The voltage difference between 12v and 5v is quite high, with a linear regulator a lot of energy would be wasted as heat with linear regulators.
Ex with a 7805, to get 5v @ 1A you'd waste (12-5) x 1 = 7w as heat to produce 5w of power. That's less than 50% efficiency.
A LM2576 will give you about 70-80% efficiency so it would be less heat produced, less waste. Instead of 7w wasted (for the 5v@1a example), you'll have maybe 1-2w of waste.
In the case of 3.3v rail, there's two possibilities. You could go with a 12v->3.3v converter using that lym2576 but this kinda makes sense only if you need a lot of current on 3.3v
For example, if you need less than 0.5-0.75A on 3.3v, it would be much cheaper and not much less efficient to use a linear regulator to get 3.3v from the 5v produced by the first dc converter. A basic, generic, cheap 1117 linear regulator needs only 1-1.2v above output voltage to work properly so it would work fine to output 3.3v with only 5v at the input.
For example, to get 3.3v @ 0.5A = 1.7w, with 12v to 3.3v dc converter you'd use about 2-2.1w to do it... with a linear regulator you'd use 5v x 0.5a = 2.5w .. so only about 0.4w difference.
In addition, by producing the 3.3v from the 5v,you would basically give the 12v to 5v converter a sort of minimum load which will increase the average efficiency - you choose the components for the 12v to 5v converter thinking of the maximum output current but the efficiency will be best at a narrow range of output current. So it would good to have the output current above some minimums almost all the time.