Sign-extension works this way: You pad with zeros for a positive number and pad with ones for a negative number, because of two's complement, and left or right depends on endianness.
Did I understand it correctly? Do I need to know more if I'm looking to understand it as it is described e.g. in Pattersson (Comp. Organization)?
Sign extension is a block that takes in your input data and append bits to it based on the MSB(most significant bit) value to maintain sign integrity. It has nothing to do with endianness. Endianness is related to how data is stored in memory and varies from one architecture to another. Also sign extension usually appends bits to the left of the input data.
Sign extension has nothing to do with endianess. The reason you perform sign extension is usually to fit an otherwise smaller bit-size into the required size , to maintain sign etc.
The concept behind sign extension is very simple, any twos-complement representation can be thought of having an infinite number of insignificant digits in the front after the actual MSB. For a positive number leading zeros are insignificant, ie having any number of them does not change the value of what you are trying to store . In the same way leading 1s are insignificant for a negative number stored using twos complement. Try it out yourself.
