Answer from cs61c-el (David Tung 16479725) for Question 2
A subnormal is a way to represent numbers between the smallest normalized floating point number (1.0000... * 10^smallest exponent) and 0.  It does this by not assumign that the values of the significand have a leading one, thus extending the number of powers of two we can represent by the number of digits that can be stored in the significand.  This just allows us to get a more gradual progression to 0, although they are not always implemented in most software.