Answer from cs61c-cj (Peter Lau 16539384) for Question 2
A subnormal is an "improper" floating point number.  Instead of having the significand always start with a 1. (except for 0), subnormals may have significands that start with 0.____.   This stretches the range of 32-bit floating point numbers from 1.0 x 2^-126 to 1.0 x 2^-149.  For double precision floating point numbers, the range increases from 1.0 x 2^-1022 to 1.0 x 2^-1074.