forum.alglib.net

Using alglib.polynomialsolve from C# runtime version 2.0.50727:

Polynomial coefficients:
1
0
4
0
6
0
4
0
1

Actual roots are (0, +/-1) x 4 but calculated roots are:
-7.74748397136954E-05, +/-0.999922521609102
7.74739404654778E-05, +/-0.999922520701244
-7.74828498355251E-05, +/-1.00007747929019
7.74837490837565E-05, +/-1.00007747839946

This is not a completely artificial test, this polynomial cropped up when designing a 4th order Butterworth bandstop filter.

Is this the expected accuracy of the polynomial solver?