Sylvestre Ledru <sylvestre.le...@inria.fr> writes: > This code will set the precision to double: > > #include <fpu_control.h> > fpu_control_t _cw; > _FPU_GETCW(_cw); > _cw = (_cw & ~_FPU_DOUBLE) | _FPU_EXTENDED; > _FPU_SETCW(_cw); > > You should get the same result on 32 & 64 bits CPU then.
Quite off-topic, but as far as I remember from GNU Fortran days, if you want to avoid `excess precision' on an x86 system, you generally need to spill registers with (the equivalent of) GCC's -ffloat-store, and take the performance hit. Most of the time you don't, of course, and the Paranoia test, for instance, passes without -ffloat-store.
