* libm/math/e_pow.c (__ieee754_pow): Fix case whereby x is close to -1.0 and y is very large to use ax (absolute value) instead of x. * libm/math/ef_pow.c (__ieee754_powf): Ditto.