From 5cc3565cc24a614e7ebd1a30ff8f0d38b7eac4d7 Mon Sep 17 00:00:00 2001 From: RealTimeChris <40668522+RealTimeChris@users.noreply.github.com> Date: Tue, 26 Nov 2024 00:58:42 -0500 Subject: [PATCH] Updating to fix some issues. --- include/fast_float/ascii_number.h | 83 +++++++++++-------- include/fast_float/constexpr_feature_detect.h | 6 ++ 2 files changed, 53 insertions(+), 36 deletions(-) diff --git a/include/fast_float/ascii_number.h b/include/fast_float/ascii_number.h index 4ad2340e..759e46f2 100644 --- a/include/fast_float/ascii_number.h +++ b/include/fast_float/ascii_number.h @@ -118,9 +118,9 @@ uint64_t simd_read8_to_u64(UC const *) { // credit @aqrit fastfloat_really_inline FASTFLOAT_CONSTEXPR14 uint32_t parse_eight_digits_unrolled(uint64_t val) { - uint64_t const mask = 0x000000FF000000FF; - uint64_t const mul1 = 0x000F424000000064; // 100 + (1000000ULL << 32) - uint64_t const mul2 = 0x0000271000000001; // 1 + (10000ULL << 32) + constexpr uint64_t mask = 0x000000FF000000FF; + constexpr uint64_t mul1 = 0x000F424000000064; // 100 + (1000000ULL << 32) + constexpr uint64_t mul2 = 0x0000271000000001; // 1 + (10000ULL << 32) val -= 0x3030303030303030; val = (val * 10) + (val >> 8); // val = (val * 2561) >> 8; val = (((val & mask) * mul1) + (((val >> 16) & mask) * mul2)) >> 32; @@ -144,6 +144,21 @@ is_made_of_eight_digits_fast(uint64_t val) noexcept { 0x8080808080808080)); } +fastfloat_really_inline constexpr bool +is_made_of_eight_digits_no_sub(uint64_t val) noexcept { + return !((((val + 0x7676767676767676) | (val)) & 0x8080808080808080)); +} + +fastfloat_really_inline FASTFLOAT_CONSTEXPR14 uint32_t +parse_eight_digits_unrolled_no_sub(uint64_t val) { + constexpr uint64_t mask = 0x000000FF000000FF; + constexpr uint64_t mul1 = 0x000F424000000064; // 100 + (1000000ULL << 32) + constexpr uint64_t mul2 = 0x0000271000000001; // 1 + (10000ULL << 32) + val = (val * 10) + (val >> 8); // val = (val * 2561) >> 8; + val = (((val & mask) * mul1) + (((val >> 16) & mask) * mul2)) >> 32; + return uint32_t(val); +} + #ifdef FASTFLOAT_HAS_SIMD // Call this if chars might not be 8 digits. @@ -209,10 +224,10 @@ bool simd_parse_if_eight_digits_unrolled(UC const *, uint64_t &) { template ::value) = 0> fastfloat_really_inline FASTFLOAT_CONSTEXPR20 void loop_parse_if_eight_digits(UC const *&p, UC const *const pend, uint64_t &i) { - if (!has_simd_opt()) { + FASTFLOAT_IF_CONSTEXPR(!has_simd_opt()) { return; } - while ((std::distance(p, pend) >= 8) && + while (((pend - p) >= 8) && simd_parse_if_eight_digits_unrolled( p, i)) { // in rare cases, this will overflow, but that's ok p += 8; @@ -223,8 +238,7 @@ fastfloat_really_inline FASTFLOAT_CONSTEXPR20 void loop_parse_if_eight_digits(char const *&p, char const *const pend, uint64_t &i) { // optimizes better than parse_if_eight_digits_unrolled() for UC = char. - while ((std::distance(p, pend) >= 8) && - is_made_of_eight_digits_fast(read8_to_u64(p))) { + while (((pend - p) >= 8) && is_made_of_eight_digits_fast(read8_to_u64(p))) { i = i * 100000000 + parse_eight_digits_unrolled(read8_to_u64( p)); // in rare cases, this will overflow, but that's ok @@ -232,36 +246,26 @@ loop_parse_if_eight_digits(char const *&p, char const *const pend, } } -// credit @realtimechris -fastfloat_really_inline constexpr bool -parse_if_eight_digits_unrolled(const char *&string, size_t &value) { - constexpr size_t byte_mask = ~size_t(0) / 255ull; - constexpr size_t msb_mask = byte_mask * 128ull; - constexpr size_t threshold_byte_mask = byte_mask * (127ull - 9ull); - constexpr size_t mask = 0x000000FF000000FFull; - constexpr size_t mul1 = 0x000F424000000064ull; - constexpr size_t mul2 = 0x0000271000000001ull; - size_t value_new = read8_to_u64(string) - 0x3030303030303030; - if (!(((value_new + threshold_byte_mask) | value_new) & msb_mask)) { - value_new = (value_new * 10) + (value_new >> 8); - value = - value * 100000000 + - ((((value_new & mask) * mul1) + (((value_new >> 16) & mask) * mul2)) >> - 32); - string += 8; - return true; +template ::value) = 0> +fastfloat_really_inline FASTFLOAT_CONSTEXPR20 void +loop_parse_if_digits(UC const *&p, UC const *const pend, uint64_t &i) { + FASTFLOAT_IF_CONSTEXPR(!has_simd_opt()) { + return; + } + while (((pend - p) >= 8) && + simd_parse_if_eight_digits_unrolled( + p, i)) { // in rare cases, this will overflow, but that's ok + p += 8; } - return false; } -fastfloat_really_inline constexpr void -loop_parse_if_digits(const char *&p, const char *const pend, - size_t &i) noexcept { - while (pend - p >= 8 && parse_if_eight_digits_unrolled(p, i)) { - } - while (p < pend && is_integer(*p)) { - i = i * 10 + static_cast(*p - '0'); - ++p; +fastfloat_really_inline FASTFLOAT_CONSTEXPR20 void +loop_parse_if_digits(char const *&p, char const *const pend, uint64_t &i) { + uint64_t val; + while ((pend - p) >= 8 && (val = read8_to_u64(p) - 0x3030303030303030, + is_made_of_eight_digits_no_sub(val))) { + i = i * 100000000 + parse_eight_digits_unrolled_no_sub(val); + p += 8; } } @@ -381,6 +385,13 @@ parse_number_string(UC const *p, UC const *pend, // can occur at most twice without overflowing, but let it occur more, since // for integers with many digits, digit parsing is the primary bottleneck. loop_parse_if_digits(p, pend, i); + + while ((p != pend) && is_integer(*p)) { + uint8_t digit = uint8_t(*p - char('0')); + ++p; + i = i * 10 + digit; // in rare cases, this will overflow, but that's ok + } + exponent = before - p; answer.fraction = span(before, size_t(p - before)); digit_count -= exponent; @@ -473,7 +484,7 @@ parse_number_string(UC const *p, UC const *pend, i = 0; p = answer.integer.ptr; UC const *int_end = p + answer.integer.len(); - uint64_t const minimal_nineteen_digit_integer{1000000000000000000}; + constexpr uint64_t minimal_nineteen_digit_integer{1000000000000000000}; while ((i < minimal_nineteen_digit_integer) && (p != int_end)) { i = i * 10 + uint64_t(*p - UC('0')); ++p; @@ -605,4 +616,4 @@ parse_int_string(UC const *p, UC const *pend, T &value, } // namespace fast_float -#endif +#endif \ No newline at end of file diff --git a/include/fast_float/constexpr_feature_detect.h b/include/fast_float/constexpr_feature_detect.h index 7624beaf..d1b3195e 100644 --- a/include/fast_float/constexpr_feature_detect.h +++ b/include/fast_float/constexpr_feature_detect.h @@ -14,6 +14,12 @@ #define FASTFLOAT_CONSTEXPR14 #endif +#if __cplusplus >= 201703L +#define FASTFLOAT_IF_CONSTEXPR if constexpr +#else +#define FASTFLOAT_IF_CONSTEXPR if +#endif + #if defined(__cpp_lib_bit_cast) && __cpp_lib_bit_cast >= 201806L #define FASTFLOAT_HAS_BIT_CAST 1 #else