Matlab code implementing new method to evaluate efficiently the Faddeeva function of complex argument (and hence the complementary error function of complex argument) by a modified numerical quadrature which has guaranteed accuracy < 0.7 exp(-N pi) in exact arithmetic, where N+1 is the number of quadrature points.
This method is described in a paper recently accepted for SIAM Journal on Numerical Analysis. The Matlab code achieves an accuracy of <1.3e-15 relative and absolute error in the first quadrant of the complex plane with N=11, and numerical tests suggest the method is at least competitive with other existing methods. See https://github.com/sms03snc/Faddeeva#readme for details.