and ep values. Apparently, gcc -0 generates code which causes a SIGFPE
when copying a double if the double has some strange form. By initializing
the sp and ep, we are guaranted the x and y will either be 0 or some
reasonable floating point value. See Lefty for more details.