Fix a bug in LdExp(extended), and test Math.Frexp/Ldexp implementations instead of own.
This fixes a bug (rambling details): frld2.patch. My CPU does not care, but that StackOverflow user said he was getting NaNs when feeding the processor with “undefined” extended
s.
This makes the test test what it is supposed to test (Math.Frexp/Ldexp
) and prevents it from taking 18 seconds: tfrldexp.patch.
Edited by Rika