On Thu, Mar 27, 2008 at 04:27:32PM +0100, Geert Bosch wrote: > >known to get these calculations wrong. Could someone with a fully > >ieee compliant FPU try this test? > > Just use -mfpmath=sse -msse2 to force use of SSE instructions > for double precision math. Should be the default nowadays, IMO.
Thanks, that works fine. (The code I posted had the wrong size for a, though.) I was searching in the gcc --help -v output before for stuff with sse, but that didn't find the -mfpmath option, since there is no description of its possible parameters, so I though my gcc installation didn't support sse math.
