The question doesn't make sense for types that are defined in standard libraries, because Altera can't use a format at will. IEEE VHDL libraries specify explicitely two's complement for SIGNED type. In case of doubt, you can refer to the VHDL standard and the library code.
Consequently, Altera is using the same numeric representation for SIGNED numbers, that are interfaced by an untypified bit vector, as in numeric MegaFunctions:
--- Quote Start ---
The signed representation for all library of parameterized modules (LPM) megafunctions is two's complement.
--- Quote End ---
Most users think it's obvious and don't ask about.
