Is this a bug in C64x+ IQMath Library (IQmath_pc.lib)?

Y Liu said:

I found a proplem when I use the function _IQNpow()

_iq16   ta, tb, tc;
float   fa1, fa2, q, fTest;

the following code is to solve  tc = a^(1/3)

q = 1.0/3.0;
fa1 = 0.027;
fa2 = 0.028;

ta = _FtoIQ16( fa1 );   
tb = _FtoIQ16( q );      
 tc = _IQ16pow( ta, tb );
 fTest = _IQ16toF( tc );       ---------->  fTest = 3.7575836    Wrong

ta = _FtoIQ16( fa2 );   
tb = _FtoIQ16( q );      
 tc = _IQ16pow( ta, tb );
 fTest = _IQ16toF( tc );       ---------->  fTest = 0.26564026  Correct

I found that for tc = a^(1/3)
if a >= 0.028, we get right results
if a <= 0.027 we get wrong results

Anyone can help?