On Wed, Feb 27, 2008 at 06:25:36AM -0500, Neal Becker wrote: > x86_64. I guess that's the same? No. x86_64 uses SSE2 for float and double, so there are no excess precision problems on x86_64. Jakub