VOIGTL: a fast subroutine for Voigt function evaluation on vector processors. Voigt function evaluations account for a significant proportion of computer time used in calculating synthetic stellar spectra, particularly if local thermodynamic equilibrium is assumed. A vectorised Voigt function evaluation routine is presented which, in typical applications, achieves a speed of 180 Mflops on a CRAY-XMP and is accurate to 13 decimal digits. A derived formula gives an order of magnitude estimate of the Voigt function absolute accuracy required for synthetic spectrum calculation, based on the signal-to-noise ratio in observed spectra (with which comparisons would be made), the number of spectrum lines included in the synthesis at some typical wavelength and the number of lines omitted from the synthesis at this wavelength because of truncation in their far-wings. Limiting the accuracy (by simple code changes) to the minimum necessary, minimises the required number of floating point operations and maximises the speed with which Voigt functions are evaluated. (Source: http://cpc.cs.qub.ac.uk/summaries/)