Peter, On 2005 Mar 11 , at 15.08, Peter W. Draper wrote: > On Fri, 11 Mar 2005, Norman Gray wrote: > >> I'll try and work out some portable way of checking whether >> denormalized numbers generate FPEs at configure time, and >> conditionalise on that appropriately. It'll presumably mean that the >> Linux_g77 case will end up being in the minimum-denormalized case, but >> that appears to be correct anyway. I've done this, and the result appears to be correct for Alphas and ix86/Linux. > I'd guess there's nothing more significant to all this than the copied > files worked! Don't you just love magic.... I realise I still haven't addressed the issue of the number of significant figures required for floats and doubles. I'll sort that shortly, as I've got to come back and fix a self-test that I broke on g77. Norman -- ---------------------------------------------------------------------- Norman Gray : Physics & Astronomy, Glasgow University, UK http://www.astro.gla.ac.uk/users/norman/ : www.starlink.ac.uk