6.43k] ------------------ 891| 6.43k| int pos1 = 0, pos2 = 0; 892| 6.43k| uint8_t type1, type2; 893| 6.43k| uint16_t s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 894| 6.43k| uint16_t s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 895| 6.43k| while (true) { ------------------ | Branch (895:12): [True: 6.43k, Folded] ------------------ 896| 6.43k| if (s1 == s2) { ------------------ | Branch (896:13): [True: 6.36k, False: 74] ------------------ 897| 6.36k| container_t *c1 = ra_get_container_at_index(&x1->high_low_container, 898| 6.36k| (uint16_t)pos1, &type1); 899| 6.36k| container_t *c2 = ra_get_container_at_index(&x2->high_low_container, 900| 6.36k| (uint16_t)pos2, &type2); 901| 6.36k| container_t *c = container_or(c1, type1, c2, type2, &result_type); 902| | 903| | // since we assume that the initial containers are non-empty, the 904| | // result here 905| | // can only be non-empty 906| 6.36k| ra_append(&answer->high_low_container, s1, c, result_type); 907| 6.36k| ++pos1; 908| 6.36k| ++pos2; 909| 6.36k| if (pos1 == length1) break; ------------------ | Branch (909:17): [True: 6.36k, False: 0] ------------------ 910| 0| if (pos2 == length2) break; ------------------ | Branch (910:17): [True: 0, False: 0] ------------------ 911| 0| s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 912| 0| s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 913| | 914| 74| } else if (s1 < s2) { // s1 < s2 ------------------ | Branch (914:20): [True: 74, False: 0] ------------------ 915| 74| container_t *c1 = ra_get_container_at_index(&x1->high_low_container, 916| 74| (uint16_t)pos1, &type1); 917| | // c1 = container_clone(c1, type1); 918| 74| c1 = get_copy_of_container(c1, &type1, is_cow(x1)); 919| 74| if (is_cow(x1)) { ------------------ | Branch (919:17): [True: 0, False: 74] ------------------ 920| 0| ra_set_container_at_index(&x1->high_low_container, pos1, c1, 921| 0| type1); 922| 0| } 923| 74| ra_append(&answer->high_low_container, s1, c1, type1); 924| 74| pos1++; 925| 74| if (pos1 == length1) break; ------------------ | Branch (925:17): [True: 74, False: 0] ------------------ 926| 0| s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 927| | 928| 0| } else { // s1 > s2 929| 0| container_t *c2 = ra_get_container_at_index(&x2->high_low_container, 930| 0| (uint16_t)pos2, &type2); 931| | // c2 = container_clone(c2, type2); 932| 0| c2 = get_copy_of_container(c2, &type2, is_cow(x2)); 933| 0| if (is_cow(x2)) { ------------------ | Branch (933:17): [True: 0, False: 0] ------------------ 934| 0| ra_set_container_at_index(&x2->high_low_container, pos2, c2, 935| 0| type2); 936| 0| } 937| 0| ra_append(&answer->high_low_container, s2, c2, type2); 938| 0| pos2++; 939| 0| if (pos2 == length2) break; ------------------ | Branch (939:17): [True: 0, False: 0] ------------------ 940| 0| s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 941| 0| } 942| 6.43k| } 943| 6.43k| if (pos1 == length1) { ------------------ | Branch (943:9): [True: 6.43k, False: 0] ------------------ 944| 6.43k| ra_append_copy_range(&answer->high_low_container, 945| 6.43k| &x2->high_low_container, pos2, length2, 946| 6.43k| is_cow(x2)); 947| 6.43k| } else if (pos2 == length2) { ------------------ | Branch (947:16): [True: 0, False: 0] ------------------ 948| 0| ra_append_copy_range(&answer->high_low_container, 949| 0| &x1->high_low_container, pos1, length1, 950| 0| is_cow(x1)); 951| 0| } 952| 6.43k| return answer; 953| 6.48k|} roaring_bitmap_or_inplace: 1064| 6.48k| const roaring_bitmap_t *x2) { 1065| 6.48k| uint8_t result_type = 0; 1066| 6.48k| int length1 = x1->high_low_container.size; 1067| 6.48k| const int length2 = x2->high_low_container.size; 1068| | 1069| 6.48k| if (0 == length2) return; ------------------ | Branch (1069:9): [True: 0, False: 6.48k] ------------------ 1070| | 1071| 6.48k| if (0 == length1) { ------------------ | Branch (1071:9): [True: 0, False: 6.48k] ------------------ 1072| 0| roaring_bitmap_overwrite(x1, x2); 1073| 0| return; 1074| 0| } 1075| 6.48k| int pos1 = 0, pos2 = 0; 1076| 6.48k| uint8_t type1, type2; 1077| 6.48k| uint16_t s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 1078| 6.48k| uint16_t s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 1079| 6.48k| while (true) { ------------------ | Branch (1079:12): [True: 6.48k, Folded] ------------------ 1080| 6.48k| if (s1 == s2) { ------------------ | Branch (1080:13): [True: 6.48k, False: 0] ------------------ 1081| 6.48k| container_t *c1 = ra_get_container_at_index(&x1->high_low_container, 1082| 6.48k| (uint16_t)pos1, &type1); 1083| 6.48k| if (!container_is_full(c1, type1)) { ------------------ | Branch (1083:17): [True: 6.48k, False: 0] ------------------ 1084| 6.48k| container_t *c2 = ra_get_container_at_index( 1085| 6.48k| &x2->high_low_container, (uint16_t)pos2, &type2); 1086| 6.48k| container_t *c = 1087| 6.48k| (type1 == SHARED_CONTAINER_TYPE) ------------------ | | 51| 6.48k|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (1087:21): [True: 0, False: 6.48k] ------------------ 1088| 6.48k| ? container_or(c1, type1, c2, type2, &result_type) 1089| 6.48k| : container_ior(c1, type1, c2, type2, &result_type); 1090| | 1091| 6.48k| if (c != c1) { // in this instance a new container was created, ------------------ | Branch (1091:21): [True: 4.64k, False: 1.83k] ------------------ 1092| | // and we need to free the old one 1093| 4.64k| container_free(c1, type1); 1094| 4.64k| } 1095| 6.48k| ra_set_container_at_index(&x1->high_low_container, pos1, c, 1096| 6.48k| result_type); 1097| 6.48k| } 1098| 6.48k| ++pos1; 1099| 6.48k| ++pos2; 1100| 6.48k| if (pos1 == length1) break; ------------------ | Branch (1100:17): [True: 6.48k, False: 0] ------------------ 1101| 0| if (pos2 == length2) break; ------------------ | Branch (1101:17): [True: 0, False: 0] ------------------ 1102| 0| s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 1103| 0| s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 1104| | 1105| 0| } else if (s1 < s2) { // s1 < s2 ------------------ | Branch (1105:20): [True: 0, False: 0] ------------------ 1106| 0| pos1++; 1107| 0| if (pos1 == length1) break; ------------------ | Branch (1107:17): [True: 0, False: 0] ------------------ 1108| 0| s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 1109| | 1110| 0| } else { // s1 > s2 1111| 0| roaring_inplace_merge_bulk(x1, x2, pos1, pos1, pos2, false); 1112| 0| return; 1113| 0| } 1114| 6.48k| } 1115| 6.48k| if (pos1 == length1) { ------------------ | Branch (1115:9): [True: 6.48k, False: 0] ------------------ 1116| 6.48k| ra_append_copy_range(&x1->high_low_container, &x2->high_low_container, 1117| 6.48k| pos2, length2, is_cow(x2)); 1118| 6.48k| } 1119| 6.48k|} roaring_bitmap_xor: 1122| 6.48k| const roaring_bitmap_t *x2) { 1123| 6.48k| uint8_t result_type = 0; 1124| 6.48k| const int length1 = x1->high_low_container.size, 1125| 6.48k| length2 = x2->high_low_container.size; 1126| 6.48k| if (0 == length1) { ------------------ | Branch (1126:9): [True: 0, False: 6.48k] ------------------ 1127| 0| return roaring_bitmap_copy(x2); 1128| 0| } 1129| 6.48k| if (0 == length2) { ------------------ | Branch (1129:9): [True: 46, False: 6.43k] ------------------ 1130| 46| return roaring_bitmap_copy(x1); 1131| 46| } 1132| 6.43k| roaring_bitmap_t *answer = 1133| 6.43k| roaring_bitmap_create_with_capacity(length1 + length2); 1134| 6.43k| roaring_bitmap_set_copy_on_write(answer, is_cow(x1) || is_cow(x2)); ------------------ | Branch (1134:46): [True: 0, False: 6.43k] | Branch (1134:60): [True: 0, False: 6.43k] ------------------ 1135| 6.43k| int pos1 = 0, pos2 = 0; 1136| 6.43k| uint8_t type1, type2; 1137| 6.43k| uint16_t s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 1138| 6.43k| uint16_t s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 1139| 6.43k| while (true) { ------------------ | Branch (1139:12): [True: 6.43k, Folded] ------------------ 1140| 6.43k| if (s1 == s2) { ------------------ | Branch (1140:13): [True: 6.36k, False: 74] ------------------ 1141| 6.36k| container_t *c1 = ra_get_container_at_index(&x1->high_low_container, 1142| 6.36k| (uint16_t)pos1, &type1); 1143| 6.36k| container_t *c2 = ra_get_container_at_index(&x2->high_low_container, 1144| 6.36k| (uint16_t)pos2, &type2); 1145| 6.36k| container_t *c = container_xor(c1, type1, c2, type2, &result_type); 1146| | 1147| 6.36k| if (container_nonzero_cardinality(c, result_type)) { ------------------ | Branch (1147:17): [True: 6.34k, False: 17] ------------------ 1148| 6.34k| ra_append(&answer->high_low_container, s1, c, result_type); 1149| 6.34k| } else { 1150| 17| container_free(c, result_type); 1151| 17| } 1152| 6.36k| ++pos1; 1153| 6.36k| ++pos2; 1154| 6.36k| if (pos1 == length1) break; ------------------ | Branch (1154:17): [True: 6.36k, False: 0] ------------------ 1155| 0| if (pos2 == length2) break; ------------------ | Branch (1155:17): [True: 0, False: 0] ------------------ 1156| 0| s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 1157| 0| s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 1158| | 1159| 74| } else if (s1 < s2) { // s1 < s2 ------------------ | Branch (1159:20): [True: 74, False: 0] ------------------ 1160| 74| container_t *c1 = ra_get_container_at_index(&x1->high_low_container, 1161| 74| (uint16_t)pos1, &type1); 1162| 74| c1 = get_copy_of_container(c1, &type1, is_cow(x1)); 1163| 74| if (is_cow(x1)) { ------------------ | Branch (1163:17): [True: 0, False: 74] ------------------ 1164| 0| ra_set_container_at_index(&x1->high_low_container, pos1, c1, 1165| 0| type1); 1166| 0| } 1167| 74| ra_append(&answer->high_low_container, s1, c1, type1); 1168| 74| pos1++; 1169| 74| if (pos1 == length1) break; ------------------ | Branch (1169:17): [True: 74, False: 0] ------------------ 1170| 0| s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 1171| | 1172| 0| } else { // s1 > s2 1173| 0| container_t *c2 = ra_get_container_at_index(&x2->high_low_container, 1174| 0| (uint16_t)pos2, &type2); 1175| 0| c2 = get_copy_of_container(c2, &type2, is_cow(x2)); 1176| 0| if (is_cow(x2)) { ------------------ | Branch (1176:17): [True: 0, False: 0] ------------------ 1177| 0| ra_set_container_at_index(&x2->high_low_container, pos2, c2, 1178| 0| type2); 1179| 0| } 1180| 0| ra_append(&answer->high_low_container, s2, c2, type2); 1181| 0| pos2++; 1182| 0| if (pos2 == length2) break; ------------------ | Branch (1182:17): [True: 0, False: 0] ------------------ 1183| 0| s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 1184| 0| } 1185| 6.43k| } 1186| 6.43k| if (pos1 == length1) { ------------------ | Branch (1186:9): [True: 6.43k, False: 0] ------------------ 1187| 6.43k| ra_append_copy_range(&answer->high_low_container, 1188| 6.43k| &x2->high_low_container, pos2, length2, 1189| 6.43k| is_cow(x2)); 1190| 6.43k| } else if (pos2 == length2) { ------------------ | Branch (1190:16): [True: 0, False: 0] ------------------ 1191| 0| ra_append_copy_range(&answer->high_low_container, 1192| 0| &x1->high_low_container, pos1, length1, 1193| 0| is_cow(x1)); 1194| 0| } 1195| 6.43k| return answer; 1196| 6.48k|} roaring_bitmap_xor_inplace: 1201| 6.48k| const roaring_bitmap_t *x2) { 1202| 6.48k| assert(x1 != x2); 1203| 6.48k| uint8_t result_type = 0; 1204| 6.48k| int length1 = x1->high_low_container.size; 1205| 6.48k| const int length2 = x2->high_low_container.size; 1206| | 1207| 6.48k| if (0 == length2) return; ------------------ | Branch (1207:9): [True: 15, False: 6.46k] ------------------ 1208| | 1209| 6.46k| if (0 == length1) { ------------------ | Branch (1209:9): [True: 1.54k, False: 4.92k] ------------------ 1210| 1.54k| roaring_bitmap_overwrite(x1, x2); 1211| 1.54k| return; 1212| 1.54k| } 1213| | 1214| | // XOR can have new containers inserted from x2, but can also 1215| | // lose containers when x1 and x2 are nonempty and identical. 1216| | 1217| 4.92k| int pos1 = 0, pos2 = 0; 1218| 4.92k| uint8_t type1, type2; 1219| 4.92k| uint16_t s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 1220| 4.92k| uint16_t s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 1221| 7.09k| while (true) { ------------------ | Branch (1221:12): [True: 7.09k, Folded] ------------------ 1222| 7.09k| if (s1 == s2) { ------------------ | Branch (1222:13): [True: 6.98k, False: 111] ------------------ 1223| 6.98k| container_t *c1 = ra_get_container_at_index(&x1->high_low_container, 1224| 6.98k| (uint16_t)pos1, &type1); 1225| 6.98k| container_t *c2 = ra_get_container_at_index(&x2->high_low_container, 1226| 6.98k| (uint16_t)pos2, &type2); 1227| | 1228| | // We do the computation "in place" only when c1 is not a shared 1229| | // container. Rationale: using a shared container safely with in 1230| | // place computation would require making a copy and then doing the 1231| | // computation in place which is likely less efficient than avoiding 1232| | // in place entirely and always generating a new container. 1233| | 1234| 6.98k| container_t *c; 1235| 6.98k| if (type1 == SHARED_CONTAINER_TYPE) { ------------------ | | 51| 6.98k|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (1235:17): [True: 0, False: 6.98k] ------------------ 1236| 0| c = container_xor(c1, type1, c2, type2, &result_type); 1237| 0| shared_container_free(CAST_shared(c1)); // so release ------------------ | | 79| 0|#define CAST_shared(c) CAST(shared_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1238| 6.98k| } else { 1239| 6.98k| c = container_ixor(c1, type1, c2, type2, &result_type); 1240| 6.98k| } 1241| | 1242| 6.98k| if (container_nonzero_cardinality(c, result_type)) { ------------------ | Branch (1242:17): [True: 4.57k, False: 2.40k] ------------------ 1243| 4.57k| ra_set_container_at_index(&x1->high_low_container, pos1, c, 1244| 4.57k| result_type); 1245| 4.57k| ++pos1; 1246| 4.57k| } else { 1247| 2.40k| container_free(c, result_type); 1248| 2.40k| roaring_inplace_merge_bulk(x1, x2, pos1, pos1 + 1, pos2 + 1, 1249| 2.40k| true); 1250| 2.40k| return; 1251| 2.40k| } 1252| | 1253| 4.57k| ++pos2; 1254| 4.57k| if (pos1 == length1) break; ------------------ | Branch (1254:17): [True: 2.40k, False: 2.17k] ------------------ 1255| 2.17k| if (pos2 == length2) break; ------------------ | Branch (1255:17): [True: 0, False: 2.17k] ------------------ 1256| 2.17k| s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 1257| 2.17k| s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 1258| | 1259| 2.17k| } else if (s1 < s2) { // s1 < s2 ------------------ | Branch (1259:20): [True: 0, False: 111] ------------------ 1260| 0| pos1++; 1261| 0| if (pos1 == length1) break; ------------------ | Branch (1261:17): [True: 0, False: 0] ------------------ 1262| 0| s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 1263| | 1264| 111| } else { // s1 > s2 1265| 111| roaring_inplace_merge_bulk(x1, x2, pos1, pos1, pos2, true); 1266| 111| return; 1267| 111| } 1268| 7.09k| } 1269| 2.40k| if (pos1 == length1) { ------------------ | Branch (1269:9): [True: 2.40k, False: 0] ------------------ 1270| 2.40k| ra_append_copy_range(&x1->high_low_container, &x2->high_low_container, 1271| 2.40k| pos2, length2, is_cow(x2)); 1272| 2.40k| } 1273| 2.40k|} roaring_bitmap_andnot: 1276| 6.48k| const roaring_bitmap_t *x2) { 1277| 6.48k| uint8_t result_type = 0; 1278| 6.48k| const int length1 = x1->high_low_container.size, 1279| 6.48k| length2 = x2->high_low_container.size; 1280| 6.48k| if (0 == length1) { ------------------ | Branch (1280:9): [True: 0, False: 6.48k] ------------------ 1281| 0| roaring_bitmap_t *empty_bitmap = roaring_bitmap_create(); 1282| 0| roaring_bitmap_set_copy_on_write(empty_bitmap, 1283| 0| is_cow(x1) || is_cow(x2)); ------------------ | Branch (1283:42): [True: 0, False: 0] | Branch (1283:56): [True: 0, False: 0] ------------------ 1284| 0| return empty_bitmap; 1285| 0| } 1286| 6.48k| if (0 == length2) { ------------------ | Branch (1286:9): [True: 46, False: 6.43k] ------------------ 1287| 46| return roaring_bitmap_copy(x1); 1288| 46| } 1289| 6.43k| roaring_bitmap_t *answer = roaring_bitmap_create_with_capacity(length1); 1290| 6.43k| roaring_bitmap_set_copy_on_write(answer, is_cow(x1) || is_cow(x2)); ------------------ | Branch (1290:46): [True: 0, False: 6.43k] | Branch (1290:60): [True: 0, False: 6.43k] ------------------ 1291| | 1292| 6.43k| int pos1 = 0, pos2 = 0; 1293| 6.43k| uint8_t type1, type2; 1294| 6.43k| uint16_t s1 = 0; 1295| 6.43k| uint16_t s2 = 0; 1296| 6.43k| while (true) { ------------------ | Branch (1296:12): [True: 6.43k, Folded] ------------------ 1297| 6.43k| s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 1298| 6.43k| s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 1299| | 1300| 6.43k| if (s1 == s2) { ------------------ | Branch (1300:13): [True: 6.36k, False: 74] ------------------ 1301| 6.36k| container_t *c1 = ra_get_container_at_index(&x1->high_low_container, 1302| 6.36k| (uint16_t)pos1, &type1); 1303| 6.36k| container_t *c2 = ra_get_container_at_index(&x2->high_low_container, 1304| 6.36k| (uint16_t)pos2, &type2); 1305| 6.36k| container_t *c = 1306| 6.36k| container_andnot(c1, type1, c2, type2, &result_type); 1307| | 1308| 6.36k| if (container_nonzero_cardinality(c, result_type)) { ------------------ | Branch (1308:17): [True: 6.11k, False: 250] ------------------ 1309| 6.11k| ra_append(&answer->high_low_container, s1, c, result_type); 1310| 6.11k| } else { 1311| 250| container_free(c, result_type); 1312| 250| } 1313| 6.36k| ++pos1; 1314| 6.36k| ++pos2; 1315| 6.36k| if (pos1 == length1) break; ------------------ | Branch (1315:17): [True: 6.36k, False: 0] ------------------ 1316| 0| if (pos2 == length2) break; ------------------ | Branch (1316:17): [True: 0, False: 0] ------------------ 1317| 74| } else if (s1 < s2) { // s1 < s2 ------------------ | Branch (1317:20): [True: 74, False: 0] ------------------ 1318| 74| const int next_pos1 = 1319| 74| ra_advance_until(&x1->high_low_container, s2, pos1); 1320| 74| ra_append_copy_range(&answer->high_low_container, 1321| 74| &x1->high_low_container, pos1, next_pos1, 1322| 74| is_cow(x1)); 1323| | // TODO : perhaps some of the copy_on_write should be based on 1324| | // answer rather than x1 (more stringent?). Many similar cases 1325| 74| pos1 = next_pos1; 1326| 74| if (pos1 == length1) break; ------------------ | Branch (1326:17): [True: 74, False: 0] ------------------ 1327| 74| } else { // s1 > s2 1328| 0| pos2 = ra_advance_until(&x2->high_low_container, s1, pos2); 1329| 0| if (pos2 == length2) break; ------------------ | Branch (1329:17): [True: 0, False: 0] ------------------ 1330| 0| } 1331| 6.43k| } 1332| 6.43k| if (pos2 == length2) { ------------------ | Branch (1332:9): [True: 4.00k, False: 2.42k] ------------------ 1333| 4.00k| ra_append_copy_range(&answer->high_low_container, 1334| 4.00k| &x1->high_low_container, pos1, length1, 1335| 4.00k| is_cow(x1)); 1336| 4.00k| } 1337| 6.43k| return answer; 1338| 6.48k|} roaring_bitmap_andnot_inplace: 1343| 6.48k| const roaring_bitmap_t *x2) { 1344| 6.48k| assert(x1 != x2); 1345| | 1346| 6.48k| uint8_t result_type = 0; 1347| 6.48k| int length1 = x1->high_low_container.size; 1348| 6.48k| const int length2 = x2->high_low_container.size; 1349| 6.48k| int intersection_size = 0; 1350| | 1351| 6.48k| if (0 == length2) return; ------------------ | Branch (1351:9): [True: 425, False: 6.05k] ------------------ 1352| | 1353| 6.05k| if (0 == length1) { ------------------ | Branch (1353:9): [True: 0, False: 6.05k] ------------------ 1354| 0| roaring_bitmap_clear(x1); 1355| 0| return; 1356| 0| } 1357| | 1358| 6.05k| int pos1 = 0, pos2 = 0; 1359| 6.05k| uint8_t type1, type2; 1360| 6.05k| uint16_t s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 1361| 6.05k| uint16_t s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 1362| 6.05k| while (true) { ------------------ | Branch (1362:12): [True: 6.05k, Folded] ------------------ 1363| 6.05k| if (s1 == s2) { ------------------ | Branch (1363:13): [True: 6.05k, False: 0] ------------------ 1364| 6.05k| container_t *c1 = ra_get_container_at_index(&x1->high_low_container, 1365| 6.05k| (uint16_t)pos1, &type1); 1366| 6.05k| container_t *c2 = ra_get_container_at_index(&x2->high_low_container, 1367| 6.05k| (uint16_t)pos2, &type2); 1368| | 1369| | // We do the computation "in place" only when c1 is not a shared 1370| | // container. Rationale: using a shared container safely with in 1371| | // place computation would require making a copy and then doing the 1372| | // computation in place which is likely less efficient than avoiding 1373| | // in place entirely and always generating a new container. 1374| | 1375| 6.05k| container_t *c; 1376| 6.05k| if (type1 == SHARED_CONTAINER_TYPE) { ------------------ | | 51| 6.05k|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (1376:17): [True: 0, False: 6.05k] ------------------ 1377| 0| c = container_andnot(c1, type1, c2, type2, &result_type); 1378| 0| shared_container_free(CAST_shared(c1)); // release ------------------ | | 79| 0|#define CAST_shared(c) CAST(shared_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1379| 6.05k| } else { 1380| 6.05k| c = container_iandnot(c1, type1, c2, type2, &result_type); 1381| 6.05k| } 1382| | 1383| 6.05k| if (container_nonzero_cardinality(c, result_type)) { ------------------ | Branch (1383:17): [True: 4.50k, False: 1.54k] ------------------ 1384| 4.50k| ra_replace_key_and_container_at_index(&x1->high_low_container, 1385| 4.50k| intersection_size++, s1, 1386| 4.50k| c, result_type); 1387| 4.50k| } else { 1388| 1.54k| container_free(c, result_type); 1389| 1.54k| } 1390| | 1391| 6.05k| ++pos1; 1392| 6.05k| ++pos2; 1393| 6.05k| if (pos1 == length1) break; ------------------ | Branch (1393:17): [True: 3.79k, False: 2.25k] ------------------ 1394| 2.25k| if (pos2 == length2) break; ------------------ | Branch (1394:17): [True: 2.25k, False: 0] ------------------ 1395| 0| s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 1396| 0| s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 1397| | 1398| 0| } else if (s1 < s2) { // s1 < s2 ------------------ | Branch (1398:20): [True: 0, False: 0] ------------------ 1399| 0| if (pos1 != intersection_size) { ------------------ | Branch (1399:17): [True: 0, False: 0] ------------------ 1400| 0| container_t *c1 = ra_get_container_at_index( 1401| 0| &x1->high_low_container, (uint16_t)pos1, &type1); 1402| | 1403| 0| ra_replace_key_and_container_at_index( 1404| 0| &x1->high_low_container, intersection_size, s1, c1, type1); 1405| 0| } 1406| 0| intersection_size++; 1407| 0| pos1++; 1408| 0| if (pos1 == length1) break; ------------------ | Branch (1408:17): [True: 0, False: 0] ------------------ 1409| 0| s1 = ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 1410| | 1411| 0| } else { // s1 > s2 1412| 0| pos2 = ra_advance_until(&x2->high_low_container, s1, pos2); 1413| 0| if (pos2 == length2) break; ------------------ | Branch (1413:17): [True: 0, False: 0] ------------------ 1414| 0| s2 = ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 1415| 0| } 1416| 6.05k| } 1417| | 1418| 6.05k| if (pos1 < length1) { ------------------ | Branch (1418:9): [True: 2.25k, False: 3.79k] ------------------ 1419| | // all containers between intersection_size and 1420| | // pos1 are junk. However, they have either been moved 1421| | // (thus still referenced) or involved in an iandnot 1422| | // that will clean up all containers that could not be reused. 1423| | // Thus we should not free the junk containers between 1424| | // intersection_size and pos1. 1425| 2.25k| if (pos1 > intersection_size) { ------------------ | Branch (1425:13): [True: 39, False: 2.21k] ------------------ 1426| | // left slide of remaining items 1427| 39| ra_copy_range(&x1->high_low_container, pos1, length1, 1428| 39| intersection_size); 1429| 39| } 1430| | // else current placement is fine 1431| 2.25k| intersection_size += (length1 - pos1); 1432| 2.25k| } 1433| 6.05k| ra_downsize(&x1->high_low_container, intersection_size); 1434| 6.05k|} roaring_bitmap_get_cardinality: 1436| 64.8k|uint64_t roaring_bitmap_get_cardinality(const roaring_bitmap_t *r) { 1437| 64.8k| const roaring_array_t *ra = &r->high_low_container; 1438| | 1439| 64.8k| uint64_t card = 0; 1440| 443k| for (int i = 0; i < ra->size; ++i) ------------------ | Branch (1440:21): [True: 378k, False: 64.8k] ------------------ 1441| 378k| card += container_get_cardinality(ra->containers[i], ra->typecodes[i]); 1442| 64.8k| return card; 1443| 64.8k|} roaring_bitmap_is_empty: 1506| 12.9k|bool roaring_bitmap_is_empty(const roaring_bitmap_t *r) { 1507| 12.9k| return r->high_low_container.size == 0; 1508| 12.9k|} roaring_bitmap_to_uint32_array: 1510| 6.48k|void roaring_bitmap_to_uint32_array(const roaring_bitmap_t *r, uint32_t *ans) { 1511| 6.48k| ra_to_uint32_array(&r->high_low_container, ans); 1512| 6.48k|} roaring_bitmap_run_optimize: 1530| 12.9k|bool roaring_bitmap_run_optimize(roaring_bitmap_t *r) { 1531| 12.9k| bool answer = false; 1532| 25.9k| for (int i = 0; i < r->high_low_container.size; i++) { ------------------ | Branch (1532:21): [True: 12.9k, False: 12.9k] ------------------ 1533| 12.9k| uint8_t type_original, type_after; 1534| 12.9k| ra_unshare_container_at_index( 1535| 12.9k| &r->high_low_container, 1536| 12.9k| (uint16_t)i); // TODO: this introduces extra cloning! 1537| 12.9k| container_t *c = ra_get_container_at_index(&r->high_low_container, 1538| 12.9k| (uint16_t)i, &type_original); 1539| 12.9k| container_t *c1 = convert_run_optimize(c, type_original, &type_after); 1540| 12.9k| if (type_after == RUN_CONTAINER_TYPE) { ------------------ | | 50| 12.9k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (1540:13): [True: 4.10k, False: 8.85k] ------------------ 1541| | answer = true; 1542| 4.10k| } 1543| 12.9k| ra_set_container_at_index(&r->high_low_container, i, c1, type_after); 1544| 12.9k| } 1545| 12.9k| return answer; 1546| 12.9k|} roaring_bitmap_shrink_to_fit: 1548| 6.48k|size_t roaring_bitmap_shrink_to_fit(roaring_bitmap_t *r) { 1549| 6.48k| size_t answer = 0; 1550| 12.9k| for (int i = 0; i < r->high_low_container.size; i++) { ------------------ | Branch (1550:21): [True: 6.48k, False: 6.48k] ------------------ 1551| 6.48k| uint8_t type_original; 1552| 6.48k| container_t *c = ra_get_container_at_index(&r->high_low_container, 1553| 6.48k| (uint16_t)i, &type_original); 1554| 6.48k| answer += container_shrink_to_fit(c, type_original); 1555| 6.48k| } 1556| 6.48k| answer += ra_shrink_to_fit(&r->high_low_container); 1557| 6.48k| return answer; 1558| 6.48k|} roaring_bitmap_remove_run_compression: 1564| 6.48k|bool roaring_bitmap_remove_run_compression(roaring_bitmap_t *r) { 1565| 6.48k| bool answer = false; 1566| 87.6k| for (int i = 0; i < r->high_low_container.size; i++) { ------------------ | Branch (1566:21): [True: 81.1k, False: 6.48k] ------------------ 1567| 81.1k| uint8_t type_original, type_after; 1568| 81.1k| container_t *c = ra_get_container_at_index(&r->high_low_container, 1569| 81.1k| (uint16_t)i, &type_original); 1570| 81.1k| if (get_container_type(c, type_original) == RUN_CONTAINER_TYPE) { ------------------ | | 50| 81.1k|#define RUN_CONTAINER_TYPE 3 ------------------ | Branch (1570:13): [True: 73.4k, False: 7.73k] ------------------ 1571| 73.4k| answer = true; 1572| 73.4k| if (type_original == SHARED_CONTAINER_TYPE) { ------------------ | | 51| 73.4k|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (1572:17): [True: 0, False: 73.4k] ------------------ 1573| 0| run_container_t *truec = CAST_run(CAST_shared(c)->container); ------------------ | | 77| 0|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1574| 0| int32_t card = run_container_cardinality(truec); 1575| 0| container_t *c1 = convert_to_bitset_or_array_container( 1576| 0| truec, card, &type_after); 1577| 0| shared_container_free(CAST_shared(c)); // frees run as needed ------------------ | | 79| 0|#define CAST_shared(c) CAST(shared_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1578| 0| ra_set_container_at_index(&r->high_low_container, i, c1, 1579| 0| type_after); 1580| | 1581| 73.4k| } else { 1582| 73.4k| int32_t card = run_container_cardinality(CAST_run(c)); ------------------ | | 77| 73.4k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 73.4k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1583| 73.4k| container_t *c1 = convert_to_bitset_or_array_container( 1584| 73.4k| CAST_run(c), card, &type_after); ------------------ | | 77| 73.4k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 73.4k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1585| 73.4k| run_container_free(CAST_run(c)); ------------------ | | 77| 73.4k|#define CAST_run(c) CAST(run_container_t *, c) // safer downcast | | ------------------ | | | | 86| 73.4k|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1586| 73.4k| ra_set_container_at_index(&r->high_low_container, i, c1, 1587| 73.4k| type_after); 1588| 73.4k| } 1589| 73.4k| } 1590| 81.1k| } 1591| 6.48k| return answer; 1592| 6.48k|} roaring_bitmap_portable_size_in_bytes: 1627| 12.9k|size_t roaring_bitmap_portable_size_in_bytes(const roaring_bitmap_t *r) { 1628| 12.9k| return ra_portable_size_in_bytes(&r->high_low_container); 1629| 12.9k|} roaring_bitmap_portable_deserialize_safe: 1632| 12.9k| size_t maxbytes) { 1633| 12.9k| roaring_bitmap_t *ans = 1634| 12.9k| (roaring_bitmap_t *)roaring_malloc(sizeof(roaring_bitmap_t)); 1635| 12.9k| if (ans == NULL) { ------------------ | Branch (1635:9): [True: 0, False: 12.9k] ------------------ 1636| 0| return NULL; 1637| 0| } 1638| 12.9k| size_t bytesread; 1639| 12.9k| bool is_ok = ra_portable_deserialize(&ans->high_low_container, buf, 1640| 12.9k| maxbytes, &bytesread); 1641| 12.9k| if (!is_ok) { ------------------ | Branch (1641:9): [True: 6.46k, False: 6.49k] ------------------ 1642| 6.46k| roaring_free(ans); 1643| 6.46k| return NULL; 1644| 6.46k| } 1645| 6.49k| roaring_bitmap_set_copy_on_write(ans, false); 1646| 6.49k| if (!is_ok) { ------------------ | Branch (1646:9): [True: 0, False: 6.49k] ------------------ 1647| 0| roaring_free(ans); 1648| 0| return NULL; 1649| 0| } 1650| 6.49k| return ans; 1651| 6.49k|} roaring_bitmap_portable_serialize: 1662| 6.48k|size_t roaring_bitmap_portable_serialize(const roaring_bitmap_t *r, char *buf) { 1663| 6.48k| return ra_portable_serialize(&r->high_low_container, buf); 1664| 6.48k|} roaring_iterate: 1745| 6.46k| void *ptr) { 1746| 6.46k| const roaring_array_t *ra = &r->high_low_container; 1747| | 1748| 75.8k| for (int i = 0; i < ra->size; ++i) ------------------ | Branch (1748:21): [True: 69.3k, False: 6.46k] ------------------ 1749| 69.3k| if (!container_iterate(ra->containers[i], ra->typecodes[i], ------------------ | Branch (1749:13): [True: 0, False: 69.3k] ------------------ 1750| 69.3k| ((uint32_t)ra->keys[i]) << 16, iterator, ptr)) { 1751| 0| return false; 1752| 0| } 1753| 6.46k| return true; 1754| 6.46k|} roaring_iterator_init: 1861| 12.9k| roaring_uint32_iterator_t *newit) { 1862| 12.9k| newit->parent = r; 1863| 12.9k| newit->container_index = 0; 1864| 12.9k| newit->has_value = loadfirstvalue(newit); 1865| 12.9k|} roaring_uint32_iterator_move_equalorlarger: 1893| 6.48k| uint32_t val) { 1894| 6.48k| uint16_t hb = val >> 16; 1895| 6.48k| const int i = ra_get_index(&it->parent->high_low_container, hb); 1896| 6.48k| if (i >= 0) { ------------------ | Branch (1896:9): [True: 6.40k, False: 73] ------------------ 1897| 6.40k| uint32_t lowvalue = 1898| 6.40k| container_maximum(it->parent->high_low_container.containers[i], 1899| 6.40k| it->parent->high_low_container.typecodes[i]); 1900| 6.40k| uint16_t lb = val & 0xFFFF; 1901| 6.40k| if (lowvalue < lb) { ------------------ | Branch (1901:13): [True: 23, False: 6.38k] ------------------ 1902| | // will have to load first value of next container 1903| 23| it->container_index = i + 1; 1904| 6.38k| } else { 1905| | // the value is necessarily within the range of the container 1906| 6.38k| it->container_index = i; 1907| 6.38k| it->has_value = loadfirstvalue_largeorequal(it, val); 1908| 6.38k| return it->has_value; 1909| 6.38k| } 1910| 6.40k| } else { 1911| | // there is no matching, so we are going for the next container 1912| 73| it->container_index = -i - 1; 1913| 73| } 1914| 96| it->has_value = loadfirstvalue(it); 1915| 96| return it->has_value; 1916| 6.48k|} roaring_uint32_iterator_advance: 1918| 335k|bool roaring_uint32_iterator_advance(roaring_uint32_iterator_t *it) { 1919| 335k| if (it->container_index >= it->parent->high_low_container.size) { ------------------ | Branch (1919:9): [True: 0, False: 335k] ------------------ 1920| 0| return (it->has_value = false); 1921| 0| } 1922| 335k| if (it->container_index < 0) { ------------------ | Branch (1922:9): [True: 0, False: 335k] ------------------ 1923| 0| it->container_index = 0; 1924| 0| return (it->has_value = loadfirstvalue(it)); 1925| 0| } 1926| 335k| uint16_t low16 = (uint16_t)it->current_value; 1927| 335k| if (container_iterator_next(it->container, it->typecode, &it->container_it, ------------------ | Branch (1927:9): [True: 329k, False: 6.23k] ------------------ 1928| 335k| &low16)) { 1929| 329k| it->current_value = it->highbits | low16; 1930| 329k| return (it->has_value = true); 1931| 329k| } 1932| 6.23k| it->container_index++; 1933| 6.23k| return (it->has_value = loadfirstvalue(it)); 1934| 335k|} roaring_bitmap_equals: 2129| 12.9k| const roaring_bitmap_t *r2) { 2130| 12.9k| const roaring_array_t *ra1 = &r1->high_low_container; 2131| 12.9k| const roaring_array_t *ra2 = &r2->high_low_container; 2132| | 2133| 12.9k| if (ra1->size != ra2->size) { ------------------ | Branch (2133:9): [True: 2.55k, False: 10.4k] ------------------ 2134| 2.55k| return false; 2135| 2.55k| } 2136| 95.5k| for (int i = 0; i < ra1->size; ++i) { ------------------ | Branch (2136:21): [True: 85.1k, False: 10.3k] ------------------ 2137| 85.1k| if (ra1->keys[i] != ra2->keys[i]) { ------------------ | Branch (2137:13): [True: 23, False: 85.0k] ------------------ 2138| 23| return false; 2139| 23| } 2140| 85.1k| } 2141| 91.5k| for (int i = 0; i < ra1->size; ++i) { ------------------ | Branch (2141:21): [True: 85.0k, False: 6.48k] ------------------ 2142| 85.0k| bool areequal = container_equals(ra1->containers[i], ra1->typecodes[i], 2143| 85.0k| ra2->containers[i], ra2->typecodes[i]); 2144| 85.0k| if (!areequal) { ------------------ | Branch (2144:13): [True: 3.90k, False: 81.1k] ------------------ 2145| 3.90k| return false; 2146| 3.90k| } 2147| 85.0k| } 2148| 6.48k| return true; 2149| 10.3k|} roaring_bitmap_is_subset: 2152| 11.7k| const roaring_bitmap_t *r2) { 2153| 11.7k| const roaring_array_t *ra1 = &r1->high_low_container; 2154| 11.7k| const roaring_array_t *ra2 = &r2->high_low_container; 2155| | 2156| 11.7k| const int length1 = ra1->size, length2 = ra2->size; 2157| | 2158| 11.7k| int pos1 = 0, pos2 = 0; 2159| | 2160| 11.7k| while (pos1 < length1 && pos2 < length2) { ------------------ | Branch (2160:12): [True: 11.2k, False: 500] | Branch (2160:30): [True: 11.2k, False: 46] ------------------ 2161| 11.2k| const uint16_t s1 = ra_get_key_at_index(ra1, (uint16_t)pos1); 2162| 11.2k| const uint16_t s2 = ra_get_key_at_index(ra2, (uint16_t)pos2); 2163| | 2164| 11.2k| if (s1 == s2) { ------------------ | Branch (2164:13): [True: 11.1k, False: 114] ------------------ 2165| 11.1k| uint8_t type1, type2; 2166| 11.1k| container_t *c1 = 2167| 11.1k| ra_get_container_at_index(ra1, (uint16_t)pos1, &type1); 2168| 11.1k| container_t *c2 = 2169| 11.1k| ra_get_container_at_index(ra2, (uint16_t)pos2, &type2); 2170| 11.1k| if (!container_is_subset(c1, type1, c2, type2)) return false; ------------------ | Branch (2170:17): [True: 11.1k, False: 0] ------------------ 2171| 0| ++pos1; 2172| 0| ++pos2; 2173| 114| } else if (s1 < s2) { // s1 < s2 ------------------ | Branch (2173:20): [True: 114, False: 0] ------------------ 2174| 114| return false; 2175| 114| } else { // s1 > s2 2176| 0| pos2 = ra_advance_until(ra2, s1, pos2); 2177| 0| } 2178| 11.2k| } 2179| 546| if (pos1 == length1) ------------------ | Branch (2179:9): [True: 500, False: 46] ------------------ 2180| 500| return true; 2181| 46| else 2182| 46| return false; 2183| 546|} roaring_bitmap_flip_inplace: 2352| 6.48k| uint64_t range_end) { 2353| 6.48k| if (range_start >= range_end || range_start > (uint64_t)UINT32_MAX + 1) { ------------------ | Branch (2353:9): [True: 5.95k, False: 525] | Branch (2353:37): [True: 0, False: 525] ------------------ 2354| 5.95k| return; 2355| 5.95k| } 2356| 525| roaring_bitmap_flip_inplace_closed(x1, (uint32_t)range_start, 2357| 525| (uint32_t)(range_end - 1)); 2358| 525|} roaring_bitmap_flip_inplace_closed: 2362| 7.00k| uint32_t range_end) { 2363| 7.00k| if (range_start > range_end) { ------------------ | Branch (2363:9): [True: 176, False: 6.82k] ------------------ 2364| 176| return; // empty range 2365| 176| } 2366| | 2367| 6.82k| uint16_t hb_start = (uint16_t)(range_start >> 16); 2368| 6.82k| const uint16_t lb_start = (uint16_t)range_start; 2369| 6.82k| uint16_t hb_end = (uint16_t)(range_end >> 16); 2370| 6.82k| const uint16_t lb_end = (uint16_t)range_end; 2371| | 2372| 6.82k| if (hb_start == hb_end) { ------------------ | Branch (2372:9): [True: 6.36k, False: 467] ------------------ 2373| 6.36k| inplace_flip_container(&x1->high_low_container, hb_start, lb_start, 2374| 6.36k| lb_end); 2375| 6.36k| } else { 2376| | // start and end containers are distinct 2377| 467| if (lb_start > 0) { ------------------ | Branch (2377:13): [True: 356, False: 111] ------------------ 2378| | // handle first (partial) container 2379| 356| inplace_flip_container(&x1->high_low_container, hb_start, lb_start, 2380| 356| 0xFFFF); 2381| 356| ++hb_start; // for the full containers. Can't wrap. 2382| 356| } 2383| | 2384| 467| if (lb_end != 0xFFFF) --hb_end; ------------------ | Branch (2384:13): [True: 442, False: 25] ------------------ 2385| | 2386| 25.0k| for (uint32_t hb = hb_start; hb <= hb_end; ++hb) { ------------------ | Branch (2386:38): [True: 24.5k, False: 467] ------------------ 2387| 24.5k| inplace_fully_flip_container(&x1->high_low_container, (uint16_t)hb); 2388| 24.5k| } 2389| | // handle a partial final container 2390| 467| if (lb_end != 0xFFFF) { ------------------ | Branch (2390:13): [True: 442, False: 25] ------------------ 2391| 442| inplace_flip_container(&x1->high_low_container, hb_end + 1, 0, 2392| 442| lb_end); 2393| 442| ++hb_end; 2394| 442| } 2395| 467| } 2396| 6.82k|} roaring_bitmap_rank: 2862| 6.48k|uint64_t roaring_bitmap_rank(const roaring_bitmap_t *bm, uint32_t x) { 2863| 6.48k| uint64_t size = 0; 2864| 6.48k| uint32_t xhigh = x >> 16; 2865| 6.48k| for (int i = 0; i < bm->high_low_container.size; i++) { ------------------ | Branch (2865:21): [True: 6.48k, False: 0] ------------------ 2866| 6.48k| uint32_t key = bm->high_low_container.keys[i]; 2867| 6.48k| if (xhigh > key) { ------------------ | Branch (2867:13): [True: 0, False: 6.48k] ------------------ 2868| 0| size += 2869| 0| container_get_cardinality(bm->high_low_container.containers[i], 2870| 0| bm->high_low_container.typecodes[i]); 2871| 6.48k| } else if (xhigh == key) { ------------------ | Branch (2871:20): [True: 6.48k, False: 0] ------------------ 2872| 6.48k| return size + container_rank(bm->high_low_container.containers[i], 2873| 6.48k| bm->high_low_container.typecodes[i], 2874| 6.48k| x & 0xFFFF); 2875| 6.48k| } else { 2876| 0| return size; 2877| 0| } 2878| 6.48k| } 2879| 0| return size; 2880| 6.48k|} roaring_bitmap_minimum: 2945| 6.48k|uint32_t roaring_bitmap_minimum(const roaring_bitmap_t *bm) { 2946| 6.48k| if (bm->high_low_container.size > 0) { ------------------ | Branch (2946:9): [True: 6.43k, False: 46] ------------------ 2947| 6.43k| container_t *c = bm->high_low_container.containers[0]; 2948| 6.43k| uint8_t type = bm->high_low_container.typecodes[0]; 2949| 6.43k| uint32_t key = bm->high_low_container.keys[0]; 2950| 6.43k| uint32_t lowvalue = container_minimum(c, type); 2951| 6.43k| return lowvalue | (key << 16); 2952| 6.43k| } 2953| 46| return UINT32_MAX; 2954| 6.48k|} roaring_bitmap_maximum: 2960| 6.48k|uint32_t roaring_bitmap_maximum(const roaring_bitmap_t *bm) { 2961| 6.48k| if (bm->high_low_container.size > 0) { ------------------ | Branch (2961:9): [True: 6.43k, False: 46] ------------------ 2962| 6.43k| container_t *container = 2963| 6.43k| bm->high_low_container.containers[bm->high_low_container.size - 1]; 2964| 6.43k| uint8_t typecode = 2965| 6.43k| bm->high_low_container.typecodes[bm->high_low_container.size - 1]; 2966| 6.43k| uint32_t key = 2967| 6.43k| bm->high_low_container.keys[bm->high_low_container.size - 1]; 2968| 6.43k| uint32_t lowvalue = container_maximum(container, typecode); 2969| 6.43k| return lowvalue | (key << 16); 2970| 6.43k| } 2971| 46| return 0; 2972| 6.48k|} roaring_bitmap_select: 2975| 6.48k| uint32_t *element) { 2976| 6.48k| container_t *container; 2977| 6.48k| uint8_t typecode; 2978| 6.48k| uint16_t key; 2979| 6.48k| uint32_t start_rank = 0; 2980| 6.48k| int i = 0; 2981| 6.48k| bool valid = false; 2982| 12.9k| while (!valid && i < bm->high_low_container.size) { ------------------ | Branch (2982:12): [True: 6.90k, False: 6.05k] | Branch (2982:22): [True: 6.48k, False: 425] ------------------ 2983| 6.48k| container = bm->high_low_container.containers[i]; 2984| 6.48k| typecode = bm->high_low_container.typecodes[i]; 2985| 6.48k| valid = 2986| 6.48k| container_select(container, typecode, &start_rank, rank, element); 2987| 6.48k| i++; 2988| 6.48k| } 2989| | 2990| 6.48k| if (valid) { ------------------ | Branch (2990:9): [True: 6.05k, False: 425] ------------------ 2991| 6.05k| key = bm->high_low_container.keys[i - 1]; 2992| 6.05k| *element |= (((uint32_t)key) << 16); // w/o cast, key promotes signed 2993| 6.05k| return true; 2994| 6.05k| } else 2995| 425| return false; 2996| 6.48k|} roaring_bitmap_intersect: 2999| 6.48k| const roaring_bitmap_t *x2) { 3000| 6.48k| const int length1 = x1->high_low_container.size, 3001| 6.48k| length2 = x2->high_low_container.size; 3002| 6.48k| uint64_t answer = 0; 3003| 6.48k| int pos1 = 0, pos2 = 0; 3004| | 3005| 6.85k| while (pos1 < length1 && pos2 < length2) { ------------------ | Branch (3005:12): [True: 6.48k, False: 379] | Branch (3005:30): [True: 6.43k, False: 46] ------------------ 3006| 6.43k| const uint16_t s1 = 3007| 6.43k| ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 3008| 6.43k| const uint16_t s2 = 3009| 6.43k| ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 3010| | 3011| 6.43k| if (s1 == s2) { ------------------ | Branch (3011:13): [True: 6.36k, False: 74] ------------------ 3012| 6.36k| uint8_t type1, type2; 3013| 6.36k| container_t *c1 = ra_get_container_at_index(&x1->high_low_container, 3014| 6.36k| (uint16_t)pos1, &type1); 3015| 6.36k| container_t *c2 = ra_get_container_at_index(&x2->high_low_container, 3016| 6.36k| (uint16_t)pos2, &type2); 3017| 6.36k| if (container_intersect(c1, type1, c2, type2)) return true; ------------------ | Branch (3017:17): [True: 6.05k, False: 305] ------------------ 3018| 305| ++pos1; 3019| 305| ++pos2; 3020| 305| } else if (s1 < s2) { // s1 < s2 ------------------ | Branch (3020:20): [True: 74, False: 0] ------------------ 3021| 74| pos1 = ra_advance_until(&x1->high_low_container, s2, pos1); 3022| 74| } else { // s1 > s2 3023| 0| pos2 = ra_advance_until(&x2->high_low_container, s1, pos2); 3024| 0| } 3025| 6.43k| } 3026| 425| return answer != 0; 3027| 6.48k|} roaring_bitmap_and_cardinality: 3049| 25.9k| const roaring_bitmap_t *x2) { 3050| 25.9k| const int length1 = x1->high_low_container.size, 3051| 25.9k| length2 = x2->high_low_container.size; 3052| 25.9k| uint64_t answer = 0; 3053| 25.9k| int pos1 = 0, pos2 = 0; 3054| 51.6k| while (pos1 < length1 && pos2 < length2) { ------------------ | Branch (3054:12): [True: 25.9k, False: 25.7k] | Branch (3054:30): [True: 25.7k, False: 184] ------------------ 3055| 25.7k| const uint16_t s1 = 3056| 25.7k| ra_get_key_at_index(&x1->high_low_container, (uint16_t)pos1); 3057| 25.7k| const uint16_t s2 = 3058| 25.7k| ra_get_key_at_index(&x2->high_low_container, (uint16_t)pos2); 3059| | 3060| 25.7k| if (s1 == s2) { ------------------ | Branch (3060:13): [True: 25.4k, False: 296] ------------------ 3061| 25.4k| uint8_t type1, type2; 3062| 25.4k| container_t *c1 = ra_get_container_at_index(&x1->high_low_container, 3063| 25.4k| (uint16_t)pos1, &type1); 3064| 25.4k| container_t *c2 = ra_get_container_at_index(&x2->high_low_container, 3065| 25.4k| (uint16_t)pos2, &type2); 3066| 25.4k| answer += container_and_cardinality(c1, type1, c2, type2); 3067| 25.4k| ++pos1; 3068| 25.4k| ++pos2; 3069| 25.4k| } else if (s1 < s2) { // s1 < s2 ------------------ | Branch (3069:20): [True: 296, False: 0] ------------------ 3070| 296| pos1 = ra_advance_until(&x1->high_low_container, s2, pos1); 3071| 296| } else { // s1 > s2 3072| 0| pos2 = ra_advance_until(&x2->high_low_container, s1, pos2); 3073| 0| } 3074| 25.7k| } 3075| 25.9k| return answer; 3076| 25.9k|} roaring_bitmap_jaccard_index: 3079| 6.48k| const roaring_bitmap_t *x2) { 3080| 6.48k| const uint64_t c1 = roaring_bitmap_get_cardinality(x1); 3081| 6.48k| const uint64_t c2 = roaring_bitmap_get_cardinality(x2); 3082| 6.48k| const uint64_t inter = roaring_bitmap_and_cardinality(x1, x2); 3083| 6.48k| return (double)inter / (double)(c1 + c2 - inter); 3084| 6.48k|} roaring_bitmap_or_cardinality: 3087| 6.48k| const roaring_bitmap_t *x2) { 3088| 6.48k| const uint64_t c1 = roaring_bitmap_get_cardinality(x1); 3089| 6.48k| const uint64_t c2 = roaring_bitmap_get_cardinality(x2); 3090| 6.48k| const uint64_t inter = roaring_bitmap_and_cardinality(x1, x2); 3091| 6.48k| return c1 + c2 - inter; 3092| 6.48k|} roaring_bitmap_andnot_cardinality: 3095| 6.48k| const roaring_bitmap_t *x2) { 3096| 6.48k| const uint64_t c1 = roaring_bitmap_get_cardinality(x1); 3097| 6.48k| const uint64_t inter = roaring_bitmap_and_cardinality(x1, x2); 3098| 6.48k| return c1 - inter; 3099| 6.48k|} roaring_bitmap_xor_cardinality: 3102| 6.48k| const roaring_bitmap_t *x2) { 3103| 6.48k| const uint64_t c1 = roaring_bitmap_get_cardinality(x1); 3104| 6.48k| const uint64_t c2 = roaring_bitmap_get_cardinality(x2); 3105| 6.48k| const uint64_t inter = roaring_bitmap_and_cardinality(x1, x2); 3106| 6.48k| return c1 + c2 - 2 * inter; 3107| 6.48k|} roaring_bitmap_contains_range: 3114| 6.48k| uint64_t range_start, uint64_t range_end) { 3115| 6.48k| if (range_start >= range_end || range_start > (uint64_t)UINT32_MAX + 1) { ------------------ | Branch (3115:9): [True: 5.58k, False: 895] | Branch (3115:37): [True: 0, False: 895] ------------------ 3116| 5.58k| return true; 3117| 5.58k| } 3118| 895| return roaring_bitmap_contains_range_closed(r, (uint32_t)range_start, 3119| 895| (uint32_t)(range_end - 1)); 3120| 6.48k|} roaring_bitmap_contains_range_closed: 3128| 895| uint32_t range_end) { 3129| 895| if (range_start > range_end) { ------------------ | Branch (3129:9): [True: 0, False: 895] ------------------ 3130| 0| return true; 3131| 0| } // empty range are always contained! 3132| 895| if (range_end == range_start) { ------------------ | Branch (3132:9): [True: 74, False: 821] ------------------ 3133| 74| return roaring_bitmap_contains(r, (uint32_t)range_start); 3134| 74| } 3135| 821| uint16_t hb_rs = (uint16_t)(range_start >> 16); 3136| 821| uint16_t hb_re = (uint16_t)(range_end >> 16); 3137| 821| const int32_t span = hb_re - hb_rs; 3138| 821| const int32_t hlc_sz = ra_get_size(&r->high_low_container); 3139| 821| if (hlc_sz < span + 1) { ------------------ | Branch (3139:9): [True: 225, False: 596] ------------------ 3140| 225| return false; 3141| 225| } 3142| 596| int32_t is = ra_get_index(&r->high_low_container, hb_rs); 3143| 596| int32_t ie = ra_get_index(&r->high_low_container, hb_re); 3144| 596| if ((ie < 0) || (is < 0) || ((ie - is) != span) || ie >= hlc_sz) { ------------------ | Branch (3144:9): [True: 31, False: 565] | Branch (3144:21): [True: 17, False: 548] | Branch (3144:33): [True: 9, False: 539] | Branch (3144:56): [True: 0, False: 539] ------------------ 3145| 57| return false; 3146| 57| } 3147| 539| const uint32_t lb_rs = range_start & 0xFFFF; 3148| 539| const uint32_t lb_re = (range_end & 0xFFFF) + 1; 3149| 539| uint8_t type; 3150| 539| container_t *c = 3151| 539| ra_get_container_at_index(&r->high_low_container, (uint16_t)is, &type); 3152| 539| if (hb_rs == hb_re) { ------------------ | Branch (3152:9): [True: 468, False: 71] ------------------ 3153| 468| return container_contains_range(c, lb_rs, lb_re, type); 3154| 468| } 3155| 71| if (!container_contains_range(c, lb_rs, 1 << 16, type)) { ------------------ | Branch (3155:9): [True: 27, False: 44] ------------------ 3156| 27| return false; 3157| 27| } 3158| 44| c = ra_get_container_at_index(&r->high_low_container, (uint16_t)ie, &type); 3159| 44| if (!container_contains_range(c, 0, lb_re, type)) { ------------------ | Branch (3159:9): [True: 5, False: 39] ------------------ 3160| 5| return false; 3161| 5| } 3162| 885| for (int32_t i = is + 1; i < ie; ++i) { ------------------ | Branch (3162:30): [True: 861, False: 24] ------------------ 3163| 861| c = ra_get_container_at_index(&r->high_low_container, (uint16_t)i, 3164| 861| &type); 3165| 861| if (!container_is_full(c, type)) { ------------------ | Branch (3165:13): [True: 15, False: 846] ------------------ 3166| 15| return false; 3167| 15| } 3168| 861| } 3169| 24| return true; 3170| 39|} roaring_bitmap_is_strict_subset: 3173| 6.48k| const roaring_bitmap_t *r2) { 3174| 6.48k| return (roaring_bitmap_get_cardinality(r2) > ------------------ | Branch (3174:13): [True: 5.31k, False: 1.16k] ------------------ 3175| 6.48k| roaring_bitmap_get_cardinality(r1) && 3176| 5.31k| roaring_bitmap_is_subset(r1, r2)); ------------------ | Branch (3176:13): [True: 250, False: 5.06k] ------------------ 3177| 6.48k|} roaring.c:containerptr_roaring_bitmap_add: 56| 19.4k| int *index) { 57| 19.4k| roaring_array_t *ra = &r->high_low_container; 58| | 59| 19.4k| uint16_t hb = val >> 16; 60| 19.4k| const int i = ra_get_index(ra, hb); 61| 19.4k| if (i >= 0) { ------------------ | Branch (61:9): [True: 6.48k, False: 12.9k] ------------------ 62| 6.48k| ra_unshare_container_at_index(ra, (uint16_t)i); 63| 6.48k| container_t *c = ra_get_container_at_index(ra, (uint16_t)i, type); 64| 6.48k| uint8_t new_type = *type; 65| 6.48k| container_t *c2 = container_add(c, val & 0xFFFF, *type, &new_type); 66| 6.48k| *index = i; 67| 6.48k| if (c2 != c) { ------------------ | Branch (67:13): [True: 3, False: 6.47k] ------------------ 68| 3| container_free(c, *type); 69| 3| ra_set_container_at_index(ra, i, c2, new_type); 70| 3| *type = new_type; 71| 3| return c2; 72| 6.47k| } else { 73| 6.47k| return c; 74| 6.47k| } 75| 12.9k| } else { 76| 12.9k| array_container_t *new_ac = array_container_create(); 77| 12.9k| container_t *c = 78| 12.9k| container_add(new_ac, val & 0xFFFF, ARRAY_CONTAINER_TYPE, type); ------------------ | | 49| 12.9k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 79| | // we could just assume that it stays an array container 80| 12.9k| ra_insert_new_key_value_at(ra, -i - 1, hb, c, *type); 81| 12.9k| *index = -i - 1; 82| 12.9k| return c; 83| 12.9k| } 84| 19.4k|} roaring.c:add_bulk_impl: 106| 8.10M| uint32_t val) { 107| 8.10M| uint16_t key = val >> 16; 108| 8.10M| if (context->container == NULL || context->key != key) { ------------------ | Branch (108:9): [True: 0, False: 8.10M] | Branch (108:39): [True: 0, False: 8.10M] ------------------ 109| 0| uint8_t typecode; 110| 0| int idx; 111| 0| context->container = 112| 0| containerptr_roaring_bitmap_add(r, val, &typecode, &idx); 113| 0| context->typecode = typecode; 114| 0| context->idx = idx; 115| 0| context->key = key; 116| 8.10M| } else { 117| | // no need to seek the container, it is at hand 118| | // because we already have the container at hand, we can do the 119| | // insertion directly, bypassing the roaring_bitmap_add call 120| 8.10M| uint8_t new_typecode; 121| 8.10M| container_t *container2 = container_add( 122| 8.10M| context->container, val & 0xFFFF, context->typecode, &new_typecode); 123| 8.10M| if (container2 != context->container) { ------------------ | Branch (123:13): [True: 128, False: 8.09M] ------------------ 124| | // rare instance when we need to change the container type 125| 128| container_free(context->container, context->typecode); 126| 128| ra_set_container_at_index(&r->high_low_container, context->idx, 127| 128| container2, new_typecode); 128| 128| context->typecode = new_typecode; 129| 128| context->container = container2; 130| 128| } 131| 8.10M| } 132| 8.10M|} roaring.c:is_cow: 41| 109k|static inline bool is_cow(const roaring_bitmap_t *r) { 42| | return r->high_low_container.flags & ROARING_FLAG_COW; ------------------ | | 46| 109k|#define ROARING_FLAG_COW UINT8_C(0x1) ------------------ 43| 109k|} roaring.c:roaring_inplace_merge_bulk: 957| 2.51k| int left, int right, bool is_xor) { 958| 2.51k| roaring_array_t *ra1 = &x1->high_low_container; 959| 2.51k| const roaring_array_t *ra2 = &x2->high_low_container; 960| 2.51k| const bool cow2 = is_cow(x2); 961| 2.51k| const int length1 = ra1->size; 962| 2.51k| const int length2 = ra2->size; 963| | 964| 2.51k| int distinct = 0; 965| 2.51k| { 966| 2.51k| int l = left, r = right; 967| 63.3k| while (l < length1 && r < length2) { ------------------ | Branch (967:16): [True: 60.8k, False: 2.51k] | Branch (967:31): [True: 60.8k, False: 0] ------------------ 968| 60.8k| uint16_t k1 = ra1->keys[l]; 969| 60.8k| uint16_t k2 = ra2->keys[r]; 970| 60.8k| if (k1 < k2) { ------------------ | Branch (970:17): [True: 0, False: 60.8k] ------------------ 971| 0| l++; 972| 60.8k| } else if (k1 > k2) { ------------------ | Branch (972:24): [True: 111, False: 60.7k] ------------------ 973| 111| r++; 974| 60.7k| } else { 975| 60.7k| l++; 976| 60.7k| r++; 977| 60.7k| } 978| 60.8k| distinct++; 979| 60.8k| } 980| 2.51k| distinct += (length1 - l) + (length2 - r); 981| 2.51k| } 982| 2.51k| const int total = dst + distinct; 983| | 984| 2.51k| roaring_array_t merged; 985| 2.51k| ra_init_with_capacity(&merged, total > 0 ? (uint32_t)total : 1); ------------------ | Branch (985:36): [True: 2.42k, False: 92] ------------------ 986| | 987| 4.69k| for (int i = 0; i < dst; i++) { ------------------ | Branch (987:21): [True: 2.17k, False: 2.51k] ------------------ 988| 2.17k| ra_append(&merged, ra1->keys[i], ra1->containers[i], ra1->typecodes[i]); 989| 2.17k| } 990| | 991| 2.51k| uint8_t result_type = 0; 992| 63.3k| while (left < length1 && right < length2) { ------------------ | Branch (992:12): [True: 60.8k, False: 2.51k] | Branch (992:30): [True: 60.8k, False: 0] ------------------ 993| 60.8k| uint16_t k1 = ra1->keys[left]; 994| 60.8k| uint16_t k2 = ra2->keys[right]; 995| 60.8k| if (k1 < k2) { ------------------ | Branch (995:13): [True: 0, False: 60.8k] ------------------ 996| 0| ra_append(&merged, k1, ra1->containers[left], ra1->typecodes[left]); 997| 0| left++; 998| 60.8k| } else if (k1 > k2) { ------------------ | Branch (998:20): [True: 111, False: 60.7k] ------------------ 999| 111| uint8_t type2 = ra2->typecodes[right]; 1000| 111| container_t *c2 = 1001| 111| get_copy_of_container(ra2->containers[right], &type2, cow2); 1002| 111| if (cow2) { ------------------ | Branch (1002:17): [True: 0, False: 111] ------------------ 1003| 0| ra_set_container_at_index(ra2, right, c2, type2); 1004| 0| } 1005| 111| ra_append(&merged, k2, c2, type2); 1006| 111| right++; 1007| 60.7k| } else { 1008| 60.7k| uint8_t type1 = ra1->typecodes[left]; 1009| 60.7k| container_t *c1 = ra1->containers[left]; 1010| 60.7k| uint8_t type2 = ra2->typecodes[right]; 1011| 60.7k| container_t *c2 = ra2->containers[right]; 1012| 60.7k| if (is_xor) { ------------------ | Branch (1012:17): [True: 60.7k, False: 0] ------------------ 1013| 60.7k| container_t *c; 1014| 60.7k| if (type1 == SHARED_CONTAINER_TYPE) { ------------------ | | 51| 60.7k|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (1014:21): [True: 0, False: 60.7k] ------------------ 1015| 0| c = container_xor(c1, type1, c2, type2, &result_type); 1016| 0| shared_container_free(CAST_shared(c1)); ------------------ | | 79| 0|#define CAST_shared(c) CAST(shared_container_t *, c) // safer downcast | | ------------------ | | | | 86| 0|#define CAST(type, value) ((type)value) | | ------------------ ------------------ 1017| 60.7k| } else { 1018| 60.7k| c = container_ixor(c1, type1, c2, type2, &result_type); 1019| 60.7k| } 1020| 60.7k| if (container_nonzero_cardinality(c, result_type)) { ------------------ | Branch (1020:21): [True: 0, False: 60.7k] ------------------ 1021| 0| ra_append(&merged, k1, c, result_type); 1022| 60.7k| } else { 1023| 60.7k| container_free(c, result_type); 1024| 60.7k| } 1025| 60.7k| } else { 1026| 0| if (container_is_full(c1, type1)) { ------------------ | Branch (1026:21): [True: 0, False: 0] ------------------ 1027| 0| ra_append(&merged, k1, c1, type1); 1028| 0| } else { 1029| 0| container_t *c = 1030| 0| (type1 == SHARED_CONTAINER_TYPE) ------------------ | | 51| 0|#define SHARED_CONTAINER_TYPE 4 ------------------ | Branch (1030:25): [True: 0, False: 0] ------------------ 1031| 0| ? container_or(c1, type1, c2, type2, &result_type) 1032| 0| : container_ior(c1, type1, c2, type2, &result_type); 1033| 0| if (c != c1) { ------------------ | Branch (1033:25): [True: 0, False: 0] ------------------ 1034| 0| container_free(c1, type1); 1035| 0| } 1036| 0| ra_append(&merged, k1, c, result_type); 1037| 0| } 1038| 0| } 1039| 60.7k| left++; 1040| 60.7k| right++; 1041| 60.7k| } 1042| 60.8k| } 1043| 2.51k| for (; left < length1; left++) { ------------------ | Branch (1043:12): [True: 0, False: 2.51k] ------------------ 1044| 0| ra_append(&merged, ra1->keys[left], ra1->containers[left], 1045| 0| ra1->typecodes[left]); 1046| 0| } 1047| 2.51k| for (; right < length2; right++) { ------------------ | Branch (1047:12): [True: 0, False: 2.51k] ------------------ 1048| 0| uint8_t type2 = ra2->typecodes[right]; 1049| 0| container_t *c2 = 1050| 0| get_copy_of_container(ra2->containers[right], &type2, cow2); 1051| 0| if (cow2) { ------------------ | Branch (1051:13): [True: 0, False: 0] ------------------ 1052| 0| ra_set_container_at_index(ra2, right, c2, type2); 1053| 0| } 1054| 0| ra_append(&merged, ra2->keys[right], c2, type2); 1055| 0| } 1056| | 1057| 2.51k| merged.flags = ra1->flags; 1058| 2.51k| ra_clear_without_containers(ra1); 1059| 2.51k| *ra1 = merged; 1060| 2.51k|} roaring.c:loadfirstvalue: 1809| 19.2k| roaring_uint32_iterator_t *newit) { 1810| 19.2k| if (iter_new_container_partial_init(newit)) { ------------------ | Branch (1810:9): [True: 12.7k, False: 6.52k] ------------------ 1811| 12.7k| uint16_t value = 0; 1812| 12.7k| newit->container_it = 1813| 12.7k| container_init_iterator(newit->container, newit->typecode, &value); 1814| 12.7k| newit->current_value = newit->highbits | value; 1815| 12.7k| } 1816| 19.2k| return newit->has_value; 1817| 19.2k|} roaring.c:iter_new_container_partial_init: 1779| 25.6k| roaring_uint32_iterator_t *newit) { 1780| 25.6k| newit->current_value = 0; 1781| 25.6k| if (newit->container_index >= newit->parent->high_low_container.size || ------------------ | Branch (1781:9): [True: 6.52k, False: 19.1k] ------------------ 1782| 19.1k| newit->container_index < 0) { ------------------ | Branch (1782:9): [True: 0, False: 19.1k] ------------------ 1783| 6.52k| newit->current_value = UINT32_MAX; 1784| 6.52k| return (newit->has_value = false); 1785| 6.52k| } 1786| 19.1k| newit->has_value = true; 1787| | // we precompute container, typecode and highbits so that successive 1788| | // iterators do not have to grab them from odd memory locations 1789| | // and have to worry about the (easily predicted) container_unwrap_shared 1790| | // call. 1791| 19.1k| newit->container = 1792| 19.1k| newit->parent->high_low_container.containers[newit->container_index]; 1793| 19.1k| newit->typecode = 1794| 19.1k| newit->parent->high_low_container.typecodes[newit->container_index]; 1795| 19.1k| newit->highbits = 1796| 19.1k| ((uint32_t) 1797| 19.1k| newit->parent->high_low_container.keys[newit->container_index]) 1798| 19.1k| << 16; 1799| 19.1k| newit->container = 1800| 19.1k| container_unwrap_shared(newit->container, &(newit->typecode)); 1801| | return true; 1802| 25.6k|} roaring.c:loadfirstvalue_largeorequal: 1840| 6.38k| roaring_uint32_iterator_t *newit, uint32_t val) { 1841| 6.38k| bool partial_init = iter_new_container_partial_init(newit); 1842| 6.38k| assert(partial_init); 1843| 6.38k| if (!partial_init) { ------------------ | Branch (1843:9): [True: 0, False: 6.38k] ------------------ 1844| 0| return false; 1845| 0| } 1846| 6.38k| uint16_t value = 0; 1847| 6.38k| newit->container_it = 1848| 6.38k| container_init_iterator(newit->container, newit->typecode, &value); 1849| 6.38k| bool found = container_iterator_lower_bound( 1850| 6.38k| newit->container, newit->typecode, &newit->container_it, &value, 1851| 6.38k| val & 0xFFFF); 1852| 6.38k| assert(found); 1853| 6.38k| if (!found) { ------------------ | Branch (1853:9): [True: 0, False: 6.38k] ------------------ 1854| 0| return false; 1855| 0| } 1856| 6.38k| newit->current_value = newit->highbits | value; 1857| | return true; 1858| 6.38k|} roaring.c:inplace_flip_container: 2214| 7.16k| uint16_t lb_start, uint16_t lb_end) { 2215| 7.16k| const int i = ra_get_index(x1_arr, hb); 2216| 7.16k| uint8_t ctype_in, ctype_out; 2217| 7.16k| container_t *flipped_container = NULL; 2218| 7.16k| if (i >= 0) { ------------------ | Branch (2218:9): [True: 6.62k, False: 531] ------------------ 2219| 6.62k| container_t *container_to_flip = 2220| 6.62k| ra_get_container_at_index(x1_arr, (uint16_t)i, &ctype_in); 2221| 6.62k| flipped_container = container_inot_range( 2222| 6.62k| container_to_flip, ctype_in, (uint32_t)lb_start, 2223| 6.62k| (uint32_t)(lb_end + 1), &ctype_out); 2224| | // if a new container was created, the old one was already freed 2225| 6.62k| if (container_nonzero_cardinality(flipped_container, ctype_out)) { ------------------ | Branch (2225:13): [True: 6.60k, False: 22] ------------------ 2226| 6.60k| ra_set_container_at_index(x1_arr, i, flipped_container, ctype_out); 2227| 6.60k| } else { 2228| 22| container_free(flipped_container, ctype_out); 2229| 22| ra_remove_at_index(x1_arr, i); 2230| 22| } 2231| | 2232| 6.62k| } else { 2233| 531| flipped_container = container_range_of_ones( 2234| 531| (uint32_t)lb_start, (uint32_t)(lb_end + 1), &ctype_out); 2235| 531| ra_insert_new_key_value_at(x1_arr, -i - 1, hb, flipped_container, 2236| 531| ctype_out); 2237| 531| } 2238| 7.16k|} roaring.c:inplace_fully_flip_container: 2265| 24.5k|static void inplace_fully_flip_container(roaring_array_t *x1_arr, uint16_t hb) { 2266| 24.5k| const int i = ra_get_index(x1_arr, hb); 2267| 24.5k| uint8_t ctype_in, ctype_out; 2268| 24.5k| container_t *flipped_container = NULL; 2269| 24.5k| if (i >= 0) { ------------------ | Branch (2269:9): [True: 6.85k, False: 17.7k] ------------------ 2270| 6.85k| container_t *container_to_flip = 2271| 6.85k| ra_get_container_at_index(x1_arr, (uint16_t)i, &ctype_in); 2272| 6.85k| flipped_container = 2273| 6.85k| container_inot(container_to_flip, ctype_in, &ctype_out); 2274| | 2275| 6.85k| if (container_nonzero_cardinality(flipped_container, ctype_out)) { ------------------ | Branch (2275:13): [True: 539, False: 6.31k] ------------------ 2276| 539| ra_set_container_at_index(x1_arr, i, flipped_container, ctype_out); 2277| 6.31k| } else { 2278| 6.31k| container_free(flipped_container, ctype_out); 2279| 6.31k| ra_remove_at_index(x1_arr, i); 2280| 6.31k| } 2281| | 2282| 17.7k| } else { 2283| 17.7k| flipped_container = container_range_of_ones(0U, 0x10000U, &ctype_out); 2284| 17.7k| ra_insert_new_key_value_at(x1_arr, -i - 1, hb, flipped_container, 2285| 17.7k| ctype_out); 2286| 17.7k| } 2287| 24.5k|} ra_init_with_capacity: 80| 67.3k|bool ra_init_with_capacity(roaring_array_t *new_ra, uint32_t cap) { 81| 67.3k| if (!new_ra) return false; ------------------ | Branch (81:9): [True: 0, False: 67.3k] ------------------ 82| 67.3k| ra_init(new_ra); 83| | 84| | // Containers hold 64Ki elements, so 64Ki containers is enough to hold 85| | // `0x10000 * 0x10000` (all 2^32) elements 86| 67.3k| if (cap > 0x10000) { ------------------ | Branch (86:9): [True: 0, False: 67.3k] ------------------ 87| 0| cap = 0x10000; 88| 0| } 89| | 90| 67.3k| if (cap > 0) { ------------------ | Branch (90:9): [True: 34.9k, False: 32.4k] ------------------ 91| 34.9k| void *bigalloc = roaring_malloc( 92| 34.9k| cap * (sizeof(uint16_t) + sizeof(container_t *) + sizeof(uint8_t))); 93| 34.9k| if (bigalloc == NULL) return false; ------------------ | Branch (93:13): [True: 0, False: 34.9k] ------------------ 94| 34.9k| new_ra->containers = (container_t **)bigalloc; 95| 34.9k| new_ra->keys = (uint16_t *)(new_ra->containers + cap); 96| 34.9k| new_ra->typecodes = (uint8_t *)(new_ra->keys + cap); 97| | // Narrowing is safe because of above check 98| 34.9k| new_ra->allocation_size = (int32_t)cap; 99| 34.9k| } 100| 67.3k| return true; 101| 67.3k|} ra_shrink_to_fit: 103| 71.2k|int ra_shrink_to_fit(roaring_array_t *ra) { 104| 71.2k| int savings = (ra->allocation_size - ra->size) * 105| 71.2k| (sizeof(uint16_t) + sizeof(container_t *) + sizeof(uint8_t)); 106| 71.2k| if (!realloc_array(ra, ra->size)) { ------------------ | Branch (106:9): [True: 0, False: 71.2k] ------------------ 107| 0| return 0; 108| 0| } 109| 71.2k| ra->allocation_size = ra->size; 110| 71.2k| return savings; 111| 71.2k|} ra_init: 113| 67.3k|void ra_init(roaring_array_t *new_ra) { 114| 67.3k| if (!new_ra) { ------------------ | Branch (114:9): [True: 0, False: 67.3k] ------------------ 115| 0| return; 116| 0| } 117| 67.3k| new_ra->keys = NULL; 118| 67.3k| new_ra->containers = NULL; 119| 67.3k| new_ra->typecodes = NULL; 120| | 121| 67.3k| new_ra->allocation_size = 0; 122| 67.3k| new_ra->size = 0; 123| 67.3k| new_ra->flags = 0; 124| 67.3k|} ra_overwrite: 127| 14.6k| bool copy_on_write) { 128| 14.6k| ra_clear_containers(dest); // we are going to overwrite them 129| 14.6k| if (source->size == 0) { // Note: can't call memcpy(NULL), even w/size ------------------ | Branch (129:9): [True: 16, False: 14.6k] ------------------ 130| 16| dest->size = 0; // <--- This is important. 131| 16| return true; // output was just cleared, so they match 132| 16| } 133| 14.6k| if (dest->allocation_size < source->size) { ------------------ | Branch (133:9): [True: 12.9k, False: 1.67k] ------------------ 134| 12.9k| if (!realloc_array(dest, source->size)) { ------------------ | Branch (134:13): [True: 0, False: 12.9k] ------------------ 135| 0| return false; 136| 0| } 137| 12.9k| } 138| 14.6k| dest->size = source->size; 139| 14.6k| memcpy(dest->keys, source->keys, dest->size * sizeof(uint16_t)); 140| | // we go through the containers, turning them into shared containers... 141| 14.6k| if (copy_on_write) { ------------------ | Branch (141:9): [True: 0, False: 14.6k] ------------------ 142| 0| for (int32_t i = 0; i < dest->size; ++i) { ------------------ | Branch (142:29): [True: 0, False: 0] ------------------ 143| 0| source->containers[i] = get_copy_of_container( 144| 0| source->containers[i], &source->typecodes[i], copy_on_write); 145| 0| } 146| | // we do a shallow copy to the other bitmap 147| 0| memcpy(dest->containers, source->containers, 148| 0| dest->size * sizeof(container_t *)); 149| 0| memcpy(dest->typecodes, source->typecodes, 150| 0| dest->size * sizeof(uint8_t)); 151| 14.6k| } else { 152| 14.6k| memcpy(dest->typecodes, source->typecodes, 153| 14.6k| dest->size * sizeof(uint8_t)); 154| 178k| for (int32_t i = 0; i < dest->size; i++) { ------------------ | Branch (154:29): [True: 164k, False: 14.6k] ------------------ 155| 164k| dest->containers[i] = 156| 164k| container_clone(source->containers[i], source->typecodes[i]); 157| 164k| if (dest->containers[i] == NULL) { ------------------ | Branch (157:17): [True: 0, False: 164k] ------------------ 158| 0| for (int32_t j = 0; j < i; j++) { ------------------ | Branch (158:37): [True: 0, False: 0] ------------------ 159| 0| container_free(dest->containers[j], dest->typecodes[j]); 160| 0| } 161| 0| ra_clear_without_containers(dest); 162| 0| return false; 163| 0| } 164| 164k| } 165| 14.6k| } 166| 14.6k| return true; 167| 14.6k|} ra_clear_containers: 169| 79.4k|void ra_clear_containers(roaring_array_t *ra) { 170| 561k| for (int32_t i = 0; i < ra->size; ++i) { ------------------ | Branch (170:25): [True: 482k, False: 79.4k] ------------------ 171| 482k| container_free(ra->containers[i], ra->typecodes[i]); 172| 482k| } 173| 79.4k|} ra_reset: 175| 64.8k|void ra_reset(roaring_array_t *ra) { 176| 64.8k| ra_clear_containers(ra); 177| 64.8k| ra->size = 0; 178| 64.8k| ra_shrink_to_fit(ra); 179| 64.8k|} ra_clear_without_containers: 181| 2.55k|void ra_clear_without_containers(roaring_array_t *ra) { 182| 2.55k| roaring_free( 183| 2.55k| ra->containers); // keys and typecodes are allocated with containers 184| 2.55k| ra->size = 0; 185| 2.55k| ra->allocation_size = 0; 186| 2.55k| ra->containers = NULL; 187| 2.55k| ra->keys = NULL; 188| | ra->typecodes = NULL; 189| 2.55k|} ra_clear: 191| 37|void ra_clear(roaring_array_t *ra) { 192| 37| ra_clear_containers(ra); 193| 37| ra_clear_without_containers(ra); 194| 37|} extend_array: 196| 89.1k|bool extend_array(roaring_array_t *ra, int32_t k) { 197| 89.1k| int32_t desired_size = ra->size + k; 198| 89.1k| const int32_t max_containers = 65536; 199| 89.1k| assert(desired_size <= max_containers); 200| 89.1k| if (desired_size > ra->allocation_size) { ------------------ | Branch (200:9): [True: 18.5k, False: 70.6k] ------------------ 201| 18.5k| int32_t new_capacity = 202| 18.5k| (ra->size < 1024) ? 2 * desired_size : 5 * desired_size / 4; ------------------ | Branch (202:13): [True: 18.5k, False: 0] ------------------ 203| 18.5k| if (new_capacity > max_containers) { ------------------ | Branch (203:13): [True: 0, False: 18.5k] ------------------ 204| 0| new_capacity = max_containers; 205| 0| } 206| | 207| 18.5k| return realloc_array(ra, new_capacity); 208| 18.5k| } 209| 70.6k| return true; 210| 89.1k|} ra_append: 213| 27.3k| uint8_t typecode) { 214| 27.3k| extend_array(ra, 1); 215| 27.3k| const int32_t pos = ra->size; 216| | 217| 27.3k| ra->keys[pos] = key; 218| 27.3k| ra->containers[pos] = c; 219| 27.3k| ra->typecodes[pos] = typecode; 220| 27.3k| ra->size++; 221| 27.3k|} ra_append_copy_range: 254| 25.8k| bool copy_on_write) { 255| 25.8k| extend_array(ra, end_index - start_index); 256| 214k| for (int32_t i = start_index; i < end_index; ++i) { ------------------ | Branch (256:35): [True: 188k, False: 25.8k] ------------------ 257| 188k| const int32_t pos = ra->size; 258| 188k| ra->keys[pos] = sa->keys[i]; 259| 188k| if (copy_on_write) { ------------------ | Branch (259:13): [True: 0, False: 188k] ------------------ 260| 0| sa->containers[i] = get_copy_of_container( 261| 0| sa->containers[i], &sa->typecodes[i], copy_on_write); 262| 0| ra->containers[pos] = sa->containers[i]; 263| 0| ra->typecodes[pos] = sa->typecodes[i]; 264| 188k| } else { 265| 188k| ra->containers[pos] = 266| 188k| container_clone(sa->containers[i], sa->typecodes[i]); 267| 188k| ra->typecodes[pos] = sa->typecodes[i]; 268| 188k| } 269| 188k| ra->size++; 270| 188k| } 271| 25.8k|} ra_advance_until_freeing: 340| 74|int32_t ra_advance_until_freeing(roaring_array_t *ra, uint16_t x, int32_t pos) { 341| 148| while (pos < ra->size && ra->keys[pos] < x) { ------------------ | Branch (341:12): [True: 148, False: 0] | Branch (341:30): [True: 74, False: 74] ------------------ 342| 74| container_free(ra->containers[pos], ra->typecodes[pos]); 343| 74| ++pos; 344| 74| } 345| 74| return pos; 346| 74|} ra_insert_new_key_value_at: 349| 34.6k| container_t *c, uint8_t typecode) { 350| 34.6k| extend_array(ra, 1); 351| | // May be an optimization opportunity with DIY memmove 352| 34.6k| memmove(&(ra->keys[i + 1]), &(ra->keys[i]), 353| 34.6k| sizeof(uint16_t) * (ra->size - i)); 354| 34.6k| memmove(&(ra->containers[i + 1]), &(ra->containers[i]), 355| 34.6k| sizeof(container_t *) * (ra->size - i)); 356| 34.6k| memmove(&(ra->typecodes[i + 1]), &(ra->typecodes[i]), 357| 34.6k| sizeof(uint8_t) * (ra->size - i)); 358| 34.6k| ra->keys[i] = key; 359| 34.6k| ra->containers[i] = c; 360| 34.6k| ra->typecodes[i] = typecode; 361| 34.6k| ra->size++; 362| 34.6k|} ra_downsize: 369| 12.5k|void ra_downsize(roaring_array_t *ra, int32_t new_length) { 370| | assert(new_length <= ra->size); 371| 12.5k| ra->size = new_length; 372| 12.5k|} ra_remove_at_index: 374| 6.40k|void ra_remove_at_index(roaring_array_t *ra, int32_t i) { 375| 6.40k| memmove(&(ra->containers[i]), &(ra->containers[i + 1]), 376| 6.40k| sizeof(container_t *) * (ra->size - i - 1)); 377| 6.40k| memmove(&(ra->keys[i]), &(ra->keys[i + 1]), 378| 6.40k| sizeof(uint16_t) * (ra->size - i - 1)); 379| 6.40k| memmove(&(ra->typecodes[i]), &(ra->typecodes[i + 1]), 380| 6.40k| sizeof(uint8_t) * (ra->size - i - 1)); 381| 6.40k| ra->size--; 382| 6.40k|} ra_remove_at_index_and_free: 384| 69|void ra_remove_at_index_and_free(roaring_array_t *ra, int32_t i) { 385| 69| container_free(ra->containers[i], ra->typecodes[i]); 386| 69| ra_remove_at_index(ra, i); 387| 69|} ra_copy_range: 395| 39| uint32_t new_begin) { 396| 39| assert(begin <= end); 397| 39| assert(new_begin < begin); 398| | 399| 39| const int range = end - begin; 400| | 401| | // We ensure to previously have freed overwritten containers 402| | // that are not copied elsewhere 403| | 404| 39| memmove(&(ra->containers[new_begin]), &(ra->containers[begin]), 405| 39| sizeof(container_t *) * range); 406| 39| memmove(&(ra->keys[new_begin]), &(ra->keys[begin]), 407| 39| sizeof(uint16_t) * range); 408| 39| memmove(&(ra->typecodes[new_begin]), &(ra->typecodes[begin]), 409| 39| sizeof(uint8_t) * range); 410| 39|} ra_shift_tail: 412| 1.97k|void ra_shift_tail(roaring_array_t *ra, int32_t count, int32_t distance) { 413| 1.97k| if (distance > 0) { ------------------ | Branch (413:9): [True: 1.33k, False: 634] ------------------ 414| 1.33k| extend_array(ra, distance); 415| 1.33k| } 416| 1.97k| int32_t srcpos = ra->size - count; 417| 1.97k| int32_t dstpos = srcpos + distance; 418| 1.97k| memmove(&(ra->keys[dstpos]), &(ra->keys[srcpos]), sizeof(uint16_t) * count); 419| 1.97k| memmove(&(ra->containers[dstpos]), &(ra->containers[srcpos]), 420| 1.97k| sizeof(container_t *) * count); 421| 1.97k| memmove(&(ra->typecodes[dstpos]), &(ra->typecodes[srcpos]), 422| 1.97k| sizeof(uint8_t) * count); 423| 1.97k| ra->size += distance; 424| 1.97k|} ra_to_uint32_array: 426| 6.48k|void ra_to_uint32_array(const roaring_array_t *ra, uint32_t *ans) { 427| 6.48k| size_t ctr = 0; 428| 75.7k| for (int32_t i = 0; i < ra->size; ++i) { ------------------ | Branch (428:25): [True: 69.2k, False: 6.48k] ------------------ 429| 69.2k| int num_added = container_to_uint32_array( 430| 69.2k| ans + ctr, ra->containers[i], ra->typecodes[i], 431| 69.2k| ((uint32_t)ra->keys[i]) << 16); 432| 69.2k| ctr += num_added; 433| 69.2k| } 434| 6.48k|} ra_has_run_container: 436| 19.4k|bool ra_has_run_container(const roaring_array_t *ra) { 437| 185k| for (int32_t k = 0; k < ra->size; ++k) { ------------------ | Branch (437:25): [True: 168k, False: 16.7k] ------------------ 438| 168k| if (get_container_type(ra->containers[k], ra->typecodes[k]) == ------------------ | Branch (438:13): [True: 2.67k, False: 166k] ------------------ 439| 168k| RUN_CONTAINER_TYPE) ------------------ | | 50| 168k|#define RUN_CONTAINER_TYPE 3 ------------------ 440| 2.67k| return true; 441| 168k| } 442| 16.7k| return false; 443| 19.4k|} ra_portable_header_size: 445| 12.9k|uint32_t ra_portable_header_size(const roaring_array_t *ra) { 446| 12.9k| if (ra_has_run_container(ra)) { ------------------ | Branch (446:9): [True: 2.67k, False: 10.2k] ------------------ 447| 2.67k| if (ra->size < ------------------ | Branch (447:13): [True: 2.67k, False: 0] ------------------ 448| 2.67k| NO_OFFSET_THRESHOLD) { // for small bitmaps, we omit the offsets 449| 2.67k| return 4 + (ra->size + 7) / 8 + 4 * ra->size; 450| 2.67k| } 451| 0| return 4 + (ra->size + 7) / 8 + 452| 0| 8 * ra->size; // - 4 because we pack the size with the cookie 453| 10.2k| } else { 454| 10.2k| return 4 + 4 + 8 * ra->size; 455| 10.2k| } 456| 12.9k|} ra_portable_size_in_bytes: 458| 12.9k|size_t ra_portable_size_in_bytes(const roaring_array_t *ra) { 459| 12.9k| size_t count = ra_portable_header_size(ra); 460| | 461| 100k| for (int32_t k = 0; k < ra->size; ++k) { ------------------ | Branch (461:25): [True: 87.6k, False: 12.9k] ------------------ 462| 87.6k| count += container_size_in_bytes(ra->containers[k], ra->typecodes[k]); 463| 87.6k| } 464| 12.9k| return count; 465| 12.9k|} ra_portable_serialize: 469| 6.48k|size_t ra_portable_serialize(const roaring_array_t *ra, char *buf) { 470| 6.48k| char *initbuf = buf; 471| 6.48k| uint32_t startOffset = 0; 472| 6.48k| bool hasrun = ra_has_run_container(ra); 473| 6.48k| if (hasrun) { ------------------ | Branch (473:9): [True: 0, False: 6.48k] ------------------ 474| 0| uint32_t cookie = SERIAL_COOKIE | ((uint32_t)(ra->size - 1) << 16); 475| 0| uint32_t cookie_le = croaring_htole32(cookie); ------------------ | | 518| 0|#define croaring_htole32(x) (x) ------------------ 476| 0| memcpy(buf, &cookie_le, sizeof(cookie_le)); 477| 0| buf += sizeof(cookie_le); 478| 0| uint32_t s = (ra->size + 7) / 8; 479| 0| memset(buf, 0, s); 480| 0| for (int32_t i = 0; i < ra->size; ++i) { ------------------ | Branch (480:29): [True: 0, False: 0] ------------------ 481| 0| if (get_container_type(ra->containers[i], ra->typecodes[i]) == ------------------ | Branch (481:17): [True: 0, False: 0] ------------------ 482| 0| RUN_CONTAINER_TYPE) { ------------------ | | 50| 0|#define RUN_CONTAINER_TYPE 3 ------------------ 483| 0| buf[i / 8] |= 1 << (i % 8); 484| 0| } 485| 0| } 486| 0| buf += s; 487| 0| if (ra->size < NO_OFFSET_THRESHOLD) { ------------------ | Branch (487:13): [True: 0, False: 0] ------------------ 488| 0| startOffset = 4 + 4 * ra->size + s; 489| 0| } else { 490| 0| startOffset = 4 + 8 * ra->size + s; 491| 0| } 492| 6.48k| } else { // backwards compatibility 493| 6.48k| uint32_t cookie = SERIAL_COOKIE_NO_RUNCONTAINER; 494| 6.48k| uint32_t cookie_le = croaring_htole32(cookie); ------------------ | | 518| 6.48k|#define croaring_htole32(x) (x) ------------------ 495| 6.48k| memcpy(buf, &cookie_le, sizeof(cookie_le)); 496| 6.48k| buf += sizeof(cookie_le); 497| 6.48k| uint32_t size_le = croaring_htole32((uint32_t)ra->size); ------------------ | | 518| 6.48k|#define croaring_htole32(x) (x) ------------------ 498| 6.48k| memcpy(buf, &size_le, sizeof(size_le)); 499| 6.48k| buf += sizeof(size_le); 500| | 501| 6.48k| startOffset = 4 + 4 + 4 * ra->size + 4 * ra->size; 502| 6.48k| } 503| 87.6k| for (int32_t k = 0; k < ra->size; ++k) { ------------------ | Branch (503:25): [True: 81.1k, False: 6.48k] ------------------ 504| 81.1k| uint16_t key_le = croaring_htole16(ra->keys[k]); ------------------ | | 517| 81.1k|#define croaring_htole16(x) (x) ------------------ 505| 81.1k| memcpy(buf, &key_le, sizeof(key_le)); 506| 81.1k| buf += sizeof(key_le); 507| | // get_cardinality returns a value in [1,1<<16], subtracting one 508| | // we get [0,1<<16 - 1] which fits in 16 bits 509| 81.1k| uint16_t card = (uint16_t)(container_get_cardinality(ra->containers[k], 510| 81.1k| ra->typecodes[k]) - 511| 81.1k| 1); 512| 81.1k| uint16_t card_le = croaring_htole16(card); ------------------ | | 517| 81.1k|#define croaring_htole16(x) (x) ------------------ 513| 81.1k| memcpy(buf, &card_le, sizeof(card_le)); 514| 81.1k| buf += sizeof(card_le); 515| 81.1k| } 516| 6.48k| if ((!hasrun) || (ra->size >= NO_OFFSET_THRESHOLD)) { ------------------ | Branch (516:9): [True: 6.48k, False: 0] | Branch (516:22): [True: 0, False: 0] ------------------ 517| | // writing the containers offsets 518| 87.6k| for (int32_t k = 0; k < ra->size; k++) { ------------------ | Branch (518:29): [True: 81.1k, False: 6.48k] ------------------ 519| 81.1k| uint32_t off_le = croaring_htole32(startOffset); ------------------ | | 518| 81.1k|#define croaring_htole32(x) (x) ------------------ 520| 81.1k| memcpy(buf, &off_le, sizeof(off_le)); 521| 81.1k| buf += sizeof(off_le); 522| 81.1k| startOffset = 523| 81.1k| startOffset + 524| 81.1k| container_size_in_bytes(ra->containers[k], ra->typecodes[k]); 525| 81.1k| } 526| 6.48k| } 527| 87.6k| for (int32_t k = 0; k < ra->size; ++k) { ------------------ | Branch (527:25): [True: 81.1k, False: 6.48k] ------------------ 528| 81.1k| buf += container_write(ra->containers[k], ra->typecodes[k], buf); 529| 81.1k| } 530| 6.48k| return buf - initbuf; 531| 6.48k|} ra_portable_deserialize: 634| 12.9k| const size_t maxbytes, size_t *readbytes) { 635| 12.9k| *readbytes = sizeof(int32_t); // for cookie 636| 12.9k| if (*readbytes > maxbytes) { ------------------ | Branch (636:9): [True: 6.10k, False: 6.85k] ------------------ 637| | // Ran out of bytes while reading first 4 bytes. 638| 6.10k| return false; 639| 6.10k| } 640| 6.85k| uint32_t cookie; 641| 6.85k| memcpy(&cookie, buf, sizeof(int32_t)); 642| 6.85k| cookie = croaring_letoh32(cookie); ------------------ | | 524| 6.85k|#define croaring_letoh32(x) croaring_htole32(x) | | ------------------ | | | | 518| 6.85k|#define croaring_htole32(x) (x) | | ------------------ ------------------ 643| 6.85k| buf += sizeof(uint32_t); 644| 6.85k| if ((cookie & 0xFFFF) != SERIAL_COOKIE && ------------------ | Branch (644:9): [True: 6.78k, False: 66] ------------------ 645| 6.78k| cookie != SERIAL_COOKIE_NO_RUNCONTAINER) { ------------------ | Branch (645:9): [True: 252, False: 6.53k] ------------------ 646| | // "I failed to find one of the right cookies. 647| 252| return false; 648| 252| } 649| 6.60k| int32_t size; 650| | 651| 6.60k| if ((cookie & 0xFFFF) == SERIAL_COOKIE) ------------------ | Branch (651:9): [True: 66, False: 6.53k] ------------------ 652| 66| size = (cookie >> 16) + 1; 653| 6.53k| else { 654| 6.53k| *readbytes += sizeof(int32_t); 655| 6.53k| if (*readbytes > maxbytes) { ------------------ | Branch (655:13): [True: 2, False: 6.53k] ------------------ 656| | // Ran out of bytes while reading second part of the cookie. 657| 2| return false; 658| 2| } 659| 6.53k| uint32_t size_le; 660| 6.53k| memcpy(&size_le, buf, sizeof(int32_t)); 661| 6.53k| size = (int32_t)croaring_letoh32(size_le); ------------------ | | 524| 6.53k|#define croaring_letoh32(x) croaring_htole32(x) | | ------------------ | | | | 518| 6.53k|#define croaring_htole32(x) (x) | | ------------------ ------------------ 662| 6.53k| buf += sizeof(uint32_t); 663| 6.53k| } 664| 6.60k| if (size < 0) { ------------------ | Branch (664:9): [True: 19, False: 6.58k] ------------------ 665| | // You cannot have a negative number of containers, the data must be 666| | // corrupted. 667| 19| return false; 668| 19| } 669| 6.58k| if (size > (1 << 16)) { ------------------ | Branch (669:9): [True: 19, False: 6.56k] ------------------ 670| | // You cannot have so many containers, the data must be corrupted. 671| 19| return false; 672| 19| } 673| 6.56k| const char *bitmapOfRunContainers = NULL; 674| 6.56k| bool hasrun = (cookie & 0xFFFF) == SERIAL_COOKIE; 675| 6.56k| if (hasrun) { ------------------ | Branch (675:9): [True: 66, False: 6.49k] ------------------ 676| 66| int32_t s = (size + 7) / 8; 677| 66| *readbytes += s; 678| 66| if (*readbytes > maxbytes) { // data is corrupted? ------------------ | Branch (678:13): [True: 19, False: 47] ------------------ 679| | // Ran out of bytes while reading run bitmap. 680| 19| return false; 681| 19| } 682| 47| bitmapOfRunContainers = buf; 683| 47| buf += s; 684| 47| } 685| 6.54k| const char *keyscards = buf; 686| | 687| 6.54k| *readbytes += size * 2 * sizeof(uint16_t); 688| 6.54k| if (*readbytes > maxbytes) { ------------------ | Branch (688:9): [True: 13, False: 6.53k] ------------------ 689| | // Ran out of bytes while reading key-cardinality array. 690| 13| return false; 691| 13| } 692| 6.53k| buf += size * 2 * sizeof(uint16_t); 693| | 694| 6.53k| bool is_ok = ra_init_with_capacity(answer, size); 695| 6.53k| if (!is_ok) { ------------------ | Branch (695:9): [True: 0, False: 6.53k] ------------------ 696| | // Failed to allocate memory for roaring array. Bailing out. 697| 0| return false; 698| 0| } 699| | 700| 87.9k| for (int32_t k = 0; k < size; ++k) { ------------------ | Branch (700:25): [True: 81.3k, False: 6.53k] ------------------ 701| 81.3k| uint16_t tmp; 702| 81.3k| memcpy(&tmp, keyscards + 4 * k, sizeof(tmp)); 703| 81.3k| answer->keys[k] = croaring_letoh16(tmp); ------------------ | | 523| 81.3k|#define croaring_letoh16(x) croaring_htole16(x) | | ------------------ | | | | 517| 81.3k|#define croaring_htole16(x) (x) | | ------------------ ------------------ 704| 81.3k| } 705| 6.53k| if ((!hasrun) || (size >= NO_OFFSET_THRESHOLD)) { ------------------ | Branch (705:9): [True: 6.48k, False: 46] | Branch (705:22): [True: 20, False: 26] ------------------ 706| 6.50k| *readbytes += size * 4; 707| 6.50k| if (*readbytes > maxbytes) { // data is corrupted? ------------------ | Branch (707:13): [True: 5, False: 6.49k] ------------------ 708| | // Ran out of bytes while reading offsets. 709| 5| ra_clear(answer); // we need to clear the containers already 710| | // allocated, and the roaring array 711| 5| return false; 712| 5| } 713| | 714| | // skipping the offsets 715| 6.49k| buf += size * 4; 716| 6.49k| } 717| | // Reading the containers 718| 87.8k| for (int32_t k = 0; k < size; ++k) { ------------------ | Branch (718:25): [True: 81.3k, False: 6.49k] ------------------ 719| 81.3k| uint16_t tmp; 720| 81.3k| memcpy(&tmp, keyscards + 4 * k + 2, sizeof(tmp)); 721| 81.3k| tmp = croaring_letoh16(tmp); ------------------ | | 523| 81.3k|#define croaring_letoh16(x) croaring_htole16(x) | | ------------------ | | | | 517| 81.3k|#define croaring_htole16(x) (x) | | ------------------ ------------------ 722| 81.3k| uint32_t thiscard = tmp + 1; 723| 81.3k| bool isbitmap = (thiscard > DEFAULT_MAX_SIZE); 724| 81.3k| bool isrun = false; 725| 81.3k| if (hasrun) { ------------------ | Branch (725:13): [True: 135, False: 81.1k] ------------------ 726| 135| if ((bitmapOfRunContainers[k / 8] & (1 << (k % 8))) != 0) { ------------------ | Branch (726:17): [True: 66, False: 69] ------------------ 727| 66| isbitmap = false; 728| 66| isrun = true; 729| 66| } 730| 135| } 731| 81.3k| if (isbitmap) { ------------------ | Branch (731:13): [True: 74.0k, False: 7.32k] ------------------ 732| | // we check that the read is allowed 733| 74.0k| size_t containersize = 734| 74.0k| BITSET_CONTAINER_SIZE_IN_WORDS * sizeof(uint64_t); 735| 74.0k| *readbytes += containersize; 736| 74.0k| if (*readbytes > maxbytes) { ------------------ | Branch (736:17): [True: 7, False: 73.9k] ------------------ 737| | // Running out of bytes while reading a bitset container. 738| 7| ra_clear(answer); // we need to clear the containers already 739| | // allocated, and the roaring array 740| 7| return false; 741| 7| } 742| | // it is now safe to read 743| 73.9k| bitset_container_t *c = bitset_container_create(); 744| 73.9k| if (c == NULL) { // memory allocation failure ------------------ | Branch (744:17): [True: 0, False: 73.9k] ------------------ 745| | // Failed to allocate memory for a bitset container. 746| 0| ra_clear(answer); // we need to clear the containers already 747| | // allocated, and the roaring array 748| 0| return false; 749| 0| } 750| 73.9k| answer->size++; 751| 73.9k| buf += bitset_container_read(thiscard, c, buf); 752| 73.9k| answer->containers[k] = c; 753| 73.9k| answer->typecodes[k] = BITSET_CONTAINER_TYPE; ------------------ | | 48| 73.9k|#define BITSET_CONTAINER_TYPE 1 ------------------ 754| 73.9k| } else if (isrun) { ------------------ | Branch (754:20): [True: 66, False: 7.26k] ------------------ 755| | // we check that the read is allowed 756| 66| *readbytes += sizeof(uint16_t); 757| 66| if (*readbytes > maxbytes) { ------------------ | Branch (757:17): [True: 8, False: 58] ------------------ 758| | // Running out of bytes while reading a run container (header). 759| 8| ra_clear(answer); // we need to clear the containers already 760| | // allocated, and the roaring array 761| 8| return false; 762| 8| } 763| 58| uint16_t n_runs; 764| 58| memcpy(&n_runs, buf, sizeof(uint16_t)); 765| 58| n_runs = croaring_letoh16(n_runs); ------------------ | | 523| 58|#define croaring_letoh16(x) croaring_htole16(x) | | ------------------ | | | | 517| 58|#define croaring_htole16(x) (x) | | ------------------ ------------------ 766| 58| size_t containersize = n_runs * sizeof(rle16_t); 767| 58| *readbytes += containersize; 768| 58| if (*readbytes > maxbytes) { // data is corrupted? ------------------ | Branch (768:17): [True: 8, False: 50] ------------------ 769| | // Running out of bytes while reading a run container. 770| 8| ra_clear(answer); // we need to clear the containers already 771| | // allocated, and the roaring array 772| 8| return false; 773| 8| } 774| | // it is now safe to read 775| | 776| 50| run_container_t *c = run_container_create(); 777| 50| if (c == NULL) { // memory allocation failure ------------------ | Branch (777:17): [True: 0, False: 50] ------------------ 778| | // Failed to allocate memory for a run container. 779| 0| ra_clear(answer); // we need to clear the containers already 780| | // allocated, and the roaring array 781| 0| return false; 782| 0| } 783| 50| answer->size++; 784| 50| buf += run_container_read(thiscard, c, buf); 785| 50| answer->containers[k] = c; 786| 50| answer->typecodes[k] = RUN_CONTAINER_TYPE; ------------------ | | 50| 50|#define RUN_CONTAINER_TYPE 3 ------------------ 787| 7.26k| } else { 788| | // we check that the read is allowed 789| 7.26k| size_t containersize = thiscard * sizeof(uint16_t); 790| 7.26k| *readbytes += containersize; 791| 7.26k| if (*readbytes > maxbytes) { // data is corrupted? ------------------ | Branch (791:17): [True: 9, False: 7.25k] ------------------ 792| | // Running out of bytes while reading an array container. 793| 9| ra_clear(answer); // we need to clear the containers already 794| | // allocated, and the roaring array 795| 9| return false; 796| 9| } 797| | // it is now safe to read 798| 7.25k| array_container_t *c = 799| 7.25k| array_container_create_given_capacity(thiscard); 800| 7.25k| if (c == NULL) { // memory allocation failure ------------------ | Branch (800:17): [True: 0, False: 7.25k] ------------------ 801| | // Failed to allocate memory for an array container. 802| 0| ra_clear(answer); // we need to clear the containers already 803| | // allocated, and the roaring array 804| 0| return false; 805| 0| } 806| 7.25k| answer->size++; 807| 7.25k| buf += array_container_read(thiscard, c, buf); 808| 7.25k| answer->containers[k] = c; 809| 7.25k| answer->typecodes[k] = ARRAY_CONTAINER_TYPE; ------------------ | | 49| 7.25k|#define ARRAY_CONTAINER_TYPE 2 ------------------ 810| 7.25k| } 811| 81.3k| } 812| 6.49k| return true; 813| 6.52k|} roaring_array.c:realloc_array: 42| 102k|static bool realloc_array(roaring_array_t *ra, int32_t new_capacity) { 43| | // 44| | // Note: not implemented using C's realloc(), because the memory layout is 45| | // Struct-of-Arrays vs. Array-of-Structs: 46| | // https://github.com/RoaringBitmap/CRoaring/issues/256 47| | 48| 102k| if (new_capacity == 0) { ------------------ | Branch (48:9): [True: 64.8k, False: 37.9k] ------------------ 49| 64.8k| roaring_free(ra->containers); 50| 64.8k| ra->containers = NULL; 51| 64.8k| ra->keys = NULL; 52| 64.8k| ra->typecodes = NULL; 53| 64.8k| ra->allocation_size = 0; 54| 64.8k| return true; 55| 64.8k| } 56| 37.9k| const size_t memoryneeded = 57| 37.9k| new_capacity * 58| 37.9k| (sizeof(uint16_t) + sizeof(container_t *) + sizeof(uint8_t)); 59| 37.9k| void *bigalloc = roaring_malloc(memoryneeded); 60| 37.9k| if (!bigalloc) return false; ------------------ | Branch (60:9): [True: 0, False: 37.9k] ------------------ 61| 37.9k| void *oldbigalloc = ra->containers; 62| 37.9k| container_t **newcontainers = (container_t **)bigalloc; 63| 37.9k| uint16_t *newkeys = (uint16_t *)(newcontainers + new_capacity); 64| 37.9k| uint8_t *newtypecodes = (uint8_t *)(newkeys + new_capacity); 65| 37.9k| assert((char *)(newtypecodes + new_capacity) == 66| 37.9k| (char *)bigalloc + memoryneeded); 67| 37.9k| if (ra->size > 0) { ------------------ | Branch (67:9): [True: 12.0k, False: 25.9k] ------------------ 68| 12.0k| memcpy(newcontainers, ra->containers, sizeof(container_t *) * ra->size); 69| 12.0k| memcpy(newkeys, ra->keys, sizeof(uint16_t) * ra->size); 70| 12.0k| memcpy(newtypecodes, ra->typecodes, sizeof(uint8_t) * ra->size); 71| 12.0k| } 72| 37.9k| ra->containers = newcontainers; 73| 37.9k| ra->keys = newkeys; 74| 37.9k| ra->typecodes = newtypecodes; 75| 37.9k| ra->allocation_size = new_capacity; 76| 37.9k| roaring_free(oldbigalloc); 77| | return true; 78| 37.9k|} _Z17ConsumeVecInRangeR18FuzzedDataProvidermjj: 24| 12.9k| uint32_t max_value) { 25| 12.9k| std::vector result = {0}; 26| 12.9k| result.resize(length); 27| 12.9k| std::generate(result.begin(), result.end(), [&]() { 28| 12.9k| return fdp.ConsumeIntegralInRange(min_value, max_value); 29| 12.9k| }); 30| 12.9k| return result; 31| 12.9k|} LLVMFuzzerTestOneInput: 33| 6.48k|extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) { 34| | /** 35| | * A bitmap may contain up to 2**32 elements. Later this function will 36| | * output the content to an array where each element uses 32 bits of 37| | * storage. That would use 16 GB. Thus this function is bound to run out of 38| | * memory. 39| | * 40| | * Even without the full serialization to a 32-bit array, a bitmap may still 41| | * use over 512 MB in the normal course of operation: that is to be expected 42| | * since it can represent all sets of integers in [0,2**32]. This function 43| | * may hold several bitmaps in memory at once, so it can require gigabytes 44| | * of memory (without bugs). Hence, unless it has a generous memory 45| | * capacity, this function will run out of memory almost certainly. 46| | * 47| | * For sanity, we may limit the range to, say, 10,000,000 which will use 38 48| | * MB or so. With such a limited range, if we run out of memory, then we can 49| | * almost certain that it has to do with a genuine bug. 50| | */ 51| | 52| 6.48k| uint32_t range_start = 0; 53| 6.48k| uint32_t range_end = 10'000'000; 54| | 55| | /** 56| | * We are not solely dependent on the range [range_start, range_end) because 57| | * ConsumeVecInRange below produce integers in a small range starting at 0. 58| | */ 59| | 60| 6.48k| FuzzedDataProvider fdp(data, size); 61| | /** 62| | * The next line was ConsumeVecInRange(fdp, 500, 0, 1000) but it would pick 63| | * 500 values at random from 0, 1000, making almost certain that all of the 64| | * values are picked. It seems more useful to pick 500 values in the range 65| | * 0,1000. 66| | */ 67| 6.48k| std::vector bitmap_data_a = ConsumeVecInRange(fdp, 500, 0, 1000); 68| 6.48k| roaring::Roaring a(bitmap_data_a.size(), bitmap_data_a.data()); 69| 6.48k| a.runOptimize(); 70| 6.48k| a.shrinkToFit(); 71| | 72| 6.48k| std::vector bitmap_data_b = ConsumeVecInRange(fdp, 500, 0, 1000); 73| 6.48k| roaring::Roaring b(bitmap_data_b.size(), bitmap_data_b.data()); 74| 6.48k| b.runOptimize(); 75| 6.48k| b.add(fdp.ConsumeIntegralInRange(range_start, range_end)); 76| 6.48k| b.addChecked(fdp.ConsumeIntegralInRange(range_start, range_end)); 77| 6.48k| b.addRange(fdp.ConsumeIntegralInRange(range_start, range_end), 78| 6.48k| fdp.ConsumeIntegralInRange(range_start, range_end)); 79| | // add half of a to b. 80| 6.48k| b.addMany(bitmap_data_a.size() / 2, bitmap_data_a.data()); 81| 6.48k| b.remove(fdp.ConsumeIntegralInRange(range_start, range_end)); 82| 6.48k| b.removeChecked( 83| 6.48k| fdp.ConsumeIntegralInRange(range_start, range_end)); 84| 6.48k| b.removeRange(fdp.ConsumeIntegralInRange(range_start, range_end), 85| 6.48k| fdp.ConsumeIntegralInRange(range_start, range_end)); 86| 6.48k| b.removeRangeClosed( 87| 6.48k| fdp.ConsumeIntegralInRange(range_start, range_end), 88| 6.48k| fdp.ConsumeIntegralInRange(range_start, range_end)); 89| 6.48k| b.maximum(); 90| 6.48k| b.minimum(); 91| 6.48k| b.contains(fdp.ConsumeIntegralInRange(range_start, range_end)); 92| 6.48k| b.containsRange( 93| 6.48k| fdp.ConsumeIntegralInRange(range_start, range_end), 94| 6.48k| fdp.ConsumeIntegralInRange(range_start, range_end)); 95| | 96| 6.48k| uint32_t element = 0; 97| 6.48k| a.select(fdp.ConsumeIntegralInRange(0, 1000), &element); 98| 6.48k| a.intersect(b); 99| 6.48k| a.jaccard_index(b); 100| 6.48k| a.or_cardinality(b); 101| 6.48k| a.andnot_cardinality(b); 102| 6.48k| a.xor_cardinality(b); 103| 6.48k| a.rank(fdp.ConsumeIntegralInRange(0, 5000)); 104| 6.48k| a.getSizeInBytes(); 105| | 106| 6.48k| roaring::Roaring c = a & b; 107| 6.48k| roaring::Roaring d = a - b; 108| 6.48k| roaring::Roaring e = a | b; 109| 6.48k| roaring::Roaring f = a ^ b; 110| 6.48k| a |= e; 111| 6.48k| a &= b; 112| 6.48k| a -= c; 113| 6.48k| a ^= f; 114| | 115| 6.48k| volatile bool is_equal = (a == b); 116| | 117| 6.48k| std::vector b_as_array = {0}; 118| 6.48k| b_as_array.resize(b.cardinality()); 119| 6.48k| b.isEmpty(); 120| 6.48k| b.toUint32Array(b_as_array.data()); 121| | 122| 6.48k| a.isSubset(b); 123| 6.48k| a.isStrictSubset(b); 124| 6.48k| b.flip(fdp.ConsumeIntegralInRange(range_start, range_end), 125| 6.48k| fdp.ConsumeIntegralInRange(range_start, range_end)); 126| 6.48k| b.flipClosed(fdp.ConsumeIntegralInRange(range_start, range_end), 127| 6.48k| fdp.ConsumeIntegralInRange(range_start, range_end)); 128| 6.48k| b.removeRunCompression(); 129| | 130| | // Move/copy constructors 131| 6.48k| roaring::Roaring copied = b; 132| 6.48k| roaring::Roaring moved = std::move(b); 133| | 134| | // Asignment operators 135| 6.48k| b = copied; 136| 6.48k| b = std::move(moved); 137| | 138| | // Safe read from serialized 139| 6.48k| std::vector read_buffer = fdp.ConsumeBytes(100); 140| 6.48k| try { 141| 6.48k| roaring::Roaring read_safely = 142| 6.48k| roaring::Roaring::readSafe(read_buffer.data(), read_buffer.size()); 143| | // The above is guaranteed to be safe. However, read_safely is maybe 144| | // in an improper state and it cannot be used safely (including for 145| | // reserialization). 146| 6.48k| } catch (...) { 147| 6.46k| } 148| | 149| | // The bitmap b can be serialized and re-read. 150| 6.48k| std::size_t expected_size_in_bytes = b.getSizeInBytes(); 151| 6.48k| std::vector buffer(expected_size_in_bytes); 152| 6.48k| std::size_t size_in_bytes = b.write(buffer.data()); 153| 6.48k| assert(expected_size_in_bytes == size_in_bytes); ------------------ | Branch (153:5): [True: 6.48k, False: 0] ------------------ 154| 6.48k| roaring::Roaring bread = 155| 6.48k| roaring::Roaring::readSafe(buffer.data(), size_in_bytes); 156| 6.48k| assert(bread == b); ------------------ | Branch (156:5): [True: 6.48k, False: 0] ------------------ 157| | 158| 6.48k| f.toString(); 159| | 160| 6.48k| volatile int unused = 0; 161| | 162| 342k| for (roaring::Roaring::const_iterator i = a.begin(); i != a.end(); i++) { ------------------ | Branch (162:58): [True: 335k, False: 6.48k] ------------------ 163| 335k| unused++; 164| 335k| } 165| | 166| 6.48k| roaring::Roaring::const_iterator b_iter = b.begin(); 167| 6.48k| b_iter.equalorlarger( 168| 6.48k| fdp.ConsumeIntegralInRange(range_start, range_end)); 169| | 170| 6.48k| return 0; 171| 6.48k|} croaring_fuzzer_cc.cc:_ZZ17ConsumeVecInRangeR18FuzzedDataProvidermjjENK3$_0clEv: 27| 6.48M| std::generate(result.begin(), result.end(), [&]() { 28| 6.48M| return fdp.ConsumeIntegralInRange(min_value, max_value); 29| 6.48M| });