Aleksandar Donev wrote:
>
> Tim Prince wrote:
>
> > A compiler for a CPU with extra precision operations could gain that much
> > precision by optimizing inner loops.
> Sorry, I forgot that not everyone uses IA-32 processors (under Linux of course
> :-) Yes, there is the possibility that the 80-bit floating point registers
> are being used to enhance precision, thanks to several people for pointing
> this out. I will look into compiler options to see if there are some switches
> I can try to force the less-careful compiler to maximize precision...
>
Alternatively, you could try to force the "careful" compilers to use
normal precision and see if that duplicates the results of the other
compiler. That could at least tell you if you might be on or near
the right track.
Dick Hendrickson
> Thanks,
> Aleksandar
|