[Boost-users] Comparing boost::multiprecision cpp_int and mpz_int

cpp_int is more convenient but mpz_int is faster especially for long numbers? What say benchmarks? For mpz_int I must install GMP on Linux or build MPIR on WIndows (build lib_mpir_core2) ? I want make sample benchmark test but I can't convert cpp_it to mpz_int nor iport bits, whereas for cpp_int I can import_bits(x, xBytes.begin(), xBytes.end()) where vector xBytes; Question: is possible import byte array to mpz_int?