Is “long double” type on Apple M1 processor + Xcode implemented as fp64, fp80 or fp128?

Question

The new M1 processor seems quite good in emulating x86 processors, including floating point computations. The long double type on x86 is fp80, that means it uses 10 bytes of mem