|
On 20 Sep 2010, at 09:18, David Kastrup <dak@gnu.org> wrote:
There are lots of subtractions in the code.
I think the original poster might be able to explain the problem (or rule out some explanations) by looking at their floating point environment in more detail. IIRC common implementations of floor() alter the rounding mode in order to work, so you might need to test it in more than one place to be sure everything is as expected. It would be nice if you could try to help instead of flaming. |