just/Just.PreciseMath
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

add precomputed mathematical constants and specialize the second division residual
.NET Test / .NET tests (push) Successful in 2m23s
Details

Browse Source
This commit is contained in:
just
2026-09-14 21:21:48 +04:00
parent 08036a31d5
commit 4efecbb1bc
11 changed files with 875 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files
1-tests/Just.PreciseMath.Tests/DoubleDoubleFormattingTests.cs
+3 -3
View File
@@ -58,12 +58,12 @@ public class DoubleDoubleFormattingTests
public void FormattingDoesNotRouteThroughDecimalOrRestrictExponentRange()
{
// Review-1 §6 verified the former decimal-based formatter threw for 1e100,
// printed "0" for 1e-100, threw for NaN/infinity, and emitted for PI digits
// printed "0" for 1e-100, threw for NaN/infinity, and emitted for pi digits
// already wrong at binary64 precision. Expected digits are exact references:
// the PI pair rounded to 32 significant digits, ties to even
// the Pi pair rounded to 32 significant digits, ties to even
// (Python Fraction/Decimal at precision 120), and the exact binary64 values of
// the powers of ten.
DoubleDouble.PI.ToString("G32", CultureInfo.InvariantCulture).ShouldBe("3.1415926535897932384626433832795");
DoubleDouble.Pi.ToString("G32", CultureInfo.InvariantCulture).ShouldBe("3.1415926535897932384626433832795");
new DoubleDouble(1e100).ToString("G", CultureInfo.InvariantCulture).ShouldBe("1.0000000000000000159028911097599E+100");
new DoubleDouble(1e-100).ToString("G", CultureInfo.InvariantCulture).ShouldBe("1.0000000000000000199918998026029E-100");
// The binary64 values are exactly representable as sums with a zero residual,