[thesz]

Adds are cheaper only for fixed-point computations. Floating point addition needs to denormalize one of its' arguments, perform an (integer) addition and then normalize the result.

Usually FP adds take a cycle or two longer than FP multiplication.

[dahart]

Depends on what you mean by ‘cheaper’. Multiplies are still more gates. The adds are slower due to longer dependency chains, not because they cost more gates.