I haven't looked, but I think Ray is probably right. How would you simplify foo(1.0b0)+ foo(3.0b0) +foo(2.0b0) ? you need to compare the bigfloats, right?