MATHLIB: mathlib powdp() runs slower on the TDA4VM C66 than on the CPU, does it normal?

My test code:

double a = 142.224389823827;
double b = 0;
for (long i = 0; i < 100000000; i++)
{
    b = powdp(a,2); // a^2
}

This code takes 20 seconds to run on the TDA4VM C66 and 4 seconds on the TDA4VM CPU.

is this normal?