fmtlib/fmt
1
0
Fork 1
mirror of https://github.com/fmtlib/fmt.git synced 2025-11-28 21:30:00 +01:00
Code Issues Packages Projects Releases Wiki Activity

Support single precision floats in grisu formatting

Browse Source 
Fixes #1336
This commit is contained in:
Orivej Desh
2019-10-12 12:17:27 +00:00
committed by Victor Zverovich
parent 91f7619cc9
commit 599e0aef45
4 changed files with 55 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
include/fmt/format-inl.h
View File

@@ -422,6 +422,18 @@ class fp {
lower.f <<= lower.e - upper.e;
lower.e = upper.e;
}
void compute_float_boundaries(fp& lower, fp& upper) const {
constexpr int min_normal_e = std::numeric_limits<float>::min_exponent -
std::numeric_limits<double>::digits;
significand_type half_ulp = 1 << (std::numeric_limits<double>::digits -
std::numeric_limits<float>::digits - 1);
if (min_normal_e > e) half_ulp <<= min_normal_e - e;
upper = normalize<0>(fp(f + half_ulp, e));
lower = fp(f - (half_ulp >> (f == implicit_bit && e > min_normal_e)), e);
lower.f <<= lower.e - upper.e;
lower.e = upper.e;
}
};
// Returns an fp number representing x - y. Result may not be normalized.
@@ -1046,7 +1058,11 @@ bool grisu_format(Double value, buffer<char>& buf, int precision,
buf.resize(to_unsigned(handler.size));
} else {
fp lower, upper; // w^- and w^+ in the Grisu paper.
fp_value.compute_boundaries(lower, upper);
if ((options & grisu_options::binary32) != 0)
fp_value.compute_float_boundaries(lower, upper);
else
fp_value.compute_boundaries(lower, upper);
// Find a cached power of 10 such that multiplying upper by it will bring
// the exponent in the range [min_exp, -32].
const auto cached_pow = get_cached_power( // \tilde{c}_{-k} in Grisu.