From ea314d319da3eeeeb4c4766b0bd8d9e1f7af578f Mon Sep 17 00:00:00 2001 From: Andersama Date: Fri, 11 Feb 2022 12:42:53 -0800 Subject: [PATCH] patch: tentative hex float parsing --- include/fast_float/ascii_number.h | 370 ++++++++++++++++++------- include/fast_float/decimal_to_binary.h | 102 +++++++ include/fast_float/fast_float.h | 11 +- include/fast_float/float_common.h | 172 ++++++++++++ include/fast_float/parse_number.h | 6 + 5 files changed, 563 insertions(+), 98 deletions(-) diff --git a/include/fast_float/ascii_number.h b/include/fast_float/ascii_number.h index 3e6bb3e9..cb619e70 100644 --- a/include/fast_float/ascii_number.h +++ b/include/fast_float/ascii_number.h @@ -68,6 +68,21 @@ fastfloat_really_inline bool is_made_of_eight_digits_fast(const char *chars) no return is_made_of_eight_digits_fast(read_u64(chars)); } +// returns a value with the high 16 bits set if not valid +// otherwise returns the conversion of the 4 hex digits at src into the bottom +// 16 bits of the 32-bit return register +// +// see +// https://lemire.me/blog/2019/04/17/parsing-short-hexadecimal-strings-efficiently/ +fastfloat_really_inline uint32_t hex_to_u32_nocheck( + const uint8_t* src) noexcept { // strictly speaking, static inline is a C-ism + uint32_t v1 = digit_to_val32[630 + src[0]]; + uint32_t v2 = digit_to_val32[420 + src[1]]; + uint32_t v3 = digit_to_val32[210 + src[2]]; + uint32_t v4 = digit_to_val32[0 + src[3]]; + return v1 | v2 | v3 | v4; +} + typedef span byte_span; struct parsed_number_string { @@ -88,12 +103,13 @@ fastfloat_really_inline parsed_number_string parse_number_string(const char *p, const char *pend, parse_options options) noexcept { const chars_format fmt = options.format; const char decimal_point = options.decimal_point; + const adv_format fmt_adv = options.format_advanced; parsed_number_string answer; answer.valid = false; answer.too_many_digits = false; answer.negative = (*p == '-'); - if (*p == '-') { // C++17 20.19.3.(7.1) explicitly forbids '+' sign here + if (*p == '-' || (*p == '+' && (fmt_adv & adv_format::allow_signed))) { // C++17 20.19.3.(7.1) explicitly forbids '+' sign here ++p; if (p == pend) { return answer; @@ -102,128 +118,290 @@ parsed_number_string parse_number_string(const char *p, const char *pend, parse_ return answer; } } + + if (fmt_adv & adv_format::prefixed) { + if (p + 2 < pend && p[0] == '0') { + // format should be hex + if (p[1] == 'x' || p[1] == 'X') { + p += 2; + } + // octal / binary floats? + else if (p[1] == 'b' || p[1] == 'B') { + p += 2; + } + else if (p[1] >= '0' && p[1] <= '7') { + p += 2; + } + } + } const char *const start_digits = p; uint64_t i = 0; // an unsigned int avoids signed overflows (which are bad) - while ((std::distance(p, pend) >= 8) && is_made_of_eight_digits_fast(p)) { - i = i * 100000000 + parse_eight_digits_unrolled(p); // in rare cases, this will overflow, but that's ok - p += 8; - } - while ((p != pend) && is_integer(*p)) { - // a multiplication by 10 is cheaper than an arbitrary integer - // multiplication - i = 10 * i + - uint64_t(*p - '0'); // might overflow, we will handle the overflow later - ++p; - } - const char *const end_of_integer_part = p; - int64_t digit_count = int64_t(end_of_integer_part - start_digits); - answer.integer = byte_span(start_digits, size_t(digit_count)); - int64_t exponent = 0; - if ((p != pend) && (*p == decimal_point)) { - ++p; - const char* before = p; - // can occur at most twice without overflowing, but let it occur more, since - // for integers with many digits, digit parsing is the primary bottleneck. + if (!(fmt & chars_format::hex)) { while ((std::distance(p, pend) >= 8) && is_made_of_eight_digits_fast(p)) { - i = i * 100000000 + parse_eight_digits_unrolled(p); // in rare cases, this will overflow, but that's ok - p += 8; + i = i * 100000000 + parse_eight_digits_unrolled(p); // in rare cases, this will overflow, but that's ok + p += 8; } while ((p != pend) && is_integer(*p)) { - uint8_t digit = uint8_t(*p - '0'); - ++p; - i = i * 10 + digit; // in rare cases, this will overflow, but that's ok + // a multiplication by 10 is cheaper than an arbitrary integer + // multiplication + i = 10 * i + + uint64_t(*p - '0'); // might overflow, we will handle the overflow later + ++p; } - exponent = before - p; - answer.fraction = byte_span(before, size_t(p - before)); - digit_count -= exponent; - } - // we must have encountered at least one integer! - if (digit_count == 0) { - return answer; - } - int64_t exp_number = 0; // explicit exponential part - if ((fmt & chars_format::scientific) && (p != pend) && (('e' == *p) || ('E' == *p))) { - const char * location_of_e = p; - ++p; - bool neg_exp = false; - if ((p != pend) && ('-' == *p)) { - neg_exp = true; - ++p; - } else if ((p != pend) && ('+' == *p)) { // '+' on exponent is allowed by C++17 20.19.3.(7.1) + + const char *const end_of_integer_part = p; + int64_t digit_count = int64_t(end_of_integer_part - start_digits); + answer.integer = byte_span(start_digits, size_t(digit_count)); + int64_t exponent = 0; + if ((p != pend) && (*p == decimal_point)) { ++p; - } - if ((p == pend) || !is_integer(*p)) { - if(!(fmt & chars_format::fixed)) { - // We are in error. - return answer; + const char* before = p; + // can occur at most twice without overflowing, but let it occur more, since + // for integers with many digits, digit parsing is the primary bottleneck. + while ((std::distance(p, pend) >= 8) && is_made_of_eight_digits_fast(p)) { + i = i * 100000000 + parse_eight_digits_unrolled(p); // in rare cases, this will overflow, but that's ok + p += 8; } - // Otherwise, we will be ignoring the 'e'. - p = location_of_e; - } else { while ((p != pend) && is_integer(*p)) { uint8_t digit = uint8_t(*p - '0'); - if (exp_number < 0x10000000) { - exp_number = 10 * exp_number + digit; - } + ++p; + i = i * 10 + digit; // in rare cases, this will overflow, but that's ok + } + exponent = before - p; + answer.fraction = byte_span(before, size_t(p - before)); + digit_count -= exponent; + } + // we must have encountered at least one integer! + if (digit_count == 0) { + return answer; + } + int64_t exp_number = 0; // explicit exponential part + if ((fmt & chars_format::scientific) && (p != pend) && (('e' == *p) || ('E' == *p))) { + const char * location_of_e = p; + ++p; + bool neg_exp = false; + if ((p != pend) && ('-' == *p)) { + neg_exp = true; + ++p; + } else if ((p != pend) && ('+' == *p)) { // '+' on exponent is allowed by C++17 20.19.3.(7.1) ++p; } - if(neg_exp) { exp_number = - exp_number; } - exponent += exp_number; + if ((p == pend) || !is_integer(*p)) { + if(!(fmt & chars_format::fixed)) { + // We are in error. + return answer; + } + // Otherwise, we will be ignoring the 'e'. + p = location_of_e; + } else { + while ((p != pend) && is_integer(*p)) { + uint8_t digit = uint8_t(*p - '0'); + if (exp_number < 0x10000000) { + exp_number = 10 * exp_number + digit; + } + ++p; + } + if(neg_exp) { exp_number = - exp_number; } + exponent += exp_number; + } + } else { + // If it scientific and not fixed, we have to bail out. + if((fmt & chars_format::scientific) && !(fmt & chars_format::fixed)) { return answer; } } + answer.lastmatch = p; + answer.valid = true; + + // If we frequently had to deal with long strings of digits, + // we could extend our code by using a 128-bit integer instead + // of a 64-bit integer. However, this is uncommon. + // + // We can deal with up to 19 digits. + if (digit_count > 19) { // this is uncommon + // It is possible that the integer had an overflow. + // We have to handle the case where we have 0.0000somenumber. + // We need to be mindful of the case where we only have zeroes... + // E.g., 0.000000000...000. + const char *start = start_digits; + while ((start != pend) && (*start == '0' || *start == decimal_point)) { + if(*start == '0') { digit_count --; } + start++; + } + if (digit_count > 19) { + answer.too_many_digits = true; + // Let us start again, this time, avoiding overflows. + // We don't need to check if is_integer, since we use the + // pre-tokenized spans from above. + i = 0; + p = answer.integer.ptr; + const char* int_end = p + answer.integer.len(); + const uint64_t minimal_nineteen_digit_integer{1000000000000000000}; + while((i < minimal_nineteen_digit_integer) && (p != int_end)) { + i = i * 10 + uint64_t(*p - '0'); + ++p; + } + if (i >= minimal_nineteen_digit_integer) { // We have a big integers + exponent = end_of_integer_part - p + exp_number; + } else { // We have a value with a fractional component. + p = answer.fraction.ptr; + const char* frac_end = p + answer.fraction.len(); + while((i < minimal_nineteen_digit_integer) && (p != frac_end)) { + i = i * 10 + uint64_t(*p - '0'); + ++p; + } + exponent = answer.fraction.ptr - p + exp_number; + } + // We have now corrected both exponent and i, to a truncated value + } + } + answer.exponent = exponent; + answer.mantissa = i; + return answer; } else { - // If it scientific and not fixed, we have to bail out. - if((fmt & chars_format::scientific) && !(fmt & chars_format::fixed)) { return answer; } - } - answer.lastmatch = p; - answer.valid = true; - - // If we frequently had to deal with long strings of digits, - // we could extend our code by using a 128-bit integer instead - // of a 64-bit integer. However, this is uncommon. - // - // We can deal with up to 19 digits. - if (digit_count > 19) { // this is uncommon - // It is possible that the integer had an overflow. - // We have to handle the case where we have 0.0000somenumber. - // We need to be mindful of the case where we only have zeroes... - // E.g., 0.000000000...000. - const char *start = start_digits; - while ((start != pend) && (*start == '0' || *start == decimal_point)) { - if(*start == '0') { digit_count --; } - start++; + while ((std::distance(p, pend) >= 4)) { + uint32_t d = hex_to_u32_nocheck((const uint8_t*)p); + if ((d & 0xffff0000) != 0) { + break; + } + i = (i << 16) + d; // in rare cases, this will overflow, but that's ok + p += 4; + } + + while ((p != pend)) { + uint8_t c = (uint8_t)*p; + uint8_t digit = chars_to_digits[c]; + if (digit >= 16) + break; + i = 16 * i + digit; + ++p; + } + + const char* const end_of_integer_part = p; + int64_t digit_count = int64_t(end_of_integer_part - start_digits); + answer.integer = byte_span(start_digits, size_t(digit_count)); + int64_t exponent = 0; + + if ((p != pend) && (*p == decimal_point)) { + ++p; + const char* before = p; + while ((std::distance(p, pend) >= 4)) { + uint32_t d = hex_to_u32_nocheck((const uint8_t*)p); + if ((d & 0xffff0000) != 0) { + break; + } + i = (i << 16) + d; // in rare cases, this will overflow, but that's ok + p += 4; + } + // can occur at most twice without overflowing, but let it occur more, since + // for integers with many digits, digit parsing is the primary bottleneck. + while ((p != pend)) { + uint8_t c = (uint8_t)*p; + uint8_t digit = chars_to_digits[c]; + if (digit >= 16) + break; + i = 16 * i + digit; + ++p; + } + int64_t exponent_shift = (int64_t)(before - p); + exponent = exponent_shift * 4; + answer.fraction = byte_span(before, size_t(p - before)); + digit_count -= exponent_shift; + } + + // we must have encountered at least one integer! + if (digit_count == 0) { + return answer; } - if (digit_count > 19) { - answer.too_many_digits = true; - // Let us start again, this time, avoiding overflows. - // We don't need to check if is_integer, since we use the - // pre-tokenized spans from above. - i = 0; - p = answer.integer.ptr; - const char* int_end = p + answer.integer.len(); - const uint64_t minimal_nineteen_digit_integer{1000000000000000000}; - while((i < minimal_nineteen_digit_integer) && (p != int_end)) { - i = i * 10 + uint64_t(*p - '0'); + + int64_t exp_number = 0; // explicit exponential part + if ((fmt & chars_format::scientific) && (p != pend) && (('p' == *p) || ('P' == *p))) { + const char * location_of_e = p; + ++p; + bool neg_exp = false; + if ((p != pend) && ('-' == *p)) { + neg_exp = true; + ++p; + } else if ((p != pend) && ('+' == *p)) { // '+' on exponent is allowed by C++17 20.19.3.(7.1) ++p; } - if (i >= minimal_nineteen_digit_integer) { // We have a big integers - exponent = end_of_integer_part - p + exp_number; - } else { // We have a value with a fractional component. + if ((p == pend) || !is_integer(*p)) { + if(!(fmt & chars_format::fixed)) { + // We are in error. + return answer; + } + // Otherwise, we will be ignoring the 'e'. + p = location_of_e; + } else { + while ((p != pend) && is_integer(*p)) { + uint8_t digit = uint8_t(*p - '0'); + if (exp_number < 0x10000000) { + exp_number = 10 * exp_number + digit; + } + ++p; + } + if(neg_exp) { exp_number = - exp_number; } + exponent += exp_number; + } + } else { + // If it scientific and not fixed, we have to bail out. + if((fmt & chars_format::scientific) && !(fmt & chars_format::fixed)) { return answer; } + } + + answer.lastmatch = p; + answer.valid = true; + //52 bits / (4 bits/digit) = 13 + if (digit_count > 13) { // this is uncommon + // It is possible that the integer had an overflow. + // We have to handle the case where we have 0.0000somenumber. + // We need to be mindful of the case where we only have zeroes... + // E.g., 0.000000000...000. + const char *start = start_digits; + while ((start != pend) && (*start == '0' || *start == decimal_point)) { + if(*start == '0') { digit_count --; } + start++; + } + if (digit_count > 13) { + answer.too_many_digits = true; + // Let us start again, this time, avoiding overflows. + // We don't need to check if is_integer, since we use the + // pre-tokenized spans from above. + i = 0; + p = answer.integer.ptr; + const char* int_end = p + answer.integer.len(); + const uint64_t minimal_thirteen_digit_integer{0x1000000000000}; + while((i < minimal_thirteen_digit_integer) && (p != int_end)) { + uint8_t c = (uint8_t)*p; + uint8_t digit = chars_to_digits[c]; + if (digit >= 16) + break; + i = i * 16 + digit; + ++p; + } + if (i >= minimal_thirteen_digit_integer) { // We have a big integers + exponent = end_of_integer_part - p + exp_number; + } else { // We have a value with a fractional component. p = answer.fraction.ptr; const char* frac_end = p + answer.fraction.len(); - while((i < minimal_nineteen_digit_integer) && (p != frac_end)) { - i = i * 10 + uint64_t(*p - '0'); + while((i < minimal_thirteen_digit_integer) && (p != frac_end)) { + uint8_t c = (uint8_t)*p; + uint8_t digit = chars_to_digits[c]; + if (digit >= 16) + break; + i = i * 16 + digit; ++p; } exponent = answer.fraction.ptr - p + exp_number; + } + // We have now corrected both exponent and i, to a truncated value } - // We have now corrected both exponent and i, to a truncated value } + + answer.exponent = exponent; + answer.mantissa = i; + return answer; } - answer.exponent = exponent; - answer.mantissa = i; - return answer; + } } // namespace fast_float diff --git a/include/fast_float/decimal_to_binary.h b/include/fast_float/decimal_to_binary.h index 6da6c66a..2198cdbc 100644 --- a/include/fast_float/decimal_to_binary.h +++ b/include/fast_float/decimal_to_binary.h @@ -189,6 +189,108 @@ adjusted_mantissa compute_float(int64_t q, uint64_t w) noexcept { return answer; } +// w * 2 ** q +template +fastfloat_really_inline +adjusted_mantissa compute_hex_float(int64_t q, uint64_t w) noexcept { + adjusted_mantissa answer; + if (w == 0) { + answer.power2 = 0; + answer.mantissa = 0; + // result should be zero + return answer; + } + if (q >= binary::infinite_power()) { + // we want to get infinity: + answer.power2 = binary::infinite_power(); + answer.mantissa = 0; + return answer; + } + + int lz = leading_zeroes(w); + w <<= lz; + + // The required precision is binary::mantissa_explicit_bits() + 3 because + // 1. We need the implicit bit + // 2. We need an extra bit for rounding purposes + // 3. We might lose a bit due to the "upperbit" routine (result too small, requiring a shift) + + value128 product = { 0x00, w }; + /* + if (product.low == 0xFFFFFFFFFFFFFFFF) { // could guard it further + // In some very rare cases, this could happen, in which case we might need a more accurate + // computation that what we can provide cheaply. This is very, very unlikely. + // + const bool inside_safe_exponent = (q >= -27) && (q <= 55); // always good because 5**q <2**128 when q>=0, + // and otherwise, for q<0, we have 5**-q<2**64 and the 128-bit reciprocal allows for exact computation. + if (!inside_safe_exponent) { + return compute_error_scaled(q, product.high, lz); + } + } + */ + // The "compute_product_approximation" function can be slightly slower than a branchless approach: + // value128 product = compute_product(q, w); + // but in practice, we can win big with the compute_product_approximation if its additional branch + // is easily predicted. Which is best is data specific. + int upperbit = int(product.high >> 63); + + answer.mantissa = product.high >> (upperbit + 64 - binary::mantissa_explicit_bits() - 3); + + answer.power2 = int32_t(q + upperbit - lz - binary::minimum_exponent() + 62); + if (answer.power2 <= 0) { // we have a subnormal? + // Here have that answer.power2 <= 0 so -answer.power2 >= 0 + if (-answer.power2 + 1 >= 64) { // if we have more than 64 bits below the minimum exponent, you have a zero for sure. + answer.power2 = 0; + answer.mantissa = 0; + // result should be zero + return answer; + } + // next line is safe because -answer.power2 + 1 < 64 + answer.mantissa >>= -answer.power2 + 1; + // Thankfully, we can't have both "round-to-even" and subnormals because + // "round-to-even" only occurs for powers close to 0. + answer.mantissa += (answer.mantissa & 1); // round up + answer.mantissa >>= 1; + // There is a weird scenario where we don't have a subnormal but just. + // Suppose we start with 2.2250738585072013e-308, we end up + // with 0x3fffffffffffff x 2^-1023-53 which is technically subnormal + // whereas 0x40000000000000 x 2^-1023-53 is normal. Now, we need to round + // up 0x3fffffffffffff x 2^-1023-53 and once we do, we are no longer + // subnormal, but we can only know this after rounding. + // So we only declare a subnormal if we are smaller than the threshold. + answer.power2 = (answer.mantissa < (uint64_t(1) << + binary::mantissa_explicit_bits())) ? 0 : 1; + return answer; + } + + // usually, we round *up*, but if we fall right in between and and we have an + // even basis, we need to round down + // We are only concerned with the cases where 5**q fits in single 64-bit word. + if ((product.low <= 1) && (q >= binary::min_exponent_round_to_even()) && (q <= binary::max_exponent_round_to_even()) && + ((answer.mantissa & 3) == 1)) { // we may fall between two floats! + // To be in-between two floats we need that in doing + // answer.mantissa = product.high >> (upperbit + 64 - binary::mantissa_explicit_bits() - 3); + // ... we dropped out only zeroes. But if this happened, then we can go back!!! + if ((answer.mantissa << (upperbit + 64 - binary::mantissa_explicit_bits() - 3)) == product.high) { + answer.mantissa &= ~uint64_t(1); // flip it so that we do not round up + } + } + + answer.mantissa += (answer.mantissa & 1); // round up + answer.mantissa >>= 1; + if (answer.mantissa >= (uint64_t(2) << binary::mantissa_explicit_bits())) { + answer.mantissa = (uint64_t(1) << binary::mantissa_explicit_bits()); + answer.power2++; // undo previous addition + } + + answer.mantissa &= ~(uint64_t(1) << binary::mantissa_explicit_bits()); + if (answer.power2 >= binary::infinite_power()) { // infinity + answer.power2 = binary::infinite_power(); + answer.mantissa = 0; + } + return answer; +} + } // namespace fast_float #endif diff --git a/include/fast_float/fast_float.h b/include/fast_float/fast_float.h index 3c483803..6adfb8ac 100644 --- a/include/fast_float/fast_float.h +++ b/include/fast_float/fast_float.h @@ -11,6 +11,11 @@ enum chars_format { general = fixed | scientific }; +enum adv_format { + none, + prefixed = 1 << 0, + allow_signed = 1 << 1 +}; struct from_chars_result { const char *ptr; @@ -19,13 +24,15 @@ struct from_chars_result { struct parse_options { constexpr explicit parse_options(chars_format fmt = chars_format::general, - char dot = '.') - : format(fmt), decimal_point(dot) {} + char dot = '.', adv_format adv_fmt = adv_format::none) + : format(fmt), decimal_point(dot), format_advanced(adv_fmt) {} /** Which number formats are accepted */ chars_format format; /** The character used as decimal point */ char decimal_point; + /** Additional flags */ + adv_format format_advanced; }; /** diff --git a/include/fast_float/float_common.h b/include/fast_float/float_common.h index bb8580ef..efdcd98c 100644 --- a/include/fast_float/float_common.h +++ b/include/fast_float/float_common.h @@ -221,6 +221,178 @@ constexpr static double powers_of_ten_double[] = { constexpr static float powers_of_ten_float[] = {1e0, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9, 1e10}; +constexpr static uint8_t chars_to_digits[] = { + 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, + 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 255, 255, 255, 255, 255, 255, + + 255, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, + 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 255, 255, 255, 255, 255, + 255, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, + 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 255, 255, 255, 255, 255, + + 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, + + 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, +}; + +constexpr const uint32_t digit_to_val32[886] = { + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, + 0x6, 0x7, 0x8, 0x9, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xa, + 0xb, 0xc, 0xd, 0xe, 0xf, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xa, 0xb, 0xc, 0xd, 0xe, + 0xf, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0x0, 0x10, 0x20, 0x30, 0x40, 0x50, + 0x60, 0x70, 0x80, 0x90, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xa0, + 0xb0, 0xc0, 0xd0, 0xe0, 0xf0, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xa0, 0xb0, 0xc0, 0xd0, 0xe0, + 0xf0, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0x0, 0x100, 0x200, 0x300, 0x400, 0x500, + 0x600, 0x700, 0x800, 0x900, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xa00, + 0xb00, 0xc00, 0xd00, 0xe00, 0xf00, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xa00, 0xb00, 0xc00, 0xd00, 0xe00, + 0xf00, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0x0, 0x1000, 0x2000, 0x3000, 0x4000, 0x5000, + 0x6000, 0x7000, 0x8000, 0x9000, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xa000, + 0xb000, 0xc000, 0xd000, 0xe000, 0xf000, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xa000, 0xb000, 0xc000, 0xd000, 0xe000, + 0xf000, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF }; + template struct binary_format { using equiv_uint = typename std::conditional::type; diff --git a/include/fast_float/parse_number.h b/include/fast_float/parse_number.h index 62ae3b03..1c397586 100644 --- a/include/fast_float/parse_number.h +++ b/include/fast_float/parse_number.h @@ -87,6 +87,12 @@ from_chars_result from_chars_advanced(const char *first, const char *last, } answer.ec = std::errc(); // be optimistic answer.ptr = pns.lastmatch; + if (options.format & chars_format::hex) { + adjusted_mantissa am = compute_hex_float>(pns.exponent, pns.mantissa); + to_float(pns.negative, am, value); + return answer; + } + // Next is Clinger's fast path. if (binary_format::min_exponent_fast_path() <= pns.exponent && pns.exponent <= binary_format::max_exponent_fast_path() && pns.mantissa <=binary_format::max_mantissa_fast_path() && !pns.too_many_digits) { value = T(pns.mantissa);