| 1 | #ifndef SIMDJSON_WESTMERE_BITMANIPULATION_H |
| 2 | #define SIMDJSON_WESTMERE_BITMANIPULATION_H |
| 3 | |
| 4 | #include "simdjson/common_defs.h" |
| 5 | #include "simdjson/portability.h" |
| 6 | #ifdef IS_X86_64 |
| 7 | #include "westmere/intrinsics.h" |
| 8 | |
| 9 | TARGET_WESTMERE |
| 10 | namespace simdjson::westmere { |
| 11 | |
| 12 | #ifndef _MSC_VER |
| 13 | // We sometimes call trailing_zero on inputs that are zero, |
| 14 | // but the algorithms do not end up using the returned value. |
| 15 | // Sadly, sanitizers are not smart enough to figure it out. |
| 16 | __attribute__((no_sanitize("undefined" ))) // this is deliberate |
| 17 | #endif |
| 18 | /* result might be undefined when input_num is zero */ |
| 19 | really_inline int trailing_zeroes(uint64_t input_num) { |
| 20 | #ifdef _MSC_VER |
| 21 | unsigned long ret; |
| 22 | // Search the mask data from least significant bit (LSB) |
| 23 | // to the most significant bit (MSB) for a set bit (1). |
| 24 | _BitScanForward64(&ret, input_num); |
| 25 | return (int)ret; |
| 26 | #else |
| 27 | return __builtin_ctzll(input_num); |
| 28 | #endif// _MSC_VER |
| 29 | } |
| 30 | |
| 31 | /* result might be undefined when input_num is zero */ |
| 32 | really_inline uint64_t clear_lowest_bit(uint64_t input_num) { |
| 33 | return input_num & (input_num-1); |
| 34 | } |
| 35 | |
| 36 | /* result might be undefined when input_num is zero */ |
| 37 | really_inline int leading_zeroes(uint64_t input_num) { |
| 38 | #ifdef _MSC_VER |
| 39 | unsigned long leading_zero = 0; |
| 40 | // Search the mask data from most significant bit (MSB) |
| 41 | // to least significant bit (LSB) for a set bit (1). |
| 42 | if (_BitScanReverse64(&leading_zero, input_num)) |
| 43 | return (int)(63 - leading_zero); |
| 44 | else |
| 45 | return 64; |
| 46 | #else |
| 47 | return __builtin_clzll(input_num); |
| 48 | #endif// _MSC_VER |
| 49 | } |
| 50 | |
| 51 | really_inline int hamming(uint64_t input_num) { |
| 52 | #ifdef _MSC_VER |
| 53 | // note: we do not support legacy 32-bit Windows |
| 54 | return __popcnt64(input_num);// Visual Studio wants two underscores |
| 55 | #else |
| 56 | return _popcnt64(input_num); |
| 57 | #endif |
| 58 | } |
| 59 | |
| 60 | really_inline bool add_overflow(uint64_t value1, uint64_t value2, |
| 61 | uint64_t *result) { |
| 62 | #ifdef _MSC_VER |
| 63 | return _addcarry_u64(0, value1, value2, |
| 64 | reinterpret_cast<unsigned __int64 *>(result)); |
| 65 | #else |
| 66 | return __builtin_uaddll_overflow(value1, value2, |
| 67 | (unsigned long long *)result); |
| 68 | #endif |
| 69 | } |
| 70 | |
| 71 | #ifdef _MSC_VER |
| 72 | #pragma intrinsic(_umul128) |
| 73 | #endif |
| 74 | really_inline bool mul_overflow(uint64_t value1, uint64_t value2, |
| 75 | uint64_t *result) { |
| 76 | #ifdef _MSC_VER |
| 77 | uint64_t high; |
| 78 | *result = _umul128(value1, value2, &high); |
| 79 | return high; |
| 80 | #else |
| 81 | return __builtin_umulll_overflow(value1, value2, |
| 82 | (unsigned long long *)result); |
| 83 | #endif |
| 84 | } |
| 85 | |
| 86 | }// namespace simdjson::westmere |
| 87 | UNTARGET_REGION |
| 88 | |
| 89 | #endif |
| 90 | #endif // SIMDJSON_WESTMERE_BITMANIPULATION_H |
| 91 | |