This cleans up the compiler options since we added software float64 support to reduce some duplication.