How does floating point number recode in detail? Is there any document or reference or why it needs to be recoded?
ucb-bar
