This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
floating point decimal to binary conversion errors
- From: Stephen L Moshier <steve at moshier dot net>
- To: gcc at gcc dot gnu dot org
- Date: Thu, 3 Oct 2002 16:49:45 -0400 (EDT)
- Subject: floating point decimal to binary conversion errors
The method that gcc uses to convert decimal strings to floating point
binary numbers has been changed. A cursory test shows differences
that did not occur before. gcc previously had high enough accuracy
to get powers of ten correct.
Compiled value of 1.0e-4667
036f79d3 a800edaa ca757023 547080fa (quad precision, sparc)
should be
036f79d3 a800edaa ca757023 547080f9
Compiled value of 1.0e3426
6c73e648 82f8b0f7 3d9f3d45 c00ced47
should be
6c73e648 82f8b0f7 3d9f3d45 c00ced48
Compiled value of 1.0e3952
73473284 c586d770 4095c912 924c3ab8
should be
73473284 c586d770 4095c912 924c3ab9