We study several variations of the Bitran–Hax variable fixing method for the continuous quadratic knapsack problem. We close
the gaps in the convergence analysis of several existing methods and provide more efficient versions. We report encouraging
computational results for large-scale problems.