Skip to content

Commit

Permalink
make FP formatting available to be used at compile-time (#2426)
Browse files Browse the repository at this point in the history 
* works only with FMT_HEADER_ONLY
* works only with float and double types (not long double)
  • Loading branch information
@alexezeder
alexezeder authored Sep 18, 2021
1 parent d9fd695 commit b4d9d82
Show file tree
Hide file tree
Showing 6 changed files with 278 additions and 109 deletions.
4 changes: 0 additions & 4 deletions include/fmt/chrono.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -864,10 +864,6 @@ template <typename T, FMT_ENABLE_IF(std::is_integral<T>::value)>
inline bool isfinite(T) {
return true;
}
template <typename T, FMT_ENABLE_IF(std::is_floating_point<T>::value)>
inline bool isfinite(T value) {
return std::isfinite(value);
}

// Converts value to int and checks that it's in the range [0, upper).
template <typename T, FMT_ENABLE_IF(std::is_integral<T>::value)>
Expand Down
4 changes: 2 additions & 2 deletions include/fmt/core.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1178,8 +1178,8 @@ template <typename Context> class value {
constexpr FMT_INLINE value(unsigned long long val) : ulong_long_value(val) {}
FMT_INLINE value(int128_t val) : int128_value(val) {}
FMT_INLINE value(uint128_t val) : uint128_value(val) {}
FMT_INLINE value(float val) : float_value(val) {}
FMT_INLINE value(double val) : double_value(val) {}
constexpr FMT_INLINE value(float val) : float_value(val) {}
constexpr FMT_INLINE value(double val) : double_value(val) {}
FMT_INLINE value(long double val) : long_double_value(val) {}
constexpr FMT_INLINE value(bool val) : bool_value(val) {}
constexpr FMT_INLINE value(char_type val) : char_value(val) {}
Expand Down
149 changes: 85 additions & 64 deletions include/fmt/format-inl.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -152,6 +152,75 @@ template <> FMT_FUNC int count_digits<4>(detail::fallback_uintptr n) {
// log10(2) = 0x0.4d104d427de7fbcc...
static constexpr uint64_t log10_2_significand = 0x4d104d427de7fbcc;

template <typename T = void> struct basic_impl_data {
// Normalized 64-bit significands of pow(10, k), for k = -348, -340, ..., 340.
// These are generated by support/compute-powers.py.
static constexpr uint64_t pow10_significands[87] = {
0xfa8fd5a0081c0288, 0xbaaee17fa23ebf76, 0x8b16fb203055ac76,
0xcf42894a5dce35ea, 0x9a6bb0aa55653b2d, 0xe61acf033d1a45df,
0xab70fe17c79ac6ca, 0xff77b1fcbebcdc4f, 0xbe5691ef416bd60c,
0x8dd01fad907ffc3c, 0xd3515c2831559a83, 0x9d71ac8fada6c9b5,
0xea9c227723ee8bcb, 0xaecc49914078536d, 0x823c12795db6ce57,
0xc21094364dfb5637, 0x9096ea6f3848984f, 0xd77485cb25823ac7,
0xa086cfcd97bf97f4, 0xef340a98172aace5, 0xb23867fb2a35b28e,
0x84c8d4dfd2c63f3b, 0xc5dd44271ad3cdba, 0x936b9fcebb25c996,
0xdbac6c247d62a584, 0xa3ab66580d5fdaf6, 0xf3e2f893dec3f126,
0xb5b5ada8aaff80b8, 0x87625f056c7c4a8b, 0xc9bcff6034c13053,
0x964e858c91ba2655, 0xdff9772470297ebd, 0xa6dfbd9fb8e5b88f,
0xf8a95fcf88747d94, 0xb94470938fa89bcf, 0x8a08f0f8bf0f156b,
0xcdb02555653131b6, 0x993fe2c6d07b7fac, 0xe45c10c42a2b3b06,
0xaa242499697392d3, 0xfd87b5f28300ca0e, 0xbce5086492111aeb,
0x8cbccc096f5088cc, 0xd1b71758e219652c, 0x9c40000000000000,
0xe8d4a51000000000, 0xad78ebc5ac620000, 0x813f3978f8940984,
0xc097ce7bc90715b3, 0x8f7e32ce7bea5c70, 0xd5d238a4abe98068,
0x9f4f2726179a2245, 0xed63a231d4c4fb27, 0xb0de65388cc8ada8,
0x83c7088e1aab65db, 0xc45d1df942711d9a, 0x924d692ca61be758,
0xda01ee641a708dea, 0xa26da3999aef774a, 0xf209787bb47d6b85,
0xb454e4a179dd1877, 0x865b86925b9bc5c2, 0xc83553c5c8965d3d,
0x952ab45cfa97a0b3, 0xde469fbd99a05fe3, 0xa59bc234db398c25,
0xf6c69a72a3989f5c, 0xb7dcbf5354e9bece, 0x88fcf317f22241e2,
0xcc20ce9bd35c78a5, 0x98165af37b2153df, 0xe2a0b5dc971f303a,
0xa8d9d1535ce3b396, 0xfb9b7cd9a4a7443c, 0xbb764c4ca7a44410,
0x8bab8eefb6409c1a, 0xd01fef10a657842c, 0x9b10a4e5e9913129,
0xe7109bfba19c0c9d, 0xac2820d9623bf429, 0x80444b5e7aa7cf85,
0xbf21e44003acdd2d, 0x8e679c2f5e44ff8f, 0xd433179d9c8cb841,
0x9e19db92b4e31ba9, 0xeb96bf6ebadf77d9, 0xaf87023b9bf0ee6b,
};

#if FMT_GCC_VERSION && FMT_GCC_VERSION < 409
FMT_GCC_PRAGMA("GCC diagnostic push")
FMT_GCC_PRAGMA("GCC diagnostic ignored \"-Wnarrowing\"")
#endif
// Binary exponents of pow(10, k), for k = -348, -340, ..., 340, corresponding
// to significands above.
static constexpr int16_t pow10_exponents[87] = {
-1220, -1193, -1166, -1140, -1113, -1087, -1060, -1034, -1007, -980, -954,
-927, -901, -874, -847, -821, -794, -768, -741, -715, -688, -661,
-635, -608, -582, -555, -529, -502, -475, -449, -422, -396, -369,
-343, -316, -289, -263, -236, -210, -183, -157, -130, -103, -77,
-50, -24, 3, 30, 56, 83, 109, 136, 162, 189, 216,
242, 269, 295, 322, 348, 375, 402, 428, 455, 481, 508,
534, 561, 588, 614, 641, 667, 694, 720, 747, 774, 800,
827, 853, 880, 907, 933, 960, 986, 1013, 1039, 1066};
#if FMT_GCC_VERSION && FMT_GCC_VERSION < 409
FMT_GCC_PRAGMA("GCC diagnostic pop")
#endif

static constexpr uint64_t power_of_10_64[20] = {
1, FMT_POWERS_OF_10(1ULL), FMT_POWERS_OF_10(1000000000ULL),
10000000000000000000ULL};
};

// This is a struct rather than an alias to avoid shadowing warnings in gcc.
struct impl_data : basic_impl_data<> {};

#if __cplusplus < 201703L
template <typename T>
constexpr uint64_t basic_impl_data<T>::pow10_significands[];
template <typename T> constexpr int16_t basic_impl_data<T>::pow10_exponents[];
template <typename T> constexpr uint64_t basic_impl_data<T>::power_of_10_64[];
#endif

template <typename T> struct bits {
static FMT_CONSTEXPR_DECL const int value =
static_cast<int>(sizeof(T) * std::numeric_limits<unsigned char>::digits);
Expand Down Expand Up @@ -275,53 +344,8 @@ FMT_CONSTEXPR inline fp operator*(fp x, fp y) {

// Returns a cached power of 10 `c_k = c_k.f * pow(2, c_k.e)` such that its
// (binary) exponent satisfies `min_exponent <= c_k.e <= min_exponent + 28`.
inline fp get_cached_power(int min_exponent, int& pow10_exponent) {
// Normalized 64-bit significands of pow(10, k), for k = -348, -340, ..., 340.
// These are generated by support/compute-powers.py.
static constexpr const uint64_t pow10_significands[] = {
0xfa8fd5a0081c0288, 0xbaaee17fa23ebf76, 0x8b16fb203055ac76,
0xcf42894a5dce35ea, 0x9a6bb0aa55653b2d, 0xe61acf033d1a45df,
0xab70fe17c79ac6ca, 0xff77b1fcbebcdc4f, 0xbe5691ef416bd60c,
0x8dd01fad907ffc3c, 0xd3515c2831559a83, 0x9d71ac8fada6c9b5,
0xea9c227723ee8bcb, 0xaecc49914078536d, 0x823c12795db6ce57,
0xc21094364dfb5637, 0x9096ea6f3848984f, 0xd77485cb25823ac7,
0xa086cfcd97bf97f4, 0xef340a98172aace5, 0xb23867fb2a35b28e,
0x84c8d4dfd2c63f3b, 0xc5dd44271ad3cdba, 0x936b9fcebb25c996,
0xdbac6c247d62a584, 0xa3ab66580d5fdaf6, 0xf3e2f893dec3f126,
0xb5b5ada8aaff80b8, 0x87625f056c7c4a8b, 0xc9bcff6034c13053,
0x964e858c91ba2655, 0xdff9772470297ebd, 0xa6dfbd9fb8e5b88f,
0xf8a95fcf88747d94, 0xb94470938fa89bcf, 0x8a08f0f8bf0f156b,
0xcdb02555653131b6, 0x993fe2c6d07b7fac, 0xe45c10c42a2b3b06,
0xaa242499697392d3, 0xfd87b5f28300ca0e, 0xbce5086492111aeb,
0x8cbccc096f5088cc, 0xd1b71758e219652c, 0x9c40000000000000,
0xe8d4a51000000000, 0xad78ebc5ac620000, 0x813f3978f8940984,
0xc097ce7bc90715b3, 0x8f7e32ce7bea5c70, 0xd5d238a4abe98068,
0x9f4f2726179a2245, 0xed63a231d4c4fb27, 0xb0de65388cc8ada8,
0x83c7088e1aab65db, 0xc45d1df942711d9a, 0x924d692ca61be758,
0xda01ee641a708dea, 0xa26da3999aef774a, 0xf209787bb47d6b85,
0xb454e4a179dd1877, 0x865b86925b9bc5c2, 0xc83553c5c8965d3d,
0x952ab45cfa97a0b3, 0xde469fbd99a05fe3, 0xa59bc234db398c25,
0xf6c69a72a3989f5c, 0xb7dcbf5354e9bece, 0x88fcf317f22241e2,
0xcc20ce9bd35c78a5, 0x98165af37b2153df, 0xe2a0b5dc971f303a,
0xa8d9d1535ce3b396, 0xfb9b7cd9a4a7443c, 0xbb764c4ca7a44410,
0x8bab8eefb6409c1a, 0xd01fef10a657842c, 0x9b10a4e5e9913129,
0xe7109bfba19c0c9d, 0xac2820d9623bf429, 0x80444b5e7aa7cf85,
0xbf21e44003acdd2d, 0x8e679c2f5e44ff8f, 0xd433179d9c8cb841,
0x9e19db92b4e31ba9, 0xeb96bf6ebadf77d9, 0xaf87023b9bf0ee6b,
};

// Binary exponents of pow(10, k), for k = -348, -340, ..., 340, corresponding
// to significands above.
static constexpr const int16_t pow10_exponents[] = {
-1220, -1193, -1166, -1140, -1113, -1087, -1060, -1034, -1007, -980, -954,
-927, -901, -874, -847, -821, -794, -768, -741, -715, -688, -661,
-635, -608, -582, -555, -529, -502, -475, -449, -422, -396, -369,
-343, -316, -289, -263, -236, -210, -183, -157, -130, -103, -77,
-50, -24, 3, 30, 56, 83, 109, 136, 162, 189, 216,
242, 269, 295, 322, 348, 375, 402, 428, 455, 481, 508,
534, 561, 588, 614, 641, 667, 694, 720, 747, 774, 800,
827, 853, 880, 907, 933, 960, 986, 1013, 1039, 1066};

FMT_CONSTEXPR inline fp get_cached_power(int min_exponent,
int& pow10_exponent) {
const int shift = 32;
const auto significand = static_cast<int64_t>(log10_2_significand);
int index = static_cast<int>(
Expand All @@ -335,7 +359,8 @@ inline fp get_cached_power(int min_exponent, int& pow10_exponent) {
const int dec_exp_step = 8;
index = (index - first_dec_exp - 1) / dec_exp_step + 1;
pow10_exponent = first_dec_exp + index * dec_exp_step;
return {pow10_significands[index], pow10_exponents[index]};
return {impl_data::pow10_significands[index],
impl_data::pow10_exponents[index]};
}

// A simple accumulator to hold the sums of terms in bigint::square if uint128_t
Expand Down Expand Up @@ -636,13 +661,6 @@ enum result {
};
}

inline uint64_t power_of_10_64(int exp) {
static constexpr const uint64_t data[] = {1, FMT_POWERS_OF_10(1),
FMT_POWERS_OF_10(1000000000ULL),
10000000000000000000ULL};
return data[exp];
}

// Generates output using the Grisu digit-gen algorithm.
// error: the size of the region (lower, upper) outside of which numbers
// definitely do not round to value (Delta in Grisu3).
Expand All @@ -662,7 +680,7 @@ FMT_INLINE FMT_CONSTEXPR digits::result grisu_gen_digits(fp value,
uint64_t fractional = value.f & (one.f - 1);
exp = count_digits(integral); // kappa in Grisu.
// Divide by 10 to prevent overflow.
auto result = handler.on_start(power_of_10_64(exp - 1) << -one.e,
auto result = handler.on_start(impl_data::power_of_10_64[exp - 1] << -one.e,
value.f / 10, error * 10, exp);
if (result != digits::more) return result;
// Generate digits for the integral part. This can produce up to 10 digits.
Expand Down Expand Up @@ -712,8 +730,8 @@ FMT_INLINE FMT_CONSTEXPR digits::result grisu_gen_digits(fp value,
--exp;
auto remainder = (static_cast<uint64_t>(integral) << -one.e) + fractional;
result = handler.on_digit(static_cast<char>('0' + digit),
power_of_10_64(exp) << -one.e, remainder, error,
exp, true);
impl_data::power_of_10_64[exp] << -one.e,
remainder, error, exp, true);
if (result != digits::more) return result;
} while (exp > 0);
// Generate digits for the fractional part.
Expand Down Expand Up @@ -2225,8 +2243,8 @@ template <typename T> decimal_fp<T> to_decimal(T x) FMT_NOEXCEPT {
// Floating-Point Printout ((FPP)^2) algorithm by Steele & White:
// https://fmt.dev/papers/p372-steele.pdf.
template <typename Double>
void fallback_format(Double d, int num_digits, bool binary32, buffer<char>& buf,
int& exp10) {
FMT_CONSTEXPR20 void fallback_format(Double d, int num_digits, bool binary32,
buffer<char>& buf, int& exp10) {
bigint numerator; // 2 * R in (FPP)^2.
bigint denominator; // 2 * S in (FPP)^2.
// lower and upper are differences between value and corresponding boundaries.
Expand Down Expand Up @@ -2342,7 +2360,9 @@ void fallback_format(Double d, int num_digits, bool binary32, buffer<char>& buf,
}

template <typename T>
int format_float(T value, int precision, float_specs specs, buffer<char>& buf) {
FMT_HEADER_ONLY_CONSTEXPR20 int format_float(T value, int precision,
float_specs specs,
buffer<char>& buf) {
static_assert(!std::is_same<T, float>::value, "");
FMT_ASSERT(value >= 0, "value is negative");

Expand All @@ -2353,13 +2373,13 @@ int format_float(T value, int precision, float_specs specs, buffer<char>& buf) {
return 0;
}
buf.try_resize(to_unsigned(precision));
std::uninitialized_fill_n(buf.data(), precision, '0');
fill_n(buf.data(), precision, '0');
return -precision;
}

if (!specs.use_grisu) return snprintf_float(value, precision, specs, buf);

if (precision < 0) {
if (!is_constant_evaluated() && precision < 0) {
// Use Dragonbox for the shortest format.
if (specs.binary32) {
auto dec = dragonbox::to_decimal(static_cast<float>(value));
Expand All @@ -2385,7 +2405,8 @@ int format_float(T value, int precision, float_specs specs, buffer<char>& buf) {
const int max_double_digits = 767;
if (precision > max_double_digits) precision = max_double_digits;
fixed_handler handler{buf.data(), 0, precision, -cached_exp10, fixed};
if (grisu_gen_digits(normalized, 1, exp, handler) == digits::error) {
if (grisu_gen_digits(normalized, 1, exp, handler) == digits::error ||
is_constant_evaluated()) {
exp += handler.size - cached_exp10 - 1;
fallback_format(value, handler.precision, specs.binary32, buf, exp);
} else {
Expand Down
Loading

1 comment on commit b4d9d82

@alexezeder
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This commit introduced some overhead for integer formatting. Not sure why exactly, because as far as I can see, there are no changes for integers.

Please sign in to comment.